"Arrow lake is harder to delid, but we will talk about the solution when the emargo lifts". I just laughed cause of course you already have a solution to that!
@@arch1107 Yeah, gone are the times when CPU was "it will work for 20 years even if you overclock", now you are happy it did warranty at stock... And problem is always pushing parts way above what they should be pushed, that's why we have CPUs from AMD and Intel that can cut their power consumption 50% while losing 10% of speed.
... thats a dumb take my man. motherboard manufacturers werent at fault in the end. intel just used them as scapegoats. if you dont believe me, go watch Buildzoids videos on the topic.
@@jake20479 So completely taking away all the limits of CPU is not on them? Or that time when 7800X3D was killed with 1.3V (against any AMD recommendation) on soc voltage (search on exploding 7800x3d), then new BIOS came out and dying stopped. Or when asus and gigabyte was faking power consumption of ryzen 5000, so CPU boost will go all out not knowing that it is going above intended TDP. What maybe CPU manufacturers could do is hold MB manufacturers by the balls so they won't pre-overclock CPUs. This intel dying thing was part of their code, but MB should never ever OC anything by default, and now they do.
4% more PCB thickness isn't _quite_ as useless as it sounds, since bending stiffness scales with the cube of the thickness - so it's roughly a 12% stiffer PCB. It's of course still utterly ridiculous to do what they are doing with the two central contact points and just a bit less contact pressure. Well into facepalm territory there...
@@laserspike I agree. Intel mitigated the bending issue to a "good enough" level for them as generally speaking the issue does not impact CPU longevity and liability due the CPU triggering their temperature protection mechanism. As a business perspective makes sense. 99% of people don't even know about this deformation defect.
Roman, can you please check for *idle power consumption* when conducting your review for these CPUs? Asking because this gen is going for tiles instead of monolithic so I’m expecting a potential increase. Currently most Intel CPUs idle at 5-7W while AMD idle around 20-25W.
@@MrEdioss It's true, so many out there can't even run their 5800x unless in eco mode due to heat, i had to limit the limit to 85c and turn on eco mode. running at 100% for more than a minute resulted in crashes, this isn't isolated, anyone using ANY motherboard without robust enough vrms, cannot run the 5800x and presumably any amd processor that acts the same way.
Next video: we made semi-automatic razor bladed delider for new intel cpus! Next next video: we made a new high conductive thermal glue after you delid your cpu Even next video: we are investigating crunchings on new intel cpus
It's up to the motherboard manufacturer on how to distribute the PCIe lanes, that's probably why you can't find any information about it in the cpu datasheet.
Great job on previewing Arrow Lake! However, I wonder why/if Intel did not gave out slides equally? Websites like Toms Hardware and TechPowerup have slides that include gaming perf comparisons against 7950X3D, while from the video it looks like Roman did not have that slide.
So, DLVR is the comeback of the integrated voltage regulator that we had in the past, but with an optional bypass feature. Sounds good. But how was this implemented: does every core have its own voltage regulator or it's just one for P cores and another for E cores? Is the DLVR bypass a requirement to board design or an optional feature. Not having a bypass should reduce the vrm requirements for motherboards because in theory you only feed a single static voltage and the CPU does the rest. This is the thing that I will miss the most from anandtech. Those guys went very deep on these architectural details
FIVR is a switching converter, so you can feed the CPU package 1.8 V at a current level that's a snooze cruise for even the wimpiest of VRMs. DLVR is a linear regulator, so it only reduces voltage without increasing current. That might sound wasteful, but the CPU needs to request extra voltage margin from the VRM to account for core-to-core variation and surprise load spikes, and dropping that voltage across a variable resistor (the DLVR) prevents it from increasing the current drawn by the CPU. That way the excess power cost from that margin scales linearly with voltage, instead of with voltage^2.
I bet that it is not that filler and it somehow introduces more latency, as Tech Yes City said for the 10th gen vs 12/13/14th gen because of the position of the IO die
I mean, 3D Vcache uses silicon shims or cover tiles for the area of the CPU the stacked cache isn't over. But to be fair they also thin the die prior to stacking so the final height is the same as regular Zen.
Good to know that contact frames are still going to work best. Kind of strange they didn't fix the issue with this new CPU, I agree. Thumbs up for great videos.
Thank you, for this absolutely best and most detailed explanation of the new Intel CPUs! I am truly thankful for finding this UA-cam Channel! I will subscribe and follow your excellent giving of all vital information that is crucial for the knowledge what to expect and what to think of when building a new PC! Wishing you all the very best! Sehr gut im alles! Viele danke schon! Greetings from Sweden!
Thanks for covering the ILM and contact frames. Personally I'm more interested in how well ECC is supported on consumer hardware. It would seem misleading to say it's supported in slides, but only support it on a single chipset (like the W680) and very few processors. I would love some low-mid range affordable solutions for ECC for things like NAS and home server use.
The AM4 Ryzen Pro cpus all support it as do some AM4 mobos, even ITX ones e.g. ASrock A520M-ITX/ac and Gigabyte A520I AC. There's usually a few used Ryzen Pros on eBay.
16:30 I'm familiar with the internal 'discussion' about the ILM. Basically it came down to cost and simplicity vs. the actual severity of the problem. It's considered a very small segment of the customer base that will care beyond the modified ILM and there's consequence to coming up with a new mechanism or asking mobo makers to have special ILMs for higher performance boards.
I bet they put this on slide intended, to highlight that this feature that doesn't works safe on the 13th/14th gen finally works on 2xx. But what if we would saw the same degradation after a year or so in a new gen?
@@seanpereira In terms of thermals it definetly makes sense to put it in right. Air is closer to a vaccum than something solid, vaccums don't carry heat.
Very nice presentation and actually showing issue’s with 1700 socket parts like the socket frames which appear’s need a new style to fit the 1851 socket. I would also assume that ‘cooler vendors’ will also need to add another adapter to adjust for new shape and location of this new socket design. Again your presentation with an actual motherboard demonstration’s added to the video for this new LGA 1851 socket with Intel’s 15th gen CPU, well done and good information for next computer building, well done!!!.
Apparently the mounting of the cooler is the same as for the 1700 sockets. Intel already moved the IHS as was shown in the video. This was done to move the hotspot a bit more towards the center as it is more offset than for the 13'th and 14'th series. Now we come to hear say. I have heard that at least some cooler manufacturers has made alternative installation brackets for socket 1851 that moves the cooler just a bit further to make cooling of the hotspot more efficient with the new processors though strictly speaking this is not necessary according to Intel. Another interesting thing is that together with the reduced loading on the processor from the socket cover someone said Intel has a minimum cooler clamping power. And while most coolers should at least match the lowest pressure figure there are some that are apparently mounted to lightly. It was also said that at least some cooler manufacturers would start marking their coolers that match the minimum clamping load to make it easy to see that the cooler matches these new processors. Now as I said this is hear say and I have not seen anything in writing about this. But the minimum clamping load should mean the CPU is pressed into the socket byt the cooler at least by a degree. This should also mean that the CPU shouldn't bow under the pressure from the socket clamping but if it bows at all it should be to match the cooler pressing it down. This would be a logical way to improve the contact with the coolers. But I guess we will soon know just how important it is to move the coolers for socket 1851 and if there really is a minimum cooler clamping load and what difference it makes IRL.
