Saturday, March 23rd 2024
AMD 24.3.1 Drivers Unlock RX 7900 GRE Memory OC Limits, Additional Performance Boost Tested
Without making much noise, AMD lifted the memory overclocking limits of the Radeon RX 7900 GRE graphics card with its latest Adrenalin 24.3.1 WHQL drivers, TechPowerUp found. The changelog is a bit vague and states "The maximum memory tuning limit may be incorrectly reported on AMD Radeon RX 7900 GRE graphics products."—we tested it. The RX 7900 GRE has been around since mid-2023, but gained prominence as the company gave it a global launch in February 2024, to help AMD better compete with the NVIDIA GeForce RTX 4070 Super. Before this, the RX 7900 GRE had started out its lifecycle as a special edition product confined to China, and its designers had ensured that it came with just the right performance positioning that didn't end up disrupting other products in the AMD stack. One of these limitations had to do with the memory overclocking potential, which was probably put in place to ensure that the RX 7900 GRE has a near-identical total board power as the RX 7800 XT.
Shortly after the global launch of the RX 7900 GRE, and responding to drama online, AMD declared the limited memory overclocking range a bug and promised a fix. The overclocking limits are defined in the graphics card VBIOS, so increasing those limits would mean shipping BIOS updates for over a dozen SKUs from all the major vendors, and requiring users to upgrade it by themselves. Such a solution isn't very practical, so AMD implemented a clock limit override in their new drivers, which reprograms the power limits on the GPU during boot-up. Nicely done, good job AMD!
During the course of our testing of the PowerColor RX 7900 GRE Hellhound graphics card, we were playing around with overclocking using the latest 24.3.1 WHQL drivers, and found that it increased the memory overclocking slider limit in AMD Software, which can be pushed all the way up to 3000 MHz now (24 Gbps GDDR6-effective). Previously the highest possible setting was 2316 MHz. This doesn't necessarily mean that the memory will overclock all the way up to 24 Gbps, you're still limited by what the GDDR6 chips are capable of. Our PowerColor Hellhound ships with Samsung K4ZAF325BC-SC20 memory chips that are rated for 20 Gbps. With our review drivers for the RX 7900 GRE, we had managed a memory overclock of 2316 MHz (18.5 Gbps GDDR6-effective); but with the new drivers, we scored a spectacular 2604 MHz (20 Gbps), which beats the 19.5 Gbps speed that the RX 7800 XT ships with.
The increased memory speed sees our 3DMark Time Spy GT1 overclocked frame rate jump from 72.6 FPS to 77.1 FPS (GPU frequency was constant between the two runs at 2803 MHz). This brings the card's total overclocking potential to an impressive 15% real-life performance gain. It remains a mystery why AMD chose to go with a slower memory sub-system than the RX 7800 XT for the RX 7900 GRE. It may have to do with achieving an almost identical board power number to the RX 7800 XT, so that board partners could end up with the same cooler noise figures as their RX 7800 XT products; or it was just a product segmentation decision—we'll never know. With the 20 Gbps overclock, the RX 7900 GRE has a hearty 640 GB/s of memory bandwidth at its disposal, which should come in handy to keep the 80 RDNA 3 compute units better-fed.
Thanks to @Dragokar for letting us know of the driver change.
Shortly after the global launch of the RX 7900 GRE, and responding to drama online, AMD declared the limited memory overclocking range a bug and promised a fix. The overclocking limits are defined in the graphics card VBIOS, so increasing those limits would mean shipping BIOS updates for over a dozen SKUs from all the major vendors, and requiring users to upgrade it by themselves. Such a solution isn't very practical, so AMD implemented a clock limit override in their new drivers, which reprograms the power limits on the GPU during boot-up. Nicely done, good job AMD!
The increased memory speed sees our 3DMark Time Spy GT1 overclocked frame rate jump from 72.6 FPS to 77.1 FPS (GPU frequency was constant between the two runs at 2803 MHz). This brings the card's total overclocking potential to an impressive 15% real-life performance gain. It remains a mystery why AMD chose to go with a slower memory sub-system than the RX 7800 XT for the RX 7900 GRE. It may have to do with achieving an almost identical board power number to the RX 7800 XT, so that board partners could end up with the same cooler noise figures as their RX 7800 XT products; or it was just a product segmentation decision—we'll never know. With the 20 Gbps overclock, the RX 7900 GRE has a hearty 640 GB/s of memory bandwidth at its disposal, which should come in handy to keep the 80 RDNA 3 compute units better-fed.
