Friday, December 18th 2020

ASUS Enables Resizable BAR Support on First-Generation AMD Ryzen CPUs

When AMD introduced its Smart Access Memory (SAM) technology, it was used as one of the key advertising points for its 5000 series of Ryzen processors based on Zen 3 architecture. At the time of launch, it was believed that only the latest generation of Ryzen processors can support it and only AMD GPUs can see a benefit in performance. However, later on, many of the motherboard makers have been playing with BIOS updates and have found a way to enable resizable BAR, the technology used for SAM, on non-AMD platforms. Today, thanks to the Reddit user Merich98 we have found out that ASUS has enabled resizable BAR support via BIOS update.

The user has used BIOS version 2409, released just a few days ago, on ASUS B450-PLUS motherboard. The feat is no extraordinary because it works on a B450 motherboard, it has been supported for a long time, but rather the feat is impressive because it works with the first generation AMD Ryzen 7 1700 processor. This contradicts the theory that SAM only needs 5000 series AMD Ryzen processors to run. However, the gains were not that great. On average, the average frame rate number has increased by a small +0.839%. This could be attributed to some margin of error, so it seems like SAM is not giving much performance uplift in this case.
Source: WCCFTech
Add your own comment

34 Comments on ASUS Enables Resizable BAR Support on First-Generation AMD Ryzen CPUs

#1
zlobby
Now we need it on mobile. :-)
Posted on Reply
#2
Chomiq
It was only a matter of time.
Posted on Reply
#3
Nephilim666
Hah knew it! Can't trust NVIDIA, can't trust AMD, can't trust Intel.
Posted on Reply
#4
GoldenX
I also have it on my B450M-A. But with no GPU to test it with, I can't confirm anything yet.
Posted on Reply
#5
AxilomX
The article says the user tested SAM on a Gen 1 Zen processor and an RX 580. But with the limited performance uplift it would explain why they haven't taken advantage of this feature until this generation.
Posted on Reply
#6
_JP_
BAM! A matter of time!
Now MSI, do the same for both the non-AGESA-supporting 3000 series Ryzens and one that does support them :D plzkthxbye
(actually interesting ASRock didn't do it first)
Posted on Reply
#7
fancucker
AMDs attempts to market this as a ryzen 5000 exclusive feature are dangerous. They are slowly evolving into the money grubbing caricature people love to tout with Nvidia and Intel. I am hoping the latter can become competitive once more to ensure we as consumers are protected
Posted on Reply
#8
apoklyps3
wait ,so the beta bios with support for 5xxx CPU series hasn't removed support for older cpus?
Posted on Reply
#9
einard
"enabling" it in bios and it working is not the same.

Why did he not check with a title that has significant gains from SAM.
Posted on Reply
#10
Caring1
AxilomXThe article says the user tested SAM on a Gen 1 Zen processor and an RX 580. But with the limited performance uplift it would explain why they haven't taken advantage of this feature until this generation.
If that is the case then an old CPU with a current GPU should be tested.
Posted on Reply
#11
FYFI13
apoklyps3wait ,so the beta bios with support for 5xxx CPU series hasn't removed support for older cpus?
Nope, I have flashed my B450 with latest BIOS that supports 5xxx CPU's and still using R5 2600X in it :)
Posted on Reply
#12
Mussels
Moderprator
Enabling it in the BIOS doesnt mean a lot, and Asus are pretty well known for BIOS settings that do nothing
even now i have duplicates for a bunch of BIOS settings - and they often dont even match up (i have two "AMD overclocking" sections and they dont sync settings on my B450 and x570 asus boards)


Also doesnt the GPU need to support it in the drivers for it to actually enable? So he'd need a 6800 or 6900 card?
Posted on Reply
#13
INSTG8R
Vanguard Beta Tester
MusselsEnabling it in the BIOS doesnt mean a lot, and Asus are pretty well known for BIOS settings that do nothing
even now i have duplicates for a bunch of BIOS settings - and they often dont even match up (i have two "AMD overclocking" sections and they dont sync settings on my B450 and x570 asus boards)


