Prime95 Problem

[RJARRRPCGP]

going down to ddr2 400 (pc3200) allows prime95 to pass most tests with the vcore at 1.350.

At DDR 1 speeds! O_O Your motherboard may have cheap caps.

 

[lilraul]

sorry meant pc2 3200!

 

[RJARRRPCGP]

sorry meant pc2 3200!

Same clock and slower timings than DDR 1 lol.

 

[QuietIce]

Hi thanks for the reply, I set all the settings to what you suggested but couldn't lower the vDIMM as the lowest I can go is 1.95v in my bios. I don't have the BE version. The ram timing is default to what you suggested.

Prime95 failed on test2 on both cores with the max temp at 52. I've attached my SPD tab details.

1.95 vDIMM is OK, too bad the other settings aren't available.

Even though those timings were default you still set them manually when you dropped the RAM to 333 MHz, right?

52 is a little high for stock speeds even with the vCore turned up a little. What heatsink are you using on the CPU and how clean is it?

 

[lilraul]

1.95 vDIMM is OK, too bad the other settings aren't available.

Even though those timings were default you still set them manually when you dropped the RAM to 333 MHz, right?

52 is a little high for stock speeds even with the vCore turned up a little. What heatsink are you using on the CPU and how clean is it?

Hi, yeah the only options for the memory voltage is 1.95,2.00,2.05 and 2.1, the rest of the settings you advised I'm able to change. The ram timings I manually change when I lower the speed. The command rate I change to 2T as default it sets to is 1T.

The heatsink is the stock one and I've cleaned it, the case has 2 extra fans in there.

 

[lilraul]

btw did you want me to monitor coretemp while prime95 was running or after running prime95? because I ran it while running prime95.

 

[QuietIce]

You have to run them together to get load core temp, which is always what we're looking for. Idle temps mean next to nothing for OC'ing ...

 

[lilraul]

That's what I did, not sure what more I could do...

 

[QuietIce]

You might try removing the case side and pointing a room fan (on low) into the case. If temps go down you may have a case airflow problem.

Still, I'm surprised you got a P95 fail with those settings. Have you tried using OCCT instead? I've got one rig that just won't pass P95 regardless of what I do, but it's been churning out verified SETI work units (number crunching) for almost two years without problems. Give OCCT a one hour test to see how it fairs. If OCCT also fails then you've most likely got a hardware problem.

 

[lilraul]

hmmm...OCCT passed with a 1 hr test, and the temp according to OCCT didn't go above 40.

This is at 400mhz, i'll try 667 next.

 

[lilraul]

if all goes well and the test passes at 667, when I do overclock the cpu again, max I want is 3.0 ghz and the ram I don't want to overclock so I'm happy with 667, does the "Chipset Over-Voltage" and "HT Over-Voltage" need changing? at the moment both are set on 'Start-up'.

btw I set the vcore to 1.275 in bios as 1.350 seemed quite high for stock speed when I've looked at other peoples vcore on and so far OCCT is running without a problem at 667 for ram speed but in OCCT it says the vcore is 1.38, is that right?

Thanks for all the help so far.

 

[QuietIce]

I wouldn't set a max limit on the OC as much as I would temp or vCore. Many people overclock as high as stock voltage will let them, which is 1.35v for the K8's. With load core temps that low I wouldn't worry too much about stressing your CPU unless the vCore is pushed above 1.45v. I've run my Opty 165 at 1.45 vCore 24/7 for four years now, you're not going to hurt the chip at 1.35-1.40v as long as your core temps are good.

Few people need to change chipset or HT Link voltage.

OCCT is probably reporting the correct vCore. OCCT makes several graphs, one of which is vCore, and it might be that the vCore is being "corrected" as load is applied. You can check that by opening the graph (should be in the OCCT directory). Expect there to be some minor fluctuation - that's typical.

 

[RJARRRPCGP]

Still, I'm surprised you got a P95 fail with those settings. Have you tried using OCCT instead? I've got one rig that just won't pass P95 regardless of what I do

In my experience, Prime95 failing means you're gonna get crashes if you ignore Prime95.
(facepalms at people accusing Prime95 of being badly written) (It's a bad motherboard, processor or PSU!)

Prime95 giving you an error has NEVER been found to be a software problem!

And OCCT, being ran for 1 hour didn't detect a problem with my E2180 failing Prime95 at 5 hours, IIRC at 3.0 Ghz.

Please try again with LinX. And run LinX for at least 3 hours.