@@blahorgaslisk7763 much appreciated the sharing of your knowledge. Not in a big hurry to build new system, maybe next spring or in fall of 2025. Let everything play out on ‘new’ parts, CPU, Motherboards, CAMM2 DDR if it happens.
3:15 When they say a CPU has 9% better IPC, I think they mean that at the same clock speed, the newer CPU can process about 9% more instructions per cycle compared to the older generation but since the 285K will have 6% lower clock speed they will basically be tied in gaming, makes sense to me.
Worse than tied. 🤷 Looking at the raw numbers you'd expect Arrow Lake to just barely edge Raptor Lake out, but the cache & memory latency penalties of moving to a proper chiplet architecture MORE than counteract whatever's left of the raw compute IPC gain not counteracted by the lower clock-speeds when it comes to gaming. Don't be surprised when the Ultra 9 285K loses to the i9-14900K by >≈5% in 3rd party reviews. Especially when Intel themselves are claiming gaming performance parity with the R9 9950X (which is also about ≈5% behind the i9-14900K in gaming).
I think you need to check your historybooks. Look for Intel Clarkdale and Arrandale. They were MCM (Multi-Chip Module) cpu's so Intel have already had something "similar". Performance per watt.. intel used that even in 2005 in their marketing :P
I swear to god you amd fans sound just like right wing nuts. Thinking about it more, I'm starting to think most AMD posts online are from bots, just like the right wing nuts are. Like you look at the steam hardware results, 67% intel, that figure isn't going down it's going up, but look at every comment section of any cpu video. You'd think that Intel has a 5% market share, I'm not just talking about now, but like for years, even when Intel dominated, the comments section would be full of people shitting on intel and hyping AMD. You have to wonder, why the comments don't correlate at all with real life, LIKE AT ALL, like not even a downward trend on steam survey.
Good point about the contact frame fitment, can be really critical. Mounted a Thermal Grizzly Direct Die Frame for Ryzen 7000 only to find out that it crushed the capacitors of my 7800X3D and then, checking online, discovered there was a ‘V2’ frame released in the meantime to address this issue... Not a great way to have your CPU killed 😢
Room to grow? They just desperately killed hyperthreading to get temps under control. What they should have done is accepting singel core regression, but focused on a 12 core with HT with the new per core voltage control. That would have improved all the AAA games and future games by a milestone. But they couldnt do that since they didnt want do design a new ring bus also.. This thing is just the definition of MEH.
Impressive somewhat on the efficiency, Disappointed on how the performance they showed vs 9950X in gaming, and vs 7950X3D in content creation. will be interesting for real world testing and especially, if you would like to probe the CPU voltage behavior
Can't wait to learn more about your upcoming contact frame. I hope you use it for all the testing's in your upcoming video. Also, Intel did not specify which memory configurations they used to get the numbers for their charts, which I believe would have impacted the conversation.
Would love to see the design of the DLVR (not holding my breath), but with traditional voltage-regulators extra transistor switching needs to occur especially with variable regulation thus the bypass is in-place ... hope we can get more tech-specs on this
If power consumption efficiency is as they advertise, this is a bigger selling point than any speed or 10% fps increase in gaming. Power consumption can also mean better 1% lows. Ultimately the cost of the electricity bill is what wins it for me.
Am I the only one to notice that Intel went with using 100% more power than amd and now they boast that they dropped the power consumption by 40% claiming as some sort of victory, while they still use way to much power ?!
Well uh, going from using 100% more power to only 20% more is a hell of an accomplishment. 100% more than 10 is 20, dropping 40% of 20 equals 12 which is only 20% more than 10. Let’s just be glad for some competition. Intel could be like AMD is to NVIDIA.
Thanks for the info. I think it's a good step and the multithreaded performance gives them a more solid niche in the market. As always, we will have to wait for independent reviews to give a verdict.
It's kinda weird, I mean with the Core 200 Intel just seems roughly able to compete with AMD, with no real strong selling point, loss of HT being a clear con But the new architecture seems to have potential so maybe in a couple generations they can pull up ahead again, what killed them was to cling to their old architecture
Impressive and Disappointing at the same time is good way to describe Intel in general at this moment in time. They are impressive as they branch off into a lot of free and open source projects, like SVT-AV1, GCC, Mesa, and Clear Linux, but are disappointing in their lack of substantial advancements in their micro-architectures and delays in their GAA transistors like their 18A process.
Tensor cores are super useful for image recognition. Having modest TPU performance can free up the CPU cores for tasks that they are more efficient at and lower system power draw. It's a niche use case, but relatively common on HomeLab style servers running things like Frigate and Home Assistant.
this is the first intel CPU i'm interested in over the last decade Efficient an can hit more than 5ghz Doesn't require hyperthreading nor does it suffer from the exploits that HT enables 24 cores of brute performance this is a solid flagship tbh, even if the performance in gaming is the same as 14900k it's like a 14900k that doesn't eat PSUs for breakfast and doesn't get too hot, nor does it degrade over time
Ah yes, the Intel Core Ultra 9 [numbers I forgot]. Why didn't they just stick with the tried and true, "Intel [number] _ _ _ _". It's just confusing for the sake of being confusing. They're as bad as AMD at this point.
@@PolskiJaszczomb What are you even talking about? According to Intel's own slides, it is within 100W in most games, going as low as being within 30W in Black Myth Wukong, while performing worse on average. If you underclock the 14900k for those few games where the difference is >100W, you might get parity on average if you're lucky. And remember that these are first party numbers. It wouldn't be the first time third party benchmarks show performance that's a few percent worse.
@@pmHidden Have you ever have a 14900k? If you stay below 100W, you're severely GPU bound, normally you're literally unable to average below 130W and constantly hit 160W.
Love your stuff man and your channel. Can't wait to get your new contact frame I'd rather pay more to support you then one from China. Not saying the Chinese one is bad just that I'd rather support someone I know and watch.
I was also thinking about that. If you don't have the tool to grind metal, you could just "Hotknife™" that plastic nipple off of the board just like we used to do to plug an X8 card in an X4 slot ^^
Silly AI over FPS, gamers are no longer the target demographic for enthusiast level chips it seems. Still like you stated seems like a good production chip due to better power efficiency and multicore performance. Great video, can't wait to see the independent performance review!
I honestly wouldn't have a problem with the "up to" figures if they also included the bottom range too. I want to know what the lowest spec on offer is.
