What could cause such a FPS dropoff?

D

DevonK

n00b
Joined
Jan 21, 2010
Messages
49
I've been running an AMD 6970 driving a Dell 30" monitor and up until recently it's been able to run older games and racing sims at 60fps without any trouble. Suddenly one day a few weeks ago I found that the framerates in some games had taken a sudden hit, to 30-40 fps. Afterburner indicated no apparent problems - GPU temp was fine, about 60, no throttling occurring, GPU load was a steady 99% in game. Virus and malware checks (Avira, Malwarebytes) came back clean, no unusual processes running. Running the 15.7.1 driver,

Tried to no effect:

- clean uninstall (tried both AMD and DDU tools for this) and reinstall of full CCC with 15.7.1
- uninstall of 15.7.1 and install of older CCC (14.12)
- install of latest Crimson package

I'd post a list of my running and stopped services but I can't upload an image here. I had not made any changes to these, but am wondering if the last huge Windows 7 update screwed something up.

Can anyone recommend any diagnostics or things to look for? Card is in the primary 16X slot on an Asus P8P67 WS Revolution mb. (I had 2 of these cards in Xfire for a few years but the mb slot for the secondary card died.)
 
Last edited:
Since you have nothing to loose....might give ddu a try for driver cleaning after removal....first thing that pops in my head anyway. Does the crimson edition 15.11.1 not work for your card? working pretty good here. figured i would mention it since it didn't seem like you tried it yet
 
DevonK said:
Card is in the primary 16X slot on an Asus P8P67 WS Revolution mb. (I had 2 of these cards in Xfire for a few years but the mb slot for the secondary card died.)
Click to expand...

Perhaps the same thing that killed the other slot is now killing the primary slot.
 
primetime said:
Since you have nothing to loose....might give ddu a try for driver cleaning after removal....first thing that pops in my head anyway. Does the crimson edition 15.11.1 not work for your card? working pretty good here. figured i would mention it since it didn't seem like you tried it yet
Click to expand...
Sorry my post wasn't complete as your response reminded me - did run DDU after a 15.7.1 install. I also tried the Crimson version, no help. (AMD is not officially supporting the 6970 on any drivers past 15.7.1 so I was reluctant to try it but figured I had nothing to lose).
 
evilsofa said:
Perhaps the same thing that killed the other slot is now killing the primary slot.
Click to expand...
Uggh I hope not - that mb is unavailable now except at an absurdly high cost on Ebay. It's pretty much the only mb for the 2600k that gives me the slot spacing needed to also run a sound card with the 2 6970s and a big CPU cooler. IIRC there was a Gigabyte board too that was OK but it won't be available now either.
 
Anyone know if I could install an ASRock Z77 EXTREME4 into my system without having to reinstall Windows 7 and my apps etc.? Just found this board for a reasonable price and looks like it works with my CPU, should be able to handle the 2 6970s plus the sound card.

TIA.
 
DevonK said:
Anyone know if I could install an ASRock Z77 EXTREME4 into my system without having to reinstall Windows 7 and my apps etc.? Just found this board for a reasonable price and looks like it works with my CPU, should be able to handle the 2 6970s plus the sound card.

TIA.
Click to expand...

I think windows may bark at you for that, since they consider the MB to be the system. Technically, changing MB would require a new license for W7. I've never tried it, so you might be able to get away with it, though.
 
Try cat 14.4. It's the only drivers that stopped my frame rate drop in elite dangerous (in super around planets).it also fixes ac4 to boot
 
R0nnieW00d said:
Try cat 14.4. It's the only drivers that stopped my frame rate drop in elite dangerous (in super around planets).it also fixes ac4 to boot
Click to expand...
Will test thanks. UPDATE - made no difference.
 
Last edited:
Ran the Uniengine Heaven v.2 benchmark - according to a review at Kitguru should be getting 56 FPS average with this card, I am getting 36. So something is definitely wrong.
 
trandoanhung1991 said:
Check the PCI-E speed using GPU-Z.

As for Heaven, what preset did you use?
Click to expand...
Thanks - I thought of that too earlier today - was the PCI-E bus running properly? When I checked with GPUZ it was good - 2.0 X16 reported running at 2.0 X16 by GPUZ.

I used the same preset for Heaven as that review I mentioned so scores would be comparable - DX11, 3D stereo off, shaders high, tessellation normal, AS 4X, AA off, fullscreen, 1080p. All ccc settings were defaults (application controlled AA and AS).
 
Swapped out the 6970 for the other one that I had taken out when the other PCIE X16 slot died - it also runs but has the exact same framerate issues so the card is not the problem, it's got to be something in the system. I doubt it's a MB issue as the slot would either work or not work, it's not going to run slowly (unless it's running at a wrong PCIE multiplier, which it is not). Totally stumped.
 
