Nvidia working on AMD SAM GPU performance boost facsimile

[ET3D]

Has this been confirmed that its just marketing?

(Sorry for the late reply, haven't followed the thread.)

For one thing, I have no reason to disbelief NVIDIA that it's part of a standard. Or that it's working on Linux with pretty much everything.

For another, Ryzen 5000 uses the same I/O die as Ryzen 3000, so it makes no sense that something that's purely I/O-related will work only on Ryzen 5000.

Though I guess that the only real confirmation would be when AMD expands support to more platforms, which I expect to happen when NVIDIA releases its solution.

[3dcandy]

(Sorry for the late reply, haven't followed the thread.)

For one thing, I have no reason to disbelief NVIDIA that it's part of a standard. Or that it's working on Linux with pretty much everything.

For another, Ryzen 5000 uses the same I/O die as Ryzen 3000, so it makes no sense that something that's purely I/O-related will work only on Ryzen 5000.

Though I guess that the only real confirmation would be when AMD expands support to more platforms, which I expect to happen when NVIDIA releases its solution.

PCI Gen4 might be why it's currently Ryzen 5000 and 500 series chipset

[kalniel]

PCI Gen4 might be why it's currently Ryzen 5000 and 500 series chipset

I thought Nvidia claimed it could work on gen 3 as well.

[Tabbykatze]

(Sorry for the late reply, haven't followed the thread.)

For one thing, I have no reason to disbelief NVIDIA that it's part of a standard. Or that it's working on Linux with pretty much everything.

For another, Ryzen 5000 uses the same I/O die as Ryzen 3000, so it makes no sense that something that's purely I/O-related will work only on Ryzen 5000.

Though I guess that the only real confirmation would be when AMD expands support to more platforms, which I expect to happen when NVIDIA releases its solution.

As far as I understand it, Zen 3 was drop in compatible with the Zen 2 IO die but the Zen 3 IO die is more refined/modified. That's why the memory timings and infinity fabric clock are higher/different in their capable compatibilities.

I thought Nvidia claimed it could work on gen 3 as well.

They have said so, they're apparently going to enable on Turing as well.

[3dcandy]

I thought Nvidia claimed it could work on gen 3 as well.

I did say currently....

[MajorZod]

I smell BS.

Probably worth pointing out that this tech has been in their CUDA workflow for some time. Migrating it to their gaming range has probably been on the cards but as we know NVIDIA like to drip feed new technology into new ranges to give them more unique selling points.

[towell]

If it truly is a software limitation, and Nvidia truly plans to make it happen on their chips, it's gonna be up to AMD to play ball with Nvidia (or at least up to Nvidia to pony up enough £££ to persuade AMD to enable it). Would hate to see this be only possible on AMD+Radeon or Intel+Nvidia systems...

[DanceswithUnix]

If it truly is a software limitation, and Nvidia truly plans to make it happen on their chips, it's gonna be up to AMD to play ball with Nvidia (or at least up to Nvidia to pony up enough £££ to persuade AMD to enable it). Would hate to see this be only possible on AMD+Radeon or Intel+Nvidia systems...

That wouldn't make sense. If someone already has an Nvidia card, are AMD going to drive them into using an Intel CPU on their next upgrade or try and suck them into the AM4 eco system?

From the Nvidia side, they would happily work with AMD if it is a poke in the eye for Intel.