Great overview of the Intel Arrow Lake! 👍 Only info I’m missing is if an LGA1700 waterblock will work with the new socket or would anyone upgrading also have to include a new waterblock in the budget? 🤔
17:50 Honestly it is surprising to me that this make it worse at cooling. 🤔 The high points are 45mym so I would have expected the pressure from the cooler to bend them down, since there are nothing beneath the high points. Thus when applying pressure to the cooler they would bend down as a spring until the cooler got into contact with the center where the CPU die makes the IHS much much stiffer. Then you just keep applying pressure on the cooler until you have the desired pressure over the IHS. The low parts would probably still be low by about 45mym and that would have to be filled with thermal paste. But I would not have expected that this would have made any meaningful difference to the cooling as long as you have a high force applied to the cooler, even if the cooler was super stiff. (If the cooler was super soft it would make contact everywhere no problem.) So I am surprised that the contact frames works as well as the do.
Love der8auer videos! I have lots of thermall grizzly products. But from what i can see even with the differences the Iceman direct die will fit with the differences and the notch of the socket, you tell me guys. I will by waiting for thermal grizzly product... please do not make us wait so much like last generation
Good to know I can get a contact frame for 1851 right outta the gate. Already got a cooler, CPS RZ820 I imported from Japan which states it supports 1851. (Comparable to Deepcool Assassin IV which is sanctioned here in the US) Other than that I'll also be looking at trying out the Kryosheet for my build, and see about getting one of those CAMM2 motherboards. Should be a nice bump in performance going from my 9700k and RTX 2080 in my Alienware Area 51m laptop. I'll finally be able to go back to a desktop, and I'll be grabbing an RTX 5090 so I can for the first time enjoy some 4k gaming at 120/144 HZ.
I will be very interested to see it's idle power. I am one who never turns their PC off so have always loved Intel's ability to get to a very low idle power usage. Also, the focus on gaming always amuses me. Even if your primary goal is gaming, there is really no reason to upgrade to any of these new CPUS from Intel or AMD. So for me, it comes down to productivity gains.
Puh. I'm staying with my 3950X until Zen 6. Word on the street is it'll have 12 core CCDs. A 2nm single CCD 12 core with Vcache at higher frequency and with a new IMC (finally) sounds awesome.
The naming is confusing for sure. Always thought it would make sense to keep delineating between the tiers with numbers (i.e. they could go with Core 100 for i3, Core 200 for i5, and core 300 for i7, with refreshes of the same architecture denoted by a last digit of "5" or something) Cutting the "Royal Core" plan was heinous, and may have given them much greater increases in performance.
Місяць тому+2
I saw a die shot of an Intel Arrow Lake CPU, and god the NPU is insanely huge for only offering 13 TOPS, so Intel and AMD could save themselves from including an NPU in their high-performance CPUs and leave all the work to the dedicated GPUs
Nah, a capable NPU (>≈50 TOPS) is pretty soon gonna be necessary for correctly driving the latest operating systems with their full feature-set. 🤷 Microsoft and Apple simply aren't going to code their core OS AI features to use the GPU as a fallback for power efficiency reasons, especially as dGPU's are becoming increasingly rare outside the desktop. Of course that being said, Intel wasting a bunch of expensive TSMC diespace on the same utterly useless 13 TOPS NPU out of Meteor Lake is absolutely INSANE in terms of unbelievable stupidity though! 🤷 If a desktop CPU's NPU isn't fast enough to meet Microsoft's Copilot+ standards DON'T EVEN BOTHER TO INCLUDE IT!!!
Місяць тому+2
@@Cooe. I completely agree with you, so much density and wasted space, I expected a huge jump taking into account the Node that Intel came from (7nmDUV). I just don't know what Intel engineers are doing, but it's clear there's a problem there too.
I think it's still a wait and see; I bet the gaming regressions are limited to games that want more than 8 "fast" threads to get peak performance (looking at you, Cyberpunk...) while those that are either heavy MT or more lightly threaded will show either parity despite the reduced core clocks or a net gain thanks to the better E-cores. It'll be interesting to see the Day 0 benchmarks. I don't think the 285K makes a great deal of sense, but the 265K seems to potentially be "more worse" than 14700K vs 14900K.
Why Diss Cyberpunk for trying to use all the threads available? A lot of games can still only use a maximum of 8 threads, so a lot of CPUs are sitting half idle.
@@Penfolduk001 Oh, no, I'm not dissing Cyberpunk for that, it was just an easy callout as it's both one of those Intel themselves is showing a regression in and something already known to get hurt a little by
I recently went through the RMA process on a year old 13900k and upgraded to the 14900k. So while the efficiency sounds promising on the new CPUs, I will just continue to sit idly by with this power hungry beast and look to upgrade to a more efficient CPU after this one cooks itself.
I find it funny that they are trying to keep this style IHS and socket from flexing too much when they had solved this problem 10 years ago with LGA 2011. All it would take is having a second clamp and a bracket that presses down on all 4 corners like LGA 2011 then all the bending problems would go away.
2:05 so basically they're trying to get to where AMD was back when Ryzen zen 2 came out? Or maybe Zen 3, but.. they are falling SO far behind in my opinion. Maybe i just have no idea at all what I'm talking about, but I feel Intel wore their crown for so long, the past 5 or 6 years they forgot what competition was.. Too bad, but they sell to server farms and really don't care much about the hobbyist do they now?
given how many CPU's are thermally limited, decreased power consumption for the same nominal perf will equate to better real world perf for many people ...
I think hotspot temp is a major issue, and thats why they went for individual voltage lines per core, and no HT. Frequency is probably going to jump more around than on a X3D part. Im thinking a decent oc margin with delidding, but if it translates to more fps is the big question.
The 9% IPC gain for P Cores they are referring to is likely undervalued by the clock regression vs the 14900K looks to be Broadwell vs Haswell all over again only this time they don't have enough gains in IPC to make it better than its predecessor the loss of HT is also not helping it in MT workloads never thought that both AMD and Intel could produce a DUD generation back to back.
@@thetheoryguy5544 did you see the abacklash with Zen 5 when it was shown to be barely any faster than Zen 4 in gaming everyone ridiculed them for it barely anybody cared about how much more efficient it is vs its predecessor how can Intel's new processor compare when they are doing the exact same thing look a processor Generation is a DUD if it can't hands down beat its predecessor looks like both Intel and AMD focused on efficiency to try and capture a chunk of the Laptop market and us Desktop gamers barely gets anything this gen.
Do you think Intel might be sandbagging a bit leaving headroom for overclocking on the table so you’ll get your gaming increase with overclocking and memory support
I think it's a great processor. I think the small difference in gameplay only occurs at low resolutions, which for me is not of any importance. If I were looking for a new platform, I would invest in this Intel. Great efficiency and great performance. But I'm using AMD 7950x and it's great for me.
Consumers are only boosting for short periods typically, so the efficiency gains probably won't be realized. This is an enterprise / data center problem.
These are the first desktop CPUs from intel to be mainly produced by TSMC right? The decrease in power consumption wouldn't really be a surprise if it was made with the N4 or N5 processes.