Last edited:
spintroniX said:
I think windows may bark at you for that, since they consider the MB to be the system. Technically, changing MB would require a new license for W7. I've never tried it, so you might be able to get away with it, though.
Click to expand...

You just call Microsoft and they take care of the license issue. There is even an automated number that will do it for you. It's all free.
 
1. Check for high interrupt issues. Download process explorer (sysinternals). Run it on a second monitor and arrange by cpu usage, if you see interrupts go over 2% you might have something causing some weirdness like a usb device or a sound card.

2. Check for clocks on the card, if the games you are playing aren't properly spooling up the 3d clocks you will notice the downgraded performance. Nvidia had this problem with the gtx 460 its possible that the 6970 you have isn't being properly utilized by the games or by the drivers.
 
Kaos_Drem said:
1. Check for high interrupt issues. Download process explorer (sysinternals). Run it on a second monitor and arrange by cpu usage, if you see interrupts go over 2% you might have something causing some weirdness like a usb device or a sound card.

2. Check for clocks on the card, if the games you are playing aren't properly spooling up the 3d clocks you will notice the downgraded performance. Nvidia had this problem with the gtx 460 its possible that the 6970 you have isn't being properly utilized by the games or by the drivers.
Click to expand...
Thanks - will look for interrupts. Afterburner shows clock and memory frequencies maxed all the time in-game, full GPU utilization.... I have turned off both hyperthreading and core parking, neither help. Game CPU loads are typically quite low.
 
Kaos_Drem said:
1. Check for high interrupt issues. ....
Click to expand...
With a game running the interrupts are never reaching 1% CPU load. The game itself is only imposing a 20-35% CPU load. The biggest user of CPU time after the game is the "interrupts" at about .8%, with the Avira ServiceHost as well as Services jumping up to that level occasionally. The Process Explorer itself is one of the highest users.
 
trandoanhung1991 said:
Can you grab Valley and run the benchmark with the extreme preset? According to http://www.geeks3d.com/20141209/gpu-benchmark-scores-1080p/#unigine_valley you should be getting around 30FPS, around 1200 score.
Click to expand...
My card should score about a 32 FPS on that bench actually, as its an MSI Lightning, factory overclocked (940 MHZ vs. the 880 standard for a 6970). Doesn't get close though - I got 24.8 FPS, score of 1038. That is about 23% lower than it should be.

I wonder if it would be worth trying to run a Windows 7 repair? I'm thinking it has to be something in the system that is causing this - when I kill all non-MS tasks except those needed for sound and video it makes no difference. Tried a CClean of registry, no help.
 
I'm stumped. Are you forcing any settings in the drivers? Like forcing AA at a constant 16x or something?
 
spintroniX said:
I think windows may bark at you for that, since they consider the MB to be the system. Technically, changing MB would require a new license for W7. I've never tried it, so you might be able to get away with it, though.
Click to expand...

what? This was never an issue with 7. It was reported that this was going to be an issue with 10 but never heard one way or another. I have never heard any reports on that. I have changed every piece of hardware several times on Win7 home and Pro
 
SomeGuy133 said:
what? This was never an issue with 7. It was reported that this was going to be an issue with 10 but never heard one way or another. I have never heard any reports on that. I have changed every piece of hardware several times on Win7 home and Pro
Click to expand...

It should boot up like it has no issues but if you right click on my pc and go to properties there's a good chance it will say unregistered version of Windows.



Concerning the fps issues from the op. I wish I could offer a solution you haven't tried. Clean the fans to see if it's a throttling issue with the new software update?
 
trandoanhung1991 said:
I'm stumped. Are you forcing any settings in the drivers? Like forcing AA at a constant 16x or something?
Click to expand...
No forcing in CCC, everything set to "use application settings". Tried toggling morphological filtering, tessellation, vertical refresh, triple buffering, no fps difference.
 
trentchau said:
It should boot up like it has no issues but if you right click on my pc and go to properties there's a good chance it will say unregistered version of Windows.
Click to expand...
My Win 7 is actually a site license version that IIRC has no limits on install #s. My concern is more about the fact that a different mb will require different drivers - is another brand of mb even going to boot into Windows if my Windows has Asus drivers for a different chipset (P67 vs. Z77) installed?
Concerning the fps issues from the op. I wish I could offer a solution you haven't tried. Clean the fans to see if it's a throttling issue with the new software update?
Click to expand...
Cleaned fans (I do it every few months anyway so never that dirty).Temps are fine, Afterburner shows no GPU throttling, going at 99%.
 
DevonK said:
No forcing in CCC, everything set to "use application settings". Tried toggling morphological filtering, tessellation, vertical refresh, triple buffering, no fps difference.
Click to expand...

