GPU Usage, Voltage, and Performance Fluctuating

Darren

Moderator
Staff member
I've been recently dabbling in overclocking my R9 390 and have starting noticing some weird correlations.

Using Furmark and Valley benchmark for testing I got it clocked to 1125/1600 with no artifacts, overheating, and at stock voltage. I had it at 1100/1550 for a little while and it ran well with consistent performance. After I reached 1125/1600 it started acting odd. While playing GTA V I've noticed I get a few things going on. I get a frame drop from 57 (capped) to usually below 50 and sometimes as low as 40. The drop is sudden and pretty noticeable, but usually lasts a second or two then hops back up to where it was. As soon as the drop happens I also see my GPU usage graph drop from 100% to much lower, usually below 50. My GPU voltage also seems to fluctuate a decent amount. I believe it's supposed to be right around 1.25 when gaming, but it usually bounces around from 1.15 to 1.25. Sometimes this correlates with performance drop, sometimes not. Temps are all well within reasonable range.

Changes to voltage up or down don't help and even back at stock clocks it continues. I haven't thoroughly tried it with other games yet but GTA hasn't been updated since well before I started noticing this. When I overclock too far with my benchmarks I'll get artifacts or a crash, but the clocks I've been using have all been stable in them. GTA however doesn't seem to cooperate despite it not stressing the GPU nearly as much or getting as hot.

Does this sound indicative of any problem? I'm a bit worried I'm pushing my CX600M too far, but unsure. I have noticed that when gaming my lights in my room or connected bathroom will flicker faintly at a certain frequency, which I'm hoping is just poor wiring and not my PSU about to explode.
 
I can't say for sure about this as I haven't been able to monitor voltages and stuff (maybe I should have when I was running two monitors) while gaming, however newer computers like yours often fluctuate normally, to save power and adjust for performance. When I run CPU-Z the processor often changes speed and voltage. I wouldn't worry about it. If you are really concerned you can take it to a shop to get it tested however I don't think there is need for concern. Also FPS fluctuation is totally normal. Also overclocking anything can lead to unstable operation when pushed too high, that is also to be expected. If I were you I wouldn't overclock much unless you are needing the performance enhanced a lot. A note about GTA 5, I have heard that it is not very well optimized for PC. I would try other new games to more accurately test performance. Assassins Creed Syndicate is a good one to benchmark with, but I don't think buying a game just to benchmark is a good way to spend money. There are also programs like 3D mark to benchmark your system.
 
Test your 12V rail with a multimeter during benchmarking and see if it goes out of spec. Also try separating the computer to another power socket of its own if possible.
 
Test your 12V rail with a multimeter during benchmarking and see if it goes out of spec. Also try separating the computer to another power socket of its own if possible.

I finally thought about doing this. I don't have a multimeter and I generally don't have any problems with benchmarks. Only games give me weirdness, usually GTA V as that's my most demanding I play regularly. This seemed to settle down slightly or I started noticing it less.

I did some tweaking and now have my 8320 at 4.4GHz with a .025 voltage bump. Previously was automatic voltage at 4.2GHz, although I would get system freezes very rarely. Also bumped up my RAM voltage as I suspect that may have contributed. I have 2 kits, a 2x2GB rated at 1600MHz 1.5v and 2x4GB rated at 1600MHz 1.65v or 1333MHz @ 1.5v. I was running at 1600MHz at 1.5v but now I'm at 1.536v.

After all those tweaks it seems a bit more consistent, and frame drops seem to usually occur when my processor gets bogged down a bit. I used to get choppiness all over the map but now it's only when hauling ass through the city. Wondering if a bit more RAM voltage would help improve smoothness further.

I have my 390 at 1100/1500 at stock voltage btw and it seems happy. I'd downclocked it to 1075/1500 and it made no difference in smoothness so back up it went.

According to HWMonitor my 12v rail usually sits at 12.096 at idle/light usage and when gaming dipped down 11.904. Doesn't seem like much but it's not really my area of expertise. My -12v rail is all over the place from -6.7 to -8, although I think that's unrelated from looking around.
 
You need a new PSU and probably VRM cooling. If you're over clocking on a board with standard vrms without cooling you can get vdroop. The -12V rail is going out of spec. It shouldn't drop below -10.8V. The 12V+ is in spec though.
 
You need a new PSU and probably VRM cooling. If you're over clocking on a board with standard vrms without cooling you can get vdroop. The -12V rail is going out of spec. It shouldn't drop below -10.8V. The 12V+ is in spec though.

From what I understand nothing uses the -12v rail anymore and it's only here for compatability standards. Also it reads all over the place at idle load anyway.

I'm getting more consistent performance now with that I fixed my RAM timings. http://www.computerforum.com/threads/different-ram-kit-weirdness.238698/

I also bumped up my voltage and it seems happy at 4.3GHz now. I had vdroop on my MSI board but this one delivers better power. I've watched vCore readings closely and they don't droop at full load.
 
That is true, but the fact that is going out of spec makes me think the PSU is struggling.

The R9 can draw between 260 and 360W and the 8350 around 140W overclocked. That is pretty close to the 550W (with rest of system). I know this is extreme cases, but check your vrm temps.
 
That is true, but the fact that is going out of spec makes me think the PSU is struggling.

The R9 can draw between 260 and 360W and the 8350 around 140W overclocked. That is pretty close to the 550W (with rest of system). I know this is extreme cases, but check your vrm temps.

Why would it not be "in spec" at idle with no load applied? That doesn't really make sense.

My VRM temps have been fine too. They just about melted on my MSI board. :D
 
Well its not in spec based on your numbers, but this could occur if the PSU circuitary is failing/struggling in other areas.
 
Back
Top