Thanks to @Dragokar for letting us know of the driver change.
32 Comments on AMD 24.3.1 Drivers Unlock RX 7900 GRE Memory OC Limits, Additional Performance Boost Tested
Just curious why anyone should be thanking you, which is generally why someone says "your welcome" ???
Also the 3060 12GB becoming a classic for a reason.
Good uplifts still though.
So my 6700 XT might tolerate a higher VRAM speed but probably not by much. And the 6700 XT already has a relative excess of VRAM bandwidth to Core count so it means less on this card anyway.
I'm now liking this GRE though as it finally seems to be performing up to it's on-paper potential.
However AMD did this could potentially be applied to other SKUs or at least to the core OC limits (PCB/Vmod OC headroom).Everything I'm seeing on the GRE's Navi 31, makes me believe this is "mobile-derived" silicon, not 'defects'.
A. The further cut-down Navi31 in the W7800 is not the new 40x39mm(?) package.
B. GREs seem to OC well (within cooling), and with less voltage:clocks vs. 7900XT(X).
C. Only 1 company 'contracted' with AMD for 7900M. (7900M is further cut-down for Power/Heat constraints.)
I suspect AMD has/had a 'glut' of '7900M' production, and no customers.
TBQH, I wish the GRE and M's packages were called "Navi 31e".
They are a distinctly different construction versus XT/XTX's Navi 31.
Video Review of the new drivers:
Personally, I had no issues w/ 1st party* 24.3.1 on my GRE.
Seemingly made CP'77 run a lil better, and clocks seemed more stable.
*I often use R.ID/AmernimeZone/OmegaReborn drivers
2600MHz(2614) VRAM on my Hynix-equipped GRE Nitro+ was as easy as moving the slider over, and hitting apply.
DON'T try to set to 3000Mhz, out the gate. Hard Reboot while playing CyberPunk 2077 was a meta experience I didn't need :laugh:
80.26 FPS in GT1 with 2720 MHz memory clock on a Nitro+ card. Rock solid in games with 2664 MHz.
I stopped trying to push my Hynix further than ~2600Mhz (as read by GPU-Z). I was merely hoping to run the VRAM at the factory-rated 2500MHz; +100Mhz was a 'nice bonus'.
(doing F@H for EHW @TM. The HIP/Pro driver I needed for OpenCL support, does not have the VRAM OC limit 'fixed'.)
Oh... Yeah. That's also probably worth mentioning:
These GREs have some undocumented 'AMD Pro' compatibility.
'Very much looking forward to R.ID's 24.3.1's in "Enterprise Mode".
Kinda excited to see what the GRE gets 'renamed into'. W7900 or W7800? :laugh:
Also, currently the fastest 7900 GRE on the Timespy Extreme leaderboards. Keep in mind I'm on a X370 board so the card runs at PCIe 3.0 speeds. Not sure if that has any noticeable effect in Timespy tho.
I think I might have won the silicon lottery this time...
(If I remember to) after I'm done w/ the folding event, I'll re-install the 24.3.1s and do a run myself.
"Get!" :laugh:
It is not like the 7900 XT OC poorly, the AIB cards with higher power limit can pull another 12%+ OC.
Of course the XTX can also OC by a similar percentage.
Edit: One thing to note is the Hellhound and a couple other models comes with 20Gbps / 2500Mhz memory that are downclocked in the first place.
So far the report I see online are, the cards that comes with slower ICs or worse binned MCDs generally tops out at around 2400Mhz stable.
hackaday.com/2022/11/09/chinese-chips-are-being-artificially-slowed-to-dodge-us-export-regulations/Obviously this exceeds the rule so this driver might actually make AMD cards non-compliant in China.
Which would mean, that some of the worldwide stock is repurposed 7900GRE for China which didn't sell (like mine). It looks like a very confusing case, because I feel that should be refundable... It's not even silicon lottery, it's a hidden-revision-that-is-not-in-specs at all. Either purposefully locked based on BW or binned based on max stable memory controller speed.
