By: Paul (pavel.delete@this.noa-labs.com), November 3, 2020 3:19 am
Room: Moderated Discussions
Mark (nothanks.delete@this.xxx.com) on October 27, 2020 6:42 pm wrote:
>
> An avenue to pursue if trying to make sense of this merger is to consider how Versal ACAP chips (and/or
> chiplets) could be combined with Zen4 chiplets to build a server product that had very good x86 performance
> as well as very good *integrated* Deep Learning performance. The Deep Learning performance could
> be very low latency, too (in contrast to NVidia GPUs hooked up to CPUs via PCIe).
>
> Will this work? Is this even AMD's plan? I don't know.
>
> It is also possible that AMD is just using overpriced AMD stock to purchase product diversification.
>
Will this work? Technically, yes, market wise, no.
The market for GPUs outside of gaming is too small, and they will not be gaining much by releasing a product that will be considered an inferior offering to a dedicated card from the start.
Making a desktop/mobile APU reaching parity with a dedicated mid-tier GPU would be a completely different story.
I think the industry heads greatly exaggerate the buying capacity of the gaming market, and equally so the enthusiasm outside of desktop space.
Not every person buying video cards is into "X7R33M G4M1NG," or even into gaming as such. There are quite a number of uses of consumer grade cards for things other than gaming.
And outside of the desktop space, it's simple. Just to few uses, and they don't make much money. There was nothing close to the "explosive" growth that was prophesized.
>
> An avenue to pursue if trying to make sense of this merger is to consider how Versal ACAP chips (and/or
> chiplets) could be combined with Zen4 chiplets to build a server product that had very good x86 performance
> as well as very good *integrated* Deep Learning performance. The Deep Learning performance could
> be very low latency, too (in contrast to NVidia GPUs hooked up to CPUs via PCIe).
>
> Will this work? Is this even AMD's plan? I don't know.
>
> It is also possible that AMD is just using overpriced AMD stock to purchase product diversification.
>
Will this work? Technically, yes, market wise, no.
The market for GPUs outside of gaming is too small, and they will not be gaining much by releasing a product that will be considered an inferior offering to a dedicated card from the start.
Making a desktop/mobile APU reaching parity with a dedicated mid-tier GPU would be a completely different story.
I think the industry heads greatly exaggerate the buying capacity of the gaming market, and equally so the enthusiasm outside of desktop space.
Not every person buying video cards is into "X7R33M G4M1NG," or even into gaming as such. There are quite a number of uses of consumer grade cards for things other than gaming.
And outside of the desktop space, it's simple. Just to few uses, and they don't make much money. There was nothing close to the "explosive" growth that was prophesized.