FPS degradation over time

AlaskanPolarbear

Reputable
Dec 30, 2015
13
0
4,520
Hey, I'm sorry if I am posting this in the wrong topic, but I have come across an issue where in games like need for speed Rivals (and pretty much any other post 2010 need for speed game, such as the run, hot pursuit 2010 etc) my FPS is stable at the locked 30fps for about 10-15 minutes but then just drops down to 18-20 and the gameplay slows down, not sure if this is just an issue with the Engine need for speed uses, but also in other games, such as Watch dogs 2 it's really a hit and miss wheter or not my fps is stable/high enough to play the game, or just not be able to run it at proper playable FPS, now I do not have a very high end rig, but this issue has only been happening since a week or so now, at least; that's what it feels like, and I am absolutely clueless as to what might cause this, I have pretty much tried anything to see what can cause this issue;

It might also be fair to mention that if I do run a benchmark such as Cinebench, the first render, or the first 2 renders do show exeptional outcomes (for my chip, at least) but then the next renders drop significantly in score.

I do run an amd fx8320 that was overclocked at 4.8ghz (stable) initially but in the meantime has been brought down to about 4.5ghz without any changed in FPS stability, etc. The 8320 also does not show any signs of thermal throttling or VRM throttling from the motherboard. (clock speeds do not drop, neither does the voltage) (the 8320 never goes over 60 degrees)

Pretty much same story with the 270x which is the Gigabyte 4gb oc version, Initially that thing was never overclocked due to it just not being capable of overvolting or such things, but overclocking or underclocking had shown absolutely no different end results in the quote on quote stability of my games.

The ram, which is a kit of 2x Kingston HyperX Fury ddr3 1600 cl10 OC to 1800, also underclocking, overclocking, screwing aroun with the timings etc show no difference in these end results at all.

Then at last (unless I forgot something) the motherboard which is a gigabyte 970-ud3p rev2 with 1.2 volts at the Vcore (which always worked since I got it, never had any issues) and 1.6volts at the ram (also never had any issues with that)

I also ran various stresstests and stability tests, such as aida64, Prime95, etc all which show no errors of any of the sort. Which really makes me wondering what is going on here, Malwarebytes also came back with no results, not even after 3 complete scans of the whole system.

My operating system is windows 10 64 bits.

And, that's pretty much it, I suppose, thanks for anyone who might attempt to answer this.
 
Solution
I said gpu vrm not gpu core temp, same as vrm for motherboard the gpu has its own vrm.
EDIT: Also the cinebench test suggests me that it is a cpu realated problem, for just testing purpose can you put a fan to flow directly towards motherboard vrms? Also what psu do you have?
The first thing i will do is to try to monitor all resources, hdd, cpu,ram everything, when teh fps drops try to see which resource lacks. Also i would monitor all frequencies cpu and gpu (memory core and cpu). Also try to run 3dmark stress test this should stress the videocard at maximum.
 
I usually have my system monitors up on my second screen by default, nothing really seems out of the ordinary, my gpu stays at 99 percent like it always does, my gpu doesn't show any high temps or drops in clockspeeds, my ram doesn't suddenly fluctuate up to the full 8gb (I knew I forgot to mention something) it's just, everything stays normal but my fps lowers, when I reboot my system or if I am lucky just restart the game I am fine for another 10 minutes. Although, I haven't stressed my gpu with 3d mark so I will do that and report back.

Although, I myself do have a very strong feeling it is the cpu due to degrading cinebench scores after 1 or 2 runs, the thing that is weird is just, like I said before, it does not seem to throttle at all.

I also realized I screwed up my post with the voltages, the Vcore was around 1,5 with the loadline set on auto, so in reality only was at around 1.42, 1.43volts and the NB core was set at 1.2 (also tried 1.25, but no change in results whatsoever, so I just went with lower = safer)
 
Yes, I do run the FC bios, so unless there is a new beta bios I am unaware of, which I doubt concidering I have been looking for about a month to find a possible unreleased beta bios or even a modded bios without any luck.

Same story with the 270x, that thing has never gotten a bios update ever since it released, in contrast to the 2gb version which got one in 2015. Also, same story with not being able to find a modded bios for the 270x.
 