Does your MB BIOS allow you to change the PCI slot speed? Can you try changing it to 8x or 4x and see if you lose any further performance?
 
horrorshow said:
Shot in the dark, power supply issue??
Click to expand...
Thanks for the suggestion, but very doubtful - I have a Corsair AX 850, which produces a clean 850 watts according to reviews - and now with only one card in the system the total draw is 100s of watts lower than that even with the GPU running at 100% load. Also GPUZ shows stable voltages.
 
Is a clean install not an option? Personally if any problem is taking 1hr, or maybe even 30 mins, or more to diagnose I just wipe and start over.

Are you showing any abnormal CPU or GPU usage (or other measurement) readings at idle?

Try running a comparable OpenCL benchmark. Luxmark 2.0 "sale" and "ball" might be able to find comparable scores for those. Also try running them with core downclocked to stock 6970 speeds and memory by 10%.
 
Debris in the PCI-E slot or on the PCI-E card pins reducing the number of lanes? It still bugs me that the other PCI-E slot you were using appeared to die. Remove your current video card, use a flashlight and magnifying glass and inspect both slots very carefully for debris, and clean the video card slot pins even if they look clean.

Now for crazy talk: TIM between the NF200 chip and the heatsink failing, either to old age or by getting bumped? This review shows the heatsink taken off, and talks a bit about the TIM used on the NF200. Does the NF200 part of the heatsink get hot at all with the GPU at 99% load? The article comments about "how hot it gets when really pushing voltage" but I'm not sure which part of the heatsink they were talking about.
 
evilsofa said:
Debris in the PCI-E slot or on the PCI-E card pins reducing the number of lanes? It still bugs me that the other PCI-E slot you were using appeared to die. Remove your current video card, use a flashlight and magnifying glass and inspect both slots very carefully for debris, and clean the video card slot pins even if they look clean.

Now for crazy talk: TIM between the NF200 chip and the heatsink failing, either to old age or by getting bumped? This review shows the heatsink taken off, and talks a bit about the TIM used on the NF200. Does the NF200 part of the heatsink get hot at all with the GPU at 99% load? The article comments about "how hot it gets when really pushing voltage" but I'm not sure which part of the heatsink they were talking about.
Click to expand...
GPUZ has a test you can run to check if the lanes are working properly, and mine are. However your suggestion re cleaning is something worth doing on the other non-functional PCIE slot to see if that might restore it.

I'll have to check that heatsink, thanks for the suggestion.

One thing of possible relevance: my case is a Silverstone Raven model which places the mb with its connectors at the top, not the back. I wonder if the mb is not built to handle the weight of the heavy video cards properly in that unusual orientation, hence the slot failure....
 
DevonK said:
I wonder if the mb is not built to handle the weight of the heavy video cards properly in that unusual orientation, hence the slot failure....
Click to expand...

The case should carry the weight of the card in that orientation. If anything it SHOULD be easier on your motherboard....

My 2 cents, if a slot failed on your MB already, chances are the current slot is going out as well. Warranty/Replace situation sounds like the best solution.....

http://www.ebay.com/itm/like/111814867218?ul_noapp=true&chn=ps&lpid=82

Barring that, I'd try a complete wipe and pray it's a software issue.
 
As an eBay Associate, HardForum may earn from qualifying purchases.
limitedaccess said:
Is a clean install not an option? Personally if any problem is taking 1hr, or maybe even 30 mins, or more to diagnose I just wipe and start over.

Are you showing any abnormal CPU or GPU usage (or other measurement) readings at idle?

Try running a comparable OpenCL benchmark. Luxmark 2.0 "sale" and "ball" might be able to find comparable scores for those. Also try running them with core downclocked to stock 6970 speeds and memory by 10%.
Click to expand...
Idle loads on both CPU and GPU are very low - Overdrive basically shows the GPU at 0%, and downclocked to 250/150. Voltages are within spec. Will try running the unigine benchmark again with GPU downclocked, but can't lower mem speed in CCC. UPDATE - saw a small dropoff in FPS - down to 24 FPS at 880 GPU clock vs. 24.8 at 940 clock.

I don't see what running additional benchies is going to show - I know from the 2 unigine ones that my card is running about 30% slower than other tested 6970s.

The "clean install" you suggest - I've done that several times with the drivers, but are you talking about Windows? Given the number of complex apps I have on this PC reinstalling them would take several hours, plus there is all the nonsense around saving preferences, setups, saved games, etc.... a royal headache. Is there a way to reinstall Win 7 leaving apps intact?
 
