I was right on the money: cdn.mos.cms.futurecdn.net/6Yug4edRzUohX5TrjodbSW-1200-80.png.webp Just get ready for my video tonight. I've never seen a company self-own a good product so bad like AMD just has.
@@MooresLawIsDead claiming you were “on the money” is an insult to every one of your viewer’s intelligence. Absolutely biblical levels of spin zone here.
@@MooresLawIsDeadTom’s Hardware is reviewing Zen 4 with lower memory 5200 mt/s and Zen 5 with 5600, and with PBO, they use 6000, you can even see it on the same chart you are providing that they consider PBO overclocking so they turn EXPO for just Zen 5. Zen 4 is handicapped compared to Zen 5 so you shouldn’t even be using that review as to prove a point. That review you claim to be right on the money is flawed, no matter how you dice it.
The craziest number I saw was Cinebench multi-core @ 60w. Zen 5 gets about 24k, which puts it toe to toe with M3 Max @ 80w! And at 80w, Zen 5 gets 29.4k! I'm sure it pans out in Apple's favor when you compare full system power, but that a chiplet, DESKTOP x86 beats out a monolithic ARM chip WITHOUT 32-BIT support at the OS/hardware level is kind of insane. edit: fixed numbers, thanks @askmedov
@@tomhsia4354 then I guess after 2 years of blueballs, I should be glad AMD forced me to wait by only letting garbage Asusss make Zen 4 X3d laptops?....
Exactly. This guy can't be taken seriously since he has AMD shares and has admitted to that fact. Reporting according to MLID is bashing Intel in every video because the AMD fans cheer him on.
@@runninginthe90s75 Nah I wouldn't say that. Also owning AMD stocks is meaningless, he doesn't have large enough following for his words to have any effect on the stock price. But it's quite ridiculous these confident performance claims less than 1 month from product release, seeing them fall flat, and then you see a video claiming what Zen 6 is going to be 2 years from now. To be fair, Zen 5 can reach higher performance (with PBO) than what AMD gave us out of the box, but its definitely not 23% faster than Zen 4. Ignore the "tech rumors" and just wait for product release and reviews, it's the same story every product cycle.
Wow! Tom you really outdid yourself on this one. Complete miss across the board from you here. Don’t think you came close on a single performance claim you made, even after you adjusted them down substantially. Jesus Christ.
@@Dante_S550_Turbo 25% uplift in gaming 🤦♂, you might want to rewatch the video you are referring to there buddy. Infact not any of the games were close to that
'why don't they just push it harder like intel' My initial thought was a mental image of Steve from GN pointing at a headline about Intel's degredation issues on a slide and saying 'That's why!' In all seriousness, the biggest brain thing AMD has been doing, on both GPU and CPU side, is focusing on efficiency instead of Tim Tayloring their stuff as stock settings. We've seen with both Nvidia's gpus (3080, 3090, and 3080ti popping from transients) and Intel's cpus (degredation) that it tends to not work too good. Even AMD got a bit of taste of it when they were optimistic and assumed they didn't need to tell mobo manufacturers that flooding the voltage was a bad idea.
the degradation is not just power consumption. it's that plus the Intel ring bus plus their cache structure. which means AMD would never have this problem.
@@lilnapkin462 That's true but it could still give consumers a bit extra peace of mind. The "Intel pushed it too far" message has been the main narrative for a while now, and AMD deliberately not doing that is a free PR win when they can easily compete regardless.
its not the voltage that kills or degrades the cpu's its the current. also intel giving the green light on removing all power limits is absurd and honestly is just done out of laziness or rather "cost savings". i personally feel like mother board bios's are so cut short most of them are just flat out broken and do not function the way they should or even intended to. ie the customer needs to fuck around and find out what bios is the particular bios the mobo vendor actually spent time getting right and not just use the most recent bios for your board. to me thats pretty fucked up. current limits should be the ONE thing that actually works properly or at the very least the avx offset. another issue is everyone things LLC is a good thing but in reality is the WORST thing you could do to a cpu with out current limits. why? because overshoot happens. and is likely the actual cause of chips getting cooked. there are many things that use avx now and if LLC see's high load it rockets voltage to the cores when they need to be lowering the voltage for that workload. do that a few million times and ya degradation is going to happen.
@@ramair325 I don't think you understand electricity, if you are trying to claim voltage isn't involved in degredation. Amperage is how much water is moving through the pipe, Voltage is how much pressure is pushing water through it. If you can't understand how the water being under higher pressure effects amperage, well, then maybe not try to comment on electronics part degredation. Because you are currently basically showing ignorance of Ohm's law. Voltage determines amperage (or 'current' as you refer to it), modified by resistance. Literally V div R equals A.
@@jtnachos16 sorry i should have said amps is what kills. your right current is just the flow but more flow equals faster erosion, from the resistance (amps) but again voltage is not the actual killer here. its the resistance. voltage is just the speed and yes that can ofc effect the heat too. but people think high voltage kills but thats not really the whole story, its the voltage with the load that kills. voltage alone is not enough unless your going extremely out of spec and then it takes very little load to kill. as the voltage increases resistance and depending how the gates are designed will determine how much voltage can be safely used.
"People will reward AMD more if they have better efficiency". I can confirm it. In my country, no tech channels were talking about that it's worth getting Zen4, until 7500f and 7700 arrived. Now AM5 a huge hit, as for ~100-170$ you get modern CPU that you can put in any cheap a620 or b650 mobo. 7800x3d also gets talked about, but for the price of three 7500f's it is a rather tough sell to most customers
X3D has always had a premium price. It's not a bang for buck chip. AM5 was always worth it if building new system. AM4 was cheaper but was slower and then you're stuck on dead platform.
Hard to say what Intel will do. With Pat in charge (and I still think he's a good CEO for the company) and so many competitors everywhere, they really need to focus on something that can get them some wins in the market. They have so much money, but in the tech world a few lagging generations can be fatal. AMD barely survived Bulldozer. Intel has deeper pockets, but it isn't 2009 any more.
They have to focus on better yields for their process. Apparently they are wildly overbinning (worse than N3B where like 80% of M3Maxes aren't golden sample 40 cores and the 2x die aka ultra fell out) and their design team is confused on what the fuck fab is doing so they can't design a proper uarch.
@@guidedorphas10 I've always used intel CPU's but my new PC that i will be building at the end of this year will most likely be AMD CPU with things looking like this
@@reahs4815 The first computer I built for my personal use had a Cyrix processor. This was them upgraded to a Intel, same motherboard though. I'm tying to remember if there were a AMD in between. I think I still have that PC in the basement though it haven't been powered in probably 20 years. I worked a lot with AMD at the computer company I worked for. But we skipped right into the Opteron family and more or less forgot about the earlier processors as they didn't perform well enough. Opteron was an interesting processor, and the biggest server using it I built had eight of them. Dream Hack borrowed that and ran gameservers on it for one hack. Then they asked if we could give it to them, but we were a relatively small company and that single server was worth about $10,000 at the time, so that was a no go.
Still wild to think that AMD was sooo close to bakruptcy, then played the ZEN Uno reverse card and now we are talking about Intel falling more and more behind. Wild.
The truth is that Intel is not lagging behind AMD, their architectures are still competitive with AMD even though it is manufactured on 10nm++++ that's the problem, if they feel they can't compete then just put their chips in. Let TSMC process it on the 5nm process and you will see a powerful Intel beast that doesn't heat up and doesn't need to push high consumption to be on par with AMD. But doing so would mean that Intel would have to temporarily abandon its chip foundries, which would be a huge waste and this scenario is very unlikely to happen.
@@chien.nguyen. believe it or not, 10nm+++ is about the same as tsmc 5nm in size and power. nanometers just became a marketing term. The problem is that intel is pushing their cpu 's to the limit to keep up with amd. That makes them use twice the power and heat like a toaster.
@@Garcheezy Intel 10nm is comparable to TSMC's 7nm products, TSMC N7P is slightly lower density and N7+ slightly higher, the reality is that Intel Is pretty far behind, and their unreleased "4nm" process is comparable to TSMC's current 5nm.
Always was an Intel person but with the current on-going issues with their 13th and 14th gen CPUs, and AMD's super impressive X3D CPU's performance in gaming - it's going to be AMD for my next build unless Intel manages a miracle.
Yeah same here, I made a 13400 3060ti system late 2022 for a good deal but once the x3d chip comes out for zen 5 im going to do a full upgrade as i cornered my self onto a DDR4 board at the time. Edit: will upgrade to 7800xt equiv or newer
Lots of equivocation going on here. No way it makes sense for AMD to sandbag on Zen5's potential. First impressions matter, and so far Zen5 is getting slammed by reviewers. Why would AMD want that? No. If Zen5 with a bit more power can beat Zen4 by ~15%, then it would have been released in that form. Plus, where is AMD's response to all the bad Zen5 reviews? Are they hiding?
Even if Intel outpaced AMD by a full 15%. How much of that would be given up by degradation, bios, security, microcode updates. Legitimately where does the IPC for Raptor Lake launch day even stack up against IPC for Raptor lake today? Downfall, Spectre was pretty much swept under the rug never to be mentioned again.
if you watch hardware unboxed original video on the 7800x3d vs now, it went from like 1% faster in gaming to 3-5% overall now vs the 14900k and this is before intels baseline profile thing, so its prob even worse now
@@NadeemAhmed-nv2br That can also be about improvements to 7800X3D performance due to AGESA updates over time. One problem AMD STILL has but who knows maybe Zen 5 will be a first, is putting out products where the software to run them still needs work. It's not TERRIBLE on the CPU side, on the GPU side it's my biggest issue with AMD, because with their GPUs you can end up with avg uplifts sometimes of around 10%, which is nice and all but is SURE would have been good if that performance was there at launch. I mean sometimes in particular games or a particular API like DX11, Vulkan they'll put out a new driver and the uplifts are 10+%.
