MAXSUN Announces Intel Arc Pro B60 Dual 48G Turbo Featuring Two Battlemage GPUs for AI Workloads

I want this to come out. have price parity or cheaper than a 5090 and be in such numbers that businesses can not deny buying these up as fast as they are made for their AI workloads. And get out of our gaming card segment!!!
 
Grimlakin said:
I want this to come out. have price parity or cheaper than a 5090 and be in such numbers that businesses can not deny buying these up as fast as they are made for their AI workloads. And get out of our gaming card segment!!!
Click to expand...
I've said this for years.

It doesn't really matter if intel or AMD have the faster GPU for AI, they need the software to complement it. CUDA is the de facto standard for compute workloads, not just AI; they need something to compete with CUDA not just the hardware.
 
Stoly said:
I've said this for years.

It doesn't really matter if intel or AMD have the faster GPU for AI, they need the software to complement it. CUDA is the de facto standard for compute workloads, not just AI; they need something to compete with CUDA not just the hardware.
Click to expand...
I've been using my 7900xtx for AI and compute workloads where the software supports it and it just works. So i don't see an issue there other than the broader segment of developers/vendors support Cuda because it's entrenched. We all know how hard it is to shift an enterprise to a different product over the one they've been using.
 
I don't know a lot about AI development and whatnot, but I understand the fact that CUDA is dominant. On the AMD side, I hear a lot about ROCm and Microsoft's DirectML as what they are trying to push.
 
there is a program called backyard ai you can grab to run local LLM's that works with Nvidia and AMD cards so you could do some comparisons and see.
 
Depending on price I know for sure we're going to buy 1 or 20. Just for testing purposes to see if they are viable for production. If they're $3000-4000 each I know we'll be in for at least 200 of them.
 
The last time I dealt with a dual-GPU card was a GeForce GTX 690, which had 2x 680 GPUs on it. I used it in a Z68 (or Z77) system I built for a friend's dad in 2012.

DX12 and Vulkan have everything developers need to get multi-GPU working in today's games. They're just not gonna go through the trouble or effort. I think the last devs to do this were Stardock and Oxide Games, with their game Ashes of the Singularity. In that game you could even use Radeons and GeForces in the same PC, working together to render the game.
 
Brent_Justice said:
I don't know a lot about AI development and whatnot, but I understand the fact that CUDA is dominant. On the AMD side, I hear a lot about ROCm and Microsoft's DirectML as what they are trying to push.
Click to expand...

DirectML has been dead even before arrival, OpenCL is the "only contender" if you can actually call it that. After years and years AMD is finally making progress with ROCm, but support is even worse than Open CL.
 
I see, I didn't realize DirectML was so dead. OpenCL has been around forever, but I guess it still hasn't taken off. I don't know a thing about ROCm, just that I've heard the word lol
 
You must log in or register to reply here.
Become a Patron!
Back
Top