A Thought on Max Temps

StellaTerra

Weaksauce
Joined
Oct 17, 2008
Messages
67
I had an interesting realization today: Max temps for GPUs are significantly higher than max temps for CPUs, at the same time that temp deltas across water blocks are lower. This means that a GPU loop can stand significantly more than a loop that includes a CPU. So here is how my crazy thought goes:
Let's talk worst case scenarios.
I live in Santa Cruz california. Temps here will never exceed 40 degrees celsius, or if they do I'm going to be too uncomfortable to be playing a video game, so ambient air temps are 40C.
At 1000 RPMs of the yate loon the XSPC RS120 radiator has a water-air temp delta of .094 C/W according to martin's lab. Let's assume that that's a good, but average score under most circumstances with a single-120mm radiator.
Now let's bring out the real crazy. Most people on the forum probably wouldn't recommend a GTX 280 on a single rad loop, but let's say we're out of our freakin skulls, and we put two on it, at stock clock. That's ~200W of heat a piece worst case scenario, so 400W is the number. That's 37.6 temp/air delta. Frying, I know, but wait for it.
Now let's say we're using a Fuxion GFX on each of the cards, which according to martin has a water-gpu delta of always less than 15.
That's 40 + 37.6 + 15 = 92.6. Yes, that's some hot video cards, but here's the thing: They don't crap out until 105C! At absolute worst case scenario a single 120 on ultra low settings is enough to keep two GTX 280s at stock from shutting down. This is possible because the finicky CPU isn't in the loop. I'm not 100% sure that all of these stats apply the same given the extreme temps, but from what I understand the greater deltas should be in the radiators favor. In addition I don't think the GPUs would ever ACTUALLY dump 400W of power into the water.

So here's my final thought: I think you could cool a fairly easily cool an entire stock (or moderately overclocked) system on two 1x120mm fan radiators with two separate loops at nearly silent fan speeds, all from inside the case without doing any modding. Imagine what you could do with higher fan speeds! The key, I think, is keeping the CPU isolated in it's own loop.
 
You might see some temp improvements running two separate loops. I doubt load temps will be any good though on a single rad. It seems most people that run two separate loops are a bit more on the extreme side and usually have a 4870x2 CF or GTX280 Tri SLi setup. It's those guys that really benefit from running separate loops, buit I could see
people with smaller cases benefiting from dual single rads as well.

For a single gtx280 and a CPU, running two separate loops for one...isn't cost effective,
two...might require the use of higher RPM fans which gives you more noise, and three, requires a 2nd pump...at least, I think it would.

That said, probably would work just fine, and would be really interested to see your temps
if you plan on doing this.
 
Sometime in the future, I'm certain I will.

All I was really getting at here was that video cards seem to have a MUCH higher tolerance for high core temps than procs, and at the same time are easier to cool, so as long as you didn't have a CPU in the mix you could get pretty crazy with GTX 280s. In the circumstances provided above you could easily take care of 500W with a single fan radiator (according to the math and Martin's lab.)

I guess this thread was meant to be a reality check/check my math type thing.
 
At 1000RPM a single XSPC radiator, coping with 200w of heat, will have a water out delta of around 25c... With two cards, if we assume the curve is linear, the water temperature will be around 50c over ambient temperature...

This is pushing dangerously close to the 100c limit... I wouldn't do this...
 
Back
Top