Intel has a lot of money and top tier engineers. Their problem is they also have top tier arrogance. I spent a year working for them during the P4 debacle. When Athlon was moping the floor with their products. The attitudes that their management would bring to the table was insane. Most were unaware of anything but clock speed. Those that were made it sound like their profit margins were so good that it didn't matter. In the end of that era their stock took a dump. The only thing that saved Intel's bacon was a group of engineers that continued to develop an alternate core based on the P3 coppermine that evolved into the Core solo and Core Duo lines and caught them back up while AMD squandered their lead.
yall just come from a reddit group and spam this shit? der8auer got 25% increased performance in gaming when he o.c. 'ed it. How is that not major headroom/ holding their own chip back? When was the last time you could o.c. something for 25% more gaming performance and 40% increase in productivity benchmarks? Oh right never, but lets spam "aged like milk." The only thing aging like milk is amds software side, it never schedules with windows properly and their drivers are like fucking russian cruise missiles, sometimes they take themselves out.
i built my son a computer with a first gen i5 running a maxwell titan. gpu does ALL the heavy lifting on pcie 2.0 as well haha. ill let him save up to build something new, or maybe when i upgrade he'll get mine..
@ProsperinoG Never!!! The glory of AM3+ lives on. Pile driver will forever be immortalized in Linux Bazzite OS. SSE4 and lackluster IPC lives on with my janky PCIe adapter NVME SSD nutured at 1280 MBps read/write. I will never leave.
I was in a Best Buy about a month ago and had a conversation with one of the salespeople about Intel vs AMD. I definitely have the impression from what I've seen that AMD is the best value, and that Intel with its problematic big/little core stuff and typically higher power consumption was to be avoided. He was like, oh no, Intel is much better, much to my surprise. So when you show the marketing materials supplied to retailers at [14:20], maybe I understand now how this person got this idea. Or maybe they just have a lot of Intel inventory to sell....
Just keep it in mind that Intel getting sued and fined for bribing retailers was only about about preventing Intel from continuing to bribe retailers, it came with no additional stipulations or regulations on the retailers themselves to stop favoring sales of Intel products. While there's not definitive proof outside allegory and hearsay, I still have no doubt there's an unwritten "gentlemen's agreement" in place between retailers and Intel.
Seems like AMD is not only giving us double the integrated graphics grunt in laptops, but efficiency that gives the Apple M4 Pro & M4 Max and maybe even the M5 a run for the money too!
@@vyor8837No. The M4 is a significant redesign that gives better performance and better efficiency. And M4 Pro/Max haven’t even been announced nor leaked. No one outside Apple (and maybe TSMC) knows the number of cores those will contain.
He's not suggesting 20%+ IPC He's suggesting that they've only really showed us the IPC difference and are hiding the boost characteristics, which also have a solid improvement. The gains suggested are IPC + PBO boosting
@@ThexXxXxOLOxXxXx I understand that it would be better if the gains were all IPC. I'm just saying that's not what the video suggested. Also, clock speed and heat are not directly related. Wattage and heat are. Above 5GHz on current gen is often a lot of heat to get rid of. We don't know how next gen behaves yet. Early info is suggesting Zen 5 will be hitting same clocks at lower power aka lower heat released. You can take an old architecture and process node and make it hot at 1GHz. Clock speed and temp are not directly related, especially not across generations.
@@TheDarksideFNothing I'm just getting tired of 6+GHz and insane cooling requirements, a focus on IPC while keeping clock speeds high, not extreme. As for Clockspeed pretty sure that's correlated with high voltage/heat My perfects vision would be a 16 core cpu no multitreading running at stable 5GHz no need for Ecores just keep it traditional, Ecores are not for the normal consumer it's commersial server stuff
@@ThexXxXxOLOxXxXx there is a correlation with voltage and clock speed to power usage within a given architecture. But we are talking about an unreleased architecture, so that correlation will be different from what we have now. Different architecture and different process node, so it will take a different amount of power to hit the same clock speed. Power versus heat is always a direct correlation regardless of architecture or process node. If you could make a processor that hits 8 GHz at 20 watts it is still only going to produce 20 watts of heat. I think we are just having a language misinterpretation rather than either of us not understanding how it works. I'm only trying to point out that Zen 5 might hit 5.5 GHz without having an insane power target. Fortunately, you can always just buy the same CPUs as everyone else and undervolt/underclock it to hit the heat targets that you're looking for. In most modern processors They are really pushing the limits of the architecture, so you can drop half the power and lose way way less than half the performance. You give up a little bit of performance, but if you want less heat find the point on that curve that you're comfortable with. Unfortunately, It's not like in the drawing room they could have just decided to make a product the same price and same performance while using significantly less power and therefore generating less heat. I do agree that I would like to see lower power targets in general, but the reason we aren't is because both companies are fighting for performance superiority and they have already pulled every other lever. The last lever left is juice the hell out of it.
der8auer was getting 25% performance on the 9700x when o.c.'ed in gaming and up to 40% in some productivity benchmarks. that's higher than mlid results....
When AMD was behind with FX at least they could compete on price (I remember getting an FX-8320e and board for just $150, less than the cost of an Intel i5 by itself) and they had the console market as well, but I don't see how Intel is going to compete on price if AMD is able to make 'em even cheaper. Hell even the ultra-budget baller segment is locked down thanks to AM4 getting new chips until 2025 so why would I want to buy an Intel PC and risk my CPU going bye bye when I could get AMD and get lower power and great performance too?
I am more on finance side... the rationale for keeping stock voltage low for AMD could be total cost of ownership. Large corporate accounts might buy lower spec products, but imagine a small office. Lower consumption, lower consumption cooling, your electricity bills falls. AMD could simply go say "we save you in bills the whole cost of the chip".... plus, SOCKET STABILITY
and having no real VRM differences to worry about when picking a board. soo many socket 1700 boards will not run 12900k at peak because they cant provide the power...
Office PC's spend most of their time at or near idle power consumption, and every single time chiplet based AMD CPUs will lose that battle to Intel monolithic CPUs by around 8-18W depending on MCLK/FCLK frequency and vSoC. The lower the CPU utilization, the more appealing Intel CPUs become in longterm cost savings. That being said, Ryzen monolithic APUs, while often being slower than Intel counterparts (apart from iGPU performance which is something to note), are still overkill for the majority of office PC tasks while consuming the same amount of power at idle as Intel CPUs while being priced similarly. All things considered, if I were tasked with upgrading an entire floor of office PC's, I would still be looking at something like an AM4 based 5600G template for cost to performance ratio, as I prefer the flexibility that the AMD iGPU performance brings to the table, and the 5600G is still massive overkill for typical office use while sipping power like an Intel chip at idle.
Doesn't matter what Intel promises. If they bring a processor that's 10% better than the last one, it will soon get a BIOS update to fix stability issues and then another one to fix safety issues and then it's only maybe 1% better. All while consuming ridiculous amounts of power. Yet people still buy those.
agreed. that's the sad thing isn't it. Laypeople would buy the most well-known brand, even though generally speaking the well-known brands are often inferior.
More like Dell & friends still buy those. The biggest purchases are managed by tech-illiterate ancients who listen only to salespeople and MBAs, not specialists.
Can confirm. My Dell work PC has a 10700 and it started out as a 16 thread CPU, and now it's 8 thread plus runs slower. I think it actually may be slower than the PC it replaced at this point, or it's pretty marginal at any rate. And on our Intel-based hypervisor servers I've had to allocate more cores to VMs to compensate. Needless to say our newest machines are EPYC. Zero issues with those.
You are missing the bigger picture. Intel is a vertically integrated fab company that cant compete against any of the rent-a-fabs at this point. All they got now is brand, and they just killed it with these degrading chips. Within the next 10 years Intel will spin off its fabs in order for the ic design side to be acquired.
@@styleisaweapon however Intel got more handouts this year alone from Uncle Sam's corp brothers than TSCM ever got from not just Taipei, but also China and Japan. Will they be able to do something out of that or is it just a way to push lobbying in an election year? Stay tuned
99.9% of people don't even know there's an instability issue with some of the i9 chips of the past 1.5 generations. The echo chamber of the inner tech world doesn't extend far.
@@styleisaweapon Well sorta yes, sorta not really. One of the big selling points of being a fab company on top of being a chip manufacturer, is that theoretically it would allow them a major competitive edge both in terms of specific node integration to serve their own unique needs first and foremost, while also bypassing most of the logistics and supply chain aspects -all while being able to produce own chips at a cost-price rate which no rent-a-fab company can compete with. It also provides manufacturing certainty for their customers, that they know Intel chips will always have first right to the fabs even in a situation of shortages or instability. For them, security of consistent product delivery means more than a per-generation uplift. The real problem Intel has had up until now is that it takes a looong time to pay off the fabs and start seeing the positive benefits. These companies, especially when they move into manufacturing at the foundry level, they don't look at benefits over 5 years. They look at it over 10+ years. The main thing that is kicking Intel right now, is that they acted like an ignorant monopoly from 2010 to 2020 and never believed AMD could truly compete on anything except price. One of the big differences between Intel and Nvidia for example, is that Nvidia never stopped and assumed competition couldn't hurt them. Sure their prices are outrageous, but they never stopped innovating, they never pretended like they couldn't bleed. Intel has been acting like a pretentious Greek god thinking they were above it all.
