Question Your experience using OCBASE OCCT -- version 8 or earlier: What's your verdict?

[BonzaiDuck]

I probably wouldn't be poking around these tech forums so much over the past month, but for my vape-pen USB disaster. Looking at the wallet-damage, I could've been up and running for the mere price of Fed-Ex RMA and a two/three-week wait -- about $88-worth. But upgrading from the Skylake to a Kaby was still fun.

I'm always looking for reliable short-cuts to stress-testing, and this month was a special case. I've seen custom Prime-95 setups that can detect errors within an hour that could otherwise be missed. I'd also observed over recent years the value of the OCCT:CPU tests, which can be chosen for "SSE", "AVX" and "AVX2". Somehow the the earlier chip, I'd missed the opportunity to use the "Negative AVX Offset" feature, so this go-around taught me some "new stuff". OCCT has built-in a LinPack test and memory tests, with settings for "data set", "Mode" (normal or extreme) and Load type (variable or steady) -- as well as the instruction-set choices.

I don't remember ever purchasing OCCT, but I was using it as much as LinX, IBT or PRime95. This time, I noticed the test durations are limited to 1 hour unless you "subscribe" to it for a year, so I did that . . . too . . .
What do other members say about it?

 

[Ajay]

I've used it for stability testing for years. I did buy a perpetual license though. It's got a lot of features and can run multiple tests that can be helpful.

 

[BonzaiDuck]

I've used it for stability testing for years. I did buy a perpetual license though. It's got a lot of features and can run multiple tests that can be helpful.

Thanks for sharing your experience with it.

As some of our veteran colleagues here suggest, over-clocking is becoming less and less "necessary" given the performance just from recent-generation Intel CPUs. But as i also volunteered from my own experience, it's still an addiction of sorts.

Running Prime95 for 24 hours is a huge inconvenience, even if one is replacing a machine still in service and doing it at a leisurely pace. I had discovered that the OCCT:CPU tests can uncover errors in an hour that wouldn't show through ten iterations of IBT or LinX. The developer -- I think he's French -- has done a nice job with that testing suite.

I think there is still an extant opinion that enthusiasts will invariably have BSOD experiences with their over-clocking adventures. That might be true if one goes into the enterprise half-blind to full information about the CPU, but "those people" are good guinea-pigs. Eventually, someone publishes a graph for a particular processor plotting VCORE against speed settings, some showing a curve for the top-end binned processors and another curve for the "potatoes".

So just getting such guidelines for wading into the trial-and-error of it relieves the BSOD panic attacks. Better than that, OCCT seems to do a good job at trapping the errors and recognizing them in real-time, so a revisit to the BIOS screens is merely preceded by the "Shutting down" Windows screen.

 

[JoeRambo]

Nice software for testing with hw monitoring that is top of the game (due to using hwinfo64 internally ). I quite like UI as well, shiny graphs etc.

My personal experience:
CPU testing has ton of options, from casual ( think CB stable ) to full on Linpack. Good stuff, i'd run some tests and if OK run RealBench for some time and call it a day.
Memory testing is not at the level of running stressapptest on linux/WSL or memtestpro, but it is OK.
GPU testing i don't like, it is basically "shader" era stuff, with TON of heat, but can still be unstable in complex games that stress more of chip esp when undervolting.
GPU memory testing won't detect anything for me.
Power testing is nice, but care needs to be taken not to blow nearest power substation with latest Intel/Nvidia stuff.

Overall 9/10, free version with 1h limit is plenty for casual checks and then moving on to specialized tests or actual gaming.

 

[BonzaiDuck]

The discussion brought to mind something I try and do with LinX. Nobody has ever attempted to tell me whether it's a fruitless exercise. Someone, at some time, may have suggested that Windows background processes could make my "procedure" less reliable.

But it wasn't my idea. We had a member -- he may still be with us -- who was a pioneer in the de-lidding phenomenon -- IdontCare was (is?) his handle. He created and contributed almost as a legend with this thread:

Bare-die testing: A delidded 3770k, an H100, and 9 different TIMs

With maybe 249 pages of Forum archives, I'm surprised I was able to find this thread in the 15 or 20 minutes I gave to it!

Now I'm not sure if Idontcare's observations about "Affinitized Linx Testing" were in that thread or another one. Better to search on all threads created by Idontcare! Maybe? Maybe not.

Even so. He described how to load LinX and choose the affinity for the program for only every other thread of an i7 hyperthreading-enabled processor. Nothing special -- anyone can poke around in the windows pop-ups and figure it out.

Then, you run LinX to use a maximum problem size and maximum RAM. Say you run ten or twenty iterations -- they're a "block sample", and ten should do it.

Idontcare observed that settings which appear to be a stable, solid over-clock may fall short. And he explained that the processor may engage in its own error-correction, so that insufficient voltage would lead to instances or iterations wherein the GFLOPs recorded for an iteration might vary to a noticeable degree across the sample.

Now, suppose Idontcare finds this thread and chooses to respond. He may correct me! He knows more than I do, and that's a fact! But even so . . . again . . .

