Temperature and Performance

[llama]

Silverfox,

Cool.

(no pun intended. ;-))

I'm looking forward to the results.

(And maybe "annoyed" was too strong a word. Sorry. But I think you know what I mean.)

-Llama

[SilverFox]

LLama, no problem.  And I can see why my wording was incorrect.  I'm often thinking one thing and writing something else.

I have results, and I must admit that,  in the end, temperature will have a negligible effect on frame rates.   While the system scored higher on average  when cool than it did when warm, the difference in performance was not as pronounced as I would have thought before the experiment.  I had a bit of a problem keeping the system temp low during cool tests and high during the high tests.  Also, the overall temperature difference was only 24°F.  I used °F for resolution.  The GPU temp was read from the Catalyst Control panel.   As a side note. ( I didn't record them but should have) the frame rates did not vary significantly during the tests.

So in the end, having the CPU cooled properly according to manufacturer specs is the important thing.

[llama]

Silverfox,

I applaud your testing regimine; it looks like you did an extremely thorough job, up to and including multiple runs and a nice average of the results.

These are completely fascinating results. Though I'm not entirely sure that even the minor recorded differences are outside the margin of error (I can run 3dMark 10 times in a row and get results that vary by about 1.2%), I think its pretty significant that the "better cooled" values are ALWAYS faster than the "worse cooled, bit still within spec" numbers, except in one case, where there's a tie.

The "better cooled" benchmarks  ALWAYS work out to be .95% .70% better than the "warm" benchmarks.  As you said, neglible, but still (probably) a trend.

Of course, by doing the "good cooling" you can now overclock your rig to the point where the temperatures rise to the "warm" level, and get that much more performance. THAT's the point of extra cooling, I think.

Again, I applaud your work. If I were you, I'd email these results to Anandtech or Tomshardware; I think they're that interesting.

-Llama

[boxboy28]

LLama and Silverfox and i applaud you8 both! Excellent job on the temp graph  to bench mark results and a Big for even going thru the time to do it all your self if only to make a point!

Extreme cooling or extra cooling is for Over clocking and longevity of the processer and other components!

But 1 point that has been made here i think is make sure your getting things cooled to spec temps!  I myself like them  in the 30-45 Degree C range!  more than that = it too hot FOR ME! (imo)

also alot of people forget case/CPU/GPU temps are  all  tied to Ambiant temps!!!!!!!!!!!

So if you live in Texas on a hot summer day vs if you live in Michigan in the winter Whats the TEMP of the house and the air inside!

One thing every one should remember is AIR FLOW IN should = AIR FLOW OUT...... and most manufacturer heat sinks and fans SHOULD be adiquit (with a goot thermal paste) for any PC in an ambient temp range of 68- to 80 degrees F (yes F i cant convert , left that formula in HS and College)

[SilverFox]

Gents thanks for the kudos.  I wish I had been more thorough, but with a scenario yesterday and a lunch with a couple of my squadies today, KOOL and Caspr, I wanted to get it done yesterday morning.  Good thing it was a cool night.  Its actually hard to achieve those 70° temps without a cool breeze.    That goes right along with your comment about ambient temperature Boxboy.  
 
LLama, I just might submit those results, but I honestly think it deserves a better experimental design.   What the experiment really needs is the stock cooling solution coupled with ribbon type cables and other air impedance to emulate a poorly ventilated system.  My original intent involved a boost in performance for poorly ventilated systems with the application of proper  cooling.  I have seen "store bought" systems that had less than optimal cooling and that's what I was targeting.  Unfortunately for the experiment, I'm not inclined to modify my system to that extent.  I'm sure everyone would agree that  such a mod would be non-sensible.  Maybe I'll do it on the kids machine  

BTW folks, I do overclock,  which is why I went water cooled.  I've backed off in recent weeks because I had some lockups in flight.  I've learned something here, and I can say that I'm not quite as obsessed with temperature anymore.

[SilverFox]

I ran across this site that presents some research on my suppositions.  For anyone who may be interested in the mathematics of this discussion I refer you to page 4 of the pdf.  Although the researchers do present it in terms anyone can readily understand.  

http://www.sliponline.org/SLIP04/Presentations/Session2/2-1_Rajendran.pdf

[rabbidrabbit]

stop fooling around and just run with a silent water cooling solution. I run a zalman reserator on both the machines in my main bedroom and they are just about silent.  It runs about 250 by time you get the cpu cooler but the good thing is its very easy to upgrade from one system to the next since at worst your getting a new heatsink for 20-30 bucks.

[DieAz]

Originally posted by Ghosth
I cheated.

I had heat issues, tried replacing fans, reversing flow, all kinds of fun stuff.

Ended up with the side of the case off, a 8" fan blowing air in right at the cpu/video card.

It ain't purty but it works.  

(If I need more cooling I just open the window behind my box a smidge.)

8 inch ehh? I'm using a 10 inch ;-)