• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Intel current and future Lakes & Rapids thread

Page 729 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Exist50 said:
That article was in response to your claim that no one was talking about AVX512 efficiency on Alder Lake etc.

More to the point, where AVX512 is power hungry, it also provides a lot more performance, and thus, significantly better efficiency. That power consumption isn't coming from nothing. If you want to talk efficiency, you can't just ignore the performance half the equation.
Click to expand...
Except that AMD's avx-512 is very efficient. I got considerably better performance and the same wattage. The curve optimizer or the chip design works very well.
 
Markfw said:
So you say I am lying ? or the kill-a-watt is showing the wrong results ?
Click to expand...
The fact that you're using kill-a-watt already means you're not quoting CPU power. And there are any number of ways you could have arrived at that number. I don't have to know which to make some basic observations about it.
Markfw said:
Except that AMD's avx-512 is very efficient. I got considerably better performance and the same wattage.
Click to expand...
What are you comparing it to? Zen 3? Alder Lake with AVX512? And either way, irrelevant to the claim that AVX512 tanks efficiency on Intel systems. I never called AMD's implementation bad.
 
Exist50 said:
The fact that you're using kill-a-watt already means you're not quoting CPU power. And there are any number of ways you could have arrived at that number. I don't have to know which to make some basic observations about it.

What are you comparing it to? Zen 3? Alder Lake with AVX512? And either way, irrelevant to the claim that AVX512 tanks efficiency on Intel systems. I never called AMD's implementation bad.
Click to expand...
I CPU, an NVME drive, and a 3070TI (not doing anything) taking 300 watts, when seconds before it was not running an AVX-512 task and it was 175 watts. I have no idea what the bios was doing, but you can't quote BIOS and say it can't happen. Look at Raptor lake. We have seen stock bios take over 300 watts, and thats not supposed to happen.
 
Markfw said:
I CPU, an NVME drive, and a 3070TI (not doing anything) taking 300 watts, when seconds before it was not running an AVX-512 task and it was 175 watts. I have no idea what the bios was doing, but you can't quote BIOS and say it can't happen. Look at Raptor lake. We have seen stock bios take over 300 watts, and thats not supposed to happen.
Click to expand...
For Raptor Lake, motherboards are setting an unlimited PL2, and they're not hiding that fact. So this should be trivial for you to check. Go into the BIOS and see what it's set at.
 
Exist50 said:
AVX512 is power hungry, it also provides a lot more performance, and thus, significantly better efficiency. That power consumption isn't coming from nothing. If you want to talk efficiency, you can't just ignore the performance half the equation.
Click to expand...
That's undeniable for sure. The performance boost can be impressive. So you can make a valid point in saying that on apps that can take advantage of AVX-512 Intel Golden Cove is more efficient than when it's disabled . I think I lost that part on your quote... AMD implementation has so far proven to be more efficient if we measure it by the same standard(Power/Performance)
 
nicalandia said:
That's undeniable for sure. The performance boost can be impressive. So you can make a valid point in saying that on apps that can take advantage of AVX-512 Intel Golden Cove is more efficient than when it's disabled . I think I lost that part on your quote... AMD implementation has so far proven to be more efficient if we measure it by the same standard(Power/Performance)
Click to expand...
Zen 4 is, itself, overall more efficient than Golden Cove. I think to compare the two in an apples to apples fashion, it would make the most sense to compare the performance/power delta relative to the baseline architecture. So a run with only AVX2 support vs with AVX512. Can't say I've seen anyone make that comparison yet, sadly.
 
Markfw said:
Any links to any Genoa benchmarks to compare to ?
Click to expand...
I am working on that right now... The Openbenchmarking is not very user friendly when the date is has not been officially posted(as in this is a leak of an unreleased product)

Edit.

No much in common between the release date benchmarks of Genoa and the scant benchmarks that has been just leaked ahead of lunch

1671573752907.png


The Jayopenbenchmark is the 2S Sapphire Rapids.

