View Single Post
Unread 03-09-2006, 09:24 AM   #26
pHaestus
Big Player
Making Big Money
 
pHaestus's Avatar
 
Join Date: Aug 2001
Location: irc.lostgeek.com #procooling.com
Posts: 4,782
Default Re: An End to the Insanity.

Flip it around though.

Subject MY test system to the same scrutiny. It would fail. How do you know the temperature measured is actually anything like the "hot" part of the CPU? Do I have any idea about the heat spreading through the silicon itself? How do you correlate that number with die sims; with other types of CPUs; with other CPUs of the same types in different systems? Why should anyone trust it at all? You can come up with a million "deal breakers" about why a particular test method is invalid for a particular waterblock or a particular class of coolers.

But what makes you think Intel and OEMs have never tested a waterblock on the TTV? What makes you think that our forum is SO much more clever than all these engineers? Cause one guy whose waterblock didn't perform all that well said so? A guy using an onboard diode reading to take temps?

Where's the scrutiny and evaluation of commercial motivations and questioning of appropriateness of THAT?

//edit: What I'm really trying to say is that blind trust IS bad. Sure it's a bad idea to think that the results from an Intel TTV-based test platform are 100% correct for all CPUs. But um the tester providing the results never said that! However, it's also bad to throw out all testing done by people with access to OEM labs and thousands of dollars in equipment because the results didnt agree with one waterblock mfgrs expectation of performance. Everyone is throwing the baby out with the bathwater, but the issue is so muddied and poisoned with personal stuff now it isn't salvageable.

Did you read my article about testing? If you want to test with an IHS present, then the Intel TTV is the BEST platform in a lot of ways. The only problem with it in my eyes is that it's not available to me (and even if it were I couldn't afford it). Barring that platform then an Intel 775 chip is the next best thing. Yea the ttv has to be corrolated with other kinds of CPUs and other die sim results, but uh SO DOES EVERY OTHER TEST METHOD. Nobody ever bothered to bitch at me for not performing statistical correlations with JoeC or BillA; why all the sudden is the TTV held to this standard?

What it comes down to is that with a simulator (if you're smart) you'll trade some small bit of real world relevance for a large gain in accuracy, robustness, and repeatability. Look at the ttv compared to copper dies and CPUs now in that framework. The ONLY disadvantage I can see is the 0.1C res of the tcs compared to pRTDs. Everything else can be handled in the correlations I think.

Noone talks about the fact that I had to swap motherboards several times over a year of testing due to leaks, bad caps, dead parts. I had to change CPUs once, and I had to resolder wires once. I had to recalibrate a few times because of moving the system around and "tweaking" the wires on CPU diode pins. ALL changes resulted in a couple of weeks of recalibration and retesting of "reference" blocks. Every time I'd take a week off, I'd have to revalidate the entire thing. The TTV and all systems based upon case temps will not have the same problems. OK they may have other ones but what makes everyone so quick to dismiss them out of hand? You guys who haven't done the kind of testing I did on SocketA have no idea the amount of trust you're placing in a reviewer to not bias the outcomes and not work with out-of-calibration parts.
__________________
Getting paid like a biker with the best crank...
-MF DOOM

Last edited by pHaestus; 03-09-2006 at 10:44 AM.
pHaestus is offline