torture test completed - 1 error(s)

[Crawlerz246]

I admin I tried going a little a little overboard (in my particular situation)... tried clocking from 205 to 210... loaded into windows, got a BSOD.

 

VCORE was clicked one notch up to 1.275 (I really don't want to increase a lot more as I will hit 70-75 degrees under load and that seems to be a stable limit), VTT to 1.23, IOH was 1.18 (then I clicked it to auto cuz hell if I know what it was set to when it was auto, IDK where to begin.)

 

clicked the whole thing back down to see a stable machine: 205 BCLK, 1.27 (I think) VCORE, 1.23 VTT, IOH to auto.

 

I come to windows, start realtemp, prime95. wait for heat up. ......... working. all workers show THREAD COMPLETED. one (3rd) shows COMPLETED - 1 ERROR. the top load (acording to realtemp is 78% as opposed to always 100%, the next 30% as opposed to 80%)... temps are normal (tops out at 66-67). no BSODs. what happened? and is it time to panic?

 

TY

[El_Capitan]

I assume you're talking about the i7 920?

 

VERY HIGH RISK!

Your max temps can go up to 100C before reaching it's threshold. To be safe, I'd make sure it never reaches 90C. Not many instances where your CPU will reach full load except for torture tests. You won't be able to overclock higher without increasing the VCore (or you're limited by your CPU and/or motherboard).

 

Increase your VCore without raising your BCLK. See the maximum temps you're satisfied with (if not, then go watercooling). Once you reach your max temp -5C (say 90C is your limit, you want 85C to be in this case), increase your BCLK by increments of 2 instead of 5. When you also increase your BCLK, your temps will also raise a bit (which is why the max temp -5C).

Edited March 8, 2010 by El_Capitan

[RJR]

I assume you're talking about the i7 920? Your max temps can go up to 100C before reaching it's threshold. To be safe, I'd make sure it never reaches 90C.

 

OMG, please don't !!!!!!!! Just because the Tjmax of a chip is 100c does NOT mean it's OK to run the chip anywhere near that temperature.

 

Please download Intel's data sheet and see what the recommended temperatures are:

 

http://developer.intel.com/design/corei7/documentation.htm

[El_Capitan]

OMG, please don't !!!!!!!! Just because the Tjmax of a chip is 100c does NOT mean it's OK to run the chip anywhere near that temperature.

 

Please download Intel's data sheet and see what the recommended temperatures are:

 

http://developer.intel.com/design/corei7/documentation.htm

It depends on what he's going for. If he wants to overclock it to see how high he "can" overclock it, then he's got temp room to spare before CPU throttling. <-- VERY HIGH RISK!

 

If he's going for the highest stable overclock, the risk is higher at say, 90C, but depending on what he's doing, the likelihood that he'll even reach 80% full load is not likely, thus the average temps never reaching ~70C is what he should be aiming for.

 

Obviously he has a stable overclock now at safe full load temps. I'm assuming he's overclocking it to overclock it.

 

I get an idle between 39 and 43 and a max of 72 on my hottest core after around an hour of Prime 95. Under normal use or gaming my CPU has never touched 60. Highest overclock on my i7 920 was 4.3 on air with full load temps reaching 82C, but idle at ~55C, and gaming temps reaching 70C. I could stay at that my highest stable overclock, but I'm fine with my Voltages and temps at 4.0Ghz.

Edited March 8, 2010 by El_Capitan

[hardnrg]

I'm not sure what you are doing in Prime95, but you should be running Options > Torture Test... and either Blend or Small FFT... it just goes on forever until you stop it...

 

and you should let it run for at least a few hours before increasing the overclock, if you get errors it's time to take the speed(s) down, or try different voltages to gain stability (eliminate errors)

 

right now I'm still working on my overclock and I'm getting pretty near to my final OC I think, unless the voltages I tune in end up being the golden key to stability at even higher speeds

 

I'm not recommending these voltages at all, especially since I'm watercooling the CPU, have a 40mm x 20mm (?) scythe ultra mini kaze (?) on the mobo heatsink, and a lot of case cooling... it's merely an indication of voltages that can help to maintain stability at high speeds:

 

CPU: 1.35v

DRAM: 1.5v

QPI: 1.35v

PLL: 1.8v

ICH: 1.1v

ICH PCIE: 1.5v

IOH: 1.28v

IOH PCIE: 1.5v

 

CPU: 21x198=4158

Uncore: 3366 MHz

QPI Link: 3564 MHz

DRAM: 594 MHz (DDR1188)

 

update: 1.28v IOH is passing P95 Blend @ 7h32m, whereas 1.24v IOH froze the OS at 3h4m, 1.6v caused Task Manager to stop working, and 1.12v caused BSOD or Freeze in less than 2 hours of running P95, so you can see that IOH >= 1.28v is essential for my stability at this speed

(hint: it's extremely useful to make a spreadsheet on another computer to record the settings used, and the outcome of all tests)

[RJR]

It depends on what he's going for. If he wants to overclock it to see how high he "can" overclock it, then he's got temp room to spare before CPU throttling.

