OCCT 3.0.0 now available

[LOUISSSSS]

Originally posted by: TidusZ
I ran prime95 blend on an overclock on a new pc I built, and got an error on one core after 5 1/2 hours, and no errors on the other 3 cores after 12 hours. I have since ran it for 2 hours on occt and no errors (2 x 1 hour tests) and I'm running it some more now. Q9550 @3.8x 1.344v on UD3P (1.3v nb). I'm interested to see if occt will pickup an error and how long it will take. I'll try a 12 hour test overnight tonight.

-it doesn't matter if theres 1 error or 4, or how long ur remaining cpu-threads run without errors after there is one. your cpu is still unstable and thats all that matters
-finding errors depends on the coding of the program used and how optimized it is at finding cpu errors.
-2 x 1hr tests isn't the same as a 2 hour test because some of the 2 x 1hr tests can be repeated. you need to run it continuous.

-on this program: the psu tests is working 100% well on my system so far and it consists of:
linpack + crysis. (from task manager) my PSU is holding well and i'm getting 80fps (est)

-my max cpu temps is set to 80c on OCCT and so far my system hasn't reached it (well below 60)

-cpu: OCCT/LINPACK how does it notify me of errors found?

 

[LOUISSSSS]

Originally posted by: soccerballtux

Originally posted by: SickBeast
Is this program bloatware compared to P95/Orthos?

How does the GPU testing work? I've yet to find one that's accurate.

ATI-tool is the generally accepted norm?

NONONONONO. where r u getting this from?

 

[VirtualLarry]

Originally posted by: TidusZ
I ran prime95 blend on an overclock on a new pc I built, and got an error on one core after 5 1/2 hours, and no errors on the other 3 cores after 12 hours. I have since ran it for 2 hours on occt and no errors (2 x 1 hour tests) and I'm running it some more now. Q9550 @3.8x 1.344v on UD3P (1.3v nb). I'm interested to see if occt will pickup an error and how long it will take. I'll try a 12 hour test overnight tonight.

See, this is why I like Prime95. It takes a while, but it's a very good test.

 

[TidusZ]

well, I let occt run overnight, and it found an error after just over two hours. Perhaps I have underestimated Occt.

 

[Idontcare]

Originally posted by: LOUISSSSS

Originally posted by: TidusZ
I ran prime95 blend on an overclock on a new pc I built, and got an error on one core after 5 1/2 hours, and no errors on the other 3 cores after 12 hours. I have since ran it for 2 hours on occt and no errors (2 x 1 hour tests) and I'm running it some more now. Q9550 @3.8x 1.344v on UD3P (1.3v nb). I'm interested to see if occt will pickup an error and how long it will take. I'll try a 12 hour test overnight tonight.

-it doesn't matter if theres 1 error or 4, or how long ur remaining cpu-threads run without errors after there is one. your cpu is still unstable and thats all that matters
-finding errors depends on the coding of the program used and how optimized it is at finding cpu errors.
-2 x 1hr tests isn't the same as a 2 hour test because some of the 2 x 1hr tests can be repeated. you need to run it continuous.

I was going to respond and then I saw your post Louis. Precisely what I was going to post. :thumbsup:

In Prime95 when a worker thread dies due to an error from instability in the core it is pointless, absolutely pointless, to continue the existing Prime95 session with the dead worker thread as now the heat generated by the CPU is 25% less than before (the core with the error no longer produces load levels of heat once the thread dies).

In the absence of the heat from that core the other three cores now operate in a cooler environment, masking their instabilities.

In Prime95 when one error is found that is the end of the point of doing the test.

 

[Thor86]

Originally posted by: VirtualLarry

Originally posted by: TidusZ
I ran prime95 blend on an overclock on a new pc I built, and got an error on one core after 5 1/2 hours, and no errors on the other 3 cores after 12 hours. I have since ran it for 2 hours on occt and no errors (2 x 1 hour tests) and I'm running it some more now. Q9550 @3.8x 1.344v on UD3P (1.3v nb). I'm interested to see if occt will pickup an error and how long it will take. I'll try a 12 hour test overnight tonight.

See, this is why I like Prime95. It takes a while, but it's a very good test.

Exactly, so I don't know where the OP thinks a 2 hour test with OCCT means your OC settings are stable.

 

[chizow]

Originally posted by: Rubycon
GPU/PSU blocked due to DX9 files missing? Anyone else get that? Shouldn't Vista already have this stuff?

Vista doesn't ship with the DX9 runtime, surprisingly. Even more surprising is that Win7 doesn't either. You have to install them separately, however, since most games requiring DX9 include the redistributable, its usually installed with the first game.


Anyways, memory test is broken for me too. All of the other tests seem to do a nice job stressing the appropriate components.

As for Prime95 vs OCCT arguments, I don't really care for the programs that test heat so much as the ones that test stability under load. Realistically, my CPU never gets as hot as it does under Prime95 (which only reaches peak heat after 6-8 hours) or IBT in actual games/apps, so its really not an accurate gauge of stability under normal circumstances anyways. If OCCT accomplishes what Prime95 does in much less time with less heat, I'm fine with that.

 

[LOUISSSSS]

Originally posted by: Thor86

Originally posted by: VirtualLarry

Originally posted by: TidusZ
I ran prime95 blend on an overclock on a new pc I built, and got an error on one core after 5 1/2 hours, and no errors on the other 3 cores after 12 hours. I have since ran it for 2 hours on occt and no errors (2 x 1 hour tests) and I'm running it some more now. Q9550 @3.8x 1.344v on UD3P (1.3v nb). I'm interested to see if occt will pickup an error and how long it will take. I'll try a 12 hour test overnight tonight.