I followed up on that discussion, and started running ten-iteration samples of affinitized LinX after I'd "found" a stable overclock voltage setting. And I was SURE -- SURE, I say! -- that starving the voltage a bit would lead to wider GFLOP variation, while bumping up the voltage in increments would narrow the range for a sample.

You could throw out "outliers", like the first iteration, or some iteration where you "had to do something" with the computer during the test run. But you could either load the results into Excel and do an average and standard deviation, or simply measure the range (maximum less minimum) on the sample to find where it had narrowed to within maybe three or four GFLOPs, and that was your final, "tuned" setting.

These other test programs help you at least get to that point and avoid the BSODs, and overall -- you want to get through this . . . sheee-** as quickly as you can without being sloppy.

So I'll throw in my two-cents worth to those who've posted here. OCCT seems to be a darn good testing program!

 

[lobz]

Thanks for sharing your experience with it.

As some of our veteran colleagues here suggest, over-clocking is becoming less and less "necessary" given the performance just from recent-generation Intel CPUs. But as i also volunteered from my own experience, it's still an addiction of sorts.

This phrase indicates you have no concept of the past 3 years of the CPU market, overclocking is not necessary since the celeron A 300, and not too effective after sandy bridge. The software is totally fine though.

 

[BonzaiDuck]

This phrase indicates you have no concept of the past 3 years of the CPU market, overclocking is not necessary since the celeron A 300, and not too effective after sandy bridge. The software is totally fine though.

Thanks to some such as yourself, I'm acquiring the concept, as you say, although mention of the Celeron 300A seems to be more hyperbole. I remember squeezing out about 700 or 800Mhz extra from a Sandy K. My Skylake cores, as you suggest, don't offer up so much. And of course, the difference in real-world performance is much in question.

The excuse I've offered suggests that it merely becomes an addiction. To pursue the analogy further, if one intends to do it, one would like to have the means to avoid overdosing the processor on voltage, or causing the PC to go into convulsions.

As for the lapse of time, 10 or 20 years for someone my age seems much shorter. I can definitely say, though, I've been away from this and other tech forums for about the time you mention. A blink of an eye to me! Like "yesterday"! There's something I suspect and would call "senior-citizen time warp". Call it the "Rip-Van-Winkle Effect".

 

[lobz]

Thanks to some such as yourself, I'm acquiring the concept, as you say, although mention of the Celeron 300A seems to be more hyperbole. I remember squeezing out about 700 or 800Mhz extra from a Sandy K. My Skylake cores, as you suggest, don't offer up so much. And of course, the difference in real-world performance is much in question.

The excuse I've offered suggests that it merely becomes an addiction. To pursue the analogy further, if one intends to do it, one would like to have the means to avoid overdosing the processor on voltage, or causing the PC to go into convulsions.

As for the lapse of time, 10 or 20 years for someone my age seems much shorter. I can definitely say, though, I've been away from this and other tech forums for about the time you mention. A blink of an eye to me! Like "yesterday"! There's something I suspect and would call "senior-citizen time warp". Call it the "Rip-Van-Winkle Effect".

I really didn't mean it as hyperbole, quite the opposite. Getting a 50% faster CPU than you paid for and surpassing 2-3x more expensive CPUs in gaming was no small feat! Sure as hell Intel's never made that 'mistake' again

Hahah, as for the overdose, just avoid LN2, that stuff destroys lives - and organic cells really fast

 

[BonzaiDuck]

I really didn't mean it as hyperbole, quite the opposite. Getting a 50% faster CPU than you paid for and surpassing 2-3x more expensive CPUs in gaming was no small feat! Sure as hell Intel's never made that 'mistake' again

Hahah, as for the overdose, just avoid LN2, that stuff destroys lives - and organic cells really fast

My addiction began much later than the hey-day of that Celeron chip. That was a time, as a computer-programming and info-systems instructor, when I religiously avoided tweaking a BIOS in such a manner. Another old geezer that I know mentioned his OC efforts with a PC-AT 286 and a chip only slightly more recent. He was almost wistful to hear me talk about the Skylake tweaks.

It's something else to do while "sheltering-in-place". For the good part of the last year, I was in the garage tweaking my 26-year-old Trooper. I had only the year before found an honest repair shop to trust for the major work -- replacing the guy I had for 17 years and now retired. Now I discover from their e-mail announcement that they, too, have decided to go out of business. At least things are up to snuff on the vehicle that I should be able to get through another 5 or 10 years just doing my own oil changes and radiator flushes. I've brought the sound-system up to 21st century capability, with two bluetooth receivers and a zillion record albums in MP3 format! I need to drive just so I can turn the music up the way we did back in the '60s and '70s!

I can either sit here at the desk with my virtual speed-demon and test myself with Assetto-Corsa, or I can fill up the real vehicle @ $4+ per gallon and "just go for a drive". Either way, I don't need to wear a mask! Of course, there's the remark by the Chef-turned-soldier in "Apocalypse Now": "Never get out of the boat!"

Somehow, I could be mixed up about a "12-step program for Over-clockers", as a procedure for doing it as opposed to avoiding it!