916

anyone have the download link for the ITB?

TwoCables

LinX has an Auto Save for it. Check out the Settings:



For stress testing, you should use the Small FFTs test in Prime95, and also LinX using "All" the available memory. You see, LinX isn't better than Prime95, and Prime95 isn't better than LinX. They're just different from each other.
It's absolutely true that you should use both (but not simultaneously - never, ever multi-task while running a stress test). The reason you should use both is because they're different from one another.
If you're overclocking, then Auto voltages should never be used - ever. The Auto setting always sets the voltage higher than necessary, and therefore it creates more unnecessary heat. And since having too much heat can cause instability, it's best to never use Auto.
I have learned from experience that an overclock can be stable enough for 12 hours of the Small FFTs test (Orthos or Prime 95 - same thing for a dual core), but yet be absolutely unable to go longer than like 5 linpack runs. You see, I had my overclock stable enough to pass over 12 hours of the Small FFTs test, but it couldn't make it much past 5 runs. But this was no surprise to me as I was still seeing some problems due to instability - although these problems were quite infrequent. After getting my overclock stable enough to do the 37 runs as seen in my sig rig, those problems completely disappeared.
The only way that LinX is superior to the IBT is that it has a much, much better GUI. But every linpack test uses the same linpack test. LinX just provides the most control over the test as well as making linpack testing just as easy as using Notepad or something. LinX has tooltips for everything, and it has a really good and short Read Me file. And if the system doesn't lock up or blue screen due to instability during the test, then the way you can tell the overclock is stable is by seeing if that green bar is still green. If it's red, then you'll see this from across the room and know what's up. You don't get that kind of obviousness with other linpack testers. But for prime testers, you get that kind of obviousness with Orthos. But to make Orthos stress a quad, two instances must be run, and then the Affinity must be set in the Task Manager for either instance so that one instance is stressing cores 1 and 2, and the other instance is stressing cores 3 and 4 (so, cores 0 and 1, and 2 and 3).

For a quick check, run about 30 minutes to an hour of the Small FFTs test, and maybe 10-20 linpack runs. After all, 50-100 runs is considered stable - at least as far as linpack testing is concerned. It doesn't guarantee that it'll be stable enough for 12 hours or more of the Small FFTs test in Prime95/Orthos.
That doesn't mean the linpack test is better than the Small FFTs test: it just means that your overclock needed to be more stable. Many people think my stress testing recommendations are extreme, but they're really not. I mean, an overclock can pass 12 hours of the Small FFTs test, but be baked in 5 runs. So, all one has to do is increase the stability until it can pass 50 runs, and bam: your overclock has just been made way more stable than it ever could have been with the Small FFTs test alone. But at the same time, it's a bad idea to only rely on linpack testing for the same reason. I have discovered in my own experience than an overclock can pass like 20 linpack runs but be fried within a few hours of the Small FFTs test. So, the overclock has to be stable enough for both.
LinX's readme file says that passing 50-100 with the maximum amount of memory used is "pure unarguable stability". They consider values lower than 50 a quick test. It says "For quick testing, lower values are acceptable". This was right after recommending to do 50-100 runs.
Actually, here's the download link to LinX: http://www.xtremesystems.org/forums/...d.php?t=201670

__________________

Rayce185

Thanks TwoCables, very helpful!

I'll get back to testing after lapping the TRUE and applying the vdroop mod. A drop of 0.088V under load just isn't acceptable.

__________________
.

=The Q9550 Club= | We Are Gr33n Machine | ß₤ứə Çřёώ
iPhone Club | Top Gear Fan Club | OCN Forza 3 Club

TwoCables

You're welcome. And I'm done editing now. I made some changes that I feel were necessary. Some were very big changes, and some simply improved my English. I have a trigger finger.

As for vdroop, please consider not doing it. Please see this AnandTech article below. I've linked to the first page of a 2-page section of this very big article that talks about why vdroop exists, and why we should not mess with it. It's a measure of protection, and removing, or even reducing vdroop can have very, very bad results in the long run.

http://www.anandtech.com/cpuchipsets...spx?i=3184&p=5

__________________

I_dalder_I

:

you quoted me exactly from your other thread

i dunno its 6am here and ive been up...it seemed interesting to me

Rayce185

So... We need someone to develop a program that both uses the prime test and linpack test and combne it into one big stresstesting program.


Does anyone volunteer?

__________________
.

=The Q9550 Club= | We Are Gr33n Machine | ß₤ứə Çřёώ
iPhone Club | Top Gear Fan Club | OCN Forza 3 Club

TwoCables

That's been done, but poorly: OCCT. But I refuse to provide a link to it. It takes over 12 seconds to open on my system. But I can open LinX, Prime95, Orthos, Real Temp and Core Temp all in the amount of time it takes to open OCCT (much faster, actually). I can probably even include CPU-Z, but I haven't tried it (I don't have OCCT anymore since finding out how long it takes to open). So, I think OCCT is a joke, and I don't mind if anyone wants to tell me that I'm flat-out wrong. But OCCT's "Small Data Set" is the Small FFTs prime test. The "Medium Data Set" is the Blend prime test. The "Large Data Set" is the Large, in-place FFTs prime test.

But many people adore OCCT because of its One-Stop-Shopness. But I have absolutely no need for such silliness. I'm sorry.

__________________

Inktfish

My OC was both Prime as LinX stable.
F@H, however failed miserably.

__________________

Rayce185

Well yeah, that's why someone needs to make a better version, just like Linx is the superior Linpack stresser.

__________________
.

=The Q9550 Club= | We Are Gr33n Machine | ß₤ứə Çřёώ
iPhone Club | Top Gear Fan Club | OCN Forza 3 Club

TwoCables

How many hours in Prime95? Was it the Small FFTs test, Blend, or Large, in-place FFTS? How many runs in LinX? How much memory did you tell it to use?

__________________