Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
×
AI AMD

AWS is Considering AMD's New AI Chips (reuters.com) 12

Amazon Web Services, the world's largest cloud computing provider, is considering using new artificial intelligence chips from AMD, though it has not made a final decision, an AWS executive told Reuters. From the report: The remarks came during an AMD event where the chip company outlined its strategy for the AI market, which is dominated by rival Nvidia. In interviews with Reuters, AMD Chief Executive Lisa Su outlined an approach to winning over major cloud computing customers by offering a menu of all the pieces needed to build the kinds of systems to power services similar to ChatGPT, but letting customers pick and choose which they want, using industry standard connections.

While AWS has not made any public commitments to use AMD's new MI300 chips in its cloud services, Dave Brown, vice president of elastic compute cloud at Amazon, said AWS is considering them. "We're still working together on where exactly that will land between AWS and AMD, but it's something that our teams are working together on," Brown said. "That's where we've benefited from some of the work that they've done around the design that plugs into existing systems."

This discussion has been archived. No new comments can be posted.

AWS is Considering AMD's New AI Chips

Comments Filter:
  • Pretty non-story.. (Score:5, Insightful)

    by Junta ( 36770 ) on Wednesday June 14, 2023 @11:02AM (#63601840)

    An executive made a non-committal comment about using or not using a product their customers might or might not demand...

    • by gweihir ( 88907 )

      Well, that is "news" for you these days. Non-stories, hypes, panics and misinformation.

  • Most AI stacks are built on top of CUDA in the end. Until there's a bridge or support for other drivers it doesn't matter as much how many teraflops you bring to the table.
  • Everyone knows one of the next shoes to drop will be the unavailability of NVIDIA high end hardware. AMD makes some too now, this is a no-brainer. There was only one hardware manufacturer for AI, now there's gonna be two. This is not that surprising.
  • If the CPU was compelling, I would have expected a statement like "we're excited about AMD's CPU and are looking forward to partnering with them" .. that's a more positive indicator than "Err, we gonna try to see if we can fit them in somewhere. Maybe."

    • by Junta ( 36770 )

      What I've heard is that AMD basically unilaterally made a packaging that's a bit awkward to integrate into a standard solution without really consulting the market.

      As I heard one person explain "the sort of thing that *maybe* nVidia could get away with, but for AMD? That's a tough sell"

      nVidia is prepping to launch a similar sort of design, but they actually worked a bit more with partners. Which is an odd thing to say, that of AMD and nVidia, nVidia being the one to work with partners... Historically nVidi

      • More awkward than buying EPYC clusters hosting standalone CDNA-based accelerators? I think not. Who exactly is saying that about MI300?

        • by Junta ( 36770 )

          Depends on your definition of 'awkward'. It may be better software wise to use, it may have better performance. However my source was that the physical form factor and the firmware were vaguely awkward and not worth dealing with for the sake of AMD accelerators, when they could just do nVidia instead.

  • A cloud vendor on just NVIDIA is beholden to NVIDIA's whims. Diversifying their hardware, even if the AMD offering has half the performance density of the NVIDIA offering, will force competition and give them power to negotiate.
    • A cloud vendor on just NVIDIA is beholden to NVIDIA's whims. Diversifying their hardware, even if the AMD offering has half the performance density of the NVIDIA offering, will force competition and give them power to negotiate.

      This assumes that there is cloud consumer demand for AMD. If not and that lack of demand is apparent to both the cloud provider and Nvidia, then the cloud provider can buy as much AMD as they want and it wouldn't affect the price negotiation with Nvidia.

      GPUs are not drop-in commodities with a common interface, e.g., GPUs are not like disk drives. The problem is not that consumers dislike AMD. The problem is the software usability. Nvidia has more software engineers than AMD, and those software engineers

Elliptic paraboloids for sale.

Working...