News Nvidia Switches Gears, Chooses Sapphire Rapids for DGX H100

Toggle sidebar Toggle sidebar
I am not surprised. Adding HBM to the CPU creates a massive performance boost.

Although we still do not have exact benchmark numbers, but we do know up to 64GB of HBM2e and possibily 1.6TB/s of bandwidth. This is way faster than way bigger than AMD's vcache.
 
The SPR has dual avx512 SIMD per core+ the AMX tiled matrix operation unit per core + the enhanced IO feeding them: DSA + 8channels DDR5 + 80 lanes pcie5 +CXL. Alder Lake has none of this, so not really going to be close to SPR IPC, especially for the HPC / AI applications that NVDA is targeting.
 
"single threaded performance". I call total BS.

A CPU with 54 (or whatever the number is) cores, and you're planning on using just one core? Jensen must think we are all stupid. The truth is most likely AMD. Nvidia doesn't want help AMD grow by giving them money.

The problem I see with this reasoning is "power-consumption". The biggest cost for a lifetime of a supercomputer is electricity. I wonder how many customer's will walk away from DGX H100 AI becasuse it has "Intel, power hungry, inside.

I have another theory about Nvidia choosing Intel: AMD not willing to sell CPUs to Nvidia, because AMD needs them for their own CPU + Accelerator Cards offerings for upcoming supercomputers, and doesn't have any, or many, to spare.
 
dalek1234 said:
"single threaded performance". I call total BS.

A CPU with 54 (or whatever the number is) cores, and you're planning on using just one core? Jensen must think we are all stupid. The truth is most likely AMD. Nvidia doesn't want help AMD grow by giving them money.

The problem I see with this reasoning is "power-consumption". The biggest cost for a lifetime of a supercomputer is electricity. I wonder how many customer's will walk away from DGX H100 AI becasuse it has "Intel, power hungry, inside.

I have another theory about Nvidia choosing Intel: AMD not willing to sell CPUs to Nvidia, because AMD needs them for their own CPU + Accelerator Cards offerings for upcoming supercomputers, and doesn't have any, or many, to spare.
Click to expand...
No it makes sense. Nvidia uses a software scheduler probably even still with newer GPUs and those are heavily single thread dependent unlike the hardware scheduler that AMD uses since a long time now, at least R9 290 days, which works way better with CPUs that have less single thread power but a lot of cores.
 
dalek1234 said:
My takeaway from this is "go for a Radeon + Ryzen".
Click to expand...
Yea but it’s not a hidden ad or something. My takeaway is that if you have a older CPU Radeon works better, Intel is also used in the video so it’s not a Ryzen exclusive. Unless GeForce switches to a hardware scheduler as well, it will also stay like this, being more dependent on newest CPUs, or higher graphics settings to alleviate a cpu bottleneck.

i recently switched from a 3700X to a 5800X3D with a 2080 Ti, and yes it helped a lot, and the 3700X isn’t a slouch by any means. So yes, i can confirm it.
 
2Be_or_Not2Be said:
While the "higher single-threaded performance" might be the official reason for going to Intel's Sapphire Rapids, I think Nvidia also wanted to switch away from AMD from the business-competitor angle as well.

Of course, this announcement is very ironic considering Tom's just posted another article soon after this one with the headline: "Intel's Sapphire Rapids Delayed Yet Again".
Click to expand...
I don’t think so, Nvidia will take what works with their systems best, very simple. Anything else would be potentially damaging to their business. As soon as AMD is better , they will switch back from the overpriced t of intel.
 
  • Like
Reactions: adamboy64
KananX said:
I don’t think so, Nvidia will take what works with their systems best, very simple. Anything else would be potentially damaging to their business. As soon as AMD is better , they will switch back from the overpriced t of intel.
Click to expand...

Sure, that's the official reason, and undoubtedly valid from a technical perspective. But taking a bit of business from AMD was probably the extra "smile" after making the hardware adjustments from AMD to Intel.
 
2Be_or_Not2Be said:
Sure, that's the official reason, and undoubtedly valid from a technical perspective. But taking a bit of business from AMD was probably the extra "smile" after making the hardware adjustments from AMD to Intel.
Click to expand...
AMD is still best, I don’t think it will impress anyone. I’m laughing it off and they will too. This is childish and petty at best, but you’re just speculating. Nvidia never hated Ryzen or CPU side of AMD, this has nothing to do with Radeon.
 
KananX said:
I don’t think so, Nvidia will take what works with their systems best, very simple. Anything else would be potentially damaging to their business. As soon as AMD is better , they will switch back from the overpriced t of intel.
Click to expand...

don't be surprised if nvidia got them with discount. plus nvidia probably having hard time to secure latest generation epyc from AMD due to it's popularity and limited amount of supply AMD can make. so that could contribute another reason why nvidia going with intel this time around. they just need a CPU they can use before their Grace is ready. future DGX most likely going to use Grace (and it' successor) exclusively.
 
renz496 said:
don't be surprised if nvidia got them with discount. plus nvidia probably having hard time to secure latest generation epyc from AMD due to it's popularity and limited amount of supply AMD can make. so that could contribute another reason why nvidia going with intel this time around. they just need a CPU they can use before their Grace is ready. future DGX most likely going to use Grace (and it' successor) exclusively.
Click to expand...
They get discounts from AMD as well, that’s nothing special.
 
renz496 said:
AMD Eypc are very popular right now. in fact a few months ago we heard AMD going to increase the price for their Epyc.

https://www.tomshardware.com/news/a...0-to-30-intels-sapphire-rapids-delayed-report

and due to how popular AMD Epyc are right now nvidia probably did not get discount even if they buy them in big volume.
Click to expand...
However, I very much doubt that intel is cheaper, Intel is well known for astronomically high prices in the server sector. And without actual numbers, I’m not gonna speculate much here, it’s a waste of time.

This is more or less a performance decision, people going into conspiracy theories etc, I doubt that and I don’t support these sentiments either. The only thing I would believe is that it’s a availability decision, second only, to the performance reason, any other thoughts I’m not gonna support, you guys can believe whatever you want.
 
KananX said:
Unless GeForce switches to a hardware scheduler as well, it will also stay like this, being more dependent on newest CPUs, or higher graphics settings to alleviate a cpu bottleneck.

i recently switched from a 3700X to a 5800X3D with a 2080 Ti, and yes it helped a lot, and the 3700X isn’t a slouch by any means. So yes, i can confirm it.
Click to expand...
Nvidia has had hardware scheduling for a couple years. Let's try to stay remotely current.

Tested: Nvidia GPU Hardware Scheduling with AMD and Intel CPUs
 
KananX said:
However, I very much doubt that intel is cheaper, Intel is well known for astronomically high prices in the server sector. And without actual numbers, I’m not gonna speculate much here, it’s a waste of time.

This is more or less a performance decision, people going into conspiracy theories etc, I doubt that and I don’t support these sentiments either. The only thing I would believe is that it’s a availability decision, second only, to the performance reason, any other thoughts I’m not gonna support, you guys can believe whatever you want.
Click to expand...

Intel is in defending mode. They are not stupid to keep the price high when competitor keep stealing server market share from them. Plus intel are very well known to use contra revenue strategy even if they did not openly talk about it. They do that with xeon phi before where some unit get heavy discount where you can get them for as low as $400 when competitor accelerator card from AMD and nvidia can cost thousands per unit. The end result is they able to break nvidia dominance and beat AMD when it comes to accelerator market before.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom