- Joined
- May 30, 2004
Recently, for me Prime95 errored at 5 hours when I was 1 hour OCCT stable and 25-pass Linpack stable.
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
AMD OverDrive - Stability Test
- Thread starter g0dM@n
- Start date
Recently, for me Prime95 errored at 5 hours when I was 1 hour OCCT stable and 25-pass Linpack stable.
- Joined
- Jun 30, 2009
OCCT has its own tests but also comes with linpack, what IBT uses.
There is also LinX, which is linpack but is supposed to have more options
- Joined
- Oct 11, 2005
- Location
- Tau'ri
QnD = Quick n Dirty
As for why tests failed on OCCT... and not IBT, I do not know. Would double check you settings, also I do not know what code each relies on, perhpas OCCT is doing more than IBT is doing...
- Joined
- Apr 9, 2009
- Location
- Toronto, Canada
Here's my beef with OC testing and stability:
My definition of stable is: I can run ANYTHING on the computer and it will not crash on me. This INCLUDES Prime95 blend test. If you fail Prime95, then you aren't stable in my books.
In my opinion, Prime95 blend test is the ONLY test which tests the CPU, CPU-NB, and Ram extensively. OCCT is a joke with the medium test - You need to run small, medium, and large test lengths to fully test an AMD system; hence, you should just run Prime95 blend. Linx just creates alot of heat and doesn't isolate the CPU-NB. AMD Overdrive stability testing is rubbish.
I used to run Linx at 90% memory for 20 passes to "quickly" test stability. However, I've given up on it now. Instead, run Prime95 "custom" test, where you keep everything default but you change "Time for each test = 1 minute". This will quickly cycle through all the fft lengths. Observe which one you fail at.
If you fail at fft length x<100k, you're looking at CPU problem. Increase your Vcore.
If you fail at fft length 101k<x<1024k, it is the cpu-nb.
If you fail at fft length x>1024k, it is a ram problem.
If you pass the "1 minute" test, then start the Prime95 blend test and go to sleep.
My definition of stable is: I can run ANYTHING on the computer and it will not crash on me. This INCLUDES Prime95 blend test. If you fail Prime95, then you aren't stable in my books.
In my opinion, Prime95 blend test is the ONLY test which tests the CPU, CPU-NB, and Ram extensively. OCCT is a joke with the medium test - You need to run small, medium, and large test lengths to fully test an AMD system; hence, you should just run Prime95 blend. Linx just creates alot of heat and doesn't isolate the CPU-NB. AMD Overdrive stability testing is rubbish.
I used to run Linx at 90% memory for 20 passes to "quickly" test stability. However, I've given up on it now. Instead, run Prime95 "custom" test, where you keep everything default but you change "Time for each test = 1 minute". This will quickly cycle through all the fft lengths. Observe which one you fail at.
If you fail at fft length x<100k, you're looking at CPU problem. Increase your Vcore.
If you fail at fft length 101k<x<1024k, it is the cpu-nb.
If you fail at fft length x>1024k, it is a ram problem.
If you pass the "1 minute" test, then start the Prime95 blend test and go to sleep.
- Joined
- Jun 30, 2009
I agree with this statement 100%, but Prime95 isn't the be all end all of stability testing. It may pass while others fail.
QuietIce
Disabled
- Joined
- May 7, 2006
- Location
- Anywhere but there
Then you'd better start buying server hardware.
Consumer systems are going to crash eventually - it's all a matter of how long it takes ...
- Joined
- Apr 9, 2009
- Location
- Toronto, Canada
This is a 100% true statement. Sadly from pesonal experience, Prime95 > Linx/OCCT/IBT/Linpack, whatever... Pass Prime95 and you've got a very good chance of passing everything else. It is always good to check with the other programs as well, however.
If you test for stability and insist that the problem is due to Prime95, then your system should crash at stock settings. Why does it only crash when you OC? The problem lies in your OC, not Prime95.
QuietIce
Disabled
- Joined
- May 7, 2006
- Location
- Anywhere but there
Well, let me put it this way. If I can run SETI work units 24/7 for a month (that's a 720 hour 4-core load test - a total of 2880 core-hours) then it's stable enough for me. If I run that SETI "test" then run P95 blend and it crashes in less than an hour I could care less ... 
- Joined
- Jun 30, 2009
OCCT test is supposed to be similar to prime, relatively small data sets crunched over and over. Linpack is Intel Burn Test. Intel came up with linpack to torture their CPUs, IBT is a front end to make using linpack easier. The version of linpack in OCCT is older than the one in IBT2.3. Some linpack builds are not compatible with AMD CPUs.
IMO nothing can test the RAM/NB like linpack.
OP
- Joined
- Sep 27, 2003
- Thread Starter
- #31
How do I know which builds are or aren't compatible with AMD... will it let me know?
I'm still trying to decide between the OCCT or Linpack tests... thanks for the info.
Right now trying to run custom (6 hours) of linpack (with OCCT v3.1.0 interface) max test mode. I don't see why I wouldn't choose "max" for test mode...
This is a great discussion guys. I've been out of the loop for a while but will check out the programs mentioned when I get off work.
I have always just ran 3dMark and then sisoft sandra performance test for intermediate overclock stability. Once I come to the end I usually run Prime95 for a few hours and call it good.
