We have to thank the competition from AMD for there to be 14 cores at 300 bucks! Just a few years ago 14 cores cost you in the thousands of dollars range.
I find Derbauer's power limiting @ 90% testing of the 13900k to be very very interesting. I'd like to see testing at at 75% and 80/85% as well. That 1% lows was definitely worth looking into, very curious.
What I'm seeing is that behind-the-motherboard-fans will be the next modding trend... "This is amazing guys, no more back-panel melting with this awesome mod"
I used to just put doubled thermal pads on the bottom of the motherboard. Probably not very efficient, but the back of the case got really warm so I guess it worked. As a bonus it held the backplate in place in case I needed to remove the cooler.
AMD's edge is AVX-512. Cinebench R23 doesn't use it, but Blender does Also 12th gen Alder Lake chips, my 12600K included has AVX-512, but you have to disable E-cores in the BIOS in order for the AVX-512 option to come available. Intel removed AVX-512 on later releases of 12th Gen CPUs, and newer BIOS updates for Z690 boards will stealth disable AVX-512. TimeSpy / Extreme by default uses AVX, but on TimeSpy Extreme Custom, you can specify to run the benchmark with AVX, AVX2 or AVX-512.
Home "gaming grade" PCs are becoming real problem for the electrical wiring gauges and breakers for rooms in many flats. I have my workstation/gaming PC, an intel iMac and a ryzen TV computer in my living room with a TV, HIFI surround set up and some pro audio gear etc. If I upgrade to a 4090 and and a CPU like this one I would probably be dangerously near tripping the 10A 240V breaker and the wiring to the wall outlets doesn't really support more and this flat was renovated before I moved in this year, lol. I would for sure have to move my floor AC unit to an extension power cable from another room.
@@spiceyfrenchtoast9421 yes, I live in Europe which is one of the places where 220-250v is the default wall socket voltage. Bathrooms occasionally has a 115v outlet but other than than it’s 230 everywhere here. So not 240, that was a typo but around there.
@@wiziek it is high end but for sure something a lot of people will buy to play games at home with. Also, whenever 5090 is released the second hand market will get a bump.
@@spiceyfrenchtoast9421 it’s not as much as it seems. It IS more but not like it seems. Notice the 10 amps? 120v is at 15 or 20 amps. Our 240v split phase starts at 30 amps (two legs of 120v) and 50 amps is common too.
How much more you can push when you have competition. ;) Fun fact: About the same amount of time has passed from the launch of the 1800X until today than what it took Intel to get from a Sandy Bridge quad core to a Kaby Lake quad core when AMD was nowhere to be seen.
It's no longer a race between AMD and Intel in CPUs. It's an all-out brawl going on at the finish with various limbs flailing over the line at different times.
I want more rambling Wendell . I hope Wendell does a live stream with Ian Cutress soon discussing the new chips. Once all the new X3D chips are released .
26:40. First TNG reference? Though at 21:22, dual Utube(?) screens of Spock and Kirk. Though Spock with a weapon and Kirk on the ground...is this a message?
When they give me a motherboard that will last 3-5 years and a CPU that consumes normal amounts of power in the high end, and still be able to upgrade to a similar one after those years, then I will consider them.
If you built this system with a 13900k from the start I don’t think there would be a reason to upgrade anything for probably 10 years ….. especially if you aren’t doing some type of Niche computer rendering for a living where saving a few second on a render will make you more money… And gaming these new CPUs will hold up for gaming most likely for a decade at this point also . Considering you can still game at 1080p with a i7 3770k today … the new systems having Pcie5 support when pcie4 isn’t even saturated all the way with memory or GPUs … they are at a point that they are very future proof honestly. Only thing would be nice is if they had at least 40 pcie lanes but that’s asking a lot of desktop PCs I’m sure that will be the next core race that we have had since Ryzen in 2017 started the 6 core minimum up to 16 . And Intel has taken this to the max giving people way way way more threads /cores than anyone will ever need. Very few actual PC users will need a 13900k’s multithread performance that’s why the 13600k will most likely be the best seller having 8 efficiency cores that are basically 7700k non Smt cores added on to the latest 6 SMT core design is huge that’s so much power it’s really actually crazy for a i5 … That and building one with a ddr4 motherboard since ddr4 is so cheap right now that will give a nice upgrade maybe in a couple years where you can pick up a ddr5 board and ddr5 if you want to but my guess is there won’t be enough of a benefit with ddr5 to justify the upgrade for maybe 5 years … kinda like ddr3 to ddr4 if you had a high end ddr3 build like a i7 4790k you really didn’t have to upgrade your PC till DDR5 was already being rumored for release… Again that app depends on what your use cases are for your PC . The people who actually will capitalize on having the newer hardware for work will be the first adopters and the enthusiast gamers… take as old as time.
We have come to a point that for gaming at 1440p or 4K, any "new" gen CPU + a midrange to high-midrange GPU will give you about the same performance. This is now the "get the one that its in stock and its cheaper in your country", yes because prices are all over the place in other parts of the world and you can find that last gen AMD is cheaper than last gen intel and the other way around to. Or you may find that neither of those are available in your country, and only have Zen 3 and Alder Lake as options.
