Simultaneous torture testing with different programs

Also, I did my stability testing at 3.7 @ 1.256v and ram at 1.66v, which is higher clocks and lower voltage that the final specs in my sig. It's serious business! :D

DeadSkull said:
Last time a thread like this started up I found it surprising that several people replied back stating that they don't bother doing the usual 8-12hrs Prime95 to test overclock stability and if the pc handles games and other simple tasks fine then they consider the pc stable. I wonder if these people are the same ones who later up start threads about how their video card drivers suck/ cannot install properly and how other OS/software related stuff goes haywire in their system.
Click to expand...

Yeah, I've read some crazy "video card and gaming issues" over on nvnews, and then in the sig is some crazy cpu overclock, and I don't even bother replying to the thread...
 
xorbe said:
Also, I did my stability testing at 3.7 @ 1.256v and ram at 1.66v, which is higher clocks and lower voltage that the final specs in my sig.
Click to expand...

:eek: I need 1.38V to be stable at 3.6. OK... I'm beginning to think all the silent corruption I had going on impacted my "minimum truly stable" results. I'm gonna make a system image to my 2TB, unplug it, and try lower voltages.
 
MechGeek said:
:eek: I need 1.38V to be stable at 3.6. OK... I'm beginning to think all the silent corruption I had going on impacted my "minimum truly stable" results. I'm gonna make a system image to my 2TB, unplug it, and try lower voltages.
Click to expand...

Did you turn off spread sectrum? 700mV cpu diff voltage? Did you try increasing qpi voltage to 1.4v? There are so many damn clock/voltage knobs in the bios for these processors.
 
Yep, I've tried a little bit of everything. Spread spectrum off, 900mV cpu diff voltage, qpi @ 1.4, PLL @ 2.0, DRAM @ 1.66.

Maybe it's because I'm running a faster uncore (3.6GHz / 7.2GT/s). How high can I go on uncore? 1.5v?

It is prime95 *OR* linpack stable at 1.2875V-1.35V @ 3.6.
Within minutes (sometimes seconds) of the other program being started, BSOD.

The previous screwiness didn't affect the OC's min stable voltage at all. I'm back at 1.38... :(


EDIT: I'm onto something... running tests at 1.30Vcore/1.45Vtt right now and it hasn't BSOD'd yet. Temperatures are 10c cooler.

Can't believe I missed that. :eek:
 
MechGeek said:
7.2GT/s
Click to expand...

Whoa 7.2! i7 965 is only 6.4 GT/s. Is that okay? I don't know what other people have gotten to on that setting.

So you got bit by the uncore too. I needed minimum ~1.37v for 3338 uncore. I'm running a bit close for my standards at 1.40v for 3302 uncore. I didn't want to go over 1.4v because it seemed like a pretty big increase already.

So you have your uncore at 3600 MHz?
 
20x190 @ 1.35V (everything else auto), LINX and Prime95 running for 1.5 hours with max temp 82c. Hopefully it will continue and make it through till morning.

Glad I ran into this thread, this is going to be my primary comp, don't want silent corruption creeping up on me later. Thinking I am becoming a multiple stress test believer.


MechGeek said:
Yep, I've tried a little bit of everything. Spread spectrum off, 900mV cpu diff voltage, qpi @ 1.4, PLL @ 2.0, DRAM @ 1.66.

Maybe it's because I'm running a faster uncore (3.6GHz / 7.2GT/s). How high can I go on uncore? 1.5v?

It is prime95 *OR* linpack stable at 1.2875V-1.35V @ 3.6.
Within minutes (sometimes seconds) of the other program being started, BSOD.

The previous screwiness didn't affect the OC's min stable voltage at all. I'm back at 1.38... :(


EDIT: I'm onto something... running tests at 1.30Vcore/1.45Vtt right now and it hasn't BSOD'd yet. Temperatures are 10c cooler.

Can't believe I missed that. :eek:
Click to expand...

What benefit do you get from not using the auto option with all these other voltage settings. Won't the boards handle or to the usually apply too much voltage?
 
mikecoscia said:
20x190 @ 1.35V (everything else auto), LINX and Prime95 running for 1.5 hours with max temp 82c. Hopefully it will continue and make it through till morning.

Glad I ran into this thread, this is going to be my primary comp, don't want silent corruption creeping up on me later. Thinking I am becoming a multiple stress test believer.
Click to expand...

Try 30 minutes of 3D gaming while those are chugging (set LINX to low priority, Prime95 takes are of itself). The 3D gaming helps stress the uncore portion too. Then slow up just a bit to 20x185 when it passes. :D
 
xorbe said:
Whoa 7.2! i7 965 is only 6.4 GT/s. Is that okay? I don't know what other people have gotten to on that setting.