2 x Intel Xeon Platinum 8462Y+ vs Genoa
 
Last edited:
Its really hard to find the power usage in that mess, but it looks like the new cores are close to the 7773x, but Genoa trashes them both.
 
Exist50 said:
So a run with only AVX2 support vs with AVX512. Can't say I've seen anyone make that comparison yet, sadly.
Click to expand...
Phoronix had a whole article about it (I remember @nicalandia posted about it before):

AMD 4th Gen EPYC 9654 "Genoa" AVX-512 Performance Analysis - Phoronix

www.phoronix.com www.phoronix.com
tl/dr:
bild_2022-12-21_00040lvfm5.png

bild_2022-12-21_00053j9ihv.png

bild_2022-12-21_00062vai9n.png
 
Markfw said:
Its really hard to find the power usage in that mess, but it looks like the new cores are close to the 7773x, but Genoa trashes them both.
Click to expand...
To be honest, that's an impressive show for That Sapphire Rapids, it's 2S system for a total of 64 cores(The 8462Y is a 32C CPU), But it's a very good benchmark to show case the advantage of AVX-512 and Large Channel capacity

At core Parity Xeon 8462Y is 55% ahead of the Fastest Milan-X(No AVX-512 On those) and at virtual parity with 2S 9374F(32C Per CPU)

1671578231279.png


2S Platinum 8462Y vs 2S EPYC 9374F
 
nicalandia said:
To be honest, that's an impressive show for That Sapphire Rapids, it's 2S system for a total of 64 cores(The 8462Y is a 32C CPU), But it's a very good benchmark to show case the advantage of AVX-512 and Large Channel capacity

At core Parity Xeon 8462Y is 55% ahead of the Fastest Milan-X(No AVX-512 On those) and at virtual parity with 2S 9374F(32C Per CPU)

View attachment 73158


2S Platinum 8462Y vs 2S EPYC 9374F
Click to expand...
STILL, NO POWER NUMBERS ON THOSE RUNS THAT i CAN SEE.

sorry caps.
 
nicalandia said:
I don't think we will see any of them before January 10th.
Click to expand...
Dying to see. I think that Raptor lake P-cores are pretty close to Zen 4 (except no avx-512) so aside from avx-512 implmentation, the new SR chip could certainly be equal to Genoa, core for core, but I bet at a hefty power price.
 
Markfw said:
Dying to see. I think that Raptor lake P-cores are pretty close to Zen 4 (except no avx-512) so aside from avx-512 implmentation, the new SR chip could certainly be equal to Genoa, core for core, but I bet at a hefty power price.
Click to expand...
SPR is doesn't have the Raptor Lake improvements. For most things, iso power per core, Genoa should have a significant lead. SPR might be competitive on heavy vector or matrix code (ignoring the accelerators), but that's it.
 
Exist50 said:
Yeah, saw that, but is there the same for Alder Lake anywhere? And Raphael, for comparison. Though hopefully they do something similar with SPR is finally available.
Click to expand...
For ADL Michael only did cpuminer benchmarks before the feature was disabled:

Intel Core i9 12900K "Alder Lake" AVX-512 On Linux - Phoronix

www.phoronix.com www.phoronix.com

Rocket Lake is more extensive:

AVX / AVX2 / AVX-512 Performance + Power On Intel Rocket Lake - Phoronix

www.phoronix.com www.phoronix.com

A comparison between RKL and Raphael exists:

AMD Ryzen 7 7700X vs. Core i9 11900K AVX-512 Performance Analysis Review - Phoronix

www.phoronix.com www.phoronix.com
 
Exist50 said:
Not quite. I'm saying that if the intention was to have a converged architecture, it would not make sense to have the graphics group leading the effort. Contrast this with networking, where the main Xeon team makes a chip, and then Intel's networking group builds on it to make Xeon-D and the 5G basestation chips. That doesn't seem to be what's happening for Falcon Shores.
Click to expand...
I am curios , given the organization structure updates intel announced yesterday, in your opinion how likely is it for intel to execute on the falcon shores concept within 3 years from now? is that moving the needle ?

 
You must log in or register to reply here.

TRENDING THREADS

Back
Top