You’re essentially buying a CPU for the next generation of games which are supposed to rely more on physics and simulation. See nvidias explanation of its toy race car demo RacerX. The X3-D chips still have a purpose around this
Any chance of an overclocking guide for 12th/13th Gen? I know you can basically leave them alone, but i want to have a play about but don't know where to start and haven't seen any decent overclocking guides for these CPU's at all, also love your videos keep it up
Since things are so close, it really comes back to RAM for me. I want to run 128GB and wonder which platform will be able to run all 4 DIMMs filled, the fastest.
I’ve got a 7950X on a X670E ProArt with 4x32GB DIMMs, it works great. The memories are Kingston Fury 5600 2x32GB 40-40-40, but I’ve got them overclocked at 6000 36-36-36 right now. Might not be super stable in the long run, but a combined sixteen hours of memtest86 suggests I could probably tighten those timings a bit more. The board supposedly supports ECC RAM but I haven’t been able to test that yet.
Then the case is made! We need cost of ownership, performance per watt included in the value vs performance calculations. Wendel! Be the first to introduce the perf/watts/cost vein diagram!
Props for AMD for knocking a sense to Intel on their crazy prices since I can remember, they'd didn't have any competition before and they could just make prices that's so absurd and not economical for most normal users, having 6 cores before is almost very hard to get without burning your money until AMD came and just made their own way and changed the game
Regarding relative performance within individual Geekbench tests, I imagine compiler optimization is still playing a role. This is where Intel carries historical advantages, where, for example, the Intel compiler will perform better on Intel instructions. Particularly for old instruction sets.
I love your stuff but man dude I would say most people watching you are us nerds. And you bein you are one of us. I am sure I am not the only person that would like to see some Linux compile time and node / npm install times and vs and rider eclipse build times of something with known hardcore built times . Personally I would love to see this list Windows Linux compile time Linux Linux compile time Windows > Vs/Rider >Chromium compile time And being the .Net flunky I am would love to see actual comparison of generational flagships fighting it out in vs Roslyn and vs + rider build.
I might consider an upgrade. Currently I use 12 years old 8 core AMD System with 16GB RAM with Debian as my developer "workstation". It's super quiet. Everything that requires horse power runs in the Servers in the 19" Rack. Till now I never had a reason to upgrade
I once upon a time I told my friends that AMD would need the 7600x to be 12 cores and the 7700x to be 16, and the 7900x 24 cores, and then no 7950x unless AMD feels it's needed at a massive 32 cores. And they called me crazy, but largely I still think it makes sense. If AMD is trying to charge a premium tier for tier over Intel.
Don't forget that the e cores don't have threads, so at the top end it's still 32 threads all round with the 13900k vs 7950x, but the 7950x has all of those in performance cores. There is some scheduling issues on Windows that have shown up on the 2xCCD chips, so I'd like to see if that gets fixed before calling a winner.
maybe they needed to rework their CCD/CCX configuration, but literally doubling cores across the stack was probably too extreme. I think the main problem for AMD is the pricing. $300 for a traditional 6 core performant part is a bit much to ask given trend in progress. If they had come out $50-100 cheaper across the stack, they would be in excellent position against Intel's offerings instead of looking silly as they do now.
@@Knirin Not everyone thinks that, I dont think that, anyone paying attention to these reviews shouldn't think that No, AMD looks silly because their $300 and $400 6 and 8 core parts get soundly beat by Intel's current entry level Raptor Lake part that has 6 p-cores and 8 e-cores
The i5 being so much less expensive is making me consider not going high end AMD for some level of upgradability for a dev work + gaming machine that might also do some vm stuff and instead have a cheaper one that has no real path for the future but still likely will work well for quite a while
@@tacticalcenter8658 I'm going to have to rewatch the L1T video on x3D since when it came out I didn't really watch much tech youtube but does 3D V cache do that much for productivity stuff, or is it also that it lowers the prices of the normal CPUs?
@@U1TR4F0RCE X3D only helps applications that use more cache. L1T included zen3 X3D in this review, the next version should be helpful with the new architecture, cause last years x3d is making it difficult to spend the kind of money zen4 currently costs in comparison. Though depending on your needs of course.
Don't get Alder Lake or Raptor Lake for anything that uses virtualization. The entire virtualization stack requires scheduler rewrites for VCPU threads to both A) know the difference between a P Core and an E Core and B) properly schedule threads between them.
I was thinking in getting the 13700k but I currently own the noctua c14s, should I stick to that cooler or should I get something else, like a water cooler?
I got the 5800X3D after the 7K launch. It's cool to see the new processors put out all these numbers, but the X3D was the best value for me since I primarily game.
I am looking forward to seeing what cooling options are going to come out in the next 4-8 months for these new AMD and Intel CPUs. Yea the vendors are saying that high temps are the new normal… I can see cooling manufactures taking this as a challenge :)
What's the deal with "Min" FPS measurements nearly always being above 1% and 0.1% lows? Is it just a quirk with the stat maths being unintuitive, where the metric only takes into account the minimum number of frames rendered in any given second interval, despite instantaneous frametimes being much worse at times?