Exactly the right thing to do for AMD. I'll take the efficiency over the last 5% of performance every day! Also, just enabling PBO is a one click solution anyway, for when you really really need it. Comparing to Intel Baseline is fair game, as well, since they're using baseline for their CPUs, too. And one last remark: Coming up with ridiculous marketing slides has been Intels go-to reaction for years now. Doesn't surprise me at all.
agreed.. i dont think many people understand that what they do on a day to day basis will run with MUCH less power than the maximum that all these new cpus can pull. i run a 10850k that ive intentionally pulled back to 4.8ghz under boost and its plenty. fighting over the top of the line stuff really doesnt mean anything to me, and probably a ton of other people.
CPU staying under 100W in all conditions at base settings (OC is another can of worms for those who need it) is very important to me because it means you can easily buy a quality aircooler that will keep it silent for the life of the system without having to worry about it. From my limited anecdotal observations it also means less paste degradation and needing being changed over time. My 4700K lasted 11+ years on first paste without any change to perf or temperature behavior and that was also a pretty cool CPU in the grand scheme of things. My 7800X3D obv doesnt need changing but considering I got it on a fire sale I might actually be able to swap it for a 9800X3D at almost no cost nor heat negatives. I'll wait to see where the RAM situation also ends up being.
What annoys me about Intel is how they pitch a CPU as having a couple dozen cores or so, but on closer inspection it turns out only a few of those cores are proper cores, while the rest are kind of nerfed quasi cores. I think if AMD did that, the shizz would hit the fan with people online calling them out and maybe even class action law suits etc.
I don’t see a problem with heterogeneous architectures. AMD already does a variant of that with their C cores. As long as it delivers the performance and efficiency I don’t see an issue.
@@mcalin5931Atom cores do not support the full instruction set. An app that uses instructions which are supported in the big cores is unable to use the e cores - as it will just crash if it would. Not supporting the full instruction set limits the real world use cases a lot.
I guess not even the 9800X3D will be worth buying, if the standard Zen 5 chips are no faster than Zen 4 why would the X3D chips be any faster than the 7800X3D either.
@@atnfn Amd zen 5 x3d will be another flopped garbage, the facts Amd allowed the new x3d cpu for overclocking shows zen x3d cpu will be slow cpu, barely an improvement just like normal zen 5.
The 9950x is a 16-25% improvement at about the same power or a little less. The 7900xtx is around a 50% improvement at the same power draw as the 6950xt, yet that's a disappointment.
RDNA 3 wasn't a disappointment compared to RDNA2. It was a Disappointment compared to the 4090. It would also be nice if Radeon had actually marketing instead of what we got with RDNA 3 launch. If they can not over promise and actually have something that's performs well they CAN have a good generation
@@PineyJustice yeah the 7900xtx is an incredible product, even with the missed potential. I think a lot of this issue is the god-tier, apple-like, mindshare Nvidia has. Even if Radeon overtook Nvidia people would still find a way to say "nope, Nvidia better". lol
Zen4 had a super rough start because as I told when it launched, they disregarded everything that people liked from Zen products. Nobody's buying Zen if it consumes power like an FX did. I'm glad they learnt the lesson.
Zen wasn't really widely popular until 3000 and 5000 outperformed intel. Speed is what moved units, secondly the power between a 5950x and a 7950x are pretty close, within 30-40w I've got both, I've spent a fair amount of time tweaking and tuning both. Not a lot of difference in an overclocked 5950x and a PBO tweaked curve optimized 7950x power wise, but the 7950x absolutely blows it out of the water in speed.
It's not that they learned their lesson it's just that they are using a more efficient process node. They did not design Zen5 to be more power efficient all the changes they highlighted consume more power.
Idk what you are talking about The Ryzen 9 7950x pulled 220 watts at peak power The I5 14600k pulls 240 watts at peak power Where is the Zen inefficiency? They increased power by 5-10% and gave us 25-35% more performance That’s 20-25% efficiency increase
Since the Ukraine war, power consumption is a big consideration in many places in Europe. Electricity prices have gone through the roof. I actually consolidated some servers to reduce power usage.
I bought my used 5950x precisely because it is so freaking energy efficient and i am sick of hot slow intel garbage. It looks like i should consider the 9900 series sometime in the near future!
Remember that you can very easily throw a PPT limit on AMD chips in the bios. Works like a charm and can even boost FPS if you're into that. They're extremely efficient if you spend those 5 minutes in BIOS to reign them in. It's well worth it.
The mistake is ever buying the 14900K in the first place. I know this because my 14900K lies dead on my carpet, while my 7800X3D outperforms everything thrown at it.
I care about Efficiency too. Thats why i have all Amd in my house now. I started in 2017 with a cheap Asrock AB350m pro4, and a 1700X. Now that motherboard is in my daughter pc, combined with a 5600X. I have a 5800X3D, and my son has a Amd 3100. Such massive value on AM4 platform. And i will be getting AM5 in the future
There's one hole in your argument. There's no reason to sandbag when you're releasing first. The reason you sandbag is to get your competition to overcharge for their product and look foolish when you launch. Intel will know the real numbers long before their own launch anyways so AMD has absolutely nothing to gain by sandbagging unless they're trying to manipulate their stock price.
I feel like AMD are keeping all the best chips for the 9950X3D and 9800X3D. ZEN 5 CPUs use less power and run a lot cooler compared to ZEN 4 CPUs.. so maybe the X3D variants will have some even better specs than the Non X3D variants while also having 3D V-Cache!
By not pushing the power draw of the current releases of 9000 chips, it leaves AMD headroom for any refresh next year depending on what Intel manages to get out the door. One key thing is that if AMD feels it needs to release an 8+16 core Zen/Zenc chip they have the power headroom needed to make sure that this can be fully powered.
Honestly I'm just waiting for the the Zen 5 X3D chips to significantly outperform Zen 4 X3D so that Zen 3 X3D CPU prices fall further (since I'm still on AM4). I'm not in the US or Europe so a lot of those great AM5 mobo + CPU + DDR5 memory deals simply aren't available for me.
AMD wouldn't do this intentionally. AMD flexing makes Intel make even more and more power envelope mistakes like they did already 14700K is also a ticking time bomb that will fail prematurely
When you think about it, coincidentally, Zen 5 CCD is on "4nm", while Zen 4 CCD is on "5nm" production. Personally, I find it a bit misleading to market chips based on their transistor production node size, like Intel did with their "Intel 7/5/4" rebranding, however, in AMD case I think it is an interesting observation. I think there's a chance someone actually planned for it.
For anyone who's intersted and/or considering an upgrade: I purchased an RX6700XT last year (May 2023) along with a new LG TV that supports 120Hz VRR. This year (June 2024) I built a new primary PC for myself, going for a Ryzen 7900 (No-X), I decided not to wait for Zen5 as I'm not buying the top spec and it should be good enough, the potential extra ~10-15% is not a big deal. My selection reasoning was that although I do a lot of gaming that it is not the reason for the upgrade, my old Ryzen 2700 (No-X again & not overclocked) was perfectly fine for all of the games I play (I'm 61 and my reflexes have been too slow for twitch shooters for 20+ years). The thing I do that was taking some time was processing my DVDs and Bluerays (I'm English and refuse to pay the TV TAX to fund the BBC). DVDs were not too bad with the 2700, even with a few filters such as medium denoise & sharpen for older sources, it could process the raw MKV into x.264 at about 30-50FPS, so up-to twice playback speed. However recently Bluerays have been selling for the same prices as the DVDs (eg: I got "The Man From UNCLE (2015)" BR for just £3.49 Inc P&P), and with 5 to 9 times the data (DVD 3-4GB > BR 25-29GB) it was slow and could take Handbrake half a day or more. The 7900 now processes BRs about as fast as the 2700 did DVDs! I went with the 7900 due to a stock power usage of 65W, and a provided AMD heatsink/fan for £20 les than the X version which would need an additional Heatsink... So a saving of about £50 for no downsides. I may upgrade the heatsink but it's not urgent at all. The mother board is the Asrock Lighting ATX with 32GB of Corsair 6000 (CAS 30 that was only about £8 more than the basic DDR5). At stock the 7900 is about 2.8 times the speed of the 2700 (per core), with PBO (I chose the 85C & -20mv setting from bios) it is ~3 times faster. I tried the simple "enable PBO" but that had the CPU hit the 95C limit in seconds and I didn't like the idea of that with the AMD heatsink, setting the 85C limited PBO gives me about the same performance plus a little cooler running. Cinebench R23 MT scores: default "Auto" PBO was 23545
When they run 16 core Zen 5 chiplet desktop processor at 40W, it shows slightly worse performance. But at higher watts it shows much better performance. AFAIK it’s still using the same 6nm IOD, so that brings the same baseline cost, which should bring the new/old closer to each other at very low watts. The monilithic laptop chip of course doesn’t have this ”problem”, which is relevant as they need to run it at very low watt/core range. RNDA 3.5 efficiency improvements for gaming seem super nice (32% better perf/watt). AMD spent time optimizing RDNA architecture for mobile phones with Samsung and they bring those learnings to their desktop/laptop GPU architecture. RDNA 3.5 seems much better than expected from a .5 name bump. Efficiency = performance in low power devices.
