Help overclocking 7870

no... overclocking/over-volting your cpu won't touch this...

if you can't get it over 1150 you got a mediocre chip. nothing to be done about it really. just because others are getting 1250 doesn't mean you will. Its the same for overclocking cpus... some overclock better then others.

something to check however, before you give up. you might be overheating your gpu... which is causing the bsod. Try pulling the side of your case off and sticking a room fan in the opening, and see if you can overclock it higher. If you can, you've got a case airflow issue which is overheating your gpu. and you should be able to overclock that gpu better with better case airflow. This isn't likely, but i'd try it before giving up.

i'm not suggesting it because i think it's likely. I just would try it before giving up. your temp monitoring/reporting might be inaccurate. that's why i didn't bother asking what temps you're running.

mostly i've got no magic solution for you, just a hail mary. I know the only times my system ever bluescreens is when my gpu overheats. that's why i tinkered with case airflow to solve the problem. My case was venting air perfectly, and keeping the cpu/mb cool... but it wasn't moving the heated air around the gpu well. so i installed a small case fan inside the case to facilitate the airflow i wanted. worked like a charm. no more BSOD, even when the temps in my room are over 100C (i'm in phoenix and have no ac... so yeah... proper airflow is a must to even use my pc in the summer)

the only way to check this (if your gpu doesn't) is to use a temp probe or IR temp reader.

Personally i'd suggest you try with the case open and a fan blowing air on it... and see if that works. lets find out if its temps causing the bsod before we start to worry about WHAT temps are causing them.

haha!

well, you can try with the side of the case off. that will knock a few degrees off at least. It might not be conclusive but it would be something.

sounds like your system is throttling because of heat...

possibly.

or you have a defective gpu... and your cpu is throttling because of some power saving feature.

no clue. :\

i'm most certainly not running with either. i've played with them, and built them... but i don't have one to run some benches.

deankenny, i was about to start a new thread because my gpu was going up to 84C when under FurMark, and fan usage on auto at 95%.
Could you please try it out and tell me your temps because if they're not close I'll have to get my card back in the store(its the same as urs)

On Topic: I could get my card stable at 1100 core and 1400 memory, with 20+ power limit, all done in CCC.(no voltage tweak)
but as soon as i reach 1110 core or more memory i get BSOD after a while of gameplay, hell, I'm not even sure if these clocks
are stable 24/7, I can't say for sure because I'm cpu limited.

Heaven 4.0, 1 lap = 67-69C, stock clocks.
Now could you please download FurMark 1.9.2, its like 10mb, run it for 2 minutes, usually it takes that much to hit the maximum temp.

Bloodie hell, if i try to oc it to 1100/1400 i reach 90C

I know this thread is old but I wanted to post because I have the same card and was able to identify a few things that might help.

One thing you should know is that in FC3 I have recorded temps on this card as high as 83C though no single benchmark was able to push it much higher than 72C. I use HWMonitor to record the highest temp while playing so I don't have to alt-tab out as this isn't 100% accurate anyway.

Note that I have a side case fan blowing directly over the GPU.

1). I removed the GPU cooler to check the situation. I found a disgusting amount of thermal paste and I also found the cooler wasn't locked down as tight as it should have been. I cleaned away the goo and used some Cooler Master paste I had laying around from my 212 Evo. This resulted in a drop of 10C on average. It also let me get about 50MHz higher on my OC (able to run 1200 GPU and 1300 VRAM).

2). I found that the throttling feature doesn't adjust at all (by simply observing voltage levels) when you change clock speed. This means that you will inevitably under-volt the card. This is likely to cause memory errors before GPU errors. Since the memory is constantly refreshing with the movement in-game, memory errors will result in a freeze or crash nearly instantly with almost no warning. GPU errors will tend to give you a little warning like stuttering, etc.

I overcame this by simply pushing the Power Control slider in CCC up to 15%. My temps have actually dropped about 2C since doing this and my benchmarks are a tiny bit better. Voltages appear to be in line. You shouldn't have to worry about the throttling causing an over volt with these clock speeds, and the voltage is locked on this card anyway, unless you use Trixx which rewrites the bios to overcome this.

Note that if you use Trixx, you can backup your bios. I would suggest using this feature, but I wouldn't suggest it for OC as its a bit buggy in its current version. I just use CCC with no issues.