Recently though, I'm stuck. I can game for 6 hours straight with no problems but fail at the arithmetic test of sandra. AMD Overdrive stability test for 1 hour (all tests) was fine. So dunno whats going on there, but thats another thread. For me...if I can game 6 hours without any hiccups I could care less how "stable" my system is in synthetic testing.
I have always just ran 3dMark and then sisoft sandra performance test for intermediate overclock stability. Once I come to the end I usually run Prime95 for a few hours and call it good.
Recently though, I'm stuck. I can game for 6 hours straight with no problems but fail at the arithmetic test of sandra. AMD Overdrive stability test for 1 hour (all tests) was fine. So dunno whats going on there, but thats another thread. For me...if I can game 6 hours without any hiccups I could care less how "stable" my system is in synthetic testing.
OP
- Joined
- Sep 27, 2003
- Thread Starter
- #33
I agree... 6 hours gaming?! I'd care less, too... did your butt hurt from sitting that long?!
I haven't gamed straight that long in a few yrs, but I do miss being able to do that.Between last night and today I was able to take advantage of something: OCCT large data set test. I used it to test my new RAM timings. I used AOD to adjust the ram voltage. I wanted the system a bit unstable so that I can TIME how long each vdimm would make the test last. SURPRISINGLY I actually found a sweet spot that didn't fail @ all!! My ram is rated for 1.90v and has an EVP of 1.95v, so I tried all of these voltages with OCCT large data set: 1.70, 1.72, 1.74, 1.76, 1.80, 1.82, 1.86, 1.96, 1.90
Some would fail within a few minutes, and some lasted a while, but although I thought 1.70-1.72 was gonna be the sweet spot I was wrong!!
1.74vdimm just passed 2 hours of large data set on OCCT... I don't know why I didn't go with Linpack, but I guess it's because OCCT gave me the option of what data set I wanted to use... Anyway, I was able to complete 2 hours custom on large data OCCT with these settings:
709mhz memory (1418mhz ddr) 7-6-6-16-27 1T
Man am I excited (knock on wood it lasts). I dropped my CPU multi but kept the vcore high enough... 266x14 = 3724mhz
Now I'm going to go for 14.5x and see how she rolls...
^^good info.
My ram is rated at 1.9v too (Corsair XMS3 1600). I run it at 1640mhz...
Are you trying to figure out with your testing the effects of lower latency vs frequency?
I was playing AoC, the online MMO...can get lost for hours in that game. Nice to play it at full settings though. Its really got great environment graphics.
My ram is rated at 1.9v too (Corsair XMS3 1600). I run it at 1640mhz...
Are you trying to figure out with your testing the effects of lower latency vs frequency?
I was playing AoC, the online MMO...can get lost for hours in that game. Nice to play it at full settings though
. Its really got great environment graphics.
- Joined
- Oct 11, 2005
- Location
- Tau'ri
OCCT is awesome
I find my final test to be PSU test for ultimate stability. Combination of CPU test and GPU test for maximum wattage draw.
I find my final test to be PSU test for ultimate stability. Combination of CPU test and GPU test for maximum wattage draw.
OP
- Joined
- Sep 27, 2003
- Thread Starter
- #36
^^good info.
My ram is rated at 1.9v too (Corsair XMS3 1600). I run it at 1640mhz...
Are you trying to figure out with your testing the effects of lower latency vs frequency?
I was playing AoC, the online MMO...can get lost for hours in that game. Nice to play it at full settings though
I'm not testing latency vs frequency... I'm testing what voltage works best for my RAM/motherboard combo.
To get higher RAM speed, I'd need to increase my FSB more... remember, I'm on a 945, not a 955 BE, so I have to raise FSB to overclock my CPU.
I think 1420mhz 7-6-6-16 at 1T is probably better than 1600mhz 8-8-8-24 2T like the RAM is rated. I'm surprised I got 7-6-6 to run... thought I was going to be stuck on 7-7-7. Cas 6 was a no go, though... I'm happy with CAS 7.
OCCT is awesome
I find my final test to be PSU test for ultimate stability. Combination of CPU test and GPU test for maximum wattage draw.
It only lets me choose between OCCT or Linpack... why's that?
- Joined
- Jun 30, 2009
Not sure. I know that all of the newer versions of IBT are AMD compatible and the build included with OCCT is.
g0dM@n said:
Same way on my system, it bitched about the DX version the first time I ran it. I figured it was because I'm on Vista x64.
Slightly off topic, but I was wondering if you could do me a favor and let me know what the hardware monitor in the bios reports the CPU voltage being at on your board? Mine reports it as somewhat higher than what I set it for, curious if yours is the same.
OP
- Joined
- Sep 27, 2003
- Thread Starter
- #38
ASUS Probe II reports voltage higher... I'll have to double-check BIOS some other time.
QuietIce
Disabled
- Joined
- May 7, 2006
- Location
- Anywhere but there
Probe II is usually right on. I've been using it for years as a secondary source for core temps (usually ±1 from CoreTemp) and for vCore (usually matches CPU-Z). Of course, Probe II takes a little longer than CoreTemp to reach the stabilized temp when going to/from idle/load but in the end it's very close ...
OP
- Joined
- Sep 27, 2003
- Thread Starter
- #40
You just reminded me... do any of you remember with DDR that you could measure the ACTUAL voltage by using a DMM on pin #7 on the ddr slot? Well, I was wondering if you could do that with DDR3...
Similar threads