@@DerekDavis213 so by that logic only americans should buy this newer gen power hungry cpu , meanwhile the rest of the globe is crying for energy crisis
@@gunturbayu6779 My logic is this: anyone that can afford over $1000 USD to build a PC with Zen4 or Raptor Lake is not going to care about an extra $10 per month on the electricity bill. The average American salary is about $60,000. Electricity bill is cheap, nothing to worry about.
Yeah…I’m shooting for a 6+ year system. Been using my 6th gen i7 with a 1080 for 6 years with no upgrades come December. That is a record for me. I’m moving to a 13th gen i9 with a 3080ti…and moving up to a Maximus Hero (running a Sabertooth right now). I’m sure I will be happy with those upgrades for a long time to come.
Noob question: why does adding more voltage increase cpu cycles? Is there some sort of mechanism on the chip that sends out the timing codes/clocks based on the amount of voltage it receives?
The VRay is a big Quandry for me, as VRay builds for render with Intels own Embry (I think?). Have to see if it is athe avx 512 maybe? Don't know if VRay supports it.
I heard one of the big reasons for going DDR5 is the 13000 series. I recently built a MSI Unify Z690 w/ DDR5 & a 12700KF . I got a good deal on some open box products. But maybe I may upgrade to a 13700KF.
13600 vs 12700: about the same power consumption, about the same performance. But 13600 is about $50 cheaper. And we are supposed to be excited about 50 bucks?
can you please do a Hypervisor test to see how good the E-cores vs P-cores for Hyper-V and esxi perhaps? I'm very skeptical on the littleBig cores from Intel.
The answer is badly. Unless your application eats every resource available, Blender, Cinebench, video encoders, or is specially designed to keep a control thread on an E-Core and heavy processing threads only on the P-Cores, rarely tried outside of HPC workloads or some video game engines the E-Cores cost you performance.
@@Knirin I have suspected as much myself when it comes to E-cores for virtualization. But I think Wendell should have done it considering he is a networking guy at heart. But it would put Intel in a bad light.
Wendell, do you think it would be viable for Intel to sell some processors that are only e- cores for office and home server production? Something like ARM but by Intel or even AMD that focuses on more efficiency and more cores for different workloads. I feel business would prefer lower power usage than more performance for normal b office work. Plus, it might be feasible to make a 16 core lower power chip for home servers.
Technically yes. Bad, very, very bad idea from both a performance and power perspectives. It takes one of three very specific workloads to seriously load multiple cores in a modern processor. Gaming, tile based rendering, and video/audio encoding. Nothing else, even code compilation, manages to regularly use more than 90% of 8 threads. The problem is the E Cores are .5 of a P core in a single thread workload and a P core has hyper-threading. Yes a P core will pull more power at full load than 4 E Cores but it can handle single thread tasks, most apps, a lot faster and can sleep as efficiently in a computer seeing occasional mouse movements.
i think i missed the original release of the P-core E-core thing... whats the physical difference between a P-core running at say 500mhz and a E-core running at 500mhz? Yes I get that "P-cores aRe FasTeR". Is it a binning thing? are the E-cores the "bad" ones so they put a frequency cap on those? I ask here because googling this sort of thing results mostly in "theyre more optimized".
The core design is wildly different, the most obvious thing being the lack of multi threads on e cores. You can fit like 4 or 6 e cores in the die space of a single P core I think. As the other guy mentioned, they're typically aimed to be around 6th gen in performance while sipping power, so not useless by any means but a long way from what would be considered high performance these days.
24 is not a proper power of 2 ... But really it's actually a 8C/16T CPU plus a "CCX" of 16 ATOM cores ... like ARM's big.LITTLE just that big.LITTLE differentiates not by features in the available instruction sets but by ISA implementation and there get it's high power, low power capabilities. Think the way Intel did it might be a great - say: fancy - engineering way but the way ARM did it is far more pragmatic and less convoluted, never heard of a scheduler problem in Linux (effectively there is no Windows for ARM) for big.LITTLE but did so for Intel. Just for show copied that from Wikipedia (added Thumb-2 as it was missing), it show IMO very straight out that these are two time the same ISA with a different µ-architectural implementation: ARM Cortex-A53 - 8-stage pipelined processor with 2-way superscalar, in-order execution pipeline - DSP and NEON SIMD extensions are mandatory per core - VFPv4 Floating Point Unit onboard (per core) - Hardware virtualization support - Thumb-2 instruction set encoding reduces the size of 32-bit programs with little impact on performance. (actually called T32 in the documentation @ ARM developer) - TrustZone security extensions (things I deemed TL;DR and just CnP'ed): - 64-byte cache lines - 10-entry L1 TLB, and 512-entry L2 TLB - 4 KiB conditional branch predictor, 256-entry indirect branch predictor ARM Cortex-A57 - Pipelined processor with deeply out of order, speculative issue 3-way superscalar execution pipeline - DSP and NEON SIMD extensions are mandatory per core - VFPv4 Floating Point Unit onboard (per core) - Hardware virtualization support - Thumb-2 instruction set encoding reduces the size of 32-bit programs with little impact on performance. (actually called T32 in the documentation @ ARM developer) - TrustZone security extensions (things I deemed TL;DR and just CnP'ed): - Program Trace Macrocell and CoreSight Design Kit for unobtrusive tracing of instruction execution - 32 KiB data (2-way set-associative) + 48 KiB instruction (3-way set-associative) L1 cache per core - Integrated low-latency level-2 (16-way set-associative) cache controller, 512 KB, 1 MB, or 2 MB configurable size per cluster - 48-entry fully associative L1 instruction Translation Lookaside Buffer (TLB) with native support for 4 KiB, 64 KiB, and 1 MB page sizes - 4-way set-associative of 1024-entry L2 TLB - 2-level dynamic predictor with Branch Target Buffer (BTB) for fast target generation - Static branch predictor - Indirect predictor - Return stack
Please someone tell me where all of their accents area coming from? All three of them have the same accent more or less. Tennesee? Kentucky? Virginia? What?
