AMD 7900XTX High idle power/VR performance preview driver fix

[learners permit]

ATTENTION: This is a preview driver


AMD Software: Adrenalin Edition 23.10.01.41 Release Notes

Fixed issues

• Certain virtual reality games or applications may encounter suboptimal performance or occasional stuttering on Radeon™ RX 7000 series GPUs.
• Application crash or driver timeout may be observed during playback of AV1 video content using DaVinci Resolve™ Studio.
• Improvements to high idle power when using select high-resolution and high refresh rate displays on Radeon™ RX 7000 series GPUs.

Haven't tried it yet but will soon. Results posted on reddit vary but VR results are mostly positive while high idle power tends to be mixed results.
Preview driver link.

 

[Decko]

Ran it tonight, VR performance improvement seemed massive, high idle power still remains sadly.

 

[learners permit]

Yea my rough estimate is near 25% increase in VR performance and maybe more. for the driver team at AMD on me.

 

[NamelessPFG]

Maybe I should've held on to that 7900 XTX a little longer, but a driver update wouldn't have fixed the defective vapor chamber issue anyway.

I'm looking forward to some new benchmarks on this - probably won't end up getting them from Babel Tech Reviews now since the main VR reviewer/benchmarker there isn't as prominent, but I'm sure someone will pick up the torch eventually.

Anyone here care to do some DCS free flight impressions? The Su-25T and TF-51D probably have much less demanding cockpits than the F-14 and AV-8B, but it'll give at least some idea of the improvement to be had there while being totally free.

 

[Gabe3]

Ran it tonight, VR performance improvement seemed massive, high idle power still remains sadly.

I was skimming through the comments on reddit and sounded like the high idle power is from having more than one monitor

 

[Lakados]

I was skimming through the comments on reddit and sounded like the high idle power is from having more than one monitor

Based on the very small amount I have looked into this it seems to me that when you connect a second display AMD automatically enables active signaling for the outgoing connections regardless of whether or not you are doing something that requires an active signal, then simply runs those at the minimum power requirements needed to maintain an active signal. So if you then do something that requires it such as daisy chaining display port monitors or using audio or display port to USB-C it can just up the voltage rather than change from a passive to an active state. So that means there is some higher idle usage but it greatly improves compatibility and troubleshooting with some corner cases for device usage which makes life a little easier for cable management I am currently appreciating that for the few AMD GPU desktops I deal with at work.

I do hope that AMD can do something about it though as idle power consumption is something that OEMs have to contend with to meet regulatory requirements and as it currently stands the power usage puts them outside of that range. So if I were to do an RFP right now, I doubt I would see anything with an AMD GPU in there.

 

[Decko]

I was skimming through the comments on reddit and sounded like the high idle power is from having more than one monitor

Yeah I have two monitors plugged in, still shouldn't be more than 30-40 watts at idle. Not 90-100. You can plug a single monitor in on older drivers and there's no issue.

 

[Lakados]

Yeah I have two monitors plugged in, still shouldn't be more than 30-40 watts at idle. Not 90-100. You can plug a single monitor in on older drivers and there's no issue.

An Active Display Port connection can consume up to 100w, and can "idle" lower closer to 20w. So running 2 DP++ connections could see an extra 200w in power consumption if the monitors aren't signaling that they are not needing that power, and could drop then as low as 40w.
That is why some users say they got huge power savings and "fixed" the problem by changing cables or going over to HDMI as if the cable was cheap enough or the monitor not supporting active mode on the HDMI connector, it forces the output over from an active to a passive signal.
It's not so much an AMD issue to fix but a VESA one where they have allowed the power signaling to be an optional part of the specification, and I am pretty sure by now it's painfully obvious that any part of these specs that are optional aren't getting implemented.
I am pretty sure that power signaling over Display Port is mandatory in the DP 2.1 spec, but the controllers for them are kind of expensive and in short supply so there are an exceedingly small number of them on the market currently and that won't be changing any time soon. Proper power signaling is one of the functions needed to meet GSync compatibility requirements, but it is not listed in the Freesync spec but is usually found in the premium pro as any controller that can meet the other requirements there would need it as a precursor.

 

[EmerilLIVE]

Haven't really used mine much since I bought it, and haven't fired up VR yet, but I'm glad that VR performance has seen a good boost.

 

[Lakados]

The progress the 7900 series has made driver wise fills me with hope.

Here’s hoping that the next gen parts can improve thermals a little.

Something around the CPU gaming performance of a 5800x3D and a little better than a 3080ti but with a power draw closer to the 4060 and CPU in the 65w range would make me happy.

My office gets too hot while gaming and it’s starting to bother me.