Do you guys think I might have degraded the chips through that? I didn't monitor memory temps but I made sure to set the fans on the card to 60% which was already plenty seeing as my core maxed out at 50C.
A Timespy run at 950mV/2530MHz mem just made my whole system lock up.
I, for one, had forgotten the limits to specs (other than flops) wrt exports to China (which may impact some 'stock' product decisions beyond 4090D, perhaps even how certain chips are designed with that limitation in mind in the future) in ways I haven't personally looked into on a deep-enough level to comment. Thanks for the info/reminder.
That said, GDDR would be volatile memory, correct? I imagine (but am not well-versed on the current definition of I/O bw) this is in relation to external bandwidth (links/protocol) and if anything may perhaps only relate to their (capabilities of) external cache structure? While I don't know the actual bw (or limitations of the link), using it's observational impact on performance I've generally equated AMD's L3 structure to something like adding 3mhz(/gbps) over the bus in relation to core speed (as the cache is likely clocked in relation to core; haven't looked into it). So, for instance, if something like a 7800xt/7900gre were running at 2800mhz core clock, the L3 would be contributing what equates to 268.8Gbps for DRAM bw. Obviously it's faster as the size is smaller, but in terms of perf does appear to shake out if you look at it wrt bw limitations. The same is true of nVIDIA's L2 cache and roughly double at same size (similar to core clock*6mhz over the bus), which nVIDIA all-but-confirmed with one product release not so long ago (something in the 4060/4070 series iirc) to excuse the more narrow bus wrt the competition.
In the case of 4090 (with 72MB L2), for instance, that translates to something like ~2730*4.5*384/8 = ~589.68Gbps, which is indeed oddly close to the number you stated. That's internal though, so I don't know if that was part of that decision-making process of the chip and/or effected by those export rules (or rather the rules were built around them after it was completed and/or released)? Now you've got me curious, but I always had a feeling these 'rules' were constructed as a compromise between the government and nVIDIA as for them to still be able use that chip in China, but for them to be conscious of using restraint moving forward.
I'm sure someone more studied could read those last couple paragraphs and declare me a simpleton, and that's fair-enough! I respect there are people that know much more in the intricate areas of these things (and their definitions/classifications) than I do, have have kept up on it better, and/or have information most don't have. I generally focus on the real-world impact to most users, and where limitations are/how they can be improved. I don't always know the correct specific language/terminology regarding how each area is classified. I'm sure there are people around here that can shed much more light on it than myself. I don't *think* what you are implying would apply, but I absolutely could not state that as a fact as I just don't have the inherent knowledge or haven't done the reseach to declare that as fact.Keep banging that drum, buddy. I'd love for you and any other people that have similar issues to have them completely resolved. Compatability with old titles can sometimes be a tricky thing (depending on how something was programmed), and sometimes not (it's just a matter of getting ahold of the right person and them having the time to fix it/implement a change). All I can hope is that if there is an easily-enough implementable fix on AMD's side that you are able to make that connection to them and it can/will be sorted to the best of their capability and your appeasement, even if it takes them a few tries to address your (or others) particular issue(s). I wish you much luck, and appreciate you for trying to get that sorted for your community of players. That's what it's all about, right? Just trying to make things the best they can be if it's possible; calling out issues/perceptions and hoping it reaches the right ears so that both they know about it and getting someone capable to help with a solution that is applicable/agreeable to the most people, if able/possible...and doing one's best at going about those things without being offensive/negative as possible.
(The last part is a personal reminder [I mean well, but know I can sometimes come across unintendly harsh towards people when trying to make a point], your attitude towards the issue appears good-natured and on-point, and once-again I applaud your persistance. :))
These are my settings. Does the job fine, no real instability noticed.
It 'shouldn't' have let you cook it, but you may have. I'd test w/ a clean driver re-install, or another PC.
(Also, I have had issues w/ loose PCIe power causing freezing, no artifacts tho)
Could be 'cold solder joints' or thermal warping too, if the card heatsoaked too much.
What you describe is 'similar' to a lot of people's VIIs when they died (from warping and cold joints, etc).
Thankfully, the GRE's not been out long enough to be outside of warranty.
While you were 'OCing' (which is typ. outside of warranty), the card has hard thermal limits to prevent damage. Potentially, those limits did not 'save' your card, and (personally) I'd not feel bad RMAing it.