AMD can provide INTEL, NVIDIA Hardware with Smart Access Memory support


When AMD unveiled its Smart Access Memory, it sounded as if the business had finally found a way to allow Ryzen CPU’s and GPU’s to work together directly to deliver higher performance than either could accomplish together. Our performance tests proved that SAM performed pretty well, but it was not clear whether or not the future would be limited to Best AMD CPU/GPU setups.


We have a response, thanks to a recent PCWorld interview. It has people on the Ryzen team working to get SAM working on NVIDIA GPUs, according to AMD, while there are people working with Intel on the Radeon team to make the feature usable with Intel CPUs and chipsets. If AMD is comfortable making this kind of announcement, it means that in these deals there is reciprocity, which means we can see cross-platform, cross-vendor support, although we have not learned anything about NVIDIA/Intel. It only makes sense for the two companies to work together, though as the alternative amounts to giving AMD a free output advantage.

This confirms that SAM is not an AMD-specific technology as such but before anyone else did, AMD did the job of activating the functionality. Initially baked into the PCIe 2.0 standard in 2008, Re-sizable BAR Capability was updated in revisions to PCIe 3.0 in 2016. Microsoft added support for the Windows 10 functionality when it launched Windows Display Driver Model 2.0, but it was obviously not supported by any GPU vendor until now.

If this were an AMD-specific technology, one may suspect that in order to use it, the company had to design Zen 3 and/or RDNA2. The fact that support can obviously be applied to hardware from Intel and Nvidia suggests that the feature was either not seen as worth the trouble or that the companies in question were not aware that it could provide a real performance uplift until it was actually tested by anyone. The latter one would have been quite droll.

According to AMD, there’s some work required to support the feature appropriately, implying we may not see it enabled immediately on Intel and NVIDIA platforms. It’ll be interesting to see what kind of performance we see other platforms and hardware pick up from enabling this capability — Intel might benefit more than AMD and AMD GPUs might benefit more than NVIDIA cards or the reverse.