So you got bit by the uncore too. I needed minimum ~1.37v for 3338 uncore. I'm running a bit close for my standards at 1.40v for 3302 uncore. I didn't want to go over 1.4v because it seemed like a pretty big increase already.

So you have your uncore at 3600 MHz?
Click to expand...

Yep, my uncore is at 3600. Same as my CPU. I like that, because they're in lockstep. That's gotta be good for something. What, I don't know...but something. :p

I'm not sure, but perhaps the i7 965 can do 6.4 GT/s at stock uncore voltage?

I've had somewhat stable OCs at 8 GT/s. So I guess 7.2 is OK with the extra voltage. It stood up to a 13-hour IBT+Prime95 beating. Not sure why a drastic CPU voltage difference made the uncore more stable when it wasn't getting enough voltage... I'll rerun the torture tests to be sure.

Also...setting the QPI link speed in the BIOS has no effect on my board for some reason, I have to take what it gives me.
 
mikecoscia said:
What benefit do you get from not using the auto option with all these other voltage settings. Won't the boards handle or to the usually apply too much voltage?
Click to expand...

At first I thought it was because the board wasn't giving enough voltage somewhere. Except for the spread spectrum thing, that should always be disabled when overclocking.

All my voltages are on Auto now except for core and uncore.
 
xorbe said:
Try 30 minutes of 3D gaming while those are chugging (set LINX to low priority, Prime95 takes are of itself). The 3D gaming helps stress the uncore portion too. Then slow up just a bit to 20x185 when it passes. :D
Click to expand...

Just ran 20min of crysis without any issues besides a little slowdown, temps maxed out around 85c while playing. Why do I need to back down to 185 BLCK? I will NEVER task my system like this outside of these stress tests.


MechGeek said:
At first I thought it was because the board wasn't giving enough voltage somewhere. Except for the spread spectrum thing, that should always be disabled when overclocking.

All my voltages are on Auto now except for core and uncore.
Click to expand...

Is there that much issue with spread spectrum, i thought you were only supposed to disable it if you were having a hard time with your OCs
 
mikecoscia said:
Just ran 20min of crysis without any issues besides a little slowdown, temps maxed out around 85c while playing. Why do I need to back down to 185 BLCK? I will NEVER task my system like this outside of these stress tests.
Click to expand...

lol that's my extra paranoid stability recommendations kicking in :D
 
mikecoscia said:
Just ran 20min of crysis without any issues besides a little slowdown, temps maxed out around 85c while playing. Why do I need to back down to 185 BLCK? I will NEVER task my system like this outside of these stress tests.
Click to expand...

Crysis is not cpu intensive and not even multi threading intensive. Playing that game is an especially bad way to test out your cpu OC stability, if you want to use a game at least use GTA IV.
 
DeadSkull said:
Crysis is not cpu intensive and not even multi threading intensive. Playing that game is an especially bad way to test out your cpu OC stability, if you want to use a game at least use GTA IV.
Click to expand...

I know, but I was playing it while running Prime95 and LINX :).

Well for those following my 20x190 @ 1.35 made it through the night and I am going on 13 hours with LINX and Prime95 running. I think LINX was on its 58 pass and prime was all over the place depending on the core, but no calc errors in either one.

Time to reinstall windows and finally start using the dam machine...lol.
 
mikecoscia said:
Is there that much issue with spread spectrum, i thought you were only supposed to disable it if you were having a hard time with your OCs
Click to expand...

Spread Spectrum is designed to reduce peak EM interference by spreading the computer's stray radio emissions across a larger spectrum.

The problem with spread spectrum being enabled is that every single clock cycle is a tiny bit slower or faster than the previous one. My rationale for disabling spread spectrum is:
- One of the "faster" clock cycles might push the OC too far, crashing at an otherwise stable speed
- Clock jitter makes it difficult for all the clocks' receivers to stay in sync with one another, especially if they have PLLs of their own or have considerable physical distance between the source and receiver. (remember, at 4GHz, electricity in copper only travels 4" per clock cycle) - a clock triggering a bit too early or too late somewhere in the system could flip a bit and cause silent data corruption.
 
mikecoscia said:
I know, but I was playing it while running Prime95 and LINX :).

Well for those following my 20x190 @ 1.35 made it through the night and I am going on 13 hours with LINX and Prime95 running. I think LINX was on its 58 pass and prime was all over the place depending on the core, but no calc errors in either one.

Time to reinstall windows and finally start using the dam machine...lol.
Click to expand...