They aren’t hiding anything, PBO CO is not stock behavior. Never has been and once you enter that territory you could just OC Arrow Lake and then we just end up right at the same spot in the conversation.
@@nab-v1w Again it’s still not stock behavior, and a lot of the actual performance comes from curve optimization not PBO. Which curve optimizer is not easy and is really finicky.
What you said about X3D's reliability is legit, especially with the inclusion of PBO2 Curve Optimizer. Running a Ryzen 7 5700X3D with a Wraith Prism and Curve Optimizer at -15 and it runs cooler than my Ryzen 5 3600 under load.
It's great that people finally pay attention to efficiency. It's ridiculous to drive chips well past their reasonable range just to get 5 extra fps. Everyone who wants that can overclock themselves like in they have in the past
People will still buy Intel and they will still keep their market share more or less. It's sad. We have to buy Intel at work as the IT staff are ignorant about hardware. AMD needs to up their consumer advertising like Intel was in the 90s. AMD Inside!
Be careful what baseline you are comparing with. Zen4 (X variants) is tuned up last minute at launch to match 12/13 gen perf. That doesn't mean we should accept 105W tdp for a 6 core cpu (7600X) to be normal.
Yea each gen is getting better simultaneously at speeds, thermals and OC'ing ability. I love my 7950X3D it runs cool and OC's hard AF. Those performance improvements on 9000's are going to be huge additions.
AMDs Zen tech team that does OC have stated PBO curve has been heavily refined, and customers will be able to tweek it more to get better performance when doing certain jobs/tasks.
14:30 Oof Apple comparing their M3 to the M1 in presentations is one thing but saying "hey our new CPU beats an ARM chip from 2021" is not what I call confidence.
It's just not the PBO overclocking. If it can perform better at higher clocks there is good chance it can do the same as undervolted meaning you can squeeze out better performance when undervolted (it will boost higher clocks with lower voltages).
I like how alot of channles are saying amd is fudging their proformance and will be slower..and here hes showing they might be faster ..it will make life way more fun soon
Would Arrow Lake be able to OC well? If they're regressing clocks and giving it far less power than before then OC to OC, Zen 5 might come very close or possibly on top in some things.
i agree 100%. i think Zen5 is going to blow a lot of minds. can't wait for them to name this gen's threadripper the 9999x. i am absolutely buying one if they do, and probably still buying one if they don't lol.
transistor counts are hard to quantify since they're all kind of funny numbers outside of raw sram test chips. redundancy, I/O, and fast logic lowers density a lot, so what happened here is probably all of these levers got pulled at once because yields and clocks were beyond expectations.
Intel is so doomed, I can't believe they won't be able to put their sh!t together for arrow lake, and now we have the stability issues and probably a class action lawsuit when more people find out their processors are degrading and crashing, thanks Intel.
Not sure if you covered this but any idea if Zen5 X3D will actually get dual chiplet vcache? Awhile back the same birdy that hinted to me that memory settings wouldn't be locked at runtime also suggested this being a possibility. I'm getting skeptical but they have been taking more jabs at Intel's scheduling challenges. If they really are able to get the boost clocks high enough to justify it then that explains some of the sandbagging. The other reason I'm skeptical is it would probably make more sense to save this for Zen 6.
Yeah so arrow lake will be caught up with the 13 14 Gen debarkle even if it is fine.. how will people know.. amd will sell alot of 9000 cpus because of the loss of trust a cover up will damage intels rep.
I bought my first ever AMD CPU - 7800X3D and it's a totally amazing CPU. It's driving the easily best tower I've ever had. I would be perfectly happy never buying Intel again despite loving that Intel has re-entered the GPU foray at least in theory and see AMD as the PC leader for the foreseeable future.
I have been an Intel Guy for over 20 years and the few times I did work with AMD CPU (the bulldozer and phantom era) they were crap. But my next build scheduled for end of July 24 is definitely going with either the 9700x or the 9600x depending on the review and cost. It's time to give AMD a chance to prove they are better long term than INTEL and for INTEL to get their house in order and really deliver a knockout with Arrow Lake and beyond.
Very curious to see what a cpu like the 9700X can do with PBO 2 and Curve Optimiser. Add in a better memory controller and the difference in performance can be quite substantial. 7700X here heavily tweaked waiting to see if it’s worth upgrading.
[SPON: Thanks to Vanta for sponsoring this video! Get $1,000 OFF Vanta: vanta.yt.link/JFe5YbT ]
Well this aged poorly. 😅
I was right on the money: cdn.mos.cms.futurecdn.net/6Yug4edRzUohX5TrjodbSW-1200-80.png.webp
Just get ready for my video tonight. I've never seen a company self-own a good product so bad like AMD just has.
@@MooresLawIsDead Of course, you're always right, clown. 🤡
@@MooresLawIsDead claiming you were “on the money” is an insult to every one of your viewer’s intelligence. Absolutely biblical levels of spin zone here.
@@MooresLawIsDeadTom’s Hardware is reviewing Zen 4 with lower memory 5200 mt/s and Zen 5 with 5600, and with PBO, they use 6000, you can even see it on the same chart you are providing that they consider PBO overclocking so they turn EXPO for just Zen 5.
Zen 4 is handicapped compared to Zen 5 so you shouldn’t even be using that review as to prove a point. That review you claim to be right on the money is flawed, no matter how you dice it.
@@MooresLawIsDeadbeing truthful is way better than hyping up amd fanboys
The craziest number I saw was Cinebench multi-core @ 60w. Zen 5 gets about 24k, which puts it toe to toe with M3 Max @ 80w! And at 80w, Zen 5 gets 29.4k! I'm sure it pans out in Apple's favor when you compare full system power, but that a chiplet, DESKTOP x86 beats out a monolithic ARM chip WITHOUT 32-BIT support at the OS/hardware level is kind of insane.
edit: fixed numbers, thanks @askmedov
Also M3 is N3 node, whereas Zen5 is N4
And it's 29K score, not 25K
Zen 5 Mobile chips will be absolutely insane.
@@tomhsia4354 then I guess after 2 years of blueballs, I should be glad AMD forced me to wait by only letting garbage Asusss make Zen 4 X3d laptops?....
@@greebj Oh you poor thing
M3 max costs 5 times what zen 5 costs and its on the shitty closed apple eco system. Zen 5 wins all the way.
Thumbnail: 23% faster
Reality: 5% faster
Oh what would you know, MLID pulling performance claims out his ahh case No. 984984
Exactly. This guy can't be taken seriously since he has AMD shares and has admitted to that fact. Reporting according to MLID is bashing Intel in every video because the AMD fans cheer him on.
MLID is getting paid by Amd to overhype this flopped garbage Zen 5, that's why he keep making non sense claim about Amd cpu.
@@runninginthe90s75 Nah I wouldn't say that. Also owning AMD stocks is meaningless, he doesn't have large enough following for his words to have any effect on the stock price. But it's quite ridiculous these confident performance claims less than 1 month from product release, seeing them fall flat, and then you see a video claiming what Zen 6 is going to be 2 years from now. To be fair, Zen 5 can reach higher performance (with PBO) than what AMD gave us out of the box, but its definitely not 23% faster than Zen 4. Ignore the "tech rumors" and just wait for product release and reviews, it's the same story every product cycle.
@@noxar8451 True, but his motivation for hyping up AMD is clear.
Wow! Tom you really outdid yourself on this one. Complete miss across the board from you here. Don’t think you came close on a single performance claim you made, even after you adjusted them down substantially. Jesus Christ.
Spoiler they did not sandbag anything.
LOL. This video aged like fine milk
der8auer o.c. ing and getting a 25% uplift in gaming and 40% in productivity benchmarks on the 9700x says yes.
@@Dante_S550_Turbo 25% uplift in gaming 🤦♂, you might want to rewatch the video you are referring to there buddy. Infact not any of the games were close to that
I stopped watching this guy cause he's been wrong like 90% of the time
This didn't age well. 😂😂😂
'why don't they just push it harder like intel'
My initial thought was a mental image of Steve from GN pointing at a headline about Intel's degredation issues on a slide and saying 'That's why!'
In all seriousness, the biggest brain thing AMD has been doing, on both GPU and CPU side, is focusing on efficiency instead of Tim Tayloring their stuff as stock settings. We've seen with both Nvidia's gpus (3080, 3090, and 3080ti popping from transients) and Intel's cpus (degredation) that it tends to not work too good. Even AMD got a bit of taste of it when they were optimistic and assumed they didn't need to tell mobo manufacturers that flooding the voltage was a bad idea.
the degradation is not just power consumption. it's that plus the Intel ring bus plus their cache structure. which means AMD would never have this problem.
@@lilnapkin462 That's true but it could still give consumers a bit extra peace of mind. The "Intel pushed it too far" message has been the main narrative for a while now, and AMD deliberately not doing that is a free PR win when they can easily compete regardless.
its not the voltage that kills or degrades the cpu's its the current.
also intel giving the green light on removing all power limits is absurd and honestly is just done out of laziness or rather "cost savings".
i personally feel like mother board bios's are so cut short most of them are just flat out broken and do not function the way they should or even intended to.
ie the customer needs to fuck around and find out what bios is the particular bios the mobo vendor actually spent time getting right and not just use the most recent bios for your board. to me thats pretty fucked up. current limits should be the ONE thing that actually works properly or at the very least the avx offset. another issue is everyone things LLC is a good thing but in reality is the WORST thing you could do to a cpu with out current limits. why? because overshoot happens. and is likely the actual cause of chips getting cooked. there are many things that use avx now and if LLC see's high load it rockets voltage to the cores when they need to be lowering the voltage for that workload. do that a few million times and ya degradation is going to happen.