The performance is cool, but the 250W TDP is wild for anyone not using workstation apps. Everyone who gets an Optiplex or HP with a non-K CPU is gonna be shocked by how much performance they lose by going down to 65W. By comparison, 5800X3D owners will be eating good for years.
It would be nice if you label the ".1%" lows as "0.1%", it's very difficult to see the difference otherwise and it gets confusing. Great review otherwise :)
that's the problem, we have to imagine it. It could very well be that 3D cache for Zen 4 wont have as big of an impact as it did for Zen 3, especially if there are thermal constraints like what we saw with Zen 3. Hopefully that issue is ironed out with the experience they got from their first go around with it, and we don't have to give up too much of Zen 4's significant gains in clock-rate.
interaction
Agreed
interlinked
Continued
Engaged
Activated
Mobo: "How much power do you want?"
Raptor Lake: "Yes"
clever girl
@@derek400004 nice Jurassic park reference
The 13900K, yes. Ostensibly, the stock 13600K, no.
This aged like fine wine lol
We have to thank the competition from AMD for there to be 14 cores at 300 bucks! Just a few years ago 14 cores cost you in the thousands of dollars range.
@@leeloodog Agree with you, I meant to thank both.
This Channel is the Real Deal.
Our man Wendell bringing some solid research to the table. 🏆
I find Derbauer's power limiting @ 90% testing of the 13900k to be very very interesting. I'd like to see testing at at 75% and 80/85% as well.
That 1% lows was definitely worth looking into, very curious.
90W not 90%
What I'm seeing is that behind-the-motherboard-fans will be the next modding trend...
"This is amazing guys, no more back-panel melting with this awesome mod"
I mean that was a thing for fx and it worked really well. Although it was more to cool down the vrms/powerplane.
I used to just put doubled thermal pads on the bottom of the motherboard. Probably not very efficient, but the back of the case got really warm so I guess it worked. As a bonus it held the backplate in place in case I needed to remove the cooler.
Would be fun to see rear motherboard heatsinks.
This was actually a thing with old Antec cases
I think just trying to get a sale on a 5800x3d and keeping all my parts gets me so close in performance still
AMD's edge is AVX-512. Cinebench R23 doesn't use it, but Blender does
Also 12th gen Alder Lake chips, my 12600K included has AVX-512, but you have to disable E-cores in the BIOS in order for the AVX-512 option to come available.
Intel removed AVX-512 on later releases of 12th Gen CPUs, and newer BIOS updates for Z690 boards will stealth disable AVX-512.
TimeSpy / Extreme by default uses AVX, but on TimeSpy Extreme Custom, you can specify to run the benchmark with AVX, AVX2 or AVX-512.
@intel asked you not to share this information Wendell .... lmao You are like F it! 🤣👌 check and pause at 3:01
Home "gaming grade" PCs are becoming real problem for the electrical wiring gauges and breakers for rooms in many flats. I have my workstation/gaming PC, an intel iMac and a ryzen TV computer in my living room with a TV, HIFI surround set up and some pro audio gear etc. If I upgrade to a 4090 and and a CPU like this one I would probably be dangerously near tripping the 10A 240V breaker and the wiring to the wall outlets doesn't really support more and this flat was renovated before I moved in this year, lol. I would for sure have to move my floor AC unit to an extension power cable from another room.
You use 240 for your pc?
@@spiceyfrenchtoast9421 yes, I live in Europe which is one of the places where 220-250v is the default wall socket voltage. Bathrooms occasionally has a 115v outlet but other than than it’s 230 everywhere here. So not 240, that was a typo but around there.
4090 isn't home gaming rade cpu, what the hell are you talking about?
@@wiziek it is high end but for sure something a lot of people will buy to play games at home with. Also, whenever 5090 is released the second hand market will get a bump.
@@spiceyfrenchtoast9421 it’s not as much as it seems. It IS more but not like it seems. Notice the 10 amps? 120v is at 15 or 20 amps. Our 240v split phase starts at 30 amps (two legs of 120v) and 50 amps is common too.
Drink Every time Wendell says "Cores"
By the way did he mention how many cores the i5 has?
I can't believe I forgot that!
Lol, the E-cores alone smoke my old Sandy Bridge processor. I think we have forgotten how far we have come in the last decade or so.
Great breakdown Wendel! So nice to see something other than the usual barrage of gaming benchmarks
GN and Hardware unboxed appeared unimpressed with the 13900K seemed to think the 13600K was the better chip primarily due to the power useage
Can you please do a video on the big little cores and OS scheduler with the 13900k to see how bad frame pacing is for gaming?