Good deal. My new uncore voltage settings made the difference too. I ran an all-night blend with the following settings, which seemed to give Prime95 and LinX equal processor time:

LinX: 4 threads, 3GB RAM
Prime95: 4 threads, 8-32768k FFT size, 1600MB RAM

Hottest CPU core was 82c... that's a big step down from my original 97c :D
 
Interesting subject. I normally stress test my overclocks first with 3DMark06 then P95. Everything may fly in one and not the other. Typically if there's a failure in P95 it occurs within the first 20 minutes of running the program. If so I adjust in bios and try again. If it runs for more than 20 minutes then I let it go for a while and call it good. Yes most components do have a burn in period at the facilty that mfs.'s them so let them run 24 hrs. is mute in my book. In the days of old that was always a big thing 24 hr burn....fooey I say to that.
 
Interesting subject. I normally stress test my overclocks first with 3DMark06 then P95. Everything may fly in one and not the other. Typically if there's a failure in P95 it occurs within the first 20 minutes of running the program. If so I adjust in bios and try again. If it runs for more than 20 minutes then I let it go for a while and call it good. Yes most components do have a burn in period at the facilty that mfg.'s them so to let them run 24 hrs. is mute in my book. In the days of old that was always a big thing 24 hr burn....fooey I say to that. As you will never be running at 100% all the time.
 
cpufast2 said:
In the days of old that was always a big thing 24 hr burn....fooey I say to that. As you will never be running at 100% all the time.
Click to expand...

Heh, speak for yourself. I have 5 computers sitting here which run 24/7 with 100% CPU loads with the exception of a power outage or a reboot. My main system is about the only one that ever gets a reboot as well.
 
cpufast2 said:
Typically if there's a failure in P95 it occurs within the first 20 minutes of running the program.
Click to expand...

One of my P95 runs failed after 16 hours. That's when I found this P95 + linpack + 3D thing, and was able to induce the issue within 2 minutes.

I've moved on to the video card ... bought one of those "stock overclock" cards, big mistake, as it doesn't come close to passing OCCT3.0 + self-check for more than a few minutes!
 
I have always tested my machines with multi tortures for hours on end. Its how they live there lives!

Take the machine I'm on right now. Its running a instance of Folding@Home SMP with all four cores being beat on. Then I have it also running Folding@Home GPU on two GTX-260's. Both video cards are hammering two cpu cores to be fed. Now Im online and also encoding a DVD. Later I might shut down one video card thats folding and go play a game. Leaving SMP Folding and one GPU Folding.

Bottom line is. No matter whats going one. 24/7 my machines are running and running at 100% load. Doing multiple things at any given time.

If a rig is not 100% stable 100% of the time 24/7 it's not stable and a worthless build. Hammer the crap out of the boxen. If it can't take it. The trash can it goes.
 
im in the same boat psyshack except i run boinc instead of F@H cpu client.. all 4 systems run full load 24/7 ultimate stress test in my opinion..
 
xorbe said:
One of my P95 runs failed after 16 hours. That's when I found this P95 + linpack + 3D thing, and was able to induce the issue within 2 minutes.

I've moved on to the video card ... bought one of those "stock overclock" cards, big mistake, as it doesn't come close to passing OCCT3.0 + self-check for more than a few minutes!
Click to expand...

Yah same here Xorbe, and I thought my 3.8 was stable at 3.0v after 24hours of prime95. Within minutes of LinX and Prime95, my system crashed. Turned the voltage up to 3.5 and stable as can be, even played a little Crysis while LinX and Prime95 were running. This thread you made me a believer in multiple stress tests, thanks everyone :).

Here's my final results...

3254120418_6c696f76d5_o.jpg


Now need to mess with my stock 4870x2
 
sirmonkey1985 said:
im in the same boat psyshack except i run boinc instead of F@H cpu client.. all 4 systems run full load 24/7 ultimate stress test in my opinion..
Click to expand...

You will never run into a hard core DC'er with unstable or junk machines. They may not be state of the art or at the bleeding edge of a OC. But they will not tolerate a pesky box!
 
psyshack said:
You will never run into a hard core DC'er with unstable or junk machines. They may not be state of the art or at the bleeding edge of a OC. But they will not tolerate a pesky box!
Click to expand...


yup all my systems minus my 6400+ are overclocked to the bleeding edge though id like to take my x2 be-2350 to 3.5ghz but the friggin motherboard wont let me go passed 1.5v and i need 1.55v to get to 3.5ghz :(

mikecoscia said:
Yah same here Xorbe, and I thought my 3.8 was stable at 3.0v after 24hours of prime95. Within minutes of LinX and Prime95, my system crashed. Turned the voltage up to 3.5 and stable as can be, even played a little Crysis while LinX and Prime95 were running. This thread you made me a believer in multiple stress tests, thanks everyone :).
Click to expand...


man 3.5v anyone smell a cookin 45nm processor? :p lol had to get ya for that typo..
 
sirmonkey1985 said:
yup all my systems minus my 6400+ are overclocked to the bleeding edge though id like to take my x2 be-2350 to 3.5ghz but the friggin motherboard wont let me go passed 1.5v and i need 1.55v to get to 3.5ghz :(




man 3.5v anyone smell a cookin 45nm processor? :p lol had to get ya for that typo..
Click to expand...