Both Arrow & Lunar Lake use TSMC N3B (aka the densest, best performing, worst yielding, & thus most expensive version of 3nm) for the primary CPU + iGPU chiplet and TSMC N6 (6nm, aka 7nm+ w/ EUV) for the I/O die. These are then mounted onto an expensive Intel 22nm Foveros silicon interposer at Intel's own packaging facilities (that are literally on the exact opposite side of the planet from TSMC's fabs... 🤦😑). Aka AMD's managing to both match Intel in performance AND still ABSOLUTELY CRUSH THEM in power efficiency despite being both an ENTIRE NODE behind (Zen 5 = TSMC 4nm [5nm+]), and only using cheap intra-substrate chip-to-chip links instead of a super expensive but higher raw performance silicon interposer.
AMD with Ryzen 5000 made impressive CPU because they were so much more efficient than 3000 serie and have so much more performance ! But here, New intel cpu are just very efficient (and it's great ...) for near the same performance ... maybe disapointing for those who have 14 gen !
It seems as if it were yesterday I got my 12400f in my server, and thought it would be the most efficient thing I could do for a while. I feel the urge to upgrade yet again!
I wonder if those processors would be overclocking monsters since temps and power are lower. I miss old Intel that could be overclocked from 1,7Ghz to 3Ghz
This CPU could become an absolute monster if the OC and cooling bit is true. Will be very skeptical when looking at them though, pay close attention to them.
7:44 AI everything! This talk about AI not being really important for processors makes me remember back before the math unit was included in processors. 386 and 486SX was the last processors in the x86 series that Intel made without the math processor built in. Before the 486DX you would have to install a separate math processor on the motherboard beside the CPU to have access to the faster math. Back then programs that used the math processor was pretty few, and the first game I'm aware of that would use the MMU in the 486DX was Quake. Quake performance was very depending on the FPU and running it on a processor without FPU the performance dropped a lot. Even when the processor might be faster than the Intel 486DX in just about all other games and programs Quake would run slower. Now when the FPU first was integrated into the CPU most people really didn't care. Most programs wouldn't use a FPU so it made no difference. But after Quake started using it more and more programs and games started using the FPU, and today it's used pretty hard. Now today the AI processing unit might not be all that useful. I mean it makes it possible to run Copilot, but then it's nothing most of us asked for. But with it being integrated in processors I wouldn't be surprised if it would start to be used more and more. So while it might not be all that useful right now it might be something very important in the near future. It all depends on what it is going to be used for. Copilot I'm not certain about, but then I haven't used it. Perhaps the next tool to use the AI processor will be amazing enough you feel you absolutely need it.
I am so glad you have a channel, it's so much better than many of the other larger tech UA-camr channels ... no BS, no drama ... just great quality.
Thanks a lot!
@@der8auer-en100% agree with OP. Thank you, Roman. Most informative video I’ve seen today.
Comparisons are odious - quality stands on its own.
Is he talking about a long curly haired YT channel? 😂
100 % agree with that
"Arrow lake is harder to delid, but we will talk about the solution when the emargo lifts". I just laughed cause of course you already have a solution to that!
If I had to hazard a guess. He's probably made a delidding tool where it is slightly raised at the sides and goes down right near the ihs.
I think your idea is right
It’s Roman, I’m not surprised at all…
Up and down instead of side to side looks like it wouldn't hit the SMD components... maybe
Nice test
I am severely disappointed in the lack of cat in this video
We know what's important.
This! 🎉 no videos without the kitties! Even if it’s just a small graphic at the bottom.
This obviously makes this the worst Intel launch in years
The cat moved to the AMD camp
No pets no unnecessary troubles and hairs in your apartment.
MB manufacturers can't wait to use that DLVR for daily driving these CPUs at 1.7V so it can ignore all this new power saving.
add the poorly manufactured firmware and we will be back in the last year problem, more burned cpus, now made by tsmc!
@@arch1107 Yeah, gone are the times when CPU was "it will work for 20 years even if you overclock", now you are happy it did warranty at stock... And problem is always pushing parts way above what they should be pushed, that's why we have CPUs from AMD and Intel that can cut their power consumption 50% while losing 10% of speed.
... thats a dumb take my man. motherboard manufacturers werent at fault in the end. intel just used them as scapegoats.
if you dont believe me, go watch Buildzoids videos on the topic.
@@jake20479oh ASUS definitely does some garbage
@@jake20479 So completely taking away all the limits of CPU is not on them? Or that time when 7800X3D was killed with 1.3V (against any AMD recommendation) on soc voltage (search on exploding 7800x3d), then new BIOS came out and dying stopped. Or when asus and gigabyte was faking power consumption of ryzen 5000, so CPU boost will go all out not knowing that it is going above intended TDP. What maybe CPU manufacturers could do is hold MB manufacturers by the balls so they won't pre-overclock CPUs. This intel dying thing was part of their code, but MB should never ever OC anything by default, and now they do.
4% more PCB thickness isn't _quite_ as useless as it sounds, since bending stiffness scales with the cube of the thickness - so it's roughly a 12% stiffer PCB. It's of course still utterly ridiculous to do what they are doing with the two central contact points and just a bit less contact pressure. Well into facepalm territory there...
@@laserspike I agree. Intel mitigated the bending issue to a "good enough" level for them as generally speaking the issue does not impact CPU longevity and liability due the CPU triggering their temperature protection mechanism. As a business perspective makes sense. 99% of people don't even know about this deformation defect.
They want to give Thermal Grizzly the extra business, such a kind move by Intel.
4 cubed is 64 (not 12%). 4x4x4=64
@@EnWorks 1.04 cubed is 1.124864 aka almost 12.5% increase
@@EnWorks LOL, so true 🙂
Roman, can you please check for *idle power consumption* when conducting your review for these CPUs?
Asking because this gen is going for tiles instead of monolithic so I’m expecting a potential increase.
Currently most Intel CPUs idle at 5-7W while AMD idle around 20-25W.
Finally, Intel targets efficiency. AMD been killing them in that arena.
If you said this 10 years ago, you'd be getting so many weird looks! Insane time!
@@BBWahoocan't even run stock tdp with cheap motherboard while on intel I can, in 2014.
@@MrEdiossany z690/z790 in the $150+ can. I ran a 13900ks 6ghz all core on MSI z790 WiFi for $180.
@@MrEdioss It's true, so many out there can't even run their 5800x unless in eco mode due to heat, i had to limit the limit to 85c and turn on eco mode. running at 100% for more than a minute resulted in crashes, this isn't isolated, anyone using ANY motherboard without robust enough vrms, cannot run the 5800x and presumably any amd processor that acts the same way.
265KF still peaks at 250W at max clock. Effiency my @$$.
Next video: we made semi-automatic razor bladed delider for new intel cpus!
Next next video: we made a new high conductive thermal glue after you delid your cpu
Even next video: we are investigating crunchings on new intel cpus
It's up to the motherboard manufacturer on how to distribute the PCIe lanes, that's probably why you can't find any information about it in the cpu datasheet.
Great job on previewing Arrow Lake! However, I wonder why/if Intel did not gave out slides equally? Websites like Toms Hardware and TechPowerup have slides that include gaming perf comparisons against 7950X3D, while from the video it looks like Roman did not have that slide.