How much more you can push when you have competition. ;)
Fun fact: About the same amount of time has passed from the launch of the 1800X until today than what it took Intel to get from a Sandy Bridge quad core to a Kaby Lake quad core when AMD was nowhere to be seen.
Really looking forward to the i7 video woo!
Now to find a week 2 2022 engineering sample with AVX512 enabled, but all the cache increases from this generation 🏎💨
It's no longer a race between AMD and Intel in CPUs. It's an all-out brawl going on at the finish with various limbs flailing over the line at different times.
"it would be an hour long, and no one would care" You mean normies won't care. I have autism. I would happily watch an hour long processor review.
I want more rambling Wendell . I hope Wendell does a live stream with Ian Cutress soon discussing the new chips. Once all the new X3D chips are released .
Bonus points for putting Spock in the background
26:40. First TNG reference? Though at 21:22, dual Utube(?) screens of Spock and Kirk. Though Spock with a weapon and Kirk on the ground...is this a message?
Spock has the high ground under a red sky vs Kirk laying in the dirt under a blue sky.... just sayin' :P
When they give me a motherboard that will last 3-5 years and a CPU that consumes normal amounts of power in the high end, and still be able to upgrade to a similar one after those years, then I will consider them.
If you built this system with a 13900k from the start I don’t think there would be a reason to upgrade anything for probably 10 years ….. especially if you aren’t doing some type of Niche computer rendering for a living where saving a few second on a render will make you more money…
And gaming these new CPUs will hold up for gaming most likely for a decade at this point also . Considering you can still game at 1080p with a i7 3770k today … the new systems having Pcie5 support when pcie4 isn’t even saturated all the way with memory or GPUs … they are at a point that they are very future proof honestly.
Only thing would be nice is if they had at least 40 pcie lanes but that’s asking a lot of desktop PCs I’m sure that will be the next core race that we have had since Ryzen in 2017 started the 6 core minimum up to 16 . And Intel has taken this to the max giving people way way way more threads /cores than anyone will ever need. Very few actual PC users will need a 13900k’s multithread performance that’s why the 13600k will most likely be the best seller having 8 efficiency cores that are basically 7700k non Smt cores added on to the latest 6 SMT core design is huge that’s so much power it’s really actually crazy for a i5 …
That and building one with a ddr4 motherboard since ddr4 is so cheap right now that will give a nice upgrade maybe in a couple years where you can pick up a ddr5 board and ddr5 if you want to but my guess is there won’t be enough of a benefit with ddr5 to justify the upgrade for maybe 5 years … kinda like ddr3 to ddr4 if you had a high end ddr3 build like a i7 4790k you really didn’t have to upgrade your PC till DDR5 was already being rumored for release…
Again that app depends on what your use cases are for your PC .
The people who actually will capitalize on having the newer hardware for work will be the first adopters and the enthusiast gamers… take as old as time.
Great, concise info. Thanks! My go-to channel from now on.
We have come to a point that for gaming at 1440p or 4K, any "new" gen CPU + a midrange to high-midrange GPU will give you about the same performance.
This is now the "get the one that its in stock and its cheaper in your country", yes because prices are all over the place in other parts of the world and you can find that last gen AMD is cheaper than last gen intel and the other way around to. Or you may find that neither of those are available in your country, and only have Zen 3 and Alder Lake as options.
You’re essentially buying a CPU for the next generation of games which are supposed to rely more on physics and simulation. See nvidias explanation of its toy race car demo RacerX. The X3-D chips still have a purpose around this
Still remember few generations ago, when they were stuck for 4 cores max, even with i7? Thanks god for ryzen, now we have 14 cores in i5 !! :DDD
Any chance of an overclocking guide for 12th/13th Gen? I know you can basically leave them alone, but i want to have a play about but don't know where to start and haven't seen any decent overclocking guides for these CPU's at all, also love your videos keep it up
Since Intel /amd furnished guidance data, would be nice to see them along with actual data
Since things are so close, it really comes back to RAM for me. I want to run 128GB and wonder which platform will be able to run all 4 DIMMs filled, the fastest.
I’ve got a 7950X on a X670E ProArt with 4x32GB DIMMs, it works great. The memories are Kingston Fury 5600 2x32GB 40-40-40, but I’ve got them overclocked at 6000 36-36-36 right now. Might not be super stable in the long run, but a combined sixteen hours of memtest86 suggests I could probably tighten those timings a bit more.
The board supposedly supports ECC RAM but I haven’t been able to test that yet.
@@СусаннаСергеевна Thanks for the heads up on the board and RAM
Then the case is made! We need cost of ownership, performance per watt included in the value vs performance calculations. Wendel! Be the first to introduce the perf/watts/cost vein diagram!
Bottom line all these cpu's discussed are good, just pick one you like!
22:01. Ryan heading home in his Vette
Props for AMD for knocking a sense to Intel on their crazy prices since I can remember, they'd didn't have any competition before and they could just make prices that's so absurd and not economical for most normal users, having 6 cores before is almost very hard to get without burning your money until AMD came and just made their own way and changed the game
🤣🤣was that a Raptor Lake V8 cold start at 22:01?
I came back to watch this again and noticed you’ve really lost some weight. Whatever you’re doing is working.
