Can you increase GPU performance with cooling alone? I recently upgraded to an RTX 3080, and I plan on doing a Noctua upgrade on it, with Arctic MX5 thermal paste and K5 Pro to replace the thermal pads. I remember being seeing a noticeable difference in performance on my previous GT 1030 after simply changing the thermal paste to MX5. Will this be the case with my cooling upgrade on my RTX 3080?
probably won't see much if you're only just changing the paste and thermal pads on a higher end card. if you optimise your air flow (e.g. ducts) you can lower your temperature which means you'll have additional overhead for overclocking if you're into that.
>if you optimise your air flow (e.g. ducts) you can lower your temperature which means you'll have additional overhead for overclocking if you're into that.
The two 120mm Noctuas will drastically improve cooling. The specific card I got, as seen in OP, can't have it's BIOS changed to one that allows more than the default 102.9% power. I tried, and it didn't work. Web browsers were freaking out whenever I opened them, even after using DDU and reinstalling drivers, and I suspect the graphics card has a physical hard limit set on the amount of power it can draw, because I wasn't able to overclock it beyond what I was already able to overclock it at when I set the power limit to 116%. I think it's because I got the LHR version. If I wanted to overclock it, I'd need to shunt mod it, and I've heard RTX 3080s aren't worth shunt modding.
I don't know if aircooling can be sufficient, if you want to get good uplift. Better get a watercooling or some kind of waterchiller.
Unless you're thermal throttling on the GPU, extra cooling won't do shit.
> I remember being seeing a noticeable difference in performance on my previous GT 1030 after simply changing the thermal paste to MX5.
Most were just cheap cards for extra monitors only so them hitting thermal throttle was non issues for most manufacturers.
You can't realistically change the performance with any noticeable changes via cooling, your best option would be to OC the card with MSI afterburner or the like.
Unless you're thermally capped you're not going to see any benefit in the performance of the card. You'll get less noise with better cooling though and more overhead for overclocking if that's your thing.
I put an alphacool block on my 3080 and ran the pipes through a wall where the radiator fans are blasting at full speed constantly so I get really low temps and only coil whine even at 100% usage and clocks and voltage tuned as high as they go, but if I'm being completely honest, 90% of that is wasted money and effort.
NO YOU CANT
AND DONT YOU EVER SAY YOU COULD
This would be funnier with the actual clip image. I approve of this meme.
Maybe just stop playing gaymes.
and do what instead?
Masturbate
>put extra cooling on car
>doesn't go faster
>WTF Y!??!?!?!
Its not the same, electronics benefit from sub zero. Some cars won't even start. If you increase cooling capacity on an engine, its so that you can also increase airflow and fuel without destroying things.
You're better off undervolting. Stock runs pretty inefficiently.
Also it's good to test how much the extra clockspeeds actually matter. Set up several stable profiles in MSI Afterburner and then switch between them in an intensive game. I found for -2% clockspeed I saved 25% on power with my RTX 2080 and the performance impact was literally 3% less FPS.
Have you compared the FPS differential between your underclocks/volts and overclocks? Just by moving some sliders I get 5 more FPS on Kombustor. If you get 3% less FPS by underclocking and could get upwards of 5% more fps by overclocking, you're almost at a differential of 10% which is not insignificant in my book
Too be fair I've only tested a handful of games. I still have all the profiles and can switch at any time if required.
Probably a smart call honestly, I just leave the overclocks as they are because I'm too lazy to switch back and forth.
They're all hotkeyed, takes a single click.
Yeah but I'm lazy and I'm forgetful
Same, I just checked mine, must have set this 3 months ago or more and never look.
These are mine.
Yes if you manage to keep it under 50C and gain 10% clockspeed just from the lower temps.
I want to know how this works.
All nvidia gpus have a firmware/hardware level dvfs that boosts based on temperature, silicon quality, voltage and current. By lowering temps, it can directly boost higher and it'll also indirectly reduce the current pulled since leakage will be lower at low temps.
Yea if you combine good cooling with an undervolt with a custom voltage/freq curve. You will get better sustained frequencies and no thermal throttling.
Yes. You are experiencing GPU boost 3.0 and how annoying it is. They start downclock at just 45C so unless you're watercooling or have some insane cooler (Rajintek Morpheus II kept my 1080ti to just 45C), you can't stop it.
You can overclock though but then you're pulling more power and generating more heat for minimal gains.
Best to go for an "optimimized" undervolt. So similar / same clockspeeds but much less power draw.
It depends on a variety of factors. If the manufacturer did a shitty job pasting and/or padding, it could be the difference between thermal throttling or no thermal throttling. That was the case with my Dell OEM 3090.
I don't understand why a person that knows how to change thermal paste will buy a oem build. Other than shortages.
>noticeable
>1%
The only thing you did is cause it to boost a couple of speed bins higher due to the lower temperature. That's all the OP will achieve either. The performance difference will be imperceptible without monitoring tools.
That 1% with the 6% boost I already have through overclocking for a 7% boost isn't too bad. I also noticed that you're more likely to have glitches related to overclocking if your GPU or GPU RAM gets too hot, like through repeated testing, so it should also boost how much you can overclock the GPU by a little bit.
Yes that's literally margin of error, for all you know rebooting and re-seating the equipment for a better connection is what made the benchmark run faster.
I had difficulty precisely overclocking it to get it above 290 FPS max. It kept on going to 289.x. 307.9 FPS max is a relatively huge boost.
Spikes are meaningless unless you ran this thing 3-5 times, then found the average of the peak FPS. For all you know something failed to load in one pass and caused higher FPS.
I paid attention during the benchmark, and there are two points at which the FPS gets that high, and in both passes, it consistently shot up like that. I noticed throughout the whole benchmark that the FPS was a bit higher at the same points.
You are still within margin of error, this is why most places for benchmarks run things +3 times. It's because they need to ensure there wasn't a fluke and then take the average. Besides you didn't even provide before and after temps.
>Besides you didn't even provide before and after temps.
I don't recall the temperature going above the 60s, and it was in the 50s most of the time.
Amazing scientific proof of study then anon you didn't even try more than a highschooler. You need to be 18 to post here
Fucking hell. I'm going to run another benchmark just for you. I never got that thing to go above 290 until really trying to max out the overclocking with what I've got so far. I don't think I'll shunt mod it.
you can't retard I am talking about running multiple passes of the BEFORE, that way you makes a control.
Meaning you did a clean reboot, made sure all power cables properly placed, re-seating the GPU, etc. prior to doing a thermal change.
Holy fuck retard do you not understand doing an experiment?
I didn't do any thermal change, I just set the fans to run at 100% constantly.
so by ramping the fans you took power/voltage away and undervolted like other anons said? grats.
Those fans don't take up more than a watt each, on full blast. Computer fans don't use that much energy, especially not modern ones.
What makes you think that you "need" more "performance"?
Your hardware is powerful enough as it is. In fact, it is already overkill.
Don't let AAAslop video gayme companies tell you otherwise.
>Your hardware is powerful enough as it is. In fact, it is already overkill.
One game I like to play, as of recent, is Starship Troopers: Extermination. It's still in early access, and I've seen videos online of people playing it with noticeable slowdowns, but I need to play it at low settings at 4K unless I want to dip below 30 FPS at times. I've even thought of getting a 3090, but I've already spent too much on upgrading my build, and I need to get a better monitor before that. My monitor supports FreeSync, and it works with G-SYNC, but it's only 60Hz maximum.