@@ramair325 I don't think you understand electricity, if you are trying to claim voltage isn't involved in degredation. Amperage is how much water is moving through the pipe, Voltage is how much pressure is pushing water through it.
If you can't understand how the water being under higher pressure effects amperage, well, then maybe not try to comment on electronics part degredation.
Because you are currently basically showing ignorance of Ohm's law. Voltage determines amperage (or 'current' as you refer to it), modified by resistance. Literally V div R equals A.
@@jtnachos16 sorry i should have said amps is what kills. your right current is just the flow but more flow equals faster erosion, from the resistance (amps) but again voltage is not the actual killer here. its the resistance. voltage is just the speed and yes that can ofc effect the heat too. but people think high voltage kills but thats not really the whole story, its the voltage with the load that kills. voltage alone is not enough unless your going extremely out of spec and then it takes very little load to kill. as the voltage increases resistance and depending how the gates are designed will determine how much voltage can be safely used.
"People will reward AMD more if they have better efficiency".
I can confirm it. In my country, no tech channels were talking about that it's worth getting Zen4, until 7500f and 7700 arrived. Now AM5 a huge hit, as for ~100-170$ you get modern CPU that you can put in any cheap a620 or b650 mobo. 7800x3d also gets talked about, but for the price of three 7500f's it is a rather tough sell to most customers
do they have triple CPU slot motherboards now?
X3D has always had a premium price. It's not a bang for buck chip. AM5 was always worth it if building new system. AM4 was cheaper but was slower and then you're stuck on dead platform.
@@farmeunit hahah imagine calling AM4 a dead platform that has been in service since 2017. So much ignorance
@@clark85 I have AM4 too. Tell me when a new chip comes out for it...
@@farmeunit this has nothing to with a new chip. This isabout communication and perception. You lack it severely
amd doesnt need a biggest advantage when they have no disadvantages. theres absolutely no reason to go with intel.
Adobe suite. Intel tends to perform a little better than AMD there.
Intel have amazing idle power consumption. Only like 5w!
@@MrLittlecat123😂. I just turn of my PC when I'm not using it.
@@НААТ lol🤣, but alteast the laptop is getting benfit from that low power.
@@JorgenKremer you know what? Fuck adobe!
Hard to say what Intel will do. With Pat in charge (and I still think he's a good CEO for the company) and so many competitors everywhere, they really need to focus on something that can get them some wins in the market. They have so much money, but in the tech world a few lagging generations can be fatal.
AMD barely survived Bulldozer. Intel has deeper pockets, but it isn't 2009 any more.
Thank you for the sane comment. An internet rarity.
Thank you for a sane reply to a sane comment. Truly an Internet rarity.
They have to focus on better yields for their process. Apparently they are wildly overbinning (worse than N3B where like 80% of M3Maxes aren't golden sample 40 cores and the 2x die aka ultra fell out) and their design team is confused on what the fuck fab is doing so they can't design a proper uarch.
The Intel marketing slide at the end is... It's just sad.
Intel fanboys literally biting their fingers rn 😂
I wonder how the user benchmark is gonna react to this 💀
@@guidedorphas10they're going to cope harder😂
@@guidedorphas10 I've always used intel CPU's but my new PC that i will be building at the end of this year will most likely be AMD CPU with things looking like this
@@reahs4815 The first computer I built for my personal use had a Cyrix processor. This was them upgraded to a Intel, same motherboard though. I'm tying to remember if there were a AMD in between. I think I still have that PC in the basement though it haven't been powered in probably 20 years.
I worked a lot with AMD at the computer company I worked for. But we skipped right into the Opteron family and more or less forgot about the earlier processors as they didn't perform well enough. Opteron was an interesting processor, and the biggest server using it I built had eight of them. Dream Hack borrowed that and ran gameservers on it for one hack. Then they asked if we could give it to them, but we were a relatively small company and that single server was worth about $10,000 at the time, so that was a no go.
Still wild to think that AMD was sooo close to bakruptcy, then played the ZEN Uno reverse card and now we are talking about Intel falling more and more behind.
Wild.
The truth is that Intel is not lagging behind AMD, their architectures are still competitive with AMD even though it is manufactured on 10nm++++ that's the problem, if they feel they can't compete then just put their chips in. Let TSMC process it on the 5nm process and you will see a powerful Intel beast that doesn't heat up and doesn't need to push high consumption to be on par with AMD. But doing so would mean that Intel would have to temporarily abandon its chip foundries, which would be a huge waste and this scenario is very unlikely to happen.
@@chien.nguyen. believe it or not, 10nm+++ is about the same as tsmc 5nm in size and power. nanometers just became a marketing term. The problem is that intel is pushing their cpu 's to the limit to keep up with amd. That makes them use twice the power and heat like a toaster.
AMD invested every penny they had left into Zen. Lisa Su is a genius
@@Garcheezy Intel 10nm is comparable to TSMC's 7nm products, TSMC N7P is slightly lower density and N7+ slightly higher, the reality is that Intel Is pretty far behind, and their unreleased "4nm" process is comparable to TSMC's current 5nm.
Always was an Intel person but with the current on-going issues with their 13th and 14th gen CPUs, and AMD's super impressive X3D CPU's performance in gaming - it's going to be AMD for my next build unless Intel manages a miracle.
i got 2 chips that degraded on me, went to amd now no more issues
Imagine saying " I'm an Intel person '"
@@thomasfischer9259 Why is so difficult for you to imagine that favouritism anymore than being an AMD fanboy or a Chev or Ford Boy.
@@thomasfischer9259aye theres been no reason to simp for intel adding and removing socket pins again after the 9900k
Yeah same here, I made a 13400 3060ti system late 2022 for a good deal but once the x3d chip comes out for zen 5 im going to do a full upgrade as i cornered my self onto a DDR4 board at the time. Edit: will upgrade to 7800xt equiv or newer
Lots of equivocation going on here. No way it makes sense for AMD to sandbag on Zen5's potential. First impressions matter, and so far Zen5 is getting slammed by reviewers. Why would AMD want that? No. If Zen5 with a bit more power can beat Zen4 by ~15%, then it would have been released in that form.
Plus, where is AMD's response to all the bad Zen5 reviews? Are they hiding?
Ah, yes. 7W APU.
You know what, I'm gonna start working for Intel from now on. I want whatever they were smoking when they made those slides.
Ehh... So where is that peak Zen 5 performance right now?
In AVX512 lmao
This guy just bashes Intel and hypes up AMD because he has shares in AMD and has admitted to that. Hyping up amd all the time gets him views.
Zen 5 "peak performance" went to toilet.
Even if Intel outpaced AMD by a full 15%. How much of that would be given up by degradation, bios, security, microcode updates. Legitimately where does the IPC for Raptor Lake launch day even stack up against IPC for Raptor lake today? Downfall, Spectre was pretty much swept under the rug never to be mentioned again.
2%
if you watch hardware unboxed original video on the 7800x3d vs now, it went from like 1% faster in gaming to 3-5% overall now vs the 14900k and this is before intels baseline profile thing, so its prob even worse now
@@NadeemAhmed-nv2br That can also be about improvements to 7800X3D performance due to AGESA updates over time. One problem AMD STILL has but who knows maybe Zen 5 will be a first, is putting out products where the software to run them still needs work. It's not TERRIBLE on the CPU side, on the GPU side it's my biggest issue with AMD, because with their GPUs you can end up with avg uplifts sometimes of around 10%, which is nice and all but is SURE would have been good if that performance was there at launch. I mean sometimes in particular games or a particular API like DX11, Vulkan they'll put out a new driver and the uplifts are 10+%.
Intel has a lot of money and top tier engineers. Their problem is they also have top tier arrogance. I spent a year working for them during the P4 debacle. When Athlon was moping the floor with their products. The attitudes that their management would bring to the table was insane. Most were unaware of anything but clock speed. Those that were made it sound like their profit margins were so good that it didn't matter. In the end of that era their stock took a dump. The only thing that saved Intel's bacon was a group of engineers that continued to develop an alternate core based on the P3 coppermine that evolved into the Core solo and Core Duo lines and caught them back up while AMD squandered their lead.
@@DOGbackwardz Raptor Lake isn’t the part competing with this.
This aged like milk.
Haha yeah
yall just come from a reddit group and spam this shit? der8auer got 25% increased performance in gaming when he o.c. 'ed it. How is that not major headroom/ holding their own chip back? When was the last time you could o.c. something for 25% more gaming performance and 40% increase in productivity benchmarks? Oh right never, but lets spam "aged like milk." The only thing aging like milk is amds software side, it never schedules with windows properly and their drivers are like fucking russian cruise missiles, sometimes they take themselves out.
MLID, What happened?
MLID and Amd became entire circus, that's what happened here.
Watching this video from my AM3+ FX 8350 pile driver Frankenstein PC with a 6700XT running on 16 lanes of pcie 2.0
Get a Xeon chip. Leave FX where it belongs
@@ProsperinoG Xeon? Why'd you recommend that when they could just go to Zen 3 or something better?
i built my son a computer with a first gen i5 running a maxwell titan. gpu does ALL the heavy lifting on pcie 2.0 as well haha. ill let him save up to build something new, or maybe when i upgrade he'll get mine..
