Intel Announces 13th Gen "Raptor Lake" - Linux Benchmarks To Come

[piotrj3]

Power-efficiency is a set of data points (which can be visualized in a chart as a curve) - not a single data point. Making judgements about 7950X vs 5950X power efficiency based on a single data point is absurd.

All Ryzen 7000 CPUs are being sold without a cooling solution, which means that more people than before will have to learn how to properly configure their Ryzen 7000 systems for power-efficiency.

I was comparing it on typical multithread (cinebench, rendering) workload with strong cooling when 7950X seems to have unrestricted power draw. Of course when Intel was occasionally drawing a little too much over declared TDP of 241W PL in AVX512 workloads (like 270W TDP) everyone cried how Intel is fake. Meanwhile when AMD pushes power limit from 170W up to 250W on just normal Blender render test here (https://youtu.be/nRaJXZMOMPU?t=540) no one bats an eye. Considering now Intel more abids TDP I wouldn't be suprised if 13900k in 90W TDP mode will have same or greater performance then AMD 90W TDP mode (called 65W TDP).

I am not even talking about efficiency in more normal workload like gaming (frames per watt), there Intel in der8auer test already was more efficient from start (in alder lake vs zen 3) until 5800X3D came.

[ms178]

as i unterstand it they did not chanceling the future dGPU.... with their DG2 lineup they had 4 different GPU dies for this generation.
this is the point they change in the future... in the future the DG3 lineup will only be 1 single gpu die...
and they chancel the highend part means they only produce a 8gb vram version they can use for notebooks and lowend/midrange for Desktop. with this stradegy they can spend more time to optimise drivers for this 1 single DG3 chip instead if split software development resources into 3-4 chips...
intel did accept that with that driver status they can not compete in the highend... but they want to sell notebooks with an extra GPU and also want one PCIe card for developers and so one.

DG4 will also be a single chip .... they will do this until their driver and gpu tech become competive.

Yeah, we both discussed this as a plausible strategy going forward the other day, but I want to hear that straight out of Pat Gelsinger's mouth to be assured that buyers don't end up getting burned due to Intel jumping the ship prematurely.

[AdrianBc]

Raptor Lake will outperform Zen4 in multicore performance as well. The 13900K will use more power than the 7950X but the 13700K/13600K should be more efficient than the 7600X/7700X due to the higher core count. AMD’s advantage will be AVX-512.

Because now both top models of Intel and AMD have 32 threads, the difference in multi-threaded performance will be indeed much smaller than with Alder Lake. Nevertheless, at equal power consumption Zen 4 will be faster, so Raptor Lake will win only when consuming more power, as you say.

Among multi-threaded applications, it is likely that the Raptor Lake models matching a Zen 4 in the number of threads will be faster for software compilation and for similar tasks that do not benefit from AVX or AVX-512, because for such tasks the Intel efficient cores have a performance similar to that of a thread of a big Intel or AMD core.

On the other hand, for scientific computing or other floating-point applications or any other applications that can benefit from using AVX or AVX-512, Zen 4 will beat easily Raptor Lake, because the Intel efficient cores are weak at AVX.

Moreover, if the program is rewritten or recompiled to use AVX-512, the advantage of Zen 4 can increase a lot, exploiting the stupid market segmentation policies of Intel, which have prevented Intel to have the best support for an instruction set that they have developed more than 14 years ago (the first public description of the first variant of AVX-512 was in 2008, 3 years before Sandy Bridge, which used the inferior AVX instruction set).

[WannaBeOCer]

Because now both top models of Intel and AMD have 32 threads, the difference in multi-threaded performance will be indeed much smaller than with Alder Lake. Nevertheless, at equal power consumption Zen 4 will be faster, so Raptor Lake will win only when consuming more power, as you say.

Among multi-threaded applications, it is likely that the Raptor Lake models matching a Zen 4 in the number of threads will be faster for software compilation and for similar tasks that do not benefit from AVX or AVX-512, because for such tasks the Intel efficient cores have a performance similar to that of a thread of a big Intel or AMD core.

On the other hand, for scientific computing or other floating-point applications or any other applications that can benefit from using AVX or AVX-512, Zen 4 will beat easily Raptor Lake, because the Intel efficient cores are weak at AVX.

Moreover, if the program is rewritten or recompiled to use AVX-512, the advantage of Zen 4 can increase a lot, exploiting the stupid market segmentation policies of Intel, which have prevented Intel to have the best support for an instruction set that they have developed more than 14 years ago (the first public description of the first variant of AVX-512 was in 2008, 3 years before Sandy Bridge, which used the inferior AVX instruction set).

At the end of the day Intel screwed up by removing AVX-512 from consumer hardware due to haters of the instruction set for example Linus Torvalds: https://www.phoronix.com/news/Linus-Torvalds-On-AVX-512

When it comes to the actual workload I doubt AVX-512 workloads are more efficient on a non-native implement of AMD's which is "double pumping" while Intel uses a native implementation of AVX-512. I have a 12700K from the first batch which still supports AVX-512 so if I do get my hands on a 7700X I'd definitely will test the two.

