Is there anyone utilizing the Nvidia control panel for optimal performance?
Is there anyone utilizing the Nvidia control panel for optimal performance?
@Vtec 9000
The Gpu Boost algorithm plays a crucial role here. It automatically and dynamically adjusts the GPU core clock beyond the base level. [Don't forget the factory OC from Asus that was added on top.]
It switches between several boost curves depending on core temperature and how often the power limit is reached or exceeded.
The cooler the core runs, and the less often the BPL is crossed, the longer the GPU maintains its optimal performance. [From what I understand, hot spots and memory junctions behave differently – they won't impact performance until they reach their thermal limits.]
The answer isn't about using more power, thanks to the BPL and the fact that power equals heat.
Thanks for the reply. That question about BPL is a bit confusing—I'm not sure if you're referring to the board power limit. What advice would you give? I'd rather stick with the default settings.
I don't think I need to undervolt since I'm only seeing temperatures below 65°C in the core and around 82°C in the memory junction.
I appreciate the effort, but if it doesn't improve performance, I'd prefer leaving everything as it is and letting the GPU handle things.
Aye, Board Power Limit.
Experiment with the power limit as suggested by Karadjgne, or undervolt.
If you have the patience for it, I'd suggest giving it a try.
I wouldn't go that far... unlike the single power limit, there are several temperature thresholds - cooler is better...
These things - cpu and gpu - always utilize more voltage than they actually need(gpu power limits are already conservative as is), but it's done that way to guarantee stability as the product is advertised. There's also the wild card that is silicon lottery.
Boost clocks are an OC, by the factory, so there's no guarantee of any one particular speed. You'll see boosting upto a certain amount but that's governed by multiple things, not just temps in one area like memory, but also the gpu, the gpu voltages, VRM's, power limits, etc. The clocks will stop boosting up when some component reaches what the card considers its max tolerance. So even if temps are good in the nemory, voltages might be high on the gpu or in the VRM's or you might have hit physical limits in the caps or inductors etc.
With that in mind, maxing out power limits can be detrimental to performance because now power isn't as much a limited factor, which can stress other components, allow too much amperage or voltage use, create more heat, which raises resistances in inductors etc. You'll get more out of moderation across everything than pushing maximums in some things.
Strongest man in the world can't lift much while balancing on one leg, yet a kid with half his strength standing on two feet, will lift more than he can.
Some good advice there, thank you.
Im going to leave stock clocks on both core and memory, and reducing the power limit is something worth trying.
maybe even try an undervolt.
There's a "silent mode" in the Asus OC tweaker 3, which drops the clocks and power limit for me.
I can get a little carried away when it comes to pushing the limits and benchmarks, to see what score I can get, forgetting that it's already a great card right out of the box and just enjoying it! Lol.
I push most things to their limits and then ease back a bit.
I prioritize the best performance for all my folding cards.
I also enable multi-threaded driver optimization.
Folding relies on shaders for its calculations with minimal other resources.
Except the CPU providing it.
Everything seems to reach around 100mhz depending on the workload, except the 960ftw 4 gig runs stock.
GTX 1060 6GB gamer
GTX 1070 SC
RTX 3060 TI FE
All are overclocked or boosted up to 1980-2070 mhz.
I significantly modified the firmware of my Asus gtx970. It reached a 128% overclock, with Samsung memory capping just under 8k. I managed to push a 125% power limit using Asus GPU Tweak and added 400MHz to the boost clocks individually. Overall, this caused total instability beginning at 111%-113% power and any higher. Reducing memory from 7996 to 7916 while keeping the boost, but setting a 114% power limit with 0mv added worked well—any change led to immediate instability. At that level of overclocking, it remained stable for more than six years.
Throughout my OC journey, starting with an old Pentium II at 350MHz, I've consistently discovered the same thing: maximum performance never delivers unless benchmarks are used. Finding the right balance—boosting performance while ensuring it functions—is crucial.