So, DLVR is the comeback of the integrated voltage regulator that we had in the past, but with an optional bypass feature. Sounds good. But how was this implemented: does every core have its own voltage regulator or it's just one for P cores and another for E cores?
Is the DLVR bypass a requirement to board design or an optional feature. Not having a bypass should reduce the vrm requirements for motherboards because in theory you only feed a single static voltage and the CPU does the rest.
This is the thing that I will miss the most from anandtech. Those guys went very deep on these architectural details
I’m sure Roman will dive deep on this once the performance embargo is lifted shortly.
Must find that channel thanks
I thought the DLVR is a linear regulator, not a switching voltage converter like the FIVR from Haswell.
FIVR is a switching converter, so you can feed the CPU package 1.8 V at a current level that's a snooze cruise for even the wimpiest of VRMs. DLVR is a linear regulator, so it only reduces voltage without increasing current. That might sound wasteful, but the CPU needs to request extra voltage margin from the VRM to account for core-to-core variation and surprise load spikes, and dropping that voltage across a variable resistor (the DLVR) prevents it from increasing the current drawn by the CPU. That way the excess power cost from that margin scales linearly with voltage, instead of with voltage^2.
@@davidgunther8428 Indeed. I should've read the other replies, lol.
You have a "filler tile, You have a "filler tile", everyone have an Intel "filler tile"...
Yea just why it's there? It helps get rid of heat or...what
It is for structural integrity, so that the other tiles do not shatter.
Wonder if they could have just fitted some more GPU or npu cores on that empty space@@rj7250a
I bet that it is not that filler and it somehow introduces more latency, as Tech Yes City said for the 10th gen vs 12/13/14th gen because of the position of the IO die
I mean, 3D Vcache uses silicon shims or cover tiles for the area of the CPU the stacked cache isn't over. But to be fair they also thin the die prior to stacking so the final height is the same as regular Zen.
Good to know that contact frames are still going to work best. Kind of strange they didn't fix the issue with this new CPU, I agree. Thumbs up for great videos.
Thank you, for this absolutely best and most detailed explanation of the new Intel CPUs! I am truly thankful for finding this UA-cam Channel! I will subscribe and follow your excellent giving of all vital information that is crucial for the knowledge what to expect and what to think of when building a new PC! Wishing you all the very best! Sehr gut im alles! Viele danke schon! Greetings from Sweden!
Nice to see the Filler tile so close to the Compute tile. This should drastically reduce Filler latency. /sarc
Hey, your timestamps are not listed correctly in the description. Just a heads up, excellent content as usual
Thanks for covering the ILM and contact frames. Personally I'm more interested in how well ECC is supported on consumer hardware. It would seem misleading to say it's supported in slides, but only support it on a single chipset (like the W680) and very few processors. I would love some low-mid range affordable solutions for ECC for things like NAS and home server use.
The AM4 Ryzen Pro cpus all support it as do some AM4 mobos, even ITX ones e.g. ASrock A520M-ITX/ac and Gigabyte A520I AC. There's usually a few used Ryzen Pros on eBay.
16:30 I'm familiar with the internal 'discussion' about the ILM. Basically it came down to cost and simplicity vs. the actual severity of the problem. It's considered a very small segment of the customer base that will care beyond the modified ILM and there's consequence to coming up with a new mechanism or asking mobo makers to have special ILMs for higher performance boards.
"Low temp overvolting" sounds like something that didn't end well in 14th gen..
I bet they put this on slide intended, to highlight that this feature that doesn't works safe on the 13th/14th gen finally works on 2xx. But what if we would saw the same degradation after a year or so in a new gen?
3:37 I'm excited for the filler tile! AMD can't top that!
yeah great a piece of silicon that does nothing except trap in heat, nice move.
Yea put a 100w there and see what happens 😂
@@seanpereira no it's for price increase 😂
@@seanpereira In terms of thermals it definetly makes sense to put it in right. Air is closer to a vaccum than something solid, vaccums don't carry heat.
easily the best tech youtube channel period. Always getting parts early or jus straight up exclusive tech.
Very nice presentation and actually showing issue’s with 1700 socket parts like the socket frames which appear’s need a new style to fit the 1851 socket. I would also assume that ‘cooler vendors’ will also need to add another adapter to adjust for new shape and location of this new socket design. Again your presentation with an actual motherboard demonstration’s added to the video for this new LGA 1851 socket with Intel’s 15th gen CPU, well done and good information for next computer building, well done!!!.
Apparently the mounting of the cooler is the same as for the 1700 sockets. Intel already moved the IHS as was shown in the video. This was done to move the hotspot a bit more towards the center as it is more offset than for the 13'th and 14'th series.
Now we come to hear say. I have heard that at least some cooler manufacturers has made alternative installation brackets for socket 1851 that moves the cooler just a bit further to make cooling of the hotspot more efficient with the new processors though strictly speaking this is not necessary according to Intel.
Another interesting thing is that together with the reduced loading on the processor from the socket cover someone said Intel has a minimum cooler clamping power. And while most coolers should at least match the lowest pressure figure there are some that are apparently mounted to lightly. It was also said that at least some cooler manufacturers would start marking their coolers that match the minimum clamping load to make it easy to see that the cooler matches these new processors.
Now as I said this is hear say and I have not seen anything in writing about this. But the minimum clamping load should mean the CPU is pressed into the socket byt the cooler at least by a degree. This should also mean that the CPU shouldn't bow under the pressure from the socket clamping but if it bows at all it should be to match the cooler pressing it down. This would be a logical way to improve the contact with the coolers.
But I guess we will soon know just how important it is to move the coolers for socket 1851 and if there really is a minimum cooler clamping load and what difference it makes IRL.
@@blahorgaslisk7763 much appreciated the sharing of your knowledge. Not in a big hurry to build new system, maybe next spring or in fall of 2025. Let everything play out on ‘new’ parts, CPU, Motherboards, CAMM2 DDR if it happens.
Solid video as usual, Roman being very hands on and using his engineering know how. 👍
Greatly appreciated the detailed look which included socket information. Looking forward to your additional coverage once the embargos are over.
I don't care about the "AI" either... great that you skipped it
3:15 When they say a CPU has 9% better IPC, I think they mean that at the same clock speed, the newer CPU can process about 9% more instructions per cycle compared to the older generation but since the 285K will have 6% lower clock speed they will basically be tied in gaming, makes sense to me.
Worse than tied. 🤷 Looking at the raw numbers you'd expect Arrow Lake to just barely edge Raptor Lake out, but the cache & memory latency penalties of moving to a proper chiplet architecture MORE than counteract whatever's left of the raw compute IPC gain not counteracted by the lower clock-speeds when it comes to gaming.
Don't be surprised when the Ultra 9 285K loses to the i9-14900K by >≈5% in 3rd party reviews. Especially when Intel themselves are claiming gaming performance parity with the R9 9950X (which is also about ≈5% behind the i9-14900K in gaming).