Where are you getting this made up information about Intel's E cores being as quick as a single thread from HT/SMT? Gracemont E cores IPC is on par with Skylake cores used from Skylake up to Comet Lake but clocked lower and on a new node which is why Gracemont is efficient. Intel's hybrid chips aren't big.LITTLE like ARM. They're big.BIGGER, Golden Cove/Raptor Cove cores are massive compared to traditional CPU cores like Gracemont/Skylake cores.

https://chipsandcheese.com/2021/12/2...he-atom-cores/

Edit: I want to add regarding AI, Gracemont cores are efficient to run inference due to the VNNI-INT8 units

[qarium]

Yeah, we both discussed this as a plausible strategy going forward the other day, but I want to hear that straight out of Pat Gelsinger's mouth to be assured that buyers don't end up getting burned due to Intel jumping the ship prematurely.

to be honest buyers already get burned because the ARC gpus are some kind of "shit"

i don't get is why do you want to buy this ? do you think it become better in 1-2 years old because they still support the hardware ?

the A770 can not even beat my vega64 and you get a vega64 for like 250€ on ebay...

so i don't get your point... this intel ARC products are for people who like to play with tech and dont care if they lose money or not.

if you have some gpu compute OpenCL workload who need 16gb vram but no big performance you can buy the a770 it has much vram with no performance what so ever...

but some workload just need vram... i am sure people who just need much vram for their openCL will be happy with the A770...

[WannaBeOCer]

to be honest buyers already get burned because the ARC gpus are some kind of "shit"

i don't get is why do you want to buy this ? do you think it become better in 1-2 years old because they still support the hardware ?

the A770 can not even beat my vega64 and you get a vega64 for like 250€ on ebay...

so i don't get your point... this intel ARC products are for people who like to play with tech and dont care if they lose money or not.

if you have some gpu compute OpenCL workload who need 16gb vram but no big performance you can buy the a770 it has much vram with no performance what so ever...

but some workload just need vram... i am sure people who just need much vram for their openCL will be happy with the A770...

What nonsense are you making up? The A770 is faster than a RTX 3060 which is faster than a RX Vega 64. It also has dedicated hardware for ray tracing and only uses 180w while a RX Vega 64 uses around ~290w

Raja Koduri and his great team made Arc who is the founder of RDNA. Which is why AMD sent him a card when he left.

https://www.pcgamer.com/amd-reunites...graphics-card/

Unlike AMDs consumer GPUs it has tensor accelerators and with 16GB of VRAM that’s going to be a fantastic consumer GPU to be introduced to deep learning utilizing their OneAPI.

[Teggs]

Everything not on the slide hasn’t been announced yet. When Intel announced Alder Lake they also only announced the 12900K/KF, 12700K/KF, and 12600K/KF. no clue where you got this ridiculous rumor from. It wasn’t until three months later until they announced the rest of their desktop lineup.

https://www.anandtech.com/show/17162...w-laptoph-cpus

I don't know how rediculous it is, but the idea comes from analysis of a slide obtained by Igor's Lab here.

Moore's Law is Dead began building a case for that interpretation at 10:28 of this video, followed by a claim that a source at Intel told him about a delay for the Raptor Lake part that was supposed to serve the middle segment.

Because Intel had no Rocket Lake i3s, simply bumping the number on Comet Lake parts and continuing to sell them, I find the idea at least plausible. The cache analysis seems rational, and you either believe that he has a knowledgeble, truthful source or not.

Edit: misremembered how '11th gen' i3 went. (It didn't.)

[WannaBeOCer]

I don't know how rediculous it is, but the idea comes from analysis of a slide obtained by Igor's Lab here.

Moore's Law is Dead began building a case for that interpretation at 10:28 of this video, followed by a claim that a source at Intel told him about a delay for the Raptor Lake part that was supposed to serve the middle segment.

Because Intel had no Rocket Lake i3s, simply bumping the number on Comet Lake parts and continuing to sell them, I find the idea at least plausible. The cache analysis seems rational, and you either believe that he has a knowledgeble, truthful source or not.

Edit: misremembered how '11th gen' i3 went. (It didn't.)

Thanks for confirming your source, I won’t bother clicking on that video. That theorist doesn’t deserve any views.

Edit: https://youtu.be/V3wE1wMIBXs

46:25 they announced 50+ new 13th gen processors. I doubt they’d re-release chips with Alder Lake since they’re competing with Zen 4, that would be foolish.

[Keats]

Thanks for confirming your source, I won’t bother clicking on that video. That theorist doesn’t deserve any views.

Edit: https://youtu.be/V3wE1wMIBXs

46:25 they announced 50+ new 13th gen processors. I doubt they’d re-release chips with Alder Lake since they’re competing with Zen 4, that would be foolish.

Given that AMD seems to have completely given up competing on low-end desktop market, that doesn't seem foolish at all.

[WannaBeOCer]

Given that AMD seems to have completely given up competing on low-end desktop market, that doesn't seem foolish at all.

AMD hasn’t given up on competing in the low-end. AMD unfortunately announced TSMC doesn’t have the capacity to manufacture more chips for them so they prioritized high-end fabrication. Intel doesn’t have to worry about securing capacity in their own fabs. I’m sure Intel will run into that issue if they keep using TSMC for their GPUs.