AMD OverDrive - Stability Test

I'm new to quad-cores, so I'm looking for the best method of OCing...

How effective is the stability test in AOD?
Which test(s) do you use and how long do you run them?

I got sick of using P95 as it's not really a "somewhat" of a stability test. I like to run short tests when I'm doing coarse-tweaking, and then I'll use p95 for fine-tweaking. I could run P95 for hours on small FFT and then blend will fail immediately, even if my RAM is set to easy (low/loose) settings. I only like P95 when I'm getting close to the end (the wall I stop at).

With single- and dual-core machine, I would use SuperPi 4M for coarse-tweaking, 32M for a bit finer tweaking, and then P95 blend for real stability testing. This combo would get me to my end OC within 2 days usually. Right now I've been messing with my Phenom II for several days... 2 different machines (one mine, one for my bro), and I'm getting sick of using just P95.

I'd use SuperPi, but 4 instances?? I don't want to open up 4 different ones and try to click them all at once... or should I?

Anyway, I figured I'd give AOD a try, but what method (as far as tests and time interval combos) is best? Or... is it just a waste of time to use AOD's stability test?

I'm a patient person, but I will always take the quickest route out if it's within reason.

OCCT or Linpack will be quick and dirty.

EarthDog said:

OCCT or Linpack will be quick and dirty.

Click to expand...

I'll have to download both of those... have heard of OCCT many times, never used it. Maybe heard of linpack once...

Any trick to them? What tests are the quick ones, etc?

THANKS!!

I´m new too with phenom II´s... Overdrive is good to quick raise CPU voltage, multipliers, etc on the fly without rebooting. I can pass 10 rounds of Linkpack but I hit the wall on Blend test. You can play games, run 3dmaks, OCCT´s, etc, etc but, imho blend IS the important test. After all, if yr. PC is most on the unestable field the test won´t last too much

The AMD Overdrive stability test is useful. It only runs for an hour which is an adequate test for casual computer users. In my experience I have found the max temps it produces are 1-2 degrees C. cooler than P95. For some reason overdrive causes my video function to black out at a certain point in the test, but it returns when the test completes. Have not been able to figure this out as I am using a discrete video card, not the onboard video and I don't even think overdrive is testing the graphics system anyway.

moonshake said:

I´m new too with phenom II´s... Overdrive is good to quick raise CPU voltage, multipliers, etc on the fly without rebooting. I can pass 10 rounds of Linkpack but I hit the wall on Blend test. You can play games, run 3dmaks, OCCT´s, etc, etc but, imho blend IS the important test. After all, if yr. PC is most on the unestable field the test won´t last too much

Click to expand...

I will always end off with blend, but I can't afford to wait hours upon hours of blending every single mhz I go up...

for that reason, I like to use something simpler, more coarse...

like my older setups (dual and single cores) i would run superPi and then make my way to p95 when I was fine-tweaking... it's just a matter of being more EFFICIENT in OCing...

i would run superPi and then make my way to p95 when I was fine-tweaking... it's just a matter of being more EFFICIENT in OCing...

Click to expand...

Yeah godman, I agree -and I´m a pretty impacient guy to get the OC job done- but, at least for me, after trying different methods, prime blend seems to be the fastest app that crash my rig. I´m also at the same boat, losing several hours testing, trying new apps like Linx, etc. It also depends in what step of yr. OC you are (CPU, FSB, NB, memory sticks/latencies, etc). It´s always good to share experiences and what works well for you. By the way, if someone recommends the most up to date - quickest way to test stability in a Phenom II, I´ll be grateful.

moonshake said:

Yeah godman, I agree -and I´m a pretty impacient guy to get the OC job done- but, at least for me, after trying different methods, prime blend seems to be the fastest app that crash my rig. I´m also at the same boat, losing several hours testing, trying new apps like Linx, etc. It also depends in what step of yr. OC you are (CPU, FSB, NB, memory sticks/latencies, etc). It´s always good to share experiences and what works well for you. By the way, if someone recommends the most up to date - quickest way to test stability in a Phenom II, I´ll be grateful.

Click to expand...

I'll try to figure out the quickest way... but not sure if I'm going to get the chance now that my system is getting to its peak.

I got my ram stable at 1430mhz 7-7-7-16 running 1T!! NB freq just above 2400mhz, PhII 945 at just under 3.8ghz.

I'm not upping the multi one half notch, but had to drop a couple of mhz on the FSB... priming now @ 3.87ghz...

Intelburn test

First one on the page, uses Linpack, works great.....

10 pass's is supposed to be equal to hours of Prime.....

+1 for IntelBurnTest. I don't even mess around with P95 anymore. Get to what seems stable, let IBT do something wacky like 50 loops overnight. It comes up, I consider myself good to go.

The Wicker Man said:

Intelburn test

First one on the page, uses Linpack, works great.....

10 pass's is supposed to be equal to hours of Prime.....

Click to expand...

JonSimonzi said:

+1 for IntelBurnTest. I don't even mess around with P95 anymore. Get to what seems stable, let IBT do something wacky like 50 loops overnight. It comes up, I consider myself good to go.

Click to expand...

I remember using that a while back... totally overlooked it since it was "INTEL" burn test...

We're getting somewhere, slowly but surely. With everything being quad-core and beyond, I need to change my methods.

The AOD test is fine as an initial stability test, certainly better than SuperPi or the like. Nothing more though. It can generally detect <85% instability. RAM/NB instability is what it will rarely detect which is something Blend is best at on Family 10h CPUs. IBT is another level up from AOD but it depends on the length it's ran for. Generally, good for 97% stability level if ran for 30mins. Only a few errors get past that which P95 will detect but only after hours running without background tasks open.