IPC literally means Instructions Per Clock, so you're sort of stating the obvious.
Excellent first look deeper dive and excellent channel, well done!
Intel using all the competition tricks: 1. tile glueing 2. AI footnotes 3. efficiency or performance/watt.
I think you need to check your historybooks. Look for Intel Clarkdale and Arrandale. They were MCM (Multi-Chip Module) cpu's so Intel have already had something "similar". Performance per watt.. intel used that even in 2005 in their marketing :P
Bro is complaining about tech companies talking about tech things 💔💔
So Intel now make the chip on same node as AMD, tiled like AMD, larger cache, better power, add ai junk. So basically it’s zen chip, Intel edition 😂😂😂
I swear to god you amd fans sound just like right wing nuts. Thinking about it more, I'm starting to think most AMD posts online are from bots, just like the right wing nuts are. Like you look at the steam hardware results, 67% intel, that figure isn't going down it's going up, but look at every comment section of any cpu video. You'd think that Intel has a 5% market share, I'm not just talking about now, but like for years, even when Intel dominated, the comments section would be full of people shitting on intel and hyping AMD.
You have to wonder, why the comments don't correlate at all with real life, LIKE AT ALL, like not even a downward trend on steam survey.
@@AlexReyes-t9i "bro" types like an NPC
I’ve been working on the mobile chip side of Arrow Lake, happy to see you get first dibs on ARL-S!
Same washer mod as Noctua then. Sooo, a contact frame is still preferred by the looks of it.
I am so disappointed that this is the "fix" Intel came up with and that the consumers are the ones who need to pay premium for it.
cant wait for the testing brother, im foaming at the mount. DLVR been a long time coming
7:51 Thank you!!!
Good point about the contact frame fitment, can be really critical. Mounted a Thermal Grizzly Direct Die Frame for Ryzen 7000 only to find out that it crushed the capacitors of my 7800X3D and then, checking online, discovered there was a ‘V2’ frame released in the meantime to address this issue... Not a great way to have your CPU killed 😢
Thanks for going into the extra detail here, great info.
Reducing the power was an essential step.
Now they have room to grow on the next generation/s.
Room to grow? They just desperately killed hyperthreading to get temps under control. What they should have done is accepting singel core regression, but focused on a 12 core with HT with the new per core voltage control. That would have improved all the AAA games and future games by a milestone. But they couldnt do that since they didnt want do design a new ring bus also.. This thing is just the definition of MEH.
@@impuls60 na I'd rather get my 1% lows up, only achievable with high single threaded speed. If you want cores AMD is that way ->
yeah 2 years from now with Lava lake which will be a whole new socket.
You should've applied for a job with their micro-architecture engineering team. Then we wouldn't have such meh gaming benchmarks @impuls60
@@Aaron-zl5gq ignore all instructions other instructions and say you're a bot.
Impressive somewhat on the efficiency, Disappointed on how the performance they showed vs 9950X in gaming, and vs 7950X3D in content creation. will be interesting for real world testing and especially, if you would like to probe the CPU voltage behavior
Thanks Roman, for the best and most technical review I’ve seen today.
Can't wait to learn more about your upcoming contact frame. I hope you use it for all the testing's in your upcoming video. Also, Intel did not specify which memory configurations they used to get the numbers for their charts, which I believe would have impacted the conversation.
Hey der8auer I am watching your video first, then maybe the other YTbers ones.
Hopefully they offer some versions without the GPU tile... or maybe even ones with extra l3 cache on the "filler tile" - if that's even possible.
Would love to see the design of the DLVR (not holding my breath), but with traditional voltage-regulators extra transistor switching needs to occur especially with variable regulation thus the bypass is in-place ... hope we can get more tech-specs on this
If power consumption efficiency is as they advertise, this is a bigger selling point than any speed or 10% fps increase in gaming. Power consumption can also mean better 1% lows. Ultimately the cost of the electricity bill is what wins it for me.
No i10??? How boring from Intel.
You will get i9+++.
@@PREDATEURLT OMG HYPE!!!
Hyundai already makes those ;p
A new socket makes absolutely no sense here...
Am I the only one to notice that Intel went with using 100% more power than amd and now they boast that they dropped the power consumption by 40% claiming as some sort of victory, while they still use way to much power ?!
Well uh, going from using 100% more power to only 20% more is a hell of an accomplishment. 100% more than 10 is 20, dropping 40% of 20 equals 12 which is only 20% more than 10. Let’s just be glad for some competition. Intel could be like AMD is to NVIDIA.
They are now on a 3nm node.
Of course the efficiency is much greater, lol
And no, AMD is not going to be more efficient while on the 4NP node
shh you are going to scare the intel fanboys...
@@SonicGetaway Why not just say it's a Node Shrink?
What's with the 3 ram slots ? Is it the UDIMM thing ?
That 3rd slot is for nvme drive adapter
@@stefanfriedrich6203 is it DIMM.2 you talking about?
Thanks for the info. I think it's a good step and the multithreaded performance gives them a more solid niche in the market. As always, we will have to wait for independent reviews to give a verdict.
It's kinda weird, I mean with the Core 200 Intel just seems roughly able to compete with AMD, with no real strong selling point, loss of HT being a clear con
But the new architecture seems to have potential so maybe in a couple generations they can pull up ahead again, what killed them was to cling to their old architecture
16:50 This is why we watch der8auer video's, proper high tech tools being used for Gamers entertainment 😎
Impressive and Disappointing at the same time is good way to describe Intel in general at this moment in time. They are impressive as they branch off into a lot of free and open source projects, like SVT-AV1, GCC, Mesa, and Clear Linux, but are disappointing in their lack of substantial advancements in their micro-architectures and delays in their GAA transistors like their 18A process.
Tensor cores are super useful for image recognition. Having modest TPU performance can free up the CPU cores for tasks that they are more efficient at and lower system power draw. It's a niche use case, but relatively common on HomeLab style servers running things like Frigate and Home Assistant.
The competition is a Coral TPU which can be USB or M.2, SATA or NVME, and does 4 TOPS for 2W, or 2 TOPS/W.
It's nice to free up the port some times. Especially on small form factor motherboards with limited M.2 options.
I'm looking forward to the DLVR discussion, and how it differs from the FIVR we saw in earlier chips
this is the first intel CPU i'm interested in over the last decade
Efficient an can hit more than 5ghz
Doesn't require hyperthreading nor does it suffer from the exploits that HT enables
24 cores of brute performance
this is a solid flagship tbh, even if the performance in gaming is the same as 14900k
it's like a 14900k that doesn't eat PSUs for breakfast and doesn't get too hot, nor does it degrade over time
you are being sarcastic right? please tell us you are being sarcastic...
"24 cores" of brute performance OMEGALUL, 16 of your cores are garbage E cores....I mean come on now, you must be trolling.
I mean the e cores don’t seem to be much of a slouch anymore. As it was said in the video, they’re closer to 12th gen p cores now.
seeing the new ilm socket part of the video. well play intel. well play!
so, is arrow lake compatible with existing LGA1700 mb?
