The Intel Execution in [2024]

Intel has gone through some hard times as of lately, but they have interesting things like VPro and now 18A, a technology Microsoft is interested in and signed a deal with them.

www.pcgamer.com

Intel CEO admits 'I've bet the whole company on 18A'

Pet Gelsinger resisted the idea last year, now he won't deny everything is riding on Intel's most advanced node yet.
www.pcgamer.com www.pcgamer.com
 
Last edited:
I'm not surprised that TR7K is selling much better than the Xeons. But for R7K vs. Core-i I'd have thought AMD would do better.
 
The mind factory numbers frequently published on x indicate (while they don't prove) that R7K is doing better in retail.
 
CarstenS said:
I'm not surprised that TR7K is selling much better than the Xeons. But for R7K vs. Core-i I'd have thought AMD would do better.
Click to expand...
It does in DIY market but OEM/prebuilts are still overwhelmingly more Intel than AMD, TR or not
 
That's a training benchmark btw. The Wccftech writer only mentions this in the fifth paragraph for some reason.

And as Techpowerup uses Wccftech as their source for the same story and doesn't read the original source article, they manage to misunderstand it and claim that the benchmark in question is an inference benchmark. :oops:

www.techpowerup.com

Intel Gaudi2 Accelerator Beats NVIDIA H100 at Stable Diffusion 3 by 55%

Stability AI, the developers behind the popular Stable Diffusion generative AI model, have run some first-party performance benchmarks for Stable Diffusion 3 using popular data-center AI GPUs, including the NVIDIA H100 "Hopper" 80 GB, A100 "Ampere" 80 GB, and Intel's Gaudi2 96 GB accelerator...
www.techpowerup.com www.techpowerup.com
 
pharma said:
wccftech.com

Intel Gaudi 2 Accelerator Up To 55% Faster Than NVIDIA H100 In Stable Diffusion, 3x Faster Than A100 In AI Benchmark Showdown

Stability AI has published a blog post that offers an AI benchmark showdown between Intel Gaudi 2 & NVIDIA's H100 and A100 GPU accelerators.
wccftech.com wccftech.com
Click to expand...
great news that the competition is starting to wake up, 'cos having a monopoly in the AI department is certainly not interesting for the future of the industry and to develop a variety of ideas and solutions. Also this way nVidia might start caring about their traditional GPUs too.

On a different note....

www.tomshardware.com

The five best Intel CPUs of all time: Chipzilla's rise and fall and rise

Intel is kind of an underdog now.
www.tomshardware.com www.tomshardware.com

Also, there is this article showing how a chip is made. The most impressive part is the interactive real time zoom where you can compare the size of a human hair, a red blood cell, a virus, pollen, etc to a chip transistor. It¡s very well done.

ig.ft.com

Inside the miracle of modern chip manufacturing

After coming up against the limits of physics, scientists are rethinking chip architecture like never before
ig.ft.com ig.ft.com
 
dorf said:
That's a training benchmark btw. The Wccftech writer only mentions this in the fifth paragraph for some reason.

And as Techpowerup uses Wccftech as their source for the same story and doesn't read the original source article, they manage to misunderstand it and claim that the benchmark in question is an inference benchmark. :oops:

www.techpowerup.com

Intel Gaudi2 Accelerator Beats NVIDIA H100 at Stable Diffusion 3 by 55%

Stability AI, the developers behind the popular Stable Diffusion generative AI model, have run some first-party performance benchmarks for Stable Diffusion 3 using popular data-center AI GPUs, including the NVIDIA H100 "Hopper" 80 GB, A100 "Ampere" 80 GB, and Intel's Gaudi2 96 GB accelerator...
www.techpowerup.com www.techpowerup.com
Click to expand...
The original source (stability ai) has both training and inference results
 
Kaotik said:
The original source (stability ai) has both training and inference results
Click to expand...
Yes, but TPU is citing training results as generation aka inference results:

With 2 nodes, 16 accelerators, and a constant batch size of 16 per accelerator (256 in all), the Intel Gaudi2 array is able to generate 927 images per second
Click to expand...

StabilityAI in the original article:

Keeping the batch size constant at 16 per accelerator, this Gaudi 2 system processed 927 training images per second
Click to expand...
 
Would've liked to have seen battlemage by now. Entering qtr2 soon, by the time it releases could be close to next gen AMD & Nvidia.
Also hope their XeSS FG is coming along nicely, but would've been nice to see it (if progress is looking good).
Not heard anything about either.
 
You must log in or register to reply here.

Similar threads

D
Intel, AMD, Microsoft & Others Team Up For Develop The “UALink”, Direct Competitor To NVIDIA NVLink
Replies
3
Views
880
pcchen
P
P
The Nobel Prize in Physics 2024: John Hopfield and Geoffrey Hinton. They trained artificial neural networks using physics
Replies
8
Views
598
Davros
D
R
9800X3D releases Nov 7th, review embargo on Nov 6th
2 3 4
Replies
62
Views
3K
Davros
D
xpea
Datacenter GPU market supply and demand (2024)
2
Replies
28
Views
3K
Erinyes
E
Cyan
NPU (Neural Processing Unit). AI Hardware Acceleration discussion.
2
Replies
34
Views
4K
pcchen
P
Back
Top