@ProsperinoG Never!!! The glory of AM3+ lives on. Pile driver will forever be immortalized in Linux Bazzite OS. SSE4 and lackluster IPC lives on with my janky PCIe adapter NVME SSD nutured at 1280 MBps read/write.
I will never leave.
CPU bottleneck that needs checking out
I was in a Best Buy about a month ago and had a conversation with one of the salespeople about Intel vs AMD. I definitely have the impression from what I've seen that AMD is the best value, and that Intel with its problematic big/little core stuff and typically higher power consumption was to be avoided. He was like, oh no, Intel is much better, much to my surprise. So when you show the marketing materials supplied to retailers at [14:20], maybe I understand now how this person got this idea. Or maybe they just have a lot of Intel inventory to sell....
And you know, a few brown envelopes, the Intel way ;-)
Also had someone tell me I needed a 2400 watt psu for my crossfire system
Just keep it in mind that Intel getting sued and fined for bribing retailers was only about about preventing Intel from continuing to bribe retailers, it came with no additional stipulations or regulations on the retailers themselves to stop favoring sales of Intel products. While there's not definitive proof outside allegory and hearsay, I still have no doubt there's an unwritten "gentlemen's agreement" in place between retailers and Intel.
Seems like AMD is not only giving us double the integrated graphics grunt in laptops, but efficiency that gives the Apple M4 Pro & M4 Max and maybe even the M5 a run for the money too!
Mhh I wouldn’t say Apple’s chips but I’d definitely say those new Qualcomm chips. Those are over hyped beyond belief.
The M4 is just a less power efficient M3...
@@vyor8837No. The M4 is a significant redesign that gives better performance and better efficiency.
And M4 Pro/Max haven’t even been announced nor leaked. No one outside Apple (and maybe TSMC) knows the number of cores those will contain.
This video aged very poorly. Zen5 performance is no where near where you have claimed it would be.
20+ % IPC improvements would indeed be awesome!
He's not suggesting 20%+ IPC
He's suggesting that they've only really showed us the IPC difference and are hiding the boost characteristics, which also have a solid improvement.
The gains suggested are IPC + PBO boosting
@@TheDarksideFNothing IPC > any thing above 5GHz
CPU's literally need an entire system on the side just to cool them WTF
@@ThexXxXxOLOxXxXx I understand that it would be better if the gains were all IPC. I'm just saying that's not what the video suggested.
Also, clock speed and heat are not directly related. Wattage and heat are.
Above 5GHz on current gen is often a lot of heat to get rid of. We don't know how next gen behaves yet. Early info is suggesting Zen 5 will be hitting same clocks at lower power aka lower heat released.
You can take an old architecture and process node and make it hot at 1GHz. Clock speed and temp are not directly related, especially not across generations.
@@TheDarksideFNothing I'm just getting tired of 6+GHz and insane cooling requirements, a focus on IPC while keeping clock speeds high, not extreme.
As for Clockspeed pretty sure that's correlated with high voltage/heat
My perfects vision would be a 16 core cpu no multitreading running at stable 5GHz no need for Ecores just keep it traditional, Ecores are not for the normal consumer it's commersial server stuff
@@ThexXxXxOLOxXxXx there is a correlation with voltage and clock speed to power usage within a given architecture. But we are talking about an unreleased architecture, so that correlation will be different from what we have now. Different architecture and different process node, so it will take a different amount of power to hit the same clock speed. Power versus heat is always a direct correlation regardless of architecture or process node.
If you could make a processor that hits 8 GHz at 20 watts it is still only going to produce 20 watts of heat.
I think we are just having a language misinterpretation rather than either of us not understanding how it works. I'm only trying to point out that Zen 5 might hit 5.5 GHz without having an insane power target.
Fortunately, you can always just buy the same CPUs as everyone else and undervolt/underclock it to hit the heat targets that you're looking for. In most modern processors They are really pushing the limits of the architecture, so you can drop half the power and lose way way less than half the performance.
You give up a little bit of performance, but if you want less heat find the point on that curve that you're comfortable with.
Unfortunately, It's not like in the drawing room they could have just decided to make a product the same price and same performance while using significantly less power and therefore generating less heat.
I do agree that I would like to see lower power targets in general, but the reason we aren't is because both companies are fighting for performance superiority and they have already pulled every other lever. The last lever left is juice the hell out of it.
Bros pulling numbers out a hat 🤡
MLID pulls numbers based at how much Amd paid him to overhyped this flopped garbage Zen 5.
der8auer was getting 25% performance on the 9700x when o.c.'ed in gaming and up to 40% in some productivity benchmarks. that's higher than mlid results....
When AMD was behind with FX at least they could compete on price (I remember getting an FX-8320e and board for just $150, less than the cost of an Intel i5 by itself) and they had the console market as well, but I don't see how Intel is going to compete on price if AMD is able to make 'em even cheaper. Hell even the ultra-budget baller segment is locked down thanks to AM4 getting new chips until 2025 so why would I want to buy an Intel PC and risk my CPU going bye bye when I could get AMD and get lower power and great performance too?
I am more on finance side... the rationale for keeping stock voltage low for AMD could be total cost of ownership. Large corporate accounts might buy lower spec products, but imagine a small office. Lower consumption, lower consumption cooling, your electricity bills falls. AMD could simply go say "we save you in bills the whole cost of the chip".... plus, SOCKET STABILITY
and having no real VRM differences to worry about when picking a board. soo many socket 1700 boards will not run 12900k at peak because they cant provide the power...
Office PC's spend most of their time at or near idle power consumption, and every single time chiplet based AMD CPUs will lose that battle to Intel monolithic CPUs by around 8-18W depending on MCLK/FCLK frequency and vSoC. The lower the CPU utilization, the more appealing Intel CPUs become in longterm cost savings.
That being said, Ryzen monolithic APUs, while often being slower than Intel counterparts (apart from iGPU performance which is something to note), are still overkill for the majority of office PC tasks while consuming the same amount of power at idle as Intel CPUs while being priced similarly.
All things considered, if I were tasked with upgrading an entire floor of office PC's, I would still be looking at something like an AM4 based 5600G template for cost to performance ratio, as I prefer the flexibility that the AMD iGPU performance brings to the table, and the 5600G is still massive overkill for typical office use while sipping power like an Intel chip at idle.
Doesn't matter what Intel promises.
If they bring a processor that's 10% better than the last one, it will soon get a BIOS update to fix stability issues and then another one to fix safety issues and then it's only maybe 1% better. All while consuming ridiculous amounts of power.
Yet people still buy those.
Yeah bcos average consumers don't know the problem Intel is facing and would just buy whatever within their budget at Costco.
agreed. that's the sad thing isn't it. Laypeople would buy the most well-known brand, even though generally speaking the well-known brands are often inferior.
More like Dell & friends still buy those. The biggest purchases are managed by tech-illiterate ancients who listen only to salespeople and MBAs, not specialists.
@@djlim4612 ohhhhh that is soo true in the car world too....
Can confirm. My Dell work PC has a 10700 and it started out as a 16 thread CPU, and now it's 8 thread plus runs slower. I think it actually may be slower than the PC it replaced at this point, or it's pretty marginal at any rate. And on our Intel-based hypervisor servers I've had to allocate more cores to VMs to compensate. Needless to say our newest machines are EPYC. Zero issues with those.
Your Intel iGPU created a rendering error at 5:39 in the bottom right corner.
I don't see it though
@@Shahzad12357 Then check your eyes. I pop up for a split second in the right corner.
@@Noname-km3zx seems like i need to get my eyes checked
@@Noname-km3zx because i don't see it
It's not visible below 720p60
Even if intel were to win with Arrow Lake, they lose because of 13-14k gen instability. AMD won this gen round
You are missing the bigger picture. Intel is a vertically integrated fab company that cant compete against any of the rent-a-fabs at this point. All they got now is brand, and they just killed it with these degrading chips. Within the next 10 years Intel will spin off its fabs in order for the ic design side to be acquired.
@@styleisaweapon That does seem likely, but it's hard to predict what will happen over the next 10 years with certainty.
@@styleisaweapon however Intel got more handouts this year alone from Uncle Sam's corp brothers than TSCM ever got from not just Taipei, but also China and Japan. Will they be able to do something out of that or is it just a way to push lobbying in an election year?
Stay tuned
99.9% of people don't even know there's an instability issue with some of the i9 chips of the past 1.5 generations. The echo chamber of the inner tech world doesn't extend far.
@@styleisaweapon Well sorta yes, sorta not really. One of the big selling points of being a fab company on top of being a chip manufacturer, is that theoretically it would allow them a major competitive edge both in terms of specific node integration to serve their own unique needs first and foremost, while also bypassing most of the logistics and supply chain aspects -all while being able to produce own chips at a cost-price rate which no rent-a-fab company can compete with. It also provides manufacturing certainty for their customers, that they know Intel chips will always have first right to the fabs even in a situation of shortages or instability. For them, security of consistent product delivery means more than a per-generation uplift.
The real problem Intel has had up until now is that it takes a looong time to pay off the fabs and start seeing the positive benefits. These companies, especially when they move into manufacturing at the foundry level, they don't look at benefits over 5 years. They look at it over 10+ years.
