AMD Scores Another EPYC Win in Exascale Computing With DOE's "El Capitan" Two-Exaflop Supercomputer

[silentbogo]

But can it multi-virtual machine run crysis?

With raytracing!

Let's see what resources are put into ROCm now that AMD has some income to fund dev. Nv has many years (decade) lead with their better fleshed out ecosystem. With nn/AI, Dnn/Dlops will feature heavily on upcoming IHV releases.

Today it's not the case. There are GPGPU APIs that can do the same and have expansive feature set and ecosystem. Heck, before yesterday I didn't even know that OpenMP already implemented GPU offloading (last time I tinkered with it 5-6 years ago).
The main reason why CUDA ruled the HPC and GPGPU compute in general, is being fast. Other aspects are just a consequence of the first one.

 

[Dammeron]

They'll have so many of those new EPYCs, surely they won't notice one is missing, right? Cause I need it...

 

[R0H1T]

They'll have so many of those new EPYCs, surely they won't notice one is missing, right? Cause I need it...

You'll need to shift jobs to pull that off, even if temporarily

 

[Vya Domus]

CUDA is more for companies like mine where we have 10 people and make biomedical imaging devices. CUDA helps us speed up the image reconstruction on the GPU versus the CPU. We are too small to make our own APIs. Giant supercomputer projects have custom tailor made software.

Completely agree, highly specialized software for these large scale computations are probably optimized down to the lowest available level like PTX for Nvidia and assembly for AMD. Truth is not a whole lot of the critical software paths there are actually going to be written in CUDA or OpenCL.

Why do you keep saying CUDA is in a locked-in eco-system? You can run CUDA code on other hardware (even on x86 and ARM, if you're desperate) using HIP through ROCm, but you need to translate (not manual conversion) to avoid any NVIDIA extensions. This is currently a lot more efficient than what can be done in OpenCL 2.1.

CUDA really is a locked ecosystem even for customers of Nvidia hardware. For example their ISA isn't open to the public and there are instances where no matter what you write in CUDA or directly in PTX it will never be as fast as the hardware is capable of. Nvidia reserves the highest level of optimizations for themselves so in order to get the most out of the hardware you purchased you either have to use a library that was hand optimized by Nvidia or if there is none for of the sort of thing you need to do then tough luck. If that's not a locked ecosystem then I don't know what is.

 

[TheoneandonlyMrK]

They picked the cheapest but good enuff and not the absolute highest performing options.

Like they could use xeons on a DOE super computer , near double the power use on a DOE system would go down well.
Now Fujitsu's 64FX chip's seems like a contender but not intel.
As for the GGPu choice perhaps they see something in the next generation of chips that we have not yet seen, they are not comparing chips that are out are they no it's chips to be made yet.

 

[bonehead123]

Great, now we can figure out how to obliterate everyone on the planet even faster & moar better than before...get ready, 'cause the end times are now upon us !

$600 million is fairly massive

Not by government spending standards, seeins how they're spending OUR money not theirs

 

[gamefoo21]

You're right about that. Corporations create these supercomputers with a major goal in mind, so they would need custom APIs to get to that goal efficiently. But what @xkm1948 is getting at is that CUDA can scale from the basic enthusiast all the way to the [big] corporations that don't have the time (or need) to have a custom API developed for them.

If anything, those same corporations would employ researchers from these universities.



Why do you keep saying CUDA is in a locked-in eco-system? You can run CUDA code on other hardware (even on x86 and ARM, if you're desperate) using HIP through ROCm, but you need to translate (not manual conversion) to avoid any NVIDIA extensions. This is currently a lot more efficient than what can be done in OpenCL 2.1.

The investment in ROCm is an advantage for everyone since all compute APIs will use this. Thank AMD for pulling this off.



They still use Apple because of deals (think 60%+ hardware and support discounts) offered by Apple. Also hardware deployment of Mac minis and Pros depends on department use cases.

Vulkan is aimed at rendering (and why any GPGPU code using Vulkan is on the graphics pipeline), which is why it succeeds OpenGL. OpenCL is meant for GPGPU use.

Oh I know Apple gives universities crazy prices. It's a great way to keep up demand once students become workers.

I'm so deep in studying Latin and writing papers on Greek and Roman epics my brain is melting, I really should focus and my posts are suffering because of that.

It's amazing how this stuff can get so muddied when you are trying to ram different stuff into it.

 

[R0H1T]

cause the end times are now upon us !

I though that was 2012 or 1999/whatever date Nostradamus came up with?

 

[prtskg]

This is the first such exascale contract where AMD is the sole purveyor of both CPUs and GPUs, with AMD's other design win with EPYC in the Cray Shasta being paired with NVIDIA graphics cards.

@Raevenlord This is 2nd AMD win for exascale computing where both cpu and gpu is from AMD. 1st one was called Frontier.

Beyond Rome: AMD's EPYC and Radeon to Power World's Fastest Exascale Supercomputer

AMD's EPYC CPUs and Radeon GPUs have been selected to power the world's fastest exascale supercomputer.

www.tomshardware.com

 

[mtcn77]

@Raevenlord This is 2nd AMD win for exascale computing where both cpu and gpu is from AMD. 1st one was called Frontier.

Beyond Rome: AMD's EPYC and Radeon to Power World's Fastest Exascale Supercomputer

AMD's EPYC CPUs and Radeon GPUs have been selected to power the world's fastest exascale supercomputer.

www.tomshardware.com

That system uses 40MW@1.5 Exaflops. FastForward 2 project aims at 20MW@1Exaflops. %33 higher.

AMD Collaborates with US DOE to Deliver the Frontier Supercomputer

It doesn't matter if it's open or closed. And if the driver is a closed binary unavailable for your platform? Least resistance. It is arguably easier to port a driver than pay for a new closed one to integrate with a moving target (OSS kernel). I do advocate opensource, but only when it's...

www.techpowerup.com

 

[IceShroom]

Great, now we can figure out how to obliterate everyone on the planet even faster & moar better than before...get ready, 'cause the end times are now upon us !

If the new supercomputer was built with 5GHz Xeon and GTX 480, then the Govt. could have obliterate us just by truning the computer 'On'.