I love that both Nvidia and Intel are scared enough of AMD to max out their power limits. We should see pricing come down as competition flourishes.
Regarding relative performance within individual Geekbench tests, I imagine compiler optimization is still playing a role. This is where Intel carries historical advantages, where, for example, the Intel compiler will perform better on Intel instructions. Particularly for old instruction sets.
I love golden samples of 12900K! I also love to call them RaptorLake.
i have 13600k paired with 3070 strix / 32 GB DDR5 (6000mhz) RAM. i think im good for another year or two.
never been a big E core fan, this generation has swung me though, finally seems like a reasonable solution
A rambling numbers video? My favorite gib more
picked one up today at microcenter for their advertised price of $569, boy she runs HOT
I love your stuff but man dude I would say most people watching you are us nerds. And you bein you are one of us.
I am sure I am not the only person that would like to see some Linux compile time and node / npm install times and vs and rider eclipse build times of something with known hardcore built times .
Personally I would love to see this list
Windows Linux compile time
Linux Linux compile time
Windows > Vs/Rider >Chromium compile time
And being the .Net flunky I am would love to see actual comparison of generational flagships fighting it out in vs Roslyn and vs + rider build.
Are we going to see Z790 motherboard reviews?curious about the gigabyte aero g
I might consider an upgrade. Currently I use 12 years old 8 core AMD System with 16GB RAM with Debian as my developer "workstation". It's super quiet. Everything that requires horse power runs in the Servers in the 19" Rack. Till now I never had a reason to upgrade
Has anyone taken a look at the Dell and Lenovo AMD Threadripper Pro 59xxWX workstations for thermals?
I once upon a time I told my friends that AMD would need the 7600x to be 12 cores and the 7700x to be 16, and the 7900x 24 cores, and then no 7950x unless AMD feels it's needed at a massive 32 cores. And they called me crazy, but largely I still think it makes sense. If AMD is trying to charge a premium tier for tier over Intel.
Don't forget that the e cores don't have threads, so at the top end it's still 32 threads all round with the 13900k vs 7950x, but the 7950x has all of those in performance cores.
There is some scheduling issues on Windows that have shown up on the 2xCCD chips, so I'd like to see if that gets fixed before calling a winner.
maybe they needed to rework their CCD/CCX configuration, but literally doubling cores across the stack was probably too extreme.
I think the main problem for AMD is the pricing. $300 for a traditional 6 core performant part is a bit much to ask given trend in progress. If they had come out $50-100 cheaper across the stack, they would be in excellent position against Intel's offerings instead of looking silly as they do now.
@@bojinglebells AMD only looks silly because everybody thinks an E Core is just a down clocked P Core. They aren't. They are .25 of a P Core.
@@Knirin Not everyone thinks that, I dont think that, anyone paying attention to these reviews shouldn't think that
No, AMD looks silly because their $300 and $400 6 and 8 core parts get soundly beat by Intel's current entry level Raptor Lake part that has 6 p-cores and 8 e-cores
The i5 being so much less expensive is making me consider not going high end AMD for some level of upgradability for a dev work + gaming machine that might also do some vm stuff and instead have a cheaper one that has no real path for the future but still likely will work well for quite a while
Consider the future AMD zen4 x3d. Worth waiting for at these prices.
@@tacticalcenter8658 I'm going to have to rewatch the L1T video on x3D since when it came out I didn't really watch much tech youtube but does 3D V cache do that much for productivity stuff, or is it also that it lowers the prices of the normal CPUs?
@@U1TR4F0RCE X3D only helps applications that use more cache. L1T included zen3 X3D in this review, the next version should be helpful with the new architecture, cause last years x3d is making it difficult to spend the kind of money zen4 currently costs in comparison. Though depending on your needs of course.
Don't get Alder Lake or Raptor Lake for anything that uses virtualization. The entire virtualization stack requires scheduler rewrites for VCPU threads to both A) know the difference between a P Core and an E Core and B) properly schedule threads between them.
@@Knirin thank you I was not aware of that issue and definitely means I'll probably go high end new AMD instead
I was thinking in getting the 13700k but I currently own the noctua c14s, should I stick to that cooler or should I get something else, like a water cooler?
I agree. Classic Star Trek is best =D There will never be another Classic Kirk. Shatner has become eternal!
0:25 "I do have the I5 12600k"
Still wanting to upgrade to the 5800x3D after all of today's reviews.
I got the 5800X3D after the 7K launch. It's cool to see the new processors put out all these numbers, but the X3D was the best value for me since I primarily game.
If you have a board and memory already, it's a no-brainer. I did just that for a linux gaming rig.
Why? If you only game it's a waste of money and if you do multithreaded stuff and game you bought the wrong AMD cpu.
@@ImDembe what? It's still one of the best gaming CPUs, and a super cheap upgrade for me cuz it would be a drop in replacement of my 3600.
@@Dan-Simms My bad, i read the post all wrong. :)
Glad i got my 5800 x3d.
What happened to the "Productivity Testing" slides?
Sorry if this is the 100th time this has been asked.
good new competition, good for VMs ???