It's built by design to give a hint at decent stability before you move up when OC'ing so I would certainly use it initially and then move on to IBT once you start to reach within your last 100MHz/temp limits/volt limits.

Using IBT... I'm liking it so far.

How many "times to run" and what "stress level" do you guys normally work with?
I chose "maximum" and left it to run 5 times... in about 5-6 minutes it completed and mentioned that my CPU is stable... so I assume prime95 is necessary to test RAM... or is there something else out there that doesn't take 12-24 hours!!

Personally I like OCCT. I run it for about 20 min. between settings, depending on how close I'm getting to the edge. If OCCT will run an hour you're pretty stable - if it'll run overnight you're good.

And I wouldn't trust P95 blend that far with the Phenom's. More than one of us has had problems with it. I've got a Phenom rig that's been churning out validated SETI work units for six months (some of the new units take ~12 hours to complete) but still won't pass a P95 blend for more than an hour or so. Call my rig unstable if you like but I'll trust 6 months of SETI over P95 any day ...

i only use OCCT, i run it untill it hits full load and the temps even out, let it go for a little bit and see and if nothing happens well things are good. i like it too because it'll out put graphs of all sorts of info so if you do have a problem, be it temp, voltage or what ever you have something you could check up on.

the biggest way i test stability is to OC it, crack open a game and see, if it hard locks or crashes well...you gotta re boot any way so getting into the bios to tweak is no big deal.
basicly real word tests, do your most intensive thing you'd be doing normally and see what it does

QuietIce said:

Personally I like OCCT. I run it for about 20 min. between settings, depending on how close I'm getting to the edge. If OCCT will run an hour you're pretty stable - if it'll run overnight you're good.

And I wouldn't trust P95 blend that far with the Phenom's. More than one of us has had problems with it. I've got a Phenom rig that's been churning out validated SETI work units for six months (some of the new units take ~12 hours to complete) but still won't pass a P95 blend for more than an hour or so. Call my rig unstable if you like but I'll trust 6 months of SETI over P95 any day ...

Click to expand...

Yeah dude it looks like prime95 is not the thing to use with phenoms... I'm going to try out OCCT like you said. Intelburntest has worked well on one system, but my other one crashed

Niku-Sama said:

i only use OCCT, i run it untill it hits full load and the temps even out, let it go for a little bit and see and if nothing happens well things are good. i like it too because it'll out put graphs of all sorts of info so if you do have a problem, be it temp, voltage or what ever you have something you could check up on.

the biggest way i test stability is to OC it, crack open a game and see, if it hard locks or crashes well...you gotta re boot any way so getting into the bios to tweak is no big deal.
basicly real word tests, do your most intensive thing you'd be doing normally and see what it does

Click to expand...

Left4Dead... haven't played it in so long... this will give me a reason to!

g0dM@n said:

Using IBT... I'm liking it so far.

How many "times to run" and what "stress level" do you guys normally work with?
I chose "maximum" and left it to run 5 times... in about 5-6 minutes it completed and mentioned that my CPU is stable... so I assume prime95 is necessary to test RAM... or is there something else out there that doesn't take 12-24 hours!!

Click to expand...

IBT is far better at testing ram. That is what the stress level setting changes is the amount of ram being used. If you set maximum, it was using all of the free memory available for testing.

I have had my system 24 hour prime blend stable and yet IBT would fail within 15 passes. I personally don't even bother with prime anymore, I run 100 passes of IBT at maximum, takes ~6 hours.

Considering that prime95 has been around for well over a decade and its intended purpose was to find Mersenne prime numbers, the algorithms used have remained largely unchanged over the years I believe that as CPU power increases prime becomes less affective as a stability test.

ViperX800 said:

IBT is far better at testing ram. That is what the stress level setting changes is the amount of ram being used. If you set maximum, it was using all of the free memory available for testing.

I have had my system 24 hour prime blend stable and yet IBT would fail within 15 passes. I personally don't even bother with prime anymore, I run 100 passes of IBT at maximum, takes ~6 hours.

Considering that prime95 has been around for well over a decade and its intended purpose was to find Mersenne prime numbers, the algorithms used have remained largely unchanged over the years I believe that as CPU power increases prime becomes less affective as a stability test.

Click to expand...

Ya... I'm probably going to stop using prime altogether, unless it's on an older system.

I was able to run 100 passes of IBT on my brother's rig, but within 2 minutes of OCCT his system rebooted, lol.

g0dM@n said:

I'll have to download both of those... have heard of OCCT many times, never used it. Maybe heard of linpack once...

Any trick to them? What tests are the quick ones, etc?

THANKS!!

Click to expand...

No reason to download both OCCT runs linpack if you set it up to do so.

I like OCCT, for QND stability as well.


I do not trust p95 anymore, does not seem to play well with AMD imho (maybe I need to update)

I have tried the AOD stability test but ould not put any faith in it, I have never seen it fail. Crash computer yes Fail no...


As far as p95 goes, I do not think it tests everything well, I used to setup blend tests to run over night and load up a couple of videos to wathc on the same machine, no stuttering no skipping....

Neur0mancer said:

No reason to download both OCCT runs linpack if you set it up to do so.

I like OCCT, for QND stability as well.


I do not trust p95 anymore, does not seem to play well with AMD imho (maybe I need to update)

I have tried the AOD stability test but ould not put any faith in it, I have never seen it fail. Crash computer yes Fail no...


As far as p95 goes, I do not think it tests everything well, I used to setup blend tests to run over night and load up a couple of videos to wathc on the same machine, no stuttering no skipping....

Click to expand...

What's QND stability?

Also, if OCCT and IBT are that similar, how come OCCT crashed my system but IBT didn't (same settings)? I ran IBT with "maximum" on the RAM - 100 loops passed, and OCCT on Auto (rebooted system within like 2 min).