[edit] 3dmark timespy came back with a score of 0, my average fps was at 2 anyway and my "graphics drivers are not approved" which is odd since I use radeon relive 17.1.2, (Hell if I know these days, amd changes their driver name every single update) which are the newest as far as I know, but this could be my issue that I have never even thought of looking at, unless I am wrong here?

What I might do, is give windows a complete reinstall, Windows had it's own dedicated ssd so reinstalling it without any data loss shouldn't be too big of a problem and shouldn't take too long, see what that helps, go back to earlier non beta drivers and hope for the best, sounds like at least a plan to me.
 
16.2.2 is the latest whql driver but that is not the issue, but the 0 score it is, you should get at least 1-2k, ok can you monitor the vrm temp of your videocard? I use hwinfo for my 2 x r9 290x and my vrm (top card) sometimes goes over 100 degrees and i start having trouble.
 


My gpu temps never exceed 70, they get near 70 rarely but never over, not even in extreme stresstests. That should be more than alright for the windforce x3 card.

Although, I just finished somewhat setting up the new installation of windows, which took a bit longer than anticipated due to unforeseen technical issues (aka windows being crap) and I will run the same benchmark again with the 16.2 drivers and see what happens, I will also take note of my exact tempratures as requested.
 
I said gpu vrm not gpu core temp, same as vrm for motherboard the gpu has its own vrm.
EDIT: Also the cinebench test suggests me that it is a cpu realated problem, for just testing purpose can you put a fan to flow directly towards motherboard vrms? Also what psu do you have?
 
Solution


Ah yea, I didn't read correctly, my bad, Got a little too hasted after that crappy windows reinstall while doing a bunch of other stuff, but I just ran the same benchmark again:
My vrm temps would max out at 53 degrees in timespy.

I honestly have no idea what psu I have, but I'm gonna assume it's not a whole lot of good, my old one was acting up with this motherboard being able to make full use of the 8320 where my old one would throttle it so.. I pretty much went to a computer store and bought one of the first 750watt supplies I could find, also, thanks to having a h440 I'd have to pretty much rip half my pc and it's cable management apart to take a look at the model, something that sounds a little excessive for me at this point in time. I will run the cinebench tests in the same conditions as before, 4.7ghz overclock, run 3-4 tests in a row, monitor the tempratures etc and report back.

I just ran around 5 cinebench tests with a small fan aimed directly at the VRM on the motherboard, which indeed did help quite a bit. No throttling was to be seen, So.. I guess that's uhh.. A really big fall I have to take here for my stupidity for not checking the VRM cooling, just figured the VRM's were fine since the cpu speed didn't drop like it did on my old board which clearly had overheating VRM's, welp.. Luckily I'm not too shy to ask and learn something new every day.

I suppose when I remember where I left off my 2 spare 120mm fans I could mount them on top of the chassis to flow air through the vrm heatsink rather than straight at it, which I assume should be more effective.
 
Well for here on is just testing to see which makes the biggest airflow impact to your performance, it would not hurt a fan to blow directly to arms and 2 fans at top as exhaust, don't know, try different combinations and see which is the most beneficial. As a last resort I would suggest lowering the overclock (or at least the voltage) or changing the cpu cooler with top-down blower style.
 
Actually mounting a fan ghetto style right on top of the power phases fixed it completely, which is a direction I just never suspected thus never looked at, but I suspect more is going on with this motherboard.. And even if not, I always like to have a stupid reason to buy new stuff I don't need. (p.s sorry for the late reply)
 


Sorry for the late reply again, but I just ended up buying a new board with some more "fancy stuff I don't need" and beefed up heatsinks on both the VRM and chipset, it's running like a dream now with the passively cooled VRM's not even touching 55 degrees C, including it somehow got my FPS up in Dirt Rally (the only game I played so far) by quite a bit.. Thanks for the help though 😀
 


I figured that, might be that the motherboard has a bit better hardware slapped onto it. Same 970 chipset though, went from the 970a-ud3p rev 2 to an MSI 970 Gaming PRO Carbon, but I did notice that my cpu is indeed using more volts now then it did on the old board, overall, better looking board, slightly cooler pc, less noisy and more fps.. I'm happy. (by the way do I have to lock this topic now or something? Or do I just leave it as is?