The main thing that is kicking Intel right now, is that they acted like an ignorant monopoly from 2010 to 2020 and never believed AMD could truly compete on anything except price. One of the big differences between Intel and Nvidia for example, is that Nvidia never stopped and assumed competition couldn't hurt them. Sure their prices are outrageous, but they never stopped innovating, they never pretended like they couldn't bleed. Intel has been acting like a pretentious Greek god thinking they were above it all.
Exactly the right thing to do for AMD. I'll take the efficiency over the last 5% of performance every day! Also, just enabling PBO is a one click solution anyway, for when you really really need it.
Comparing to Intel Baseline is fair game, as well, since they're using baseline for their CPUs, too.
And one last remark: Coming up with ridiculous marketing slides has been Intels go-to reaction for years now. Doesn't surprise me at all.
agreed.. i dont think many people understand that what they do on a day to day basis will run with MUCH less power than the maximum that all these new cpus can pull. i run a 10850k that ive intentionally pulled back to 4.8ghz under boost and its plenty. fighting over the top of the line stuff really doesnt mean anything to me, and probably a ton of other people.
CPU staying under 100W in all conditions at base settings (OC is another can of worms for those who need it) is very important to me because it means you can easily buy a quality aircooler that will keep it silent for the life of the system without having to worry about it.
From my limited anecdotal observations it also means less paste degradation and needing being changed over time. My 4700K lasted 11+ years on first paste without any change to perf or temperature behavior and that was also a pretty cool CPU in the grand scheme of things.
My 7800X3D obv doesnt need changing but considering I got it on a fire sale I might actually be able to swap it for a 9800X3D at almost no cost nor heat negatives. I'll wait to see where the RAM situation also ends up being.
The rumour is that the X870 boards will be able to run 7,200mt/s DDR5
Same here with a 3700K, those were the days! Did you also undervolt to push the performance/heat ratio even further?
Imagine buying a 9700X and using a 35$ cooler keeping it under 60°C.
Video age like milk lmao
Well they weren’t
What annoys me about Intel is how they pitch a CPU as having a couple dozen cores or so, but on closer inspection it turns out only a few of those cores are proper cores, while the rest are kind of nerfed quasi cores. I think if AMD did that, the shizz would hit the fan with people online calling them out and maybe even class action law suits etc.
ATOM CPU cores are efficiency cores
Well, AMD is rocking efficient performance cores. @@mcalin5931
I don’t see a problem with heterogeneous architectures. AMD already does a variant of that with their C cores. As long as it delivers the performance and efficiency I don’t see an issue.
@@roccociccone597 AMD's C cores are full instruction set cores do, not comparable to e-cores in the slightest...
@@mcalin5931Atom cores do not support the full instruction set. An app that uses instructions which are supported in the big cores is unable to use the e cores - as it will just crash if it would.
Not supporting the full instruction set limits the real world use cases a lot.
If you turn off Intel Baseline the Intel products will eventually shit themselves, so there is that too..
Poor Amd after releasing this flopped garbage Zen 5. Intel gonna slap them hard with Arrow Lake easily.
I guess not even the 9800X3D will be worth buying, if the standard Zen 5 chips are no faster than Zen 4 why would the X3D chips be any faster than the 7800X3D either.
@@atnfn Amd zen 5 x3d will be another flopped garbage, the facts Amd allowed the new x3d cpu for overclocking shows zen x3d cpu will be slow cpu, barely an improvement just like normal zen 5.
Now if only the Radeon GPUs could even claim half the wins the CPU division has.
The 9950x is a 16-25% improvement at about the same power or a little less. The 7900xtx is around a 50% improvement at the same power draw as the 6950xt, yet that's a disappointment.
RDNA 3 wasn't a disappointment compared to RDNA2. It was a Disappointment compared to the 4090.
It would also be nice if Radeon had actually marketing instead of what we got with RDNA 3 launch. If they can not over promise and actually have something that's performs well they CAN have a good generation
Radeon already has a lot of wins against nvidia, they just don't have that performance crown people meatride nvidia for.
@@PineyJustice yeah the 7900xtx is an incredible product, even with the missed potential. I think a lot of this issue is the god-tier, apple-like, mindshare Nvidia has. Even if Radeon overtook Nvidia people would still find a way to say "nope, Nvidia better". lol
Nvidia is riding the high of CUDA/AI acceleration. Performance per dollar, Radeon has them beat for half a decade.
Zen4 had a super rough start because as I told when it launched, they disregarded everything that people liked from Zen products. Nobody's buying Zen if it consumes power like an FX did. I'm glad they learnt the lesson.
Zen wasn't really widely popular until 3000 and 5000 outperformed intel. Speed is what moved units, secondly the power between a 5950x and a 7950x are pretty close, within 30-40w
I've got both, I've spent a fair amount of time tweaking and tuning both. Not a lot of difference in an overclocked 5950x and a PBO tweaked curve optimized 7950x power wise, but the 7950x absolutely blows it out of the water in speed.
My Ryzen 7700X was pretty power hungry out of the box, but with some PBO tuning the power efficiency isn't terrible.
It's not that they learned their lesson it's just that they are using a more efficient process node. They did not design Zen5 to be more power efficient all the changes they highlighted consume more power.
even the exploding X3D chips
Idk what you are talking about
The Ryzen 9 7950x pulled 220 watts at peak power
The I5 14600k pulls 240 watts at peak power
Where is the Zen inefficiency?
They increased power by 5-10% and gave us 25-35% more performance
That’s 20-25% efficiency increase
Tom im waiting for your Doctor Eggman arc with the stache
6:00 bro who is out here turbocharging their cpu with 1 kW 😭😭😭
It's 10kW, not 1 MW. Obviously that's still way more than the CPU was actually using, that was probably just to have no power limit essentially.
@@CutoutClipsdammit now I look stupid
@@blasiankxng With your edit you now said 1kW when the chart says 10kW lol
@@CutoutClips L after L lol
Intel
Intel is probably shitting bricks....
more like bricking shits
Since the Ukraine war, power consumption is a big consideration in many places in Europe. Electricity prices have gone through the roof. I actually consolidated some servers to reduce power usage.
Excited for reviewers to give us their info in a few weeks!
The ol hide the horse shoe in the 🥊 trick.
First!!! Love your videos Tom!!!
Also, I just checked and I subbed to your channel in march 2019. How many subs did you have at that time?
Cheers!
He at time have Cheers!😂
Iam since 6/18 ! And this guy put all the ARC stuuf and Golden cove (Alder Lake ) way wwwaaayyy sooner than anyone!
PBO on Zen 3 definitely helped multi threaded performance. My 5950x is 3.6 all thread stock. PBO lifts to 4.2-4.4. Single Threaded only a slight bump.
I bought my used 5950x precisely because it is so freaking energy efficient and i am sick of hot slow intel garbage. It looks like i should consider the 9900 series sometime in the near future!
Remember that you can very easily throw a PPT limit on AMD chips in the bios. Works like a charm and can even boost FPS if you're into that. They're extremely efficient if you spend those 5 minutes in BIOS to reign them in. It's well worth it.
I don't think it's sensible pushing the 14900k past baseline.
There's some evidence that points out if it's already damaged, running at baseline might not even guarantee a stable system
The mistake is ever buying the 14900K in the first place. I know this because my 14900K lies dead on my carpet, while my 7800X3D outperforms everything thrown at it.
I care about Efficiency too.
Thats why i have all Amd in my house now.
I started in 2017 with a cheap Asrock AB350m pro4, and a 1700X. Now that motherboard is in my daughter pc, combined with a 5600X.
I have a 5800X3D, and my son has a Amd 3100.
Such massive value on AM4 platform. And i will be getting AM5 in the future
I bought a ryzen 5600 in January and put it in a b550. I looked at the temps and cost of zen 5 and said naaaa.
I thought the ad was going to be for Pied Piper! 😂🤣
Silicon Valley was such a great show. I was hoping for the same.
he flopped
So does Amd.
@@runninginthe90s75 true!
Big ups mate. We crave for those x3D scoops
I was just wondering today if AMD was sandbagging again
Perfect timing Tom!
Shouldn't have to have to overclock just to get the CPU to perform better than last gen. AMD messed this up.
I said 2 months ago that AMD was sandbagging.
😆
There's one hole in your argument. There's no reason to sandbag when you're releasing first. The reason you sandbag is to get your competition to overcharge for their product and look foolish when you launch. Intel will know the real numbers long before their own launch anyways so AMD has absolutely nothing to gain by sandbagging unless they're trying to manipulate their stock price.
Creating some doubt around a launch prevents competitors responding perfectly.
I feel like AMD are keeping all the best chips for the 9950X3D and 9800X3D. ZEN 5 CPUs use less power and run a lot cooler compared to ZEN 4 CPUs.. so maybe the X3D variants will have some even better specs than the Non X3D variants while also having 3D V-Cache!
Even if AMD "only" wins by 5%, they still get the "we aren't burning chips and denying it" win. That's worth at least another 5%
By not pushing the power draw of the current releases of 9000 chips, it leaves AMD headroom for any refresh next year depending on what Intel manages to get out the door. One key thing is that if AMD feels it needs to release an 8+16 core Zen/Zenc chip they have the power headroom needed to make sure that this can be fully powered.
Honestly I'm just waiting for the the Zen 5 X3D chips to significantly outperform Zen 4 X3D so that Zen 3 X3D CPU prices fall further (since I'm still on AM4). I'm not in the US or Europe so a lot of those great AM5 mobo + CPU + DDR5 memory deals simply aren't available for me.