So am I wrong, 1 m.2 before your cards go x8? Not the biggest deal if true but still
I am looking forward to seeing what cooling options are going to come out in the next 4-8 months for these new AMD and Intel CPUs. Yea the vendors are saying that high temps are the new normal… I can see cooling manufactures taking this as a challenge :)
If the limitation is in the design of the core to heat spreader, there 's not really much room for improvement AFTER the heat spreader.
What's the deal with "Min" FPS measurements nearly always being above 1% and 0.1% lows? Is it just a quirk with the stat maths being unintuitive, where the metric only takes into account the minimum number of frames rendered in any given second interval, despite instantaneous frametimes being much worse at times?
Fuzzy focus Wendell is the best Wendell.
With great power comes great ... electricity bill.
So true! xD
Many Americans spend $15 every day for Starbucks and gasoline to drive to work. Electricity for the home computer is nothing.
@@DerekDavis213 so by that logic only americans should buy this newer gen power hungry cpu , meanwhile the rest of the globe is crying for energy crisis
@@gunturbayu6779 My logic is this: anyone that can afford over $1000 USD to build a PC with Zen4 or Raptor Lake is not going to care about an extra $10 per month on the electricity bill.
The average American salary is about $60,000. Electricity bill is cheap, nothing to worry about.
If you're going to update your computer in a couple years, you'll be on an entirely new motherboard which kind of shoots the price savings.
Yeah…I’m shooting for a 6+ year system. Been using my 6th gen i7 with a 1080 for 6 years with no upgrades come December. That is a record for me. I’m moving to a 13th gen i9 with a 3080ti…and moving up to a Maximus Hero (running a Sabertooth right now). I’m sure I will be happy with those upgrades for a long time to come.
Noob question: why does adding more voltage increase cpu cycles? Is there some sort of mechanism on the chip that sends out the timing codes/clocks based on the amount of voltage it receives?
The VRay is a big Quandry for me, as VRay builds for render with Intels own Embry (I think?). Have to see if it is athe avx 512 maybe? Don't know if VRay supports it.
NUTS!!!!! 14 CORE I5!!!!!! MIND BLOWN!!!! my 12 core 5900x seems inadequate, lmbo!!!!
Not all cores are made equal. Especially not big little. I am amused how much people trying to ignore that these e-cores are not so e.
Hmm the KF models dosn't come with ECC support...
14 CORES BABY!!!!!!!!!!
But 8 of them are *WEAK* .
@@DerekDavis213 i know just love the way he says 14 core alot lol
What is the go to board chipset for a productivity value build with i5-13600 and ddr4? B660 or Z690? Does one or the other affect upgrade path?
I heard one of the big reasons for going DDR5 is the 13000 series. I recently built a MSI Unify Z690 w/ DDR5 & a 12700KF . I got a good deal on some open box products. But maybe I may upgrade to a 13700KF.
What upgrade path? There is no upgrade path, you can only get 12 and 13 gen cpu on these boards.
Would be interesting to see 13600 vs 12700
13600 vs 12700: about the same power consumption, about the same performance.
But 13600 is about $50 cheaper. And we are supposed to be excited about 50 bucks?
@@DerekDavis213 he's a seller.
@@user-xx8tn4kx4l Who is a seller?
How is your min frame rate higher than your 1% low?
can you please do a Hypervisor test to see how good the E-cores vs P-cores for Hyper-V and esxi perhaps? I'm very skeptical on the littleBig cores from Intel.
The answer is badly. Unless your application eats every resource available, Blender, Cinebench, video encoders, or is specially designed to keep a control thread on an E-Core and heavy processing threads only on the P-Cores, rarely tried outside of HPC workloads or some video game engines the E-Cores cost you performance.
@@Knirin I have suspected as much myself when it comes to E-cores for virtualization. But I think Wendell should have done it considering he is a networking guy at heart. But it would put Intel in a bad light.
that shirt collar
Wendell, do you think it would be viable for Intel to sell some processors that are only e- cores for office and home server production? Something like ARM but by Intel or even AMD that focuses on more efficiency and more cores for different workloads. I feel business would prefer lower power usage than more performance for normal b office work. Plus, it might be feasible to make a 16 core lower power chip for home servers.
Technically yes. Bad, very, very bad idea from both a performance and power perspectives. It takes one of three very specific workloads to seriously load multiple cores in a modern processor. Gaming, tile based rendering, and video/audio encoding. Nothing else, even code compilation, manages to regularly use more than 90% of 8 threads.
The problem is the E Cores are .5 of a P core in a single thread workload and a P core has hyper-threading. Yes a P core will pull more power at full load than 4 E Cores but it can handle single thread tasks, most apps, a lot faster and can sleep as efficiently in a computer seeing occasional mouse movements.
Nice review thank you! Definitely want a 13900k with my next upgrade. Can't wait to use this for virtualization.
i think i missed the original release of the P-core E-core thing...
whats the physical difference between a P-core running at say 500mhz and a E-core running at 500mhz?
Yes I get that "P-cores aRe FasTeR". Is it a binning thing? are the E-cores the "bad" ones so they put a frequency cap on those?
I ask here because googling this sort of thing results mostly in "theyre more optimized".
E cores are like Skylake cores in performance but atom cores in power efficiency. Basically a phone or laptop like efficiency modified to desktop
The core design is wildly different, the most obvious thing being the lack of multi threads on e cores. You can fit like 4 or 6 e cores in the die space of a single P core I think.