 

If the OP is asking what the problem is with an error in Prime95, this is most likely his first OC. Telling a first time OCer that he can go until it throttles (Tjmax)or just below is just irresponsible to say the least. If you want to run your computer at up to 90c or even Tjmax, that's great knock yourself out, nobody is stopping you. Stating anything beyond Intel's recommended max thermal specification is putting his computer at risk and since he is a beginner it shouldn't be done.

[El_Capitan]

If the OP is asking what the problem is with an error in Prime95, this is most likely his first OC. Telling a first time OCer that he can go until it throttles (Tjmax)or just below is just irresponsible to say the least. If you want to run your computer at up to 90c or even Tjmax, that's great knock yourself out, nobody is stopping you. Stating anything beyond Intel's recommended max thermal specification is putting his computer at risk and since he is a beginner it shouldn't be done.

True, but there's a first time for everything.

 

I should have stated the risk factor, I didn't see the beginning of his sig that it was his first build.

[El_Capitan]

@Crawlerz246 - Run the test again and see if the error is repeated. If it still occurs, isolate the issue by seeing if it's your CPU or your Memory. Go to Options -> Torture Test... -> set to Small FFTs

 

Run to see if the error is still occurring. If it is, it's probably your CPU.

 

If that's the case, bump up the VCore up one notch or your BCLK down one and repeat the test.

 

If the error no longer occurs from running Small FFTs, try adjusting your memory speed and timings (loosen them first) and repeat the original test again (Blend test).

[Crawlerz246]

stop guys stop! I know about temps and I've made it my mission not to cross 70 under full load! nowhwere NEAR 100!

[Crawlerz246]

today I went to realtemp.exe and pressed TEST SENSOR to launch P95. it revved up OK to 100%. topped out at a bearable 66 degress (as before).

 

IN p95's window and I quote...

 

[Mar 8 09:14] Worker starting

[Mar 8 09:14] Setting affinity to run worker on logical CPU #2

[Mar 8 09:14] Beginning a continuous self-test to check your computer.

[Mar 8 09:14] Please read stress.txt. Choose Test/Stop to end this test.

[Mar 8 09:14] Test 1, 4000 Lucas-Lehmer iterations of M19922945 using FFT length 1024K.

[Mar 8 09:15] Test 2, 4000 Lucas-Lehmer iterations of M19922943 using FFT length 1024K.

[Mar 8 09:16] Test 3, 4000 Lucas-Lehmer iterations of M19374367 using FFT length 1024K.

[Mar 8 09:17] Test 4, 4000 Lucas-Lehmer iterations of M19174369 using FFT length 1024K.

[Mar 8 09:18] FATAL ERROR: Rounding was 0.5, expected less than 0.4

[Mar 8 09:18] Hardware failure detected, consult stress.txt file.

[Mar 8 09:18] Torture Test completed 3 tests in 3 minutes - 1 errors, 0 warnings.

[Mar 8 09:18] Worker stopped.

 

p.s. how does one UN lock the RAM multiplier?

Edited March 8, 2010 by Crawlerz246

[Crawlerz246]

@Crawlerz246 - Run the test again and see if the error is repeated. If it still occurs, isolate the issue by seeing if it's your CPU or your Memory. Go to Options -> Torture Test... -> set to Small FFTs

 

Run to see if the error is still occurring. If it is, it's probably your CPU.

 

If that's the case, bump up the VCore up one notch or your BCLK down one and repeat the test.

 

If the error no longer occurs from running Small FFTs, try adjusting your memory speed and timings (loosen them first) and repeat the original test again (Blend test).

 

ran P95 blend test for nearly three hours (not starting from realtemp)... no errors

Edited March 8, 2010 by Crawlerz246

[spitter]

ran P95 blend test for over an hour so far (not starting from realtemp)... no errors

 

 

So the sensor test finds an error in less than 30mins and you run p95 to find out if it's for real?? The sensor test is very good at finding a stable machine.