4 ms·
It will be interesting to see how "discrete" thses GPUs will be. I'm assuming they will only be "descrete" in the sense that they are not on the same chip, but
by gbrown_ 9y ago
It will be interesting to see how "discrete" thses GPUs will be. I'm assuming they will only be "descrete" in the sense that they are not on the same chip, but rather same package (via EMIB).
Either way surely this is a move by Intel to take away from Nvidia's consumer share (which makes up the vast majority of their income) as Nvidia make inroads into the data center market?
- diab0lic 9y agoPart of me feels it would be very awesome to see onboard GPU hardware outputting similar performance to discrete chips. Of course it would change the size of the socket, or at least the piece of hardware in the socket and cooling requirements. This would have downsides in terms of consumer choice, mind you, or even the fact that upgrading a chip would involve upgrading both. It definitely has merit in the server or small form factor space though.
- wyldfire 9y agoThe big win that discrete GPUs provide to the cloud/backend marketplace (that Intel sorta plays in via Xeon Phi) is from large banks of VERY fast memory coupled with fast-clocked vector processors. But without a bunch of HBM or something similar, the discrete GPU won't be able to do training at the scale that NVIDIA and AMD do.
- gbrown_ 9y agoOne would assume that in the data center for discrete cards Intel would do something with their Nervana acquisition and HBM, or possibly (but less likely) MCDRAM.
- hackerfromthefu 9y agoMy guess is they are competing for the nascent consumer vr/ar market, which may not require top tier gpu performance for that much longer. Microsoft's Mixed Reality platform has the stated goal of running on integrated graphics and even a mid tier card in a year or two should do fine for usable vr/ar.