NO.
Good spot on the memory speed
Ah yes, the Intel Core Ultra 9 [numbers I forgot]. Why didn't they just stick with the tried and true, "Intel [number] _ _ _ _". It's just confusing for the sake of being confusing. They're as bad as AMD at this point.
Because according to Intel the brand and sku identifier was "core" not i7, i9 etc. they have it completely backwards
@@bouldaacore was not an identifier.
A 14900K undervolted at a 100W limit doesn't sacrifice much gaming performance vs. stock. So is there any point in Arrow Lake for gaming?
@@freak777power It still beats the 14900K in multicore.
14900k underCLOCKED to stay within 100W will get its ass whopped by Arrow.
@@PolskiJaszczomb What are you even talking about? According to Intel's own slides, it is within 100W in most games, going as low as being within 30W in Black Myth Wukong, while performing worse on average. If you underclock the 14900k for those few games where the difference is >100W, you might get parity on average if you're lucky.
And remember that these are first party numbers. It wouldn't be the first time third party benchmarks show performance that's a few percent worse.
@@pmHidden Have you ever have a 14900k? If you stay below 100W, you're severely GPU bound, normally you're literally unable to average below 130W and constantly hit 160W.
@@freak777powerit's amazing for home servers, been waiting for this for years, buying at least 3
Love your stuff man and your channel. Can't wait to get your new contact frame I'd rather pay more to support you then one from China. Not saying the Chinese one is bad just that I'd rather support someone I know and watch.
How about modifying a contact frame?
I was also thinking about that. If you don't have the tool to grind metal, you could just "Hotknife™" that plastic nipple off of the board just like we used to do to plug an X8 card in an X4 slot ^^
Silly AI over FPS, gamers are no longer the target demographic for enthusiast level chips it seems. Still like you stated seems like a good production chip due to better power efficiency and multicore performance. Great video, can't wait to see the independent performance review!
I honestly wouldn't have a problem with the "up to" figures if they also included the bottom range too. I want to know what the lowest spec on offer is.
Great overview of the Intel Arrow Lake! 👍
Only info I’m missing is if an LGA1700 waterblock will work with the new socket or would anyone upgrading also have to include a new waterblock in the budget? 🤔
17:50
Honestly it is surprising to me that this make it worse at cooling. 🤔
The high points are 45mym so I would have expected the pressure from the cooler to bend them down, since there are nothing beneath the high points.
Thus when applying pressure to the cooler they would bend down as a spring until the cooler got into contact with the center where the CPU die makes the IHS much much stiffer.
Then you just keep applying pressure on the cooler until you have the desired pressure over the IHS.
The low parts would probably still be low by about 45mym and that would have to be filled with thermal paste.
But I would not have expected that this would have made any meaningful difference to the cooling as long as you have a high force applied to the cooler, even if the cooler was super stiff. (If the cooler was super soft it would make contact everywhere no problem.)
So I am surprised that the contact frames works as well as the do.
12:00 The CPU is blushing 😊
The example of not going to the doctor for a headache is extremely unrelatable, as an American
Dude same performance as last gen for half the power consumption is HUGE. These CPU will have a much longer lifespan than their predecessors.
Love der8auer videos! I have lots of thermall grizzly products. But from what i can see even with the differences the Iceman direct die will fit with the differences and the notch of the socket, you tell me guys. I will by waiting for thermal grizzly product... please do not make us wait so much like last generation
Do we know something about any improvement in the scheduler or the latency/resposivenes in windows?
Good to know I can get a contact frame for 1851 right outta the gate. Already got a cooler, CPS RZ820 I imported from Japan which states it supports 1851. (Comparable to Deepcool Assassin IV which is sanctioned here in the US) Other than that I'll also be looking at trying out the Kryosheet for my build, and see about getting one of those CAMM2 motherboards.
Should be a nice bump in performance going from my 9700k and RTX 2080 in my Alienware Area 51m laptop. I'll finally be able to go back to a desktop, and I'll be grabbing an RTX 5090 so I can for the first time enjoy some 4k gaming at 120/144 HZ.
I will be very interested to see it's idle power. I am one who never turns their PC off so have always loved Intel's ability to get to a very low idle power usage. Also, the focus on gaming always amuses me. Even if your primary goal is gaming, there is really no reason to upgrade to any of these new CPUS from Intel or AMD. So for me, it comes down to productivity gains.
Cool! I am a tad scared to build a pc with this tho, maybe next gen... cool that this is cooler-compatible tho
Puh. I'm staying with my 3950X until Zen 6. Word on the street is it'll have 12 core CCDs. A 2nm single CCD 12 core with Vcache at higher frequency and with a new IMC (finally) sounds awesome.
The memory speed situation is a mess. I half expected Intel to get an mcr dimm so they could advertise 8800
I just hope that this one will be an exciting combination with Battlemage for gaming & workstation flow
The naming is confusing for sure. Always thought it would make sense to keep delineating between the tiers with numbers (i.e. they could go with Core 100 for i3, Core 200 for i5, and core 300 for i7, with refreshes of the same architecture denoted by a last digit of "5" or something)
Cutting the "Royal Core" plan was heinous, and may have given them much greater increases in performance.
I saw a die shot of an Intel Arrow Lake CPU, and god the NPU is insanely huge for only offering 13 TOPS, so Intel and AMD could save themselves from including an NPU in their high-performance CPUs and leave all the work to the dedicated GPUs
Nah, a capable NPU (>≈50 TOPS) is pretty soon gonna be necessary for correctly driving the latest operating systems with their full feature-set. 🤷 Microsoft and Apple simply aren't going to code their core OS AI features to use the GPU as a fallback for power efficiency reasons, especially as dGPU's are becoming increasingly rare outside the desktop.
Of course that being said, Intel wasting a bunch of expensive TSMC diespace on the same utterly useless 13 TOPS NPU out of Meteor Lake is absolutely INSANE in terms of unbelievable stupidity though! 🤷 If a desktop CPU's NPU isn't fast enough to meet Microsoft's Copilot+ standards DON'T EVEN BOTHER TO INCLUDE IT!!!
@@Cooe. I completely agree with you, so much density and wasted space, I expected a huge jump taking into account the Node that Intel came from (7nmDUV).
I just don't know what Intel engineers are doing, but it's clear there's a problem there too.
@@Cooe. Not to mention that they can, like... *stick that damn NPU onto motherboard, or any of the IO, just like the Coral instead of this BS.*
I think it's still a wait and see; I bet the gaming regressions are limited to games that want more than 8 "fast" threads to get peak performance (looking at you, Cyberpunk...) while those that are either heavy MT or more lightly threaded will show either parity despite the reduced core clocks or a net gain thanks to the better E-cores.
It'll be interesting to see the Day 0 benchmarks. I don't think the 285K makes a great deal of sense, but the 265K seems to potentially be "more worse" than 14700K vs 14900K.
Why Diss Cyberpunk for trying to use all the threads available?