Lol that's thats the second time I did that, happened in anothe thread too. I really need to proof read my posts...hahah.

1.35 volts not 3.5v :D
 
Well my room was getting to hot so need a lower OC. I was prime95 and Linx stable at 3.2ghz/1.25v. I tried throwing ATI tool into the mix and Linx was getting errors after a few passes. Running everything at stock speeds now with the same mixture of tests for the hell of it. Maybe its a software conflict.
 
Not a software conflict ... haha you discovered how much heat it can really put down!
It's always a good idea to run the tools at stock clocks and voltage first to know it works, and then go from there.
 
xorbe said:
Not a software conflict ... haha you discovered how much heat it can really put down!
It's always a good idea to run the tools at stock clocks and voltage first to know it works, and then go from there.
Click to expand...

Yah just ran for 5 hours and it made it through 15 passes with out any of errors. It's not the heat because it never went over 65c with the 3.2ghz. I think it the mem speed is to high (Corsair 1600 @1604) or there was not enough voltage. Gonna continue messing with it today and see what I find.
 
Try loosening your memory timings to 9-9-9-24? I've heard that memory timings have to be relaxed a bit in 6-DIMM configurations. Is yours a matched set of 6 or two kits of 3?
 
@mikecoscia - I see you have 6 sticks of ram? Try 1333 MHz, you're really loading the electrical bus to the ram modules.
 
OK so anyway.... I ran prime95 and IBT maxing out my memory for over a hour then on a whim added in OCCT GPU test for 20 mins.

No errors, temps stayed safe. SO I am done right? Or is a hour not enough? is there something going that happens at 2 or 3 or 4 hours that is not happening at 1 hour?

Can I just use my machine now? :eek: Geez first AIDS then silent corruption. If computers can get acne I am done. :eek:
 
I think the problem with adding more and more programs onto your stability testing is that every time you add a program and fail, you're "proving" how much using more programs helps. Right?

Not quite. Why stop at 4? Your system might not actually be stable, and that 5th one might just topple you over. Should you really risk it?

I say, stress test your system somewhere beyond where you'll expect it to need to go, but not so far out that you'll only ever experience that when something's gone wrong anyway.
 
No, at some point the machine is basically 100% loaded, and adding more programs would do nothing except cause the existing ones to run slower and slower.

The point is to get it 100% loaded with cpu functional unit, cache, memory and PCIe activity. With 8 threads, that's a challenge.

We're saying IBT and Prime95 do not achieve full utilization on i7 by themselves.

@lens_pirate : if you want to be a complete stability nut like I am, test at frequency F and voltage V. After you find the apparent minimum stable point, run at 0.95*F and 1.0247*V -- that increases power through voltage by 5%, then decreases frequency by 5% to counter. But a lot of people around here seem to scoff at true stability and peace of mind. That means you need to keep your voltage in mind as you do your initial testing so you can add 2.47% later. That means testing at 1.31250v would result in 1.35000v at the end this way.
 
Took my box to work today.

Ran IBT
prime95
and OCCT GPU test

7 hours. It actually heated my office up! No kidding it made the room stuffy. No issues. I am bored with this game. I am going to just start using my PC for fun.

Still I think this was worth doing. I have no lingering doubts what so ever about my O/C. So thanks for suggesting this.
 
I haven't gotten around to OCing the system in my sig yet so maybe I will try the Linx+Prime95 test tonight to see how stable my non-OC'd system is.

Edit: My wife just threw a wrench into this idea by making plans for us to be out tonight.
 
I was able to get a 40 minute run in with both Prime95 in 64bit mode and Linx in 64bit mode without any problems other than 2 of my cores periodically hitting 62C according to Realtemp.
 
Ryland said:
I was able to get a 40 minute run in with both Prime95 in 64bit mode and Linx in 64bit mode without any problems other than 2 of my cores periodically hitting 62 according to Realtemp.
Click to expand...

Wow 62c, what are you using to cool water?
 
You must log in or register to reply here.
Back
Top