Clock Speeds Not Resetting - 7950 Boost

J

just1tree

Weaksauce
Joined
Apr 4, 2007
Messages
83
Hey guys, I have a sapphire 7950 Boost that is failing to return to it's stock clocks. Running MSI afterburner to monitor my clocks, it idles in windows at 300 Mhz core, 150 Mhz memory. But when I launch a 3d application the clocks are stuck at 460 core and 625 Mhz memory. To fix this, I have to go into AMD overdrive settings and change the clock to 1100 Mhz core (overclocked from 925) and 1400 Mhz memory (overclocked from 1200). I am running catalyst 15.5 beta for the Witcher 3 and I initially thought it was the driver, but I rolled back to 14.12 and the same situation was occurring. Has anyone run into this and or heard of it occurring? Any thoughts would be appreciated.
 
The only time I've seen my clocks drop from what I've set it at, is when I overclock and have set a voltage that is too low to sustain the overclock.

Having said that, did you try on both bios settings on your card - hopefully at least one of them is still the factory default.
What about temps on the card - did they seem normal?
 
Last edited:
Temps were reasonable. Maybe 70c max, I'll try swapping vbios to see if that fixes the downclocks. Maybe the vbios values were overwritten somehow? I would normally run trixx for my over clock and msi afterburner for on screen monitoring. And these clocks were low following a cold boot so not like it was hot then slowed itself.
 
Is msi afterburner setting the clocks there? Does it still do it if msi afterburner isn't running?
 
I did try testing it with afterburner off, used gpuz to display max clocks and it showed the same values. I also Uninstalled trixx to make sure it didn't interfere. The only way I could adjust the clocks was through the driver.
 
Well I didn't end up moving the vbios switch, I went through uninstalling the driver 15.5 and found that 14.12 was still around. So I ran driversweeper, followed by CCleaner and finally reinstalled 15.5. So my clocks are restoring properly. But it seems that my card is now having problems maintaining it's boost clock, let alone it's default speeds of 850. I started up shadow of mordor, and watched as it went from 850 down to like 700. Temps were about 65C so don't believe it to be temp related but i've considered repasting it in the past. I've had it for about 2 years, maybe it's just telling me to get a gtx 970...
 
just1tree said:
Well I didn't end up moving the vbios switch, I went through uninstalling the driver 15.5 and found that 14.12 was still around. So I ran driversweeper, followed by CCleaner and finally reinstalled 15.5. So my clocks are restoring properly. But it seems that my card is now having problems maintaining it's boost clock, let alone it's default speeds of 850. I started up shadow of mordor, and watched as it went from 850 down to like 700. Temps were about 65C so don't believe it to be temp related but i've considered repasting it in the past. I've had it for about 2 years, maybe it's just telling me to get a gtx 970...
Click to expand...

Actually I went from the exact same drivers and I haven't had an issue. Which brings up a point I have noticed over the last 2 years, especially with Mantle. Seems AMD driver updates have issues with Intel CPUs. Don't know why exactly, just noticing the bulk of issues seem to have Intel CPUs. Other guys with AMD CPUs back at the introduction of Mantle did just one update and no problems. The Intel guys had to update the driver at least 3 times to get it to take.

Anyway just noticing it and always wondered what could be causing it.
 
I'd suspect the card is going into current limiting mode. If the core clock dropping way down happens when the GPU is heavily loaded and the GPU isn't overheating, it's most likely a current limit kicking in.

There are a few reasons why a card would have a current limit:

- The card manufacturer set a hard limit in the video card BIOS, either "just because" or because going over x limit can cause the card to self destruct.
- AMD set a current limit in the drivers/GPU itself.
- The VRMs on the video card are crap, or are overheating and going into current limiting mode to prevent self destruction.
 
GiGaBiTe said:
I'd suspect the card is going into current limiting mode. If the core clock dropping way down happens when the GPU is heavily loaded and the GPU isn't overheating, it's most likely a current limit kicking in.

There are a few reasons why a card would have a current limit:

- The card manufacturer set a hard limit in the video card BIOS, either "just because" or because going over x limit can cause the card to self destruct.
- AMD set a current limit in the drivers/GPU itself.
- The VRMs on the video card are crap, or are overheating and going into current limiting mode to prevent self destruction.
Click to expand...

But he said it was doing it straight after startup so not likely any of those scenarios at work here.
 
GiGaBiTe said:
I'd suspect the card is going into current limiting mode. If the core clock dropping way down happens when the GPU is heavily loaded and the GPU isn't overheating, it's most likely a current limit kicking in.

There are a few reasons why a card would have a current limit:

- The card manufacturer set a hard limit in the video card BIOS, either "just because" or because going over x limit can cause the card to self destruct.
- AMD set a current limit in the drivers/GPU itself.
- The VRMs on the video card are crap, or are overheating and going into current limiting mode to prevent self destruction.
Click to expand...

I'm not sure if there is anything that I can do to fix any of those issues, and I'm guessing that would mean the card is near end of life.

I dumped my vbios using GPU-Z yesterday, and saw that there are like 4 clock speed settings in the performance profile. There is the 300, a 500, the 850 and a 925 clock speed setting. So I saw no evidence of that 460 value I was seeing before.

I did some more testing this morning, I had GPU-Z up with current sensors active and was logging the results as well. I fired up furmark so I could watch the sensors while it was running, and I saw that the GPU clock would run at 850 (not even boost 925) and after maybe 30 seconds it would start swapping clocks between 850 and 500. This was continuous until I stopped furmark. Temps at most were 60c for the gpu and the vrm's. I checked and it looks like sapphire has only a 2 year warranty so i'm likely out if this is a hardware failure. Would be nice if I could find a software solution or way to identify with certainty the symptoms you had identified. I may swap this card to my wifes pc tonight to see if it continues to have the same issues with a different power supply and windows install.
 
just1tree said:
and after maybe 30 seconds it would start swapping clocks between 850 and 500. This was continuous until I stopped furmark.
Click to expand...

This really makes me suspicious of current limiting, like for some reason the VRMs are going into pulse mode because of some failure condition. Either that, or one or a few of the VRMs have died and the rest can't keep up with the demand.

If you haven't already pulled the card apart to re-paste it, I'd do that and pay special attention to the VRM area.
 
Well I pulled the card, dusted, re-pasted. Installed the card in a separate pc to isolate from software issues. Found that after the card heated up to 55c or so it throttled back from 850 core. So essentially it seems to be a hardware failure. There are some portions of the PCB that are not blue but were maybe white, I see some tinge to them, maybe a little coppery but not like burnt or anything of that nature. So if I'm out of warranty, and it's hardware there's not much else I can do with it I would imagine.
 
I would say to post a picture of the suspect areas, but it'd be a pain to take the card apart again just for pictures.

If there are some white/copper looking areas on the PCB, it's pretty certain something blew up.
 
Had a 6770 that had a VRM memory issue. Ran fine till you ran something then it would ramp up fast. Blew a PSU, fortunately it was 3 days old so got a replacement. Card sits quietly in the dark of my closet likely never to run again.
 
I think I've found that if the temp of the gpu stays under that 55c barrier then it will not down-clock to the 500 Mhz clock I was seeing in furmark. I think re-pasting and removing some dust on my intake fans on my case helped a little bit as I played some tonight on it and it was stable enough at stock clocks.
 
You must log in or register to reply here.
Back
Top