• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

SuperPI or Prime95?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

martinjon666

Member
Joined
Oct 10, 2005
Location
Idaho
New to superPI, i have been using prime95. Which is better, more accurate? What I do like about SuperPI is that it takes A LOT of time less but i am worried that it may be less accurate.
 
It's easy to pass any length of SuperPI calculations with an unstable system...not that I'm a huge fan of Prime either. Usually though, if you can get through Prime, you can get through nearly anything.
 
what would you suggest other than prime, i kinda get sick of using the same program day in and day out, maybe there is something better and faster.
 
martinjon666 said:
what would you suggest other than prime, i kinda get sick of using the same program day in and day out, maybe there is something better and faster.

Folding@Home (see the forum section) is a great way to test for stability, and I've been told it will find stability problems a heck of a lot faster than p95, so check that out.
 
For me Pi has been a nice tool to find memory errors in windows. running the 32meg calc usualy error cus of memory timmings at high OC. ALso long runs of Pi aren't great for finding CPU stability because Pi stops CPU use when it needs to page to the HDD, resulting in much less CPU stress. Prime has always been better IMHo for plucking out my CPU instabilities as its relentless in its CPU stress. never stopping, and always checking what its calculating.

A couple others that find instabilities that arent always apparent are PCmark04 , and 3dmark 03 Sometimes, and this is subjective. they just find crap the others miss, and error out.
 
Apparently Pcmark04 is very good for finding errors. 3dmark05 is also popular for this.

I personally like CPU burn in with error checking enabled, although the funnest way to find errors is load up a game and play until for about 4hours if that passes I am pretty sure nothing else will make your system crash.
 
speed bump said:
Apparently Pcmark04 is very good for finding errors. 3dmark05 is also popular for this.

I personally like CPU burn in with error checking enabled, although the funnest way to find errors is load up a game and play until for about 4hours if that passes I am pretty sure nothing else will make your system crash.

PCMark04 is great, I was 15 hours Prime95 stable, and PCMark04 crashed after 30 seconds.
 
If i am testing for stability on my entire system, i usually will run 32m PI, all 3 3dmarks, am3, and pcmock04. If i can pass those w/o a problem i know i am for the most part safe and save myself the 20+ hours you need to know if your prime stable :)

As a bencher normally i will just run the 3dmarks a few times, and if they pass i know i am atleast gaming/benching stable which is all i care about :p
 
cool, thanks for all the input, it does seem like superPI is not meant as a stand alone stress test, but i will definitely keep it around for more choices when stressing.
 
I've fallen in love with OCCT. Seems to be just as stressful as Prime95 but you get all the info you need in 30 minutes. SuperPI is good also but not as stressful as Prime95 or OCCT.
 
jcw122 said:
Folding@Home (see the forum section) is a great way to test for stability, and I've been told it will find stability problems a heck of a lot faster than p95, so check that out.

I always thought folding is by far the worst to test stability, because if you are unstable you send back bad data which isnt a good thing, correct me if im wrong?
 
Folding can be used for FINAL stability testing IFF you take precautions. Of course, if you are a folder like me, you will be folding 724 forever anyway :)

But before folding I run:
- pcmark04 (make sure you get a score! else it failed). This doesn't seem to work on the newer conroes/mobos so i have started using pcmark05. This requires at least some minimal level of video capability so i use something like an x600 to do my OC with, then later put in el cheapo 2mb pci video for just folding rigs. PCMark05 takes quite a bit longer to run than PCMark04 unfortunately.

- AND at least the first 2 fft test sizes in P95 torture test blended. (about 10 iterations?) You could run more p95 or Orthos if you don't fold 724 on your rig like I do. OC settings cannot be too near the max or my rigs will later produce errors if the ambient temps fluctuate, etc. Orthos is an easy way to run dual prime95 on the dual cores. Most of my rigs have good mem, and I have never had a bad stick of ddr/ddr2 memory so I don't use memtest or lengthy p95 tests unless i am struggling with an OC or suspect mobo issues.

I used to think dual p95's produce less heat than dual folding but since i have started using orthos on the conroes, I had noticed that is definitely not the case. Gautam also pointed this out in a conroe thread in the folding forum. Orthos on DC's will produce higher cpu temps. So testing with it for a few hours is a good idea especially if you are on air, since if it passes it gives you a bit of buffer on your temps in case (when) ambients go up later on, or your hsf/rad gets dust/dirt clogged.

Then folding is the FINAL stability test .... NEVER the FIRST test.

- after a WU downloads, stop the folding instance and backup the downloaded WU. To backup the WU copy the 'WORK' subdir AND QUEUE.DAT file to a backup directory (i have backup subdirs in all my folding directories).
- then disable the lan - i prefer to disable it via network properties and leave the window open to remind myself of its state. You can also unplug the lan cable if it is easier for you ... i have too many rigs and cables so i prefer the disable approach.
- then start the folding instance.

If it EUE's:
- remember what % it was at (or look in the fahlog.txt)
- stop the folding instance
- delete the work subdir and queue.dat from the folding directory.
- copy the backed up WORK subdir AND QUEUE.dat file (these files are the WU) from the BACKUP directory into the folding directory.
- fix your OC - my usual approach is to lower FSB on my intels by 5. You might want to raise VC a bit but since my rigs fold 724 i find lowering fsb is usually better since it gets better temps and is more likely to be long term stable.
If you are tweaking memory you may need to increase divider/reduce multiplier, loosen timings or give it more vddr, or all of the above :D

Some WU's will EUE even on stabile hardware ... it is very infrequent for non-beta testing, less frequent on advmethods, and very infrequent on public/normal WU's. If you restart the wu form the backup and it EUE's again at the same % completion, then it is not likely due to your OC. Stanford re-assigns EUE'ed wus to several other computers to check the result ... if they eue at the same point on multiple computers, then it is a valid result. So you can do the same on your own rig before you let the WU be submitted back to stanford.

If it continues to eue at the same % completion, enable the lan so it can submit the results (since it is a 'valid' EUE), and let it download a new wu. Then go back to the beginning of the process (ie. stop the folding instance, backup the wu, disable the lan .....)

I prefer the 5.02 or 5.04 text consoles running from the desktop, not as a service, while i am doing final stability testing since it makes it easier to see what is going on, and stoping the instance is easy but make sure you use ctrl-c to stop it so you don't corrupt the wu. If it gets corrupted it will start at the beginning of the wu (ie starts at 0%).

If you normally run folding as a service, just disable the service and stop it, create a shortcut to the console.exe in the folding directory with whatever flags you run with eg -local -forceasm -verbosity 9 etc.

You should NOT use the advmethods flag during stability testing.

The advmethods class of wu's are not usually good candidates for final stability testing since they have a higher probability of EUE'ing than the non advmethods wu's.

After a WU or 2 have been successfully completed, your cpu, mem are stabile.

This does not stress your video so that needs games, 3d testers etc. I am a noob when it comes to video :eek:

If you have any questions, pm me or post in the folding forum,
 
Last edited:
Back