See, this is why I like Prime95. It takes a while, but it's a very good test.

Exactly, so I don't know where the OP thinks a 2 hour test with OCCT means your OC settings are stable.

exactly exactly exactly. you're free to do whatever you want with your computer. you can run p95 for 800 hours for more stability, but why dont u go ahead and give OCCT a few shots. try to push your cpu a bit more and test 2 hours w/ occt.

in MY test findings, 2 hours of occt test tells me that my CPU is stable (for anything i do.) Keep in mind that nothing is ever 100% stable. but feel free to ignore this statement and go on with your 24 hour testing.

As for the program itself, i ran the GPU:Memtest for 10 passes yesterday and it went error free.
My system also finished 1 hour of the power supply test error free.

 

[TidusZ]

Originally posted by: Thor86

Originally posted by: VirtualLarry

Originally posted by: TidusZ
I ran prime95 blend on an overclock on a new pc I built, and got an error on one core after 5 1/2 hours, and no errors on the other 3 cores after 12 hours. I have since ran it for 2 hours on occt and no errors (2 x 1 hour tests) and I'm running it some more now. Q9550 @3.8x 1.344v on UD3P (1.3v nb). I'm interested to see if occt will pickup an error and how long it will take. I'll try a 12 hour test overnight tonight.

See, this is why I like Prime95. It takes a while, but it's a very good test.

Exactly, so I don't know where the OP thinks a 2 hour test with OCCT means your OC settings are stable.

I don't think you guys understood the point of my post. I'm trying to compare how quickly prime95 takes to find an error when its close to stable (but not stable) compared to occt. I found occt was quicker at finding an error in this one particular instance. If other people provide such examples we can figure out the characteristics of OCCT, which for many people here (it seems), is a newly added program to add to the list for testing stability.

So to add to that, I ran prime95 again with slightly higher volts, found an error after 7 hours, have been running occt for 5h so far and no error found yet.

 

[LOUISSSSS]

OCCT is now a new stability testing program. its been around for years.
iirc, it uses a similar algorithm as p95, but only MORE OPTIMIZED so there isn't a any way it can be worse off.

not only once but more than 10 times i've found that using occt finds errors faster on average than prime 95, which often takes 12-24 hours to find an error. OCCT usually stops me under 2 hours.

i've also found that 2 hours of OCCT testing gives me a stable enough CPU for anything i do: some video encoding for my home viewing + for my iphone and some gaming. i've never encountered any crashes that were cpu related. only occasional crashes in game with blue screens due to nv4.disp nvidia drivers.

 

[Idontcare]

Originally posted by: LOUISSSSS
OCCT is now a new stability testing program. its been around for years.
iirc, it uses a similar algorithm as p95, but only MORE OPTIMIZED so there isn't a any way it can be worse off.

An Achilles heel of these stability programs though is that they don't stress test all the instructions that a processor can be tasked with processing. They focus on a very narrow set class of computations, albeit a class that is well represented by the applications generally used as well, in an effort to maximize IPC so power consumption can be pushed to high values.

What we really need is a program that (A) does exactly what these programs do, and (B) once the CPU is operating in that elevated temperature environment a test suite that iteratively executes 100% of all possible processor instructions to ensure that signal/noise for all logic loops is still acceptable to deliver stable behavior.

 

[LOUISSSSS]

you guys know that warning that OCCT gives you when you have cpu-z open when you try and open OCCT after? do you guys just ignore that warning or do you close up all cpu-z and CPUID Hardware Monitor Programs before using OCCT?

whats the effects on it?

 

[Idontcare]

Yeah, I close CPUz. I don't know what the problem would be as you can have multiple instances of CPUz open at the same time and it doesn't cause CPUz to go crazy or anything. But I close it as OCCT asks nonetheless.

Maybe email the author and ask? I'd be curious what he says if you do get a response.

 

[nyker96]

i love the new occt, got almost everything in one package, except their e21xx M0 temperature was reported 15C less than the actual. I think that is a mistake, but you can always run realtemp along side of it. I actually think their monitoring capability is adequate.

 

[faxon]

been getting the upgrade itch again lately. figured i would have some fun with this to see if i can find any stability issues which would give me a reason to put out for stuff. my only complaint is that you cant run the GPU test in windowed mode at higher than 1280x1024, which is an unfair representation of my normal gaming habits, since i spend a lot of time playing windowed in some games due to needing to constantly access other apps while im playing (browser, combat tracker, ventrilo). still, looks like the test puts a larger load on my GPU than everquest2 does, which is the app i play windowed, so i can live with it

 

[taltamir]

Originally posted by: Rubycon

Originally posted by: taltamir

Originally posted by: Rubycon

Originally posted by: WobbleWobble
I get a ton of errors in the GPU Memtest

I got a stock BFG 260 GTX and I'm running my system at stock. Anyone else able to run the GPU memtest successfully?

Me too and it quacked at me! GTX280 factory OC from EVGA here.

maybe nvidia engineers KNEW what they were doing when they set those default clocks...

The tests give the same results regardless of how the clocks were set. This program is most likely producing bogus results. I never have any issues with the card otherwise.

I can run the memtest fine, dozens of tests all passed without a single hitch on my eVGA GTX260 SC

But if underclocking your card it still gives errors... well it could be bogus results or it could be a deeper issue with your card

 

[LOUISSSSS]

what can go wrong with leaving CPU-z open with OCCT at the same time? will that mess with the test results? or will it only mess with their display results?

 

[Rubycon]

FYI 3.0.1 is available and I get NO ERRORS on the GPU memory test. The older (3.0.0) version - as suspected - was borked.