AMD3700+ 2.2 ghz@2.9 ghz OC Prime95 fails after 10h what would you do

Xzar

Distinguished
Jul 20, 2006
54
0
18,630
San Diego Core
Asus A8N5X
this fan:
http://www.tigerdirect.ca/applications/searchtools/item-details.asp?Sku=S457-1088

Antec Neo Power 430 HE (I heard seasonic built it) with 384 watts on the 12v rail

I changed the voltage from auto to manual and set it at 1.4v in the BIOS, and it peaks in cpu-z at 1.46v Oddly enough the auto mode for voltage control of the BIOS sometimes has my cpu-z voltage at 1.47 if I recall correctly

I was about to stop the stress test and call the OC a success then a few minutes after 10 hours passed the error happened.

Temps are under full load of Prime95 with open side-case (I don't mind leaving my case open when I use the PC)
60-62 degrees for the core
50 degrees for the next highest temp of speedfan

264 mhz FSB
RAM is at 2.5-3-3-7 2.5 GB (3 sticks total) of Kingston RAM DDR400, it says 2.6v on the sticker on the RAM, voltage for RAM is on auto
4x multiplier on the HT Link for 1056 mhz (max for the HT Link was 1225 mhz on a previous test)

I removed 512 MB of Xerox RAM rated at ddr333 to allow for a higher FSB use, it used to crash in memtest very early at 261 mhz bus.

Would you guy say that the system is stable enough for gaming (which isn't nearly as stressful for the PC as Prime95) or would you either:

-Up the voltage to 1.425 volt in the BIOS and try again? The problem with this is that the BIOS says 1.425 but cpu-z can sometimes show it at 1.52, which I over the 1.5v another thread in this forum said was the limit for socket 939 cpu.

-I only ran the memtest86+ for 30 mins, test 5 and test 8 one pass, should I run the memtest86 for like 4 hours? How long should you run Memtest anyway?

-Or give up and drop down to a lower OC? Yesterday, at 2.809 ghz I ran Prime95 for 8h30 hours successfully (at the lower 1.375 v bios voltage), however, who knows if it would have failed too after 10 hours.

The duration of the stress-test on Prime95 was a tough choice to make, what went on in my mind was 10 hours of test, then 1 more test, but that last test was the one to fail.

If I only aimed for 9 hours for example the OC would have looked like a complete success.

I was planning on buying a new PC in November anyway, so I don't mind trying new stuff and OCing even if it breaks, cuz OCing is fun. But if I can get my OC stable then I'll delay buying the new PC until Starcraft 2 comes out.
 
Solution
Stability is a subjective call. It's really up to you.

However, I use several tests to determine stability.

Prime95, 4-8 hours of Small FTT
Prime95, 8-10 hours of Blend
LinX, 20-30 passes
Memtest 86+, 3 -5 passes
3D Mark Vantage, 3 loop(all tests)

flyin15sec

Distinguished
Apr 16, 2008
985
0
19,010
Stability is a subjective call. It's really up to you.

However, I use several tests to determine stability.

Prime95, 4-8 hours of Small FTT
Prime95, 8-10 hours of Blend
LinX, 20-30 passes
Memtest 86+, 3 -5 passes
3D Mark Vantage, 3 loop(all tests)
 
Solution
Do you need it on for 10hrs at a time or more? If the answer is yes, then yeah you should lower it a bit or find a way to increase ventilation in your case. If the answer is no then I fail to see the issue. Prime 95 is a stress test, so it is likely your PC will be stable under most conditions for much longer. If you turn it on to do stuff and play games, and then turn it off after less than 10hrs, then it's stable for the duration you use it and I would be trying to overclock it even more.
 

Xzar

Distinguished
Jul 20, 2006
54
0
18,630


I don't know where to view/modify the chipset voltage for my mobo (A8N5X). A brief research seems to indicate that Asus mobos of my generation cannot change their chipset voltage without extensive mods.

Earlier tonight I ran Memtest for 1 full pass, and OCCT Large Data Set CPU test which focuses on RAM for 4h15 without errors, so I guess my RAM can work fine at the current stock RAM voltage and 264 mhz bus. Maybe the problem was my CPU.

Posted by flyin15sec


Stability is a subjective call. It's really up to you.

However, I use several tests to determine stability.

Prime95, 4-8 hours of Small FTT
Prime95, 8-10 hours of Blend
LinX, 20-30 passes
Memtest 86+, 3 -5 passes
3D Mark Vantage, 3 loop(all tests)

That's a nice battery of tests to confirm stability. It reminds me that I should also test my GPU too to see if it will function correctly with the new FSB that is 32% faster. Plus I'm curious about my 3DMark score, never done one of those before.

Since my PC failed a 10h Prime95 test, I'll try to run another 12h one tonight now that I upped the BIOS cpu voltage from 1.4v to 1.425v. Since the voltage bump I did a 2h15 small Data Set OCCT which did not detect any errors.

If my next 12h Prime95 run is stable, I'll know for sure that my system is stable at 1.425v. After that I'll re-drop voltage to 1.4v and run small data set OCCT for 1 hour, to see if it can detect the problem that Prime95 needed 10h to detect. There are a lot of rumors that OCCT can detect problems faster than Prime95, and since I'm a first time OCer I'm curious about that.

Posted by megamanx00
Do you need it on for 10hrs at a time or more? If the answer is yes, then yeah you should lower it a bit or find a way to increase ventilation in your case. If the answer is no then I fail to see the issue. Prime 95 is a stress test, so it is likely your PC will be stable under most conditions for much longer. If you turn it on to do stuff and play games, and then turn it off after less than 10hrs, then it's stable for the duration you use it and I would be trying to overclock it even more.

Yeah I might need it on for 10 hours, but yeah I agree that nothing I do will be as intensive as the Prime95 stress test. This is my first OC and I want to get stable results to know that I OCed it completety right lol

Since I upped the voltage, temps went way up, but still within acceptable maxs

According to:
http://www.amdcompare.com/us-en/desktop/details.aspx?opn=ADA3700DKA5CF

max temp for San Diego Core 3700+ is 54 degrees to 71 degrees
I wonder why the max temp is a range of numbers instead of 1 absolute number

At 1.375v Prime95 max temp was 58 degrees, most of the test it was 56
1.4v max was 62 degrees, most of the time of the Prime95 test it was 60
1.425v max was 68 briefly, usually at 66, so this voltage which I am currently using is pretty much the max I can go

Now, even though my BIOS voltage is 1.425v, the OCCT voltage display (which borrows the display of CPU-Z) shows the vcore at around 1.502 volt usually and occasionally at 1.52v. Is this too much voltage? Here is the graph of the Vcore variation during my OCCT 2h15 small data set cpu test.