Also doesnt the GPU need to support it in the drivers for it to actually enable? So he'd need a 6800 or 6900 card?
Gigabyte BIOS is the same with “duplicate settings“ the Tweaker section and the AMD Overclocking section have most of the same adjustments available.
Posted on Reply
#14
InVasMani
Caring1If that is the case then an old CPU with a current GPU should be tested.
Consider this a 2080Ti on averages only bottlenecks 6-9% at 1080 to 4K at x4 link speed while x8 is 2-3%. Where this could actually really help a fair bit is for integrated graphics at lower resolutions. It would actually also help mGPU since they split PCIE lanes on a x16 slot into two x8/x8 configurations on the motherboard I think in 3-way and QUAD SLI/CF the gains on scaling in those configurations might be decent even if you enabled SAM support on those setups. It's really something that would be great to investigated. Think about that for a moment if mGPU setups and scaling could be tangibly improved through the use of SAM support. Going from a MMIO size of 256MB to 4GB is huge in the context of mGPU you could have in theory the equivalent of 16 card mGPU setup with the same aperture MMIO bandwidth and I/O pretty significant to me. That said weaker GPU's aren't too PCIE bottlenecked even at x4 link speed and x8 link speeds on PCIE 3.0, but still it could amount to a night and day difference in the 0.1% and 1% FPS averages and micro stutter especially in tandem with faster M.2 storage the also helps in that area.

Another aspect with mGPU and DX12 could it be leveraged with this to have each GPU to stripe data to VRAM across them? I feel like mGPU could be better with data striping VRAM and GPU processing potentially think lucid hydra type approach. I really feel they were onto something in concept, but it never quite materialized for consumers. To me I see a huge potential on data stripe parity VRAM across multiple GPU's with DX12 because it can pool the VRAM resources from what I recall. I don't know if it can do data parity VRAM allocation across them though it would be enormous with SAM and the increased aperture BAR size across multiple PCIE slots and GPU's!!? Something else as well the way Radeon Boost functions it plays right into SAM's strengths. This really could impact the landscape of mGPU with VRAM allocation data parity across numerous cards and PCIE buses like RAID-5 for the GPU's BAR size basically and with dynamic variable resolution in the case of Radeon Boost that helps make heavier usage out of SAM.

In theory with the data parity across 4 GPU's with interlacing and MFAA you could render it all at the same speed as a single 1 frame, but actually be rendering 4 separate frames across 4 different GPU's and utilizing the resources striped across each. The interlacing wouldn't be as ideal on resolution clarity, but the temporal rendering from MFAA could possibly lessen the impact of it plus spreading it across 4 frames would reduce some of the shimmer negatives of interlacing at the same refresh rates and MFAA itself reduces shimmer as well so a win/win you end up with 4 GPU's doing the work and at 1/4 the resource overhead of MSAA with progressive scan by using interlacing and MFAA.
MusselsEnabling it in the BIOS doesnt mean a lot, and Asus are pretty well known for BIOS settings that do nothing
even now i have duplicates for a bunch of BIOS settings - and they often dont even match up (i have two "AMD overclocking" sections and they dont sync settings on my B450 and x570 asus boards)


Also doesnt the GPU need to support it in the drivers for it to actually enable? So he'd need a 6800 or 6900 card?
If I'm not mistaken you can define that right in the INF file itself or in the windows registry. I think the BIOS firmware support is the real hurdle it's getting MB makers to adopt and enable it to be used in the first place. Until the MB makers enable support for it the GPU makers certainly can't enable support.
fancuckerAMDs attempts to market this as a ryzen 5000 exclusive feature are dangerous. They are slowly evolving into the money grubbing caricature people love to tout with Nvidia and Intel. I am hoping the latter can become competitive once more to ensure we as consumers are protected
They are, but do you blame them for following in their footsteps with the examples they set and used weren't a positive thing for AMD nor consumers that suffered in the same manner!!? What goes around comes around correct? I'm not saying it's great for consumers, but I at least see AMD's perspective they kind of took advantage of a situation a bit in the same manner the others have in the past. It's no more a problem that AMD did it than if it had been Intel or Nvidia and looking at objectively it's hard to fault them doing in turn to them to their competitors what's been done to them in these situations leveraging a perk they were able to discover to exploit and manipulate so to speak. It's how the corporate world operates unfortunately for us consumers. Bottom line though AMD kick started the push for SAM support to the benefit of every much like they did with Mantle and it's influence on DX12 and other things along the way.
Posted on Reply
#15
Vayra86
fancuckerAMDs attempts to market this as a ryzen 5000 exclusive feature are dangerous. They are slowly evolving into the money grubbing caricature people love to tout with Nvidia and Intel. I am hoping the latter can become competitive once more to ensure we as consumers are protected
Protected like they are in the current competitive GPU environment, you mean?
Posted on Reply
#16
Jism
fancuckerAMDs attempts to market this as a ryzen 5000 exclusive feature are dangerous. They are slowly evolving into the money grubbing caricature people love to tout with Nvidia and Intel. I am hoping the latter can become competitive once more to ensure we as consumers are protected
Nope.

