We first noticed AMD's problems with high idle power draw earlier this year, and it was (sort of, kind of) 'fixed' back in July. As the RDNA 3 GPUs are among the best graphics cards, an idle power draw of up to 90W on the RX 7900 XTX was not a good look. But after various driver updates over the intervening months, it looks like AMD has finally solved the problem.
I've tested the five RX 7000-series GPUs with the latest AMD 23.12.1 drivers, as well as the previous 23.11.1 drivers. Actually, I had previously started logging idle power last month, and when looking at the finally implemented hardware accelerated GPU scheduling, I noticed things had changed. So I commenced retesting all of the GPUs.
Update: But mistakes were made. Sigh. It appears some of the cards defaulted to a 4K 60Hz refresh rate, which lowered idle power draw substantially. Some GPUs do fine at 4K and 144Hz. Others are fine at 4K and 120Hz. But many GPUs still have relatively high power draw at anything above 4K and 60Hz. The text below hasn't been updated, as I'm having to retest things. There will be a future article on idle and low use power draw, using more extensive testing.
Looking at the 23.11.1 results should immediately illustrate why I suggested the idle power issues were not yet completely fixed with the previous drivers. Now, part of the issue may be that the PC was connected to an Acer X27 Predator display — a 4K 144 Hz G-Sync monitor. One of AMD's previous "workarounds" was to enable FreeSync support if available, which it's not on G-Sync displays like the X27. Except, a static refresh rate shouldn't be a problem.
With the last drivers (and the RX 7800 XT / RX 7700 XT launch drivers), I noticed idle power draw, measured using an Nvidia PCAT v2 device, was quite high at around 50W. Power use dropped a bit with the monitor set to a 60 Hz refresh rate, but it was still above 40W. The RX 7900 XTX/XT meanwhile were down in the sub-30W range, and with the 23.11.1 drivers they dropped to less than 20W. In fact, the bigger GPUs were using less power than even the smallest Navi 33 GPU.
Now, with the 23.12.1 drivers installed, things look far more sensible. The big Navi 31 GPUs are pulling 15–16 watts of power, the medium Navi 32 GPUs are in the 12–13 watts range, and the smallest Navi 33 GPU sips just 8W.
To be clear, this is when the PC is sitting fully idle at the Windows desktop, still, with the monitor set to 4K and 144Hz. Setting the display to a 60 Hz refresh rate dropped the power another 30%, but any activity beyond moving the mouse cursor around will start to increase the power draw.
For example, having the PCAT window open and visible, which shows a constantly changing real-time power draw, increased the power use to over 30W on the 7900 XTX. Opening a web page in Chrome — it's surprising how much power Chrome can use — depending on the content, saw power use jump to anywhere between 30W and 90W. But if your PC is just sitting there, showing static content, you should get similar power use to what's shown in the chart.
Of course if your PC is fully idle and the monitor is set to go to sleep, you can expect even lower power use. We didn't check every GPU, but the RX 7900 XTX as an example dropped to just ~11W of power use when the display went to sleep. Or just put the PC to sleep and then power use drops to basically zero.
Stay on the Cutting Edge
Join the experts who read Tom's Hardware for the inside track on enthusiast PC tech news — and have for over 25 years. We'll send breaking news and in-depth reviews of CPUs, GPUs, AI, maker hardware and more straight to your inbox.
Jarred Walton is a senior editor at Tom's Hardware focusing on everything GPU. He has been working as a tech journalist since 2004, writing for AnandTech, Maximum PC, and PC Gamer. From the first S3 Virge '3D decelerators' to today's GPUs, Jarred keeps up with all the latest graphics trends and is the one to ask about game performance.
Thats nice. I wish my 3080 would ilde at less than 34.5 or 91 watts depending on how many of my monitors I have up...Reply
I'm glad they finally fixed that.Reply
I didn't even know this was an issue until now considering I have an RX 6800.Reply
Huh? This was a 7000 series issue.Order 66 said:I didn't even know this was an issue until now considering I have an RX 6800.
Oh, I missed this one!Reply
Thanks Jarred for testing this!
AMD are such saints for taking months to fix this, meanwhile if it was anyone else they'd be crucified here lmaoReply
You're absolutely, 100% wrong on that one.Freestyle80 said:AMD are such saints for taking months to fix this, meanwhile if it was anyone else they'd be crucified here lmao
It's been a known and very nudged issue with the 7000 series cards since day one and everyone has been annoyed by it. As it's IDLE power, most people probably don't care as much, but it is still an issue. Specially on laptops.
So no: no one I know of and reading here at the forums and the Discord has given AMD a pass on this one.
That’s my point, the fact that I had a 6800 meant that I didn’t experience it firsthand, thus I didn’t know it was a thing.P1nky said:Huh? This was a 7000 series issue.
Thank you very much for testing this, @JarredWaltonGPU .Reply
any activity beyond moving the mouse cursor around will start to increase the power draw.I'm going to hazard a guess that this could be tamed quite effectively by using a custom frequency curve (or something like that). I expect the GPU is just cranking up its clock speeds too aggressively, at the first sign of activity.
For example, having the PCAT window open and visible, which shows a constantly changing real-time power draw, increased the power use to over 30W on the 7900 XTX. Opening a web page in Chrome — it's surprising how much power Chrome can use — depending on the content, saw power use jump to anywhere between 30W and 90W.
Alchemist has a "high idle" problem that was mitigated somewhat, but remains an ongoing issue and I've read will require a hardware change to fully fix. You could poll the public sentiment on that, for comparison.Freestyle80 said:AMD are such saints for taking months to fix this, meanwhile if it was anyone else they'd be crucified here lmao
Also, the MTT S80 reportedly had completely massive idle power draw, at launch - I forget exactly how much, but into the hundreds of Watts. I haven't heard updates on that issue, perhaps because its ongoing performance saga has captured too much attention.