Friday, November 13th 2020
NVIDIA is Working on Technology Similar to AMD's Smart Access Memory
AMD's Smart Access Memory (SAM) is a new technology that AMD decided to launch with its Ryzen 5000 series CPUs and Radeon RX 6000 series GPUs. The technology aims to solve the problem where a CPU can only access a fraction of GPU VRAM at once, making some bottlenecks in the system. By utilizing the bandwidth of PCIe, the SAM expands its data channels and uses all the speed that the PCIe connection offers. However, it appears that AMD might not be the only company offering such technology. Thanks to Gamer's Nexus, they got a reply from NVIDIA regarding a technology similar to AMD's SAM.
NVIDIA responded that: "The capability for resizable BAR is part of the PCI Express spec. NVIDIA hardware supports this functionality and will enable it on Ampere GPUs through future software updates. We have it working internally and are seeing similar performance results." And indeed, it has been a part of the PCIe specification since 2008. This document dating to 2008 says that "This optional ECN adds a capability for Functions with BARs to report various options for sizes of their memory mapped resources that will operate properly. Also added is an ability for software to program the size to configure the BAR to." Every PCIe compatible device can enable it with the driver update through the software.
AMD's SAM implementation currently requires a Ryzen 5000 series CPU, 500 series motherboard chipset, and the latest Radeon RX 6000 series GPU. NVIDIA's implementation could cover a wide range of hardware, including Intel and AMD CPUs and their respective platforms. That means that even the PCIe 3.0 standard will get some love, as the current Intel desktop platforms are limited to the PCIe 3.0. NVIDIA will use a driver update to enable such a feature, however, it may take some time to arrive as the feature is still being developed.
Source:
Gamer's Nexus (Twitter)
NVIDIA responded that: "The capability for resizable BAR is part of the PCI Express spec. NVIDIA hardware supports this functionality and will enable it on Ampere GPUs through future software updates. We have it working internally and are seeing similar performance results." And indeed, it has been a part of the PCIe specification since 2008. This document dating to 2008 says that "This optional ECN adds a capability for Functions with BARs to report various options for sizes of their memory mapped resources that will operate properly. Also added is an ability for software to program the size to configure the BAR to." Every PCIe compatible device can enable it with the driver update through the software.
64 Comments on NVIDIA is Working on Technology Similar to AMD's Smart Access Memory
high clock DDR4, PCI-E 4.0 and the high infinity fabric speed on ryzen are probably key here
For AMD it's just good marketing; pairing a 6x00 SERIES GPU with a 5x00 SERIES CPU. And be honest; the 5x00 on it's own already beats intels best offering. Intel is no longer the "gaming" king or platform.
But i can jack up the PCI-E bus from 100 to 106Mhz, and see some simular extra performance increase as well; alot of computer simply have alot of untapped potential.
Edit: composter.com.ua/documents/ECN_Resizable_BAR.pdf
Sponsors: AMD & HP. Not a sign of Nvidia there when it was introduced. But at the other end; back then cards did'nt have 16GB of VRAM so the functionality was kind of useless back then. Cards came with ~ 256 up to 1GB of VRAM.Finally? Lol. It's bin paying off in both consumer and enterprise section(s) for years. They have the complete console market, being sold in the millions each generation.
I've always told people that loved AMD simply because they were the underdog: wait for them to compete with Intel on equal footing and you'll they behave the same.
And this isn't even about taking sides or anything. It's simply how businesses work.
And no one thought about this before AMD, or did I miss one of your pre Ryzen 5000 threads about no one increasing the accessable memory between CPU and GPU.
It literally wasn't a thing even though it was allowed before.
Good times though ,Nvidia grabs another idea ,we all get some free performance eh Even my 2060 ,this guy said so.
I think AMD just happened to make it work first. Quite possibly because they stuck with one GPU and one CPU series (which diesn't have to stay like this forever), where Nvidia couldn't.
literally wasn't a thing even though it was allowed before"So I knew it was already a thing before.
And the Nvidia passing it to my 2060 was the sarky part.
The bit about AMD having the idea was not sarcasm, before they spoke the the words SAM, as I said no one was talking about it or asking for it.
Or did I miss a thread?!.
My only quarrel with AMD is their closed-source PSP and stuff.