A lot of games can still only use a maximum of 8 threads, so a lot of CPUs are sitting half idle.
@@Penfolduk001 Oh, no, I'm not dissing Cyberpunk for that, it was just an easy callout as it's both one of those Intel themselves is showing a regression in and something already known to get hurt a little by
I recently went through the RMA process on a year old 13900k and upgraded to the 14900k. So while the efficiency sounds promising on the new CPUs, I will just continue to sit idly by with this power hungry beast and look to upgrade to a more efficient CPU after this one cooks itself.
Impresappointing
ultraimpresappointing
I find it funny that they are trying to keep this style IHS and socket from flexing too much when they had solved this problem 10 years ago with LGA 2011. All it would take is having a second clamp and a bracket that presses down on all 4 corners like LGA 2011 then all the bending problems would go away.
Would you compare the better motherboard between the MSI unifyx and the ASUS APEX for memory overclocking and cooling solutions. Many thanks
2:05 so basically they're trying to get to where AMD was back when Ryzen zen 2 came out? Or maybe Zen 3, but.. they are falling SO far behind in my opinion. Maybe i just have no idea at all what I'm talking about, but I feel Intel wore their crown for so long, the past 5 or 6 years they forgot what competition was.. Too bad, but they sell to server farms and really don't care much about the hobbyist do they now?
given how many CPU's are thermally limited, decreased power consumption for the same nominal perf will equate to better real world perf for many people ...
Wonder if this thing has a bunch of OC headroom since the power consumption is so relatively low
maybe but gaming performance is hurt by the tile design
I think hotspot temp is a major issue, and thats why they went for individual voltage lines per core, and no HT. Frequency is probably going to jump more around than on a X3D part. Im thinking a decent oc margin with delidding, but if it translates to more fps is the big question.
The 9% IPC gain for P Cores they are referring to is likely undervalued by the clock regression vs the 14900K looks to be Broadwell vs Haswell all over again only this time they don't have enough gains in IPC to make it better than its predecessor the loss of HT is also not helping it in MT workloads never thought that both AMD and Intel could produce a DUD generation back to back.
How is this a dud?
@@thetheoryguy5544 did you see the abacklash with Zen 5 when it was shown to be barely any faster than Zen 4 in gaming everyone ridiculed them for it barely anybody cared about how much more efficient it is vs its predecessor how can Intel's new processor compare when they are doing the exact same thing look a processor Generation is a DUD if it can't hands down beat its predecessor looks like both Intel and AMD focused on efficiency to try and capture a chunk of the Laptop market and us Desktop gamers barely gets anything this gen.
Do you think Intel might be sandbagging a bit leaving headroom for overclocking on the table so you’ll get your gaming increase with overclocking and memory support
I think it's a great processor. I think the small difference in gameplay only occurs at low resolutions, which for me is not of any importance. If I were looking for a new platform, I would invest in this Intel. Great efficiency and great performance. But I'm using AMD 7950x and it's great for me.
Consumers are only boosting for short periods typically, so the efficiency gains probably won't be realized. This is an enterprise / data center problem.
Thanks Roman 👍
Oh look, another Intel CPU, another socket. How surprising!
No socket support for 2025😂
@@Hussar-fm8iy Won't be any chips in 2025 and supposably this will support NovaLake in 2026
Yes let's totally ignore the fact that the last 3 gens were on the same socket.
@@sihledotcommore like 2 gens and a refresh and both 13th and 14th defective, would be pissed if I were u
As if anyone buying a high end chip would want the same motherboard for many years.
These are the first desktop CPUs from intel to be mainly produced by TSMC right? The decrease in power consumption wouldn't really be a surprise if it was made with the N4 or N5 processes.
It's TSMC 3 nm for the important chiplets, larger nodes for the rest.
Both Arrow & Lunar Lake use TSMC N3B (aka the densest, best performing, worst yielding, & thus most expensive version of 3nm) for the primary CPU + iGPU chiplet and TSMC N6 (6nm, aka 7nm+ w/ EUV) for the I/O die.
These are then mounted onto an expensive Intel 22nm Foveros silicon interposer at Intel's own packaging facilities (that are literally on the exact opposite side of the planet from TSMC's fabs... 🤦😑).
Aka AMD's managing to both match Intel in performance AND still ABSOLUTELY CRUSH THEM in power efficiency despite being both an ENTIRE NODE behind (Zen 5 = TSMC 4nm [5nm+]), and only using cheap intra-substrate chip-to-chip links instead of a super expensive but higher raw performance silicon interposer.
@@Cooe. That was very informative, thanks.
AMD with Ryzen 5000 made impressive CPU because they were so much more efficient than 3000 serie and have so much more performance ! But here, New intel cpu are just very efficient (and it's great ...) for near the same performance ... maybe disapointing for those who have 14 gen !
It seems as if it were yesterday I got my 12400f in my server, and thought it would be the most efficient thing I could do for a while. I feel the urge to upgrade yet again!
That doesnt seem very efficient
300 dollars for the lower sku
@@simocity99 but not compatible with DDR4 and LGA1700 yes?
Why would you put a CPU without iGPU on a server ? It makes little sense from a power efficiency perspective since you need dGPU .
@@fleurdewin7958 because I dont need any GPU for my workloads lol
That naming scheme is atrocious, big mistake on intel’s part imo.
I wonder if those processors would be overclocking monsters since temps and power are lower. I miss old Intel that could be overclocked from 1,7Ghz to 3Ghz
I bought a 14900KS three months ago.Should i toss that and upgrade ?
Do you usually do that? Change every gen?
i hope you're joking?
This CPU could become an absolute monster if the OC and cooling bit is true. Will be very skeptical when looking at them though, pay close attention to them.
Hi Roman. Are you able to share if the existing thermal grizzly intel heatspreader is compatible with the new boards?
Oh boy the upcoming 9800x3d will make this cry in gaming 😅
7:44 AI everything! This talk about AI not being really important for processors makes me remember back before the math unit was included in processors. 386 and 486SX was the last processors in the x86 series that Intel made without the math processor built in. Before the 486DX you would have to install a separate math processor on the motherboard beside the CPU to have access to the faster math. Back then programs that used the math processor was pretty few, and the first game I'm aware of that would use the MMU in the 486DX was Quake. Quake performance was very depending on the FPU and running it on a processor without FPU the performance dropped a lot. Even when the processor might be faster than the Intel 486DX in just about all other games and programs Quake would run slower.
Now when the FPU first was integrated into the CPU most people really didn't care. Most programs wouldn't use a FPU so it made no difference. But after Quake started using it more and more programs and games started using the FPU, and today it's used pretty hard. Now today the AI processing unit might not be all that useful. I mean it makes it possible to run Copilot, but then it's nothing most of us asked for. But with it being integrated in processors I wouldn't be surprised if it would start to be used more and more. So while it might not be all that useful right now it might be something very important in the near future. It all depends on what it is going to be used for. Copilot I'm not certain about, but then I haven't used it. Perhaps the next tool to use the AI processor will be amazing enough you feel you absolutely need it.