Last edited:
DevonK said:
The "clean install" you suggest - I've done that several times with the drivers, but are you talking about Windows? Given the number of complex apps I have on this PC reinstalling them would take several hours, plus there is all the nonsense around saving preferences, setups, saved games, etc.... a royal headache. ?
Click to expand...

Hours as opposed to the days spent trying to diagnose this issue?

I simply don't follow your logic....

Format and start fresh so we can eliminate the "software wildcard".

Yes, it's a b*tch but it boggles the mind you didn't try this after say, day 3.
 
DevonK said:
Idle loads on both CPU and GPU are very low - Overdrive basically shows the GPU at 0%, and downclocked to 250/150. Voltages are within spec. Will try running the unigine benchmark again with GPU downclocked, but can't lower mem speed in CCC. UPDATE - saw a small dropoff in FPS - down to 24 FPS at 880 GPU clock vs. 24.8 at 940 clock.

I don't see what running additional benchies is going to show - I know from the 2 unigine ones that my card is running about 30% slower than other tested 6970s.

The "clean install" you suggest - I've done that several times with the drivers, but are you talking about Windows? Given the number of complex apps I have on this PC reinstalling them would take several hours, plus there is all the nonsense around saving preferences, setups, saved games, etc.... a royal headache. Is there a way to reinstall Win 7 leaving apps intact?
Click to expand...

You are getting somewhere, it has to be the OS or the motherboard at this point, imho. A clean install of the OS is a pain, except it may be the only way to isolate the problem. Electronics can work in a sort of pseudo working way, it is not always works great or dead. Part of the lines to the PCIe slot may have problems, causing reduced performance. The video card doesn't seem to be the culprit since you tested an identical card and didn't get any change.
 
DevonK said:
Idle loads on both CPU and GPU are very low - Overdrive basically shows the GPU at 0%, and downclocked to 250/150. Voltages are within spec. Will try running the unigine benchmark again with GPU downclocked, but can't lower mem speed in CCC. UPDATE - saw a small dropoff in FPS - down to 24 FPS at 880 GPU clock vs. 24.8 at 940 clock.

I don't see what running additional benchies is going to show - I know from the 2 unigine ones that my card is running about 30% slower than other tested 6970s.

The "clean install" you suggest - I've done that several times with the drivers, but are you talking about Windows? Given the number of complex apps I have on this PC reinstalling them would take several hours, plus there is all the nonsense around saving preferences, setups, saved games, etc.... a royal headache. Is there a way to reinstall Win 7 leaving apps intact?
Click to expand...

Can you not downclock the memory using afterburner? The reason I suggested this is that memory instability may result in lower performance from error correction and not exhibit crashing or artifacts.

Do you have a spare drive or able to make a partition? This way you can try a clean install while still keeping your existing install just to troubleshoot whether or not that is the issue.
 
horrorshow said:
Hours as opposed to the days spent trying to diagnose this issue?

I simply don't follow your logic....

Format and start fresh so we can eliminate the "software wildcard".

Yes, it's a b*tch but it boggles the mind you didn't try this after say, day 3.
Click to expand...
I've only spent 3-4 hours working on this. Replacing all my apps and settings will take 2 days. 100s of gigs of apps.
 
limitedaccess said:
Can you not downclock the memory using afterburner? The reason I suggested this is that memory instability may result in lower performance from error correction and not exhibit crashing or artifacts.

Do you have a spare drive or able to make a partition? This way you can try a clean install while still keeping your existing install just to troubleshoot whether or not that is the issue.
Click to expand...
I'll try the memory downclock, although the fact that 2 cards have the identical problem doesn't give me much hope for that solution. I'll see if I can partition enough space out of my drive to allow a clean test install, good idea there. Will apps still run after booting from that install?
 
Maddening problem, for sure.
Disable or remove the sound card and test FPS again.
Disable or remove the network card and test.
Remove extra USB devices (ie. all but KB/mouse) and test.
The idea here is to eliminate possible causes of slowdown. If you have other devices connected, disable/remove them one by one and retest.
Double check BIOS settings for CPU clocks, memory clocks, multipliers and voltages.
Did you update any device drivers around the time you noticed the speed change? Think hard, here. Something changed. If it was due to an OS update, however, that would really suck though I'd be surprised if it was a direct cause.
The only other things I can think of are to make sure all devices are running the newest versions of drivers. Motherboard chipset drivers included - sometimes they get updated periodically.
Alternatively, install an older version of video drivers that you know worked fast before. See if it makes a difference.
After that list is exhausted, a reinstall of the OS on a fresh/spare drive would be my last resort. Do a minimalist install (OS w/updates, chipset drivers, video drivers, benchmarks/games) and test. If it is still slow after that, then you know that you are the crazy one. ;)
 
You must log in or register to reply here.
Back
Top