 

[QuietIce]

In my experience, Prime95 failing means you're gonna get crashes if you ignore Prime95.
(facepalms at people accusing Prime95 of being badly written) (It's a bad motherboard, processor or PSU!)

I've never said P95 is badly written and in all honesty I don't know why one of my machines has a problem with it. It won't pass a P95 test for an hour, yet it's been churning out verified work units (24/7 number crunching) for over a year-and-a-half. You can say whatever you want about P95 being the end-all of stability testing but I'll see your 24 hour P95 test and raise you about 10,000 verified work units over a period of 20 months. Go ask the other Crunchers and Folders which is the better test - I already know what their answer will be.


BTW - You can run OCCT as long as you like, just like P95. After fritzing around with P95 for three days I decided to test with OCCT and it went through 19 hours before I switched to Crunching. Until then I was as adamant as you about P95 testing but three wasted days changed my mind.

 

[lilraul]

A update on this issue, seems like something is wrong hardware wise.

This is testing with OCCT

cpu freq 200mhz 2.6 clock speed
400mhz ddr2, vcore at 1.275 - PASS

cpu freq 200mhz 2.6 clock speed
667mhz ddr2 - FAIL

cpu freq 200mhz 2.6 clock speed
667mhz ddr2 increased vcore to 1.4 - FAIL

cpu freq 200mhz 2.6 clock speed
667mhz ddr2 kept vcore at 1.4 and increased vDIMM to 2.00 - FAIL

cpu freq 200mhz 2.6 clock speed
533mhz ddr2 vcore at 1.375, lowered vDIMM back to 1.95 - FAIL

cpu freq 200mhz 2.6 clock speed
533mhz ddr2, changed HT Link to 5x - FAIL

cpu freq 200mhz 2.6 clock speed
533mhz ddr2, Increased "Chipset Over-Voltage" and "HT Over-Voltage and decreased HT Link back to 4x - FAIL

cpu freq 230mhz 3.0 clock speed
400mhz ddr2, vcore at 1.375 - FAIL
Ram overclocks to 460 ddr2

cpu freq 220mhz 2.8 clock speed
400mhz ddr2, vcore at 1.375 - PASS
Ram overclocks to 440 ddr2

all the FAILS happen within 3 mins of OCCT running.

It seems like if the ram speed is anything above 440, OCCT will fail. It seems like a memory problem but memest86 passes it but I would want to check by trying other sticks, don't want to buy them tho just incase the sticks are not at fault so need to borrow them from someone..

 

[RJARRRPCGP]

I've never said P95 is badly written and in all honesty I don't know why one of my machines has a problem with it. It won't pass a P95 test for an hour, yet it's been churning out verified work units (24/7 number crunching) for over a year-and-a-half. You can say whatever you want about P95 being the end-all of stability testing but I'll see your 24 hour P95 test and raise you about 10,000 verified work units over a period of 20 months. Go ask the other Crunchers and Folders which is the better test - I already know what their answer will be.


BTW - You can run OCCT as long as you like, just like P95. After fritzing around with P95 for three days I decided to test with OCCT and it went through 19 hours before I switched to Crunching. Until then I was as adamant as you about P95 testing but three wasted days changed my mind.

That's why you should maybe try LinX or IntelBurnTest.

 

[QuietIce]

A update on this issue, seems like something is wrong hardware wise.

This is testing with OCCT

cpu freq 200mhz 2.6 clock speed
400mhz ddr2, vcore at 1.275 - PASS

[trim]

all the FAILS happen within 3 mins of OCCT running.

It seems like if the ram speed is anything above 440, OCCT will fail. It seems like a memory problem but memest86 passes it but I would want to check by trying other sticks, don't want to buy them tho just incase the sticks are not at fault so need to borrow them from someone..

Good idea. As it is it could be the RAM, board, or CPU - no way to really know, though it is odd that several hours of MemTest86+ doesn't show any errors.

That's why you should maybe try LinX or IntelBurnTest.

The more recent versions of OCCT have a LinX option, if that's what you need to sustain your illusions of stability.

 

[lilraul]

Speaking of LinX, on CPU:OCCT the test passed on 220 cpu freq and 400 mhz ram speed but failed on the linpack, what does that mean?

 

[AlabamaCajun]

Reading over this I suspect instability in the northbridge. On some MOBOs it was possible to tweak the settings for it but on others it had no effect are was not available. I think it's 1.25V on that chip which going up a notch or two might help.

Another option is to run the RAM at 5-6-5-22 just to give the controller a little more head room.

 

[lilraul]

would that be the Chipset Over-Voltage setting?