GPU overclocking with reduced core voltage – stability concerns?
GPU overclocking with reduced core voltage – stability concerns?
I own an HD7990 that is struggling to keep up with newer games such as Witcher 3 in its later stages. I've experimented with various methods to boost frame rate and minimize microstuttering. The card operates at standard speeds (1000MHz/1500MHz, core/mem) and with a reduced voltage for some time. Voltage was lowered from 1200mV to 1125mV, which significantly reduced temperatures and even improved performance slightly. The two cores alternated between 950MHz and 1GHz, but stabilizing the voltage resolved the issue. Previously, peak temperatures could reach up to 92 degrees Celsius. Now I've also overclocked it to 1025/1525, though the voltage remains at 1125mV.
I've been using Heaven for some time now and have noticed consistent improvements. Voltage and clock speed remain stable, and no noticeable artifacts appear in Heaven. My main concern is whether this approach will be sustainable. I'm planning a modest overclock with reduced voltage and have set the power limit to +20%. After adjusting the power slider, stability has improved greatly—previously the system would fluctuate in power and clock speed by about 10MHz. Now everything is rock steady.
I've read that the power slider affects the GPU's power handling, but I'm curious how this interacts with the lowered voltage. Will this increase the amount of current or wattage? The temperatures are lower, yet stability has improved. Where does the extra power come from when the voltage is reduced from its stock level?
I look forward to any suggestions from the community and appreciate your help in between.
Gamer-Potatoes :
fan speed stays below 49% that’s an issue. The power slider is there for situations where you boost the voltage so the GPU reaches its true capacity. And regarding the TIM adjustment, don’t stress too much. It’s similar to changing the CPU TIM, but with more screws involved (which is why I suggest using a tray instead of just working on it). You typically need more TIM than on a CPU. As long as you apply it correctly—usually also to memory chips and VRMs, generally matching the manufacturer’s instructions—there’s nothing seriously wrong.
Power limits start to become a concern when voltage is increased excessively. Usually it doesn’t matter as long as you don’t reduce it or exceed AMD’s maximum increase. For overclocking, stick to the recommended steps: boost the clock, if instability appears, raise the voltage. Keep repeating until you achieve a stable high clock at a suitable voltage. By this stage, the GPU is likely out of warranty, so you might want to upgrade its thermal paste for improved temperatures—an older card running hot will almost certainly have worn TIM, making this a worthwhile option. Apart from that, if you want better cooling, watercooling could help, but for this card a full cover waterblock would be necessary, which would allow you to purchase a higher-performance model at the next price level.
Mr Kagouris:
Power limits become a concern when voltage is increased excessively. Usually it doesn’t matter as long as you don’t reduce it or exceed AMD’s maximum. Regarding overclocking, just stick to the usual steps: increase clock speed if needed, then adjust voltage. Keep going until you achieve a stable high clock at a suitable voltage. At that stage, I think the GPU is likely out of warranty, so you might want to upgrade its thermal paste for better performance—especially since an older card running hot will almost certainly have worn TIM. Apart from this, improving heat dissipation could mean switching to watercooling, but for this card a full cover waterblock would be necessary, which would require a significant investment and could lead to a better alternative.
Thank you for your response. But my main question is about combining overclocking with undervolting. Temperatures have improved a lot—peak temps dropped from 92°C to around 81°C, averaging 76°C, and the fan speed never goes above 49%. My worry is how raising the power slider will impact reliability, given that voltage was reduced by about 75mV compared to the stock value, even with only a small overclock.
P.S
I’m not confident enough to change the TIM—I haven’t touched any GPU hardware and can’t afford another one if things go wrong. Times have been tough lately.
Gamer-Potatoes :
fan speed stays below 49% that's an issue. Regarding the power slider, it's only relevant when you boost the voltage too much and the GPU reaches its maximum capacity. As for adjusting the TIM, don't stress. It's similar to changing the CPU TIM, but with more screws involved (which is why I suggest using a tray to keep them organized). You typically need more TIM than on a CPU. As long as you apply it correctly—usually following the manufacturer's instructions and ensuring all screws and washers are intact—it won't cause problems. In fact, this can significantly improve your thermal performance. While you're concerned about potential issues, remember that lower temperatures help extend component life.
Mr Kagouris:
Gamer-Potatoes :
The fan speed never goes beyond 49%, which is a concern. Regarding the power slider, it’s only relevant when increasing voltage to push the GPU near its maximum capacity. As for adjusting the TIM, don’t worry—it behaves similarly to changing the CPU TIM, though you’ll need more screws (which is why I suggest using a tray, unlike my first attempt that caused a mess). You typically apply more TIM than on a CPU, and it’s important to keep all components in place and avoid losing screws or washers. This adjustment can significantly improve thermal performance.
While you’re concerned about potential issues, remember that lower temperatures enhance component lifespan.
Regarding the fan speed, before undervolting it would spin faster, but I’m using the original profile, so it maintains its current speed at 75°C as it did before. The only difference is that now it doesn’t reach that threshold as quickly.
As for the power slider, I found some unusual behavior. My OC runs at 25MHz on the core and memory, but with 0% powerslider it would throttle like 8%, while at 12% throttling happens after a longer delay. At 20% it never throttles. The temperatures remain consistent regardless of the slider setting. However, if I revert to stock settings and voltage with the slider at 0%, it only starts throttling in the high 80s, reaching around 50MHz.
I’m still puzzled by this. Any suggestions?
Thanks again for your quick response.