Might be waiting a long time
AMD wouldn't do this intentionally. AMD flexing makes Intel make even more and more power envelope mistakes like they did already
14700K is also a ticking time bomb that will fail prematurely
When you think about it, coincidentally, Zen 5 CCD is on "4nm", while Zen 4 CCD is on "5nm" production.
Personally, I find it a bit misleading to market chips based on their transistor production node size, like Intel did with their "Intel 7/5/4" rebranding,
however, in AMD case I think it is an interesting observation.
I think there's a chance someone actually planned for it.
I think its clear that amd will release higher frequency cpus later to get more cash from people
For anyone who's intersted and/or considering an upgrade:
I purchased an RX6700XT last year (May 2023) along with a new LG TV that supports 120Hz VRR.
This year (June 2024) I built a new primary PC for myself, going for a Ryzen 7900 (No-X), I decided not to wait for Zen5 as I'm not buying the top spec and it should be good enough, the potential extra ~10-15% is not a big deal.
My selection reasoning was that although I do a lot of gaming that it is not the reason for the upgrade, my old Ryzen 2700 (No-X again & not overclocked) was perfectly fine for all of the games I play (I'm 61 and my reflexes have been too slow for twitch shooters for 20+ years).
The thing I do that was taking some time was processing my DVDs and Bluerays (I'm English and refuse to pay the TV TAX to fund the BBC). DVDs were not too bad with the 2700, even with a few filters such as medium denoise & sharpen for older sources, it could process the raw MKV into x.264 at about 30-50FPS, so up-to twice playback speed. However recently Bluerays have been selling for the same prices as the DVDs (eg: I got "The Man From UNCLE (2015)" BR for just £3.49 Inc P&P), and with 5 to 9 times the data (DVD 3-4GB > BR 25-29GB) it was slow and could take Handbrake half a day or more. The 7900 now processes BRs about as fast as the 2700 did DVDs!
I went with the 7900 due to a stock power usage of 65W, and a provided AMD heatsink/fan for £20 les than the X version which would need an additional Heatsink... So a saving of about £50 for no downsides. I may upgrade the heatsink but it's not urgent at all. The mother board is the Asrock Lighting ATX with 32GB of Corsair 6000 (CAS 30 that was only about £8 more than the basic DDR5).
At stock the 7900 is about 2.8 times the speed of the 2700 (per core), with PBO (I chose the 85C & -20mv setting from bios) it is ~3 times faster. I tried the simple "enable PBO" but that had the CPU hit the 95C limit in seconds and I didn't like the idea of that with the AMD heatsink, setting the 85C limited PBO gives me about the same performance plus a little cooler running.
Cinebench R23 MT scores: default "Auto" PBO was 23545
When they run 16 core Zen 5 chiplet desktop processor at 40W, it shows slightly worse performance. But at higher watts it shows much better performance. AFAIK it’s still using the same 6nm IOD, so that brings the same baseline cost, which should bring the new/old closer to each other at very low watts. The monilithic laptop chip of course doesn’t have this ”problem”, which is relevant as they need to run it at very low watt/core range.
RNDA 3.5 efficiency improvements for gaming seem super nice (32% better perf/watt). AMD spent time optimizing RDNA architecture for mobile phones with Samsung and they bring those learnings to their desktop/laptop GPU architecture. RDNA 3.5 seems much better than expected from a .5 name bump. Efficiency = performance in low power devices.
They aren’t hiding anything, PBO CO is not stock behavior. Never has been and once you enter that territory you could just OC Arrow Lake and then we just end up right at the same spot in the conversation.
I would say u would undervolt arrow lake but these days that’s just overclocking. I agree with u tho
@@WillFuIYou’re undervolting with curve optimizer, it just doesn’t change the target voltages
Pbo is easy and for all cpu and boards
PBO isn't an OC. It's the same as Intel with no power limits which Intel claims is in spec.
@@nab-v1w Again it’s still not stock behavior, and a lot of the actual performance comes from curve optimization not PBO. Which curve optimizer is not easy and is really finicky.
What you said about X3D's reliability is legit, especially with the inclusion of PBO2 Curve Optimizer. Running a Ryzen 7 5700X3D with a Wraith Prism and Curve Optimizer at -15 and it runs cooler than my Ryzen 5 3600 under load.
This is making overcloking good again
It's great that people finally pay attention to efficiency. It's ridiculous to drive chips well past their reasonable range just to get 5 extra fps. Everyone who wants that can overclock themselves like in they have in the past
I didn't know you were a French cowboy 😂😂 nice stache
I had to watch this video in the background, the stache was too distracting to pay attention.
People will still buy Intel and they will still keep their market share more or less. It's sad. We have to buy Intel at work as the IT staff are ignorant about hardware. AMD needs to up their consumer advertising like Intel was in the 90s. AMD Inside!
Be careful what baseline you are comparing with.
Zen4 (X variants) is tuned up last minute at launch to match 12/13 gen perf. That doesn't mean we should accept 105W tdp for a 6 core cpu (7600X) to be normal.
Yea each gen is getting better simultaneously at speeds, thermals and OC'ing ability. I love my 7950X3D it runs cool and OC's hard AF. Those performance improvements on 9000's are going to be huge additions.
AMDs Zen tech team that does OC have stated PBO curve has been heavily refined, and customers will be able to tweek it more to get better performance when doing certain jobs/tasks.
14:30 Oof
Apple comparing their M3 to the M1 in presentations is one thing but saying "hey our new CPU beats an ARM chip from 2021" is not what I call confidence.
It's just not the PBO overclocking. If it can perform better at higher clocks there is good chance it can do the same as undervolted meaning you can squeeze out better performance when undervolted (it will boost higher clocks with lower voltages).
I like how alot of channles are saying amd is fudging their proformance and will be slower..and here hes showing they might be faster ..it will make life way more fun soon
I mainly avoided an upgrade from Zen3 because Zen4 and 12-14th gen are more cooking ware than computing.
Would Arrow Lake be able to OC well? If they're regressing clocks and giving it far less power than before then OC to OC, Zen 5 might come very close or possibly on top in some things.
Tom: Gomez Addams edition gets +35% more brain IPC
i agree 100%. i think Zen5 is going to blow a lot of minds.
can't wait for them to name this gen's threadripper the 9999x. i am absolutely buying one if they do, and probably still buying one if they don't lol.
transistor counts are hard to quantify since they're all kind of funny numbers outside of raw sram test chips.
redundancy, I/O, and fast logic lowers density a lot, so what happened here is probably all of these levers got pulled at once because yields and clocks were beyond expectations.
We will find out how fast is new amd and new intel CPU on real world benchmarks when they come out.
Intel is so doomed, I can't believe they won't be able to put their sh!t together for arrow lake, and now we have the stability issues and probably a class action lawsuit when more people find out their processors are degrading and crashing, thanks Intel.
I havent seen Smokey and the Bandidt for years. Im going to give it a view this weekend.
holds up just saw it first time
Not sure if you covered this but any idea if Zen5 X3D will actually get dual chiplet vcache? Awhile back the same birdy that hinted to me that memory settings wouldn't be locked at runtime also suggested this being a possibility. I'm getting skeptical but they have been taking more jabs at Intel's scheduling challenges.
If they really are able to get the boost clocks high enough to justify it then that explains some of the sandbagging. The other reason I'm skeptical is it would probably make more sense to save this for Zen 6.
Zen 5 9950x3d will still only have the x3d cache on one chiplet
Sorry ccd
Intel is now in the same position AMD was before Zen CPUs. How times have changed !
I haven't regretted getting a Ryzen 7 7800X3D at all. It's an amazing and stable CPU, while being more power efficient than Intel.
Exciting times. Looking at those scores, my geriatric 3600X is now really starting to show its age...
looks like the pbo insanity was true huh
Here's the real question... Will Arrow Lake CPUs be REALIABLE? Isn't there 100% failure rate with gen 13 and 14 Intel processors?
No becuase the ring bus was only meant to support up to 10 cores. They have to change to the mesh bus instead.
Yeah so arrow lake will be caught up with the 13 14 Gen debarkle even if it is fine.. how will people know.. amd will sell alot of 9000 cpus because of the loss of trust a cover up will damage intels rep.
Nowhere near 100% failure rate, RMAs are less than 0.01%. MANY people are trying to snag free CPUs which of course the channels are not covering.
I'm guessing the Zen 5 X3D gonna be great coz of the general improvement in the architecture like wider front end, better branch predictor etc.
I bought my first ever AMD CPU - 7800X3D and it's a totally amazing CPU. It's driving the easily best tower I've ever had. I would be perfectly happy never buying Intel again despite loving that Intel has re-entered the GPU foray at least in theory and see AMD as the PC leader for the foreseeable future.
I have been an Intel Guy for over 20 years and the few times I did work with AMD CPU (the bulldozer and phantom era) they were crap. But my next build scheduled for end of July 24 is definitely going with either the 9700x or the 9600x depending on the review and cost. It's time to give AMD a chance to prove they are better long term than INTEL and for INTEL to get their house in order and really deliver a knockout with Arrow Lake and beyond.
Does it matter how Arrow Lake performs and what it costs, when we must question whether Intel can produce a product that works as advertised?
Very curious to see what a cpu like the 9700X can do with PBO 2 and Curve Optimiser. Add in a better memory controller and the difference in performance can be quite substantial. 7700X here heavily tweaked waiting to see if it’s worth upgrading.
Can't find zen 5 performance..
Intel’s goose may be cooked. With Epyc now available on AM5, they’re giving so many customers a real option to move to AMD.