Performance % lost on air vs WC

Epyon

[H]ard|Gawd
Joined
Oct 25, 2001
Messages
1,182
On average what do you think the performance lost is on air vs WC? I am getting into Red shift 3d and Keyshot hard now that Gpu rendering is getting fast. I googled but could not really fine a answer. With the new cards coming and being so powerful wanted to know if i should stick to blower type and run 2x or water cool them if the difference would be huge.
 
FYI last I used Keyshot (v5 or v6) it was only cpu rendering, so a monster AMD Ryzen 8c16t(or yummy threadripper 32c64t) would do it wonders. To answer your question though would be to look at your gpus usage during a render to see if the frequency limits itself as the heat increases, as well as if that has a tangible effect on your render times. I think Nvidia cards throttle at 90c ish at least in the consumer silicon, so they generally frequency boost under heavy load until they either hit a package power wall or a heat limit. You might have a situation that one of you cards is starved for air and water might be able to help that depending on your config.
 
You could always try air and see what happens. When I watercool I usually bios mod and/or hard PL mod depending. A good custom cooler would probably be around 2-5% of watercooling if I had to guess. Maybe 10% if it's a blower.
 
With Boost 2.0 and 3.0 on NVIDIA cards the maximum clock bin is directly tied to GPU temperature, so the cooler you can maintain it the faster it will be. With two cards it would be worth it to do a custom loop since space is already cramped with air coolers.
 
If you already have a case with tons of fans, it's under 2% at stock voltages, and up to 5% at MASSIVELY higher voltages. We're talking a 30C difference between a good OEM air cooler and a good water cooler, which doesn't mean shit. See here:

http://techbuyersguru.com/benching-geforce-gtx-1080-ti-air-vs-liquid-cooling?page=2

Just like I say for ALL water cooling: not worth the investment, unless your dream is chasing benchmark scores for a particular card, since you can just buy a faster video card for the cost of the card plus WC.

The only thing water cooling would help you with is if you bought the stupidest case you could have, with zero airflow, and you decided to pack SLI cards into it :)
 
Last edited:
I believe was 80 or 85c for Nvidia(max redline 90c if caps are rated as used to, not want to push up there as kill life of product granted, Nv did very well with Pascal as an example to keep temps down performance maxed out as much possible often as can)

I could absolutely be wrong (would not be first, nor the last)
I am basing my thought on they have predominantely chose to use cps/Vreg etc rated for 85 and 105c vs AMD / ATi who tended to use 100 (115) and 125c instead.

lower temp available does not mean auto "terrible design"(nor a guarantee max possible speed)
then again, at the cost of a few cents per unit (which consumer absolutely pays a premium for in first place...NVDA especially

I do agree with ~10% averaged (least I am say as such) any higher you likely got a real nice setup / "cherry" unit

(provided similar cooling ability) liquid over air, especially on GPU (full block style) ... more often at or near max clocks little need to throttle vs air often enough
cooler makers for even 3rd party do not seem to "bother" much at all to flesh out make best possible "cool all parts" on the GPU (some rare exception to this rule of course)

No one right answer for every possible build, Air usually is "prefered" for most folks as little can "go wrong"
compared to liquid of any sort of which many factors can "make or break"

^.^
 
If you already have a case with tons of fans, it's under 2% at stock voltages, and up to 5% at MASSIVELY higher voltages. We're talking a 30C difference between a good OEM air cooler and a good water cooler, which doesn't mean shit. See here:

http://techbuyersguru.com/benching-geforce-gtx-1080-ti-air-vs-liquid-cooling?page=2

Just like I say for ALL water cooling: not worth the investment, unless your dream is chasing benchmark scores for a particular card, since you can just buy a faster video card for the cost of the card plus WC.

The only thing water cooling would help you with is if you bought the stupidest case you could have, with zero airflow, and you decided to pack SLI cards into it :)

Well I think he did say 2x blowers so I guessed earlier around 10%. I watercool to heat exchange with my pool and for funsies.

You’re right it’s not really worth it from a value perspective. It’s more of a hobby. You also have that pesky problem that water and electronics don’t mix. Assuming this guy is talking about 2080tis that’s $2,400 you have on the line.
 
Last edited:
I reckon around 2 to 4% extra performance on a CPU.
I use an Accelero cooler on my GPU which is near water performance (+very quiet but is a pita to fit) and also gets around 2% benefit.
The biggest bonus is how quiet you can run a water cooler and your equipment should last longer with the lower temps.
 
I reckon around 2 to 4% extra performance on a CPU.
I use an Accelero cooler on my GPU which is near water performance (+very quiet but is a pita to fit) and also gets around 2% benefit.
The biggest bonus is how quiet you can run a water cooler and your equipment should last longer with the lower temps.

Last time I checked, most 3rd-party video card coolers had comparable or better noise levesls to water cooling. They even turn the fans off during idle.


If you're not stressing it 24/7, the "increased lifetime" is purely in your head. I've run air cooled GPUs for two decades, and all of them have outlasted their usefulness. Even toasty cards like the stock HD 4850 single-slot blower (running at 85C!) But I've also never been one to run my cards 24/7, just a few hours of gaming every night.

If you're stressing it 24/7, a water cooler might buy you a year or two more before it fails. But even server GPUS are air cooled, because of the complexity of doing water cooling, so it's not a panacea for high workloads.

Because even in a server, the GPU is rarely stressed 24/7, they don't bother with water cooling because the difference in lifetime is pretty small.
 
Last edited:
Last time I checked, most 3rd-party video card coolers had comparable or better noise levesls to water cooling. They even turn the fans off during idle.


If you're not stressing it 24/7, the "increased lifetime" is purely in your head. I've run air cooled GPUs for two decades, and all of them have outlasted their usefulness. Even toasty cards like the stock HD 4850 single-slot blower (running at 85C!) But I've also never been one to run my cards 24/7, just a few hours of gaming every night.

If you're stressing it 24/7, a water cooler might buy you a year or two more before it fails. But even server GPUS are air cooled, because of the complexity of doing water cooling, so it's not a panacea for high workloads.

Because even in a server, the GPU is rarely stressed 24/7, they don't bother with water cooling because the difference in lifetime is pretty small.
I'm not sure why you are telling me about 3rd party coolers, I stated I use an Accelero with my GPU instead of a water block.

The lifetime matters if you wish to run high clocks and pass the card on to someone else to use later.
High overclocks without good cooling can cause silicon to degrade enough to cause problems at stock clocks, done it a few times.
VRMs and capacitors can suffer rapid ageing too.
 
Back in the day I cut 3 x 120mm fans into the top of my case (2005 -6 ) flipped the power supply as to get all that air going over it without ever needing the factory fan and a way for the mod to give back as we were looking for everything that could be cooled down ,
 
On the 20xx series I'd say that watercooling makes a bigger difference in clock stability than the max clocks you can get. It will throttle less. Obviously noise is greatly reduced and that's my #1 reason to slap an AIO on my GPU.
 
Back
Top