3). If you're overclocking your GPU make sure you're using the default factor of 100MHz FSB. Raising this results in overclocking the PCI bus, too, which will make the PCIe components potentially unstable.

4). Overclocking the memory to 5.5GHz resulted in higher memory bandwidth, but would not run stable in games. However, I was able to run 3D Mark and saw less than a 5% improvement in my score. Watching the frame rate counter during the benchmark showed no noticeable increase in average FPS. My conclusion is that its not worth the trouble trying to get this stable at the higher memory speed.

5). For some unknown reason that makes sense only to my GPU, my best overall performance in games and benchmarks comes from having the memory at 5.1GHz and CPU at 1.15-1.2GHz. My guess is its a combination of the memory bus on the card combined with the memory timings. 5.1GHz might be an ideal timings:speed ratio for this VRAM, and the bus speed between GPU and ram might be closest to no bottleneck on either end at these rates.

Hope that helps!

I know this thread is old but I wanted to post because I have the same card and was able to identify a few things that might help.

One thing you should know is that in FC3 I have recorded temps on this card as high as 83C though no single benchmark was able to push it much higher than 72C. I use HWMonitor to record the highest temp while playing so I don't have to alt-tab out as this isn't 100% accurate anyway.

Note that I have a side case fan blowing directly over the GPU.

1). I removed the GPU cooler to check the situation. I found a disgusting amount of thermal paste and I also found the cooler wasn't locked down as tight as it should have been. I cleaned away the goo and used some Cooler Master paste I had laying around from my 212 Evo. I used a tiny, literally rice grain sized drop.

I also found the copper surface on the cooler to be less than ideal. There were noticeable grooves in this. I corrected this with some 1000 gauge steel wool. I probably spent about 30 minutes doing this and it resulted in a reasonably flat, shiny surface with much better GPU contact.

Note that there are two sets of screws on the GPU cooler: one set hold it to the card, the other hold the heatsink to the fan. Both were not as snug as they really should be. Careful not to tighten the fan screws to much, but make them snug. The screws that attach the unit to the card actually have a torque point where they suddenly take more effort to turn. I went about 1 1/2 revolutions past this point. Don't worry, you won't hurt the card; there's rubber compression pads that the screws are working against.

This resulted in a drop of 10C on average. It also let me get about 50MHz higher on my OC (able to run 1200 GPU and 1300 VRAM).

2). I found that the throttling feature doesn't adjust at all (by simply observing voltage levels) when you change clock speed. This means that you will inevitably under-volt the card. This is likely to cause memory errors before GPU errors. Since the memory is constantly refreshing with the movement in-game, memory errors will result in a freeze or crash nearly instantly with almost no warning. GPU errors will tend to give you a little warning like stuttering, etc.

I overcame this by simply pushing the Power Control slider in CCC up to 15%. My temps have actually dropped about 2C since doing this and my benchmarks are a tiny bit better. Voltages appear to be in line. You shouldn't have to worry about the throttling causing an over volt with these clock speeds, and the voltage is locked on this card anyway, unless you use Trixx which rewrites the bios to overcome this.

Note that if you use Trixx, you can backup your bios. I would suggest using this feature, but I wouldn't suggest it for OC as its a bit buggy in its current version. I just use CCC with no issues.

3). If you're overclocking your *CPU* make sure you're using the default factor of 100MHz FSB. Raising this results in overclocking the PCI bus, too, which will make the PCIe components potentially unstable.

4). Overclocking the memory to 5.5GHz resulted in higher memory bandwidth, but would not run stable in games. However, I was able to run 3D Mark and saw less than a 5% improvement in my score. Watching the frame rate counter during the benchmark showed no noticeable increase in average FPS. My conclusion is that its not worth the trouble trying to get this stable at the higher memory speed.

5). For some unknown reason that makes sense only to my GPU, my best overall performance in games and benchmarks comes from having the memory at 5.1GHz and CPU at 1.15-1.2GHz. My guess is its a combination of the memory bus on the card combined with the memory timings. 5.1GHz might be an ideal timings:speed ratio for this VRAM, and the bus speed between GPU and ram might be closest to no bottleneck on either end at these rates.

Hope that helps!

thanks for the update

glad you got it worked out.