Your just not getting support for older generation(s). Simple as that. You will get support on the 5x00 generation(s).

If you have this option in the 3x0 series it does'nt mean it offers anything beneficial. It was said that it works best with a 6800 class GPU that contains 16GB of ram. Not anything below.
Posted on Reply
#17
InVasMani
JismNope.

Your just not getting support for older generation(s). Simple as that. You will get support on the 5x00 generation(s).

If you have this option in the 3x0 series it does'nt mean it offers anything beneficial. It was said that it works best with a 6800 class GPU that contains 16GB of ram. Not anything below.
It should work most ideally at lower resolutions so high refresh rate gaming and well mobile/integrated and bandwidth PCIE bandwidth link speed restricted mGPU actually.
Posted on Reply
#18
dicktracy
AMD got caught with their pants down by third parties. Go Asus go!
Posted on Reply
#19
InVasMani
AMD got caught making use of a feature that's technically been around for a long while the industry didn't make use of and overlooked until AMD decided it was useful to them.
Posted on Reply
#20
Jism
InVasManiAMD got caught making use of a feature that's technically been around for a long while the industry didn't make use of and overlooked until AMD decided it was useful to them.
Well it was AMD and HP if i'm correct that "invented" the feature...
Posted on Reply
#21
InVasMani
AMD and HP funded research into I think is closer to it.
Posted on Reply
#22
Lightofhonor
Have it on my board, but enabling it breaks NVME boot support so I can't go past the BIOS.

Might work if I had a SATA drive. Sadly, my 6800 will be SAM-less for now.
Posted on Reply
#23
InVasMani
Now I wonder if the NVME works fine if you boot off SATA with SAM enabled and if it helps the NVME device or not speed wise. For me the NVME aspect of aperture bar size is the most interesting part about it. I'd really curious about what a M.2 DRAM storage device with SAM could potentially be like if it can make a big difference to NVME performance. I still really am inclined to a think a M.2 DRAM volatile storage could be great though pair it with a CMOS battery and PCIE wired microSD slot to back up to. It's all plenty slim and the battery would be enough to make a emergency write to the microSD card in a power outage I believe. I think you could give a good 4 DRAM chips along with the other things w/o much trouble along with a chip controller. Forget optane I think that's a better option and could somewhat get around 2-channel memory bandwidth on consumer platforms a bit as well using a handful of volatile high speed DRAM based M.2 devices could perk up performance a lot.
Posted on Reply
#24
lexluthermiester
Nephilim666Hah knew it! Can't trust NVIDIA, can't trust AMD, can't trust Intel.
There's a difference between a company saying they're releasing a new kind of tech for one series of products & a third party enabling it for another set of similar products and a company doing something shady. This doesn't qualify as shady, not even remotely.
Posted on Reply
#25
Mussels
Moderprator
LightofhonorHave it on my board, but enabling it breaks NVME boot support so I can't go past the BIOS.

Might work if I had a SATA drive. Sadly, my 6800 will be SAM-less for now.
well, thats poop. I can enable it but see no difference here.

I believe we'll find out that the smaller the VRAM size on a card, the less of a difference this makes - so that performance gain seen on 6x00 cards will be smaller on anything else to the point of not mattering.
Posted on Reply
Add your own comment
Copyright © 2004-2021 www.techpowerup.com. All rights reserved.
All trademarks used are properties of their respective owners.