Multiple significant inquiries concerning overclocking aspects like clocks and temperatures.
Multiple significant inquiries concerning overclocking aspects like clocks and temperatures.
Hello everyone. I have a few key questions about overclocking the GPU. I'm currently doing it on my system.
1. If I push the GPU too hard by increasing core or memory clock, can that cause damage? If not, why do people recommend making small adjustments when raising clocks? Shouldn't we start from the top and go down instead of from the bottom and adding incrementally?
2. I understand that increasing core or memory clock slightly raises heat a bit, but how significant is the temperature increase during overvolting? I'm curious about this when I boost my 1070 model.
3. I've heard that pushing the memory clock too much can damage it. Is that accurate? I recently got a checkered screen and had to reboot hard after overclocking it too much. I'm hoping the answer is no, or else I'm in trouble.
Thanks to anyone who can help with these points.
each unique gpu design has varying temperature limits, especially when it comes to heat. individual board manufacturers such as msi, evga, sapphire gigabyte, and gigabyte each set their own thermal goals. in practice, these advanced gpu components can manage temperatures around 90 degrees without problems, but that’s significantly higher than what’s considered optimal. i wouldn’t suggest running the card excessively hot except perhaps for a short benchmark test. most systems maintain a temperature range of 78-83 degrees, and ideally staying below 80 degrees would be best. if you overclock and stay within 2-3 degrees of that target during gameplay, it’s wise to reduce speed slightly to prevent throttling.
I prefer gradual adjustments, aiming for an optimal mix of clock, memory, and voltage settings. As you progress, evaluate performance and compare to earlier configurations; adjusting one parameter while others stay constant can affect results. For example, if clock runs at 100 and memory at 60 with good performance, increasing memory to 80 might lower performance slightly. This suggests needing a higher clock speed to maintain improvement, possibly requiring an increase from 125 to achieve better results. It's about patience and trial-and-error.
1. regarding the stock GPU BIOS... it's safe, you won't likely cause any harm. there are numerous safeguards in place that prevent the core from reaching hazardous voltage levels or overheating. feel free to attempt top-down adjustments; starting from the bottom is simpler.
2. voltage increases temperatures significantly more than clock speeds.
3. once again no. the memory will simply encounter errors. it can handle much higher temperatures than other components, making damage improbable. you won't be able to boost the DRAM voltage beyond a few high-end overclocking cards such as the kingpin, so damage remains unlikely. in practice, memory overclocking won't provide substantial performance improvements. better manage the heat and stress on your core or reduce overall temperatures and fan noise.
the stock gpu bios... no, you won't probably harm anything. there are so many safeguards in place that the core will never hit dangerous voltage levels or temperatures that could cause damage. feel free to try top down but bottom up is simpler.
voltage increases temperatures much quicker than clocks.
again no. the memory will just throw errors. the memory can handle much higher temps than the rest of the card, so damaging it is unlikely. since you can't boost the dram voltage on anything except a few top-tier overclocking cards like the kingpin, damage is unlikely. in reality, memory overclocking won't provide significant performance gains. save the extra heat and stress for your core or lower overall temps and fan noise.
for answer #3, the response was unexpected. several youtubers mention that the vrm's (the parts handling memory) are very sensitive to temperature. this likely explains why people warn about memory clocks potentially causing failure. i think they meant the vrm's are most affected by voltage changes, not memory clocks. as memory clock temperatures don't usually drop much, can you confirm if vrm's are indeed the most temperature-sensitive components? thanks for your prompt reply.
vrm is a voltage regulating module that manages the vcore on the core. It seems related to the memory too, but the core requires much higher amperage and current, making it more critical than the memory. A high vcore here would likely lead to vrm failure well before gddr5 issues. I don’t think about memory clocks much since they’re already sufficient to keep the core saturated.
I prefer a gradual approach, ensuring the best harmony between clock, memory, and voltage settings. As you progress, evaluate performance and compare to previous configurations; adjusting one parameter while increasing another might actually hurt performance. For instance, if clock runs at 100 and memory at 60, raising memory to 80 could lower performance slightly. This suggests you may need a higher clock speed to maintain true improvement, possibly reaching 125 for better results with memory at 80. It's all about patience and trial-and-error.
Secondly, my temperatures remain in the 70s when clocks are adjusted and power limits are set to maximum. Mostly I stay below 72. Do you think overvolting is necessary, or should I focus on improving cooling first? Nvidia has capped voltage at around 1.093 volts with Pascal, which is close to 1.050 volts. This gives me confidence I can proceed without reaching dangerous temperatures. Could you confirm this, as it's my first time experimenting with overvoltage?
I can share the information from the stock bios regarding the 1070 model. The card is expected to increase its frequency to around 2100mhz ±, and it will automatically lower its clock speed when it hits those limits. Operating within a 25-50mhz range should still provide the same boosted maximum. I might be mistaken, but it seems you could start experiencing thermal throttling before the core receives insufficient voltage, long before reaching unsafe vcore levels since you're limited by the stock bios. You might try increasing fan speed to handle the noise, but this would compromise cooling performance and isn't worth the marginal gain in MHz.
I can share what I see in the stock bios for the 1070 model—it says the card will increase its frequency up to around 2100mhz ± and will automatically slow down when it hits those limits. Staying within a 25-50mhz range should still give you the same boosted maximum. I’m not sure, but I think you’ll start throttling up due to heat before the core gets too low, especially since the stock bios cap you out. You might try increasing fan speed to handle the noise, but it won’t really help if you’re not getting enough cooling. If you want better performance, you could consider adding more fans or switching to an aftermarket case for lower temperatures. Since your current setup is a mid-tower with extra fan spots, or you’re okay with sacrificing a few MHz for better cooling, that’s your choice. But if the temps are still too high and wear is increasing, upgrading cooling would be wise.
Safety might be involved too, I've noticed many people and discussions on forums where someone watches a video of someone doing an OC and takes it as absolute truth, then they crank up their GPU settings and start using it—sometimes causing damage, usually because of a weak or low-quality power supply, though other factors can also play a role