As the other guy mentioned, they're typically aimed to be around 6th gen in performance while sipping power, so not useless by any means but a long way from what would be considered high performance these days.
One block of 8 E Cores is an overclocked i7 6970 consuming about 35W of power.
24 is not a proper power of 2 ...
But really it's actually a 8C/16T CPU plus a "CCX" of 16 ATOM cores ... like ARM's big.LITTLE just that big.LITTLE differentiates not by features in the available instruction sets but by ISA implementation and there get it's high power, low power capabilities. Think the way Intel did it might be a great - say: fancy - engineering way but the way ARM did it is far more pragmatic and less convoluted, never heard of a scheduler problem in Linux (effectively there is no Windows for ARM) for big.LITTLE but did so for Intel.
Just for show copied that from Wikipedia (added Thumb-2 as it was missing), it show IMO very straight out that these are two time the same ISA with a different µ-architectural implementation:
ARM Cortex-A53
- 8-stage pipelined processor with 2-way superscalar, in-order execution pipeline
- DSP and NEON SIMD extensions are mandatory per core
- VFPv4 Floating Point Unit onboard (per core)
- Hardware virtualization support
- Thumb-2 instruction set encoding reduces the size of 32-bit programs with little impact on performance.
(actually called T32 in the documentation @ ARM developer)
- TrustZone security extensions
(things I deemed TL;DR and just CnP'ed):
- 64-byte cache lines
- 10-entry L1 TLB, and 512-entry L2 TLB
- 4 KiB conditional branch predictor, 256-entry indirect branch predictor
ARM Cortex-A57
- Pipelined processor with deeply out of order, speculative issue 3-way superscalar execution pipeline
- DSP and NEON SIMD extensions are mandatory per core
- VFPv4 Floating Point Unit onboard (per core)
- Hardware virtualization support
- Thumb-2 instruction set encoding reduces the size of 32-bit programs with little impact on performance.
(actually called T32 in the documentation @ ARM developer)
- TrustZone security extensions
(things I deemed TL;DR and just CnP'ed):
- Program Trace Macrocell and CoreSight Design Kit for unobtrusive tracing of instruction execution
- 32 KiB data (2-way set-associative) + 48 KiB instruction (3-way set-associative) L1 cache per core
- Integrated low-latency level-2 (16-way set-associative) cache controller, 512 KB, 1 MB, or 2 MB configurable size per cluster
- 48-entry fully associative L1 instruction Translation Lookaside Buffer (TLB) with native support for 4 KiB, 64 KiB, and 1 MB page sizes
- 4-way set-associative of 1024-entry L2 TLB
- 2-level dynamic predictor with Branch Target Buffer (BTB) for fast target generation
- Static branch predictor
- Indirect predictor
- Return stack
intel chip thing is power and hungry with wendell? tune in next time to boost channel.
Please someone tell me where all of their accents area coming from? All three of them have the same accent more or less. Tennesee? Kentucky? Virginia? What?
good stuff
holy cow, i get it, it has 14 or whatever cores
Yeah, but did you know the i5 has 14 cores. ~Editor Autumn
🤣
7-10 nm Lithography in 2022 yah and no one even mentioned that.
The 13600K has great performance/$. It would be my choice if I wouldn't have a 5900X already.
13500 will be mostly the same for less if you can wait for it and dont need unlocked.
@@trignite wrong 13500 wont have high clocks
My new 12700k i7 only has 12 cores... ;( Damn these yearly "never the right time to buy" days...
How many cores in the i5?
Don't even think he brought that up. Very strange.
Manage to cool 13900k and it will be powerful
mfw 14 cores in an i5
WTF is plucked supposed to mean?
Woooooooo!
engagement
The performance is cool, but the 250W TDP is wild for anyone not using workstation apps. Everyone who gets an Optiplex or HP with a non-K CPU is gonna be shocked by how much performance they lose by going down to 65W. By comparison, 5800X3D owners will be eating good for years.
Wait, how many cores?
Waiting for the 13400. :)
i5-13400, i5-13500, i5-13600 (non-K)
it's literally a 12600 non-K rebranded. ALL Raptor Lake i5 and lower non-K's are rebranded 12th gen. Go buy it right now.
@@namyun2743 they're all rebranded 12th gen i5's, just buy the current i5's now.
it will be interesting if 13th gen smashes 5800x3d on 4090 as LTT video showed.
my 13600k is sooooo good 🤤
How many cores in an i5? It wasn't clear
6 P-cores + 8 E-cores = 14 total physical cores and 20 threads.
It would be nice if you label the ".1%" lows as "0.1%", it's very difficult to see the difference otherwise and it gets confusing. Great review otherwise :)
Blessed be the holy Algorithm
Just imagine what the upcoming ryzen 7 series with 3dv cache will be like!!!! Intel may be in trouble again!!!
that's the problem, we have to imagine it. It could very well be that 3D cache for Zen 4 wont have as big of an impact as it did for Zen 3, especially if there are thermal constraints like what we saw with Zen 3. Hopefully that issue is ironed out with the experience they got from their first go around with it, and we don't have to give up too much of Zen 4's significant gains in clock-rate.
Engram entered