Normal XP 2600 Speed?

[ajbrun]

I can usually, but rarely I get a blue screen. Usually when playing UT: 2004, so when stressing the system I guess.

I ran it again just now and got 0.5 expected 0.4
Ran for 7 minutes.

[javalord]

Don't rely on prime95 as good as it is on testing stability its not always stable on some systems

Myself I can run countless loops of sandra sys burn in, 3D Mark, Commanche 4 bench, F@H and video encoding at 2.8GHz but prime dies after 10-15mins with errors. Now its been proven by people such as marci at O-cUk that primne is not the be all and end all of system stability.

Basically if your system is stable for everything you do ie gaming etc but can not complete prime then don't worry. Real world testings far better

Prime95 not stable = system is not 100% stable. Real world tests dont show fine discrepencies which will bite you in the ass one day when you need it most.

Video encoding can have small block errors in the encoding results, you just wont notice it. 3D Mark? The GPU does most the work. And as for F@H, you'll be sending bad results back, which you dont even use yourself.

The moment you do something intensive such as hours of gaming, you'll wonder why your game drops out.

[ajbrun]

Yeah - I thought it was important, or at least useful. Anyway, I want to get no errors to make it 100% stable. Any more ideas how?

[Applecrusher]

push the voltage up a tad??

[ajbrun]

OK, I can change the voltage, but what to? It's currently at "H/W". I'm not sure what that means, but I can change it too other values too. Would 1.60 be about right, or should I go higher?

[Firelord]

Prime95 not stable = system is not 100% stable. Real world tests dont show fine discrepencies which will bite you in the ass one day when you need it most.

Video encoding can have small block errors in the encoding results, you just wont notice it. 3D Mark? The GPU does most the work. And as for F@H, you'll be sending bad results back, which you dont even use yourself.

The moment you do something intensive such as hours of gaming, you'll wonder why your game drops out.

Totally disagree. You can't reley on a single program to say whether your system is stable or not, you wouldn't run one test in a lab and say ohh thats fine you run multiple different tests to make sure that something works or is stable.

I have had prime stable rigs that crashed in 3D mode or restarted when video encoding - would you class them as stable?

See threads here -http://www.over-clock.com/ivb/index.php?showtopic=7682&hl=

At the end of the day if your rig will play games hours on end and do what you need it to do but its not stable in a paticular non real world program what does it matter?

My old A64 setup (newqy 3000+) would run 3D'01SE for 3 days on end looping on a realtime stress under task manager and that does test everything from GPU to CPU and mem but would fail prime after half hour. I could then encode video for 8 hrs but prime would still fail. Does this mean my rig is unstable?

NO!

[Firelord]

Click me

This is a quote from Marci at Over-clock and he is extremely knowledgable and properly one of the best clockers in the uk. What he don't know isn't worth knowing...

Severely doubt passive cooling is feasible really... even on that CPU. If the bios doesn't let you decrease then you can't do it basically... s'a limitation of your motherboard

Now, to get it stable at 2.4Ghz, just up the voltage one notch. Get into windows and check stability. If it ain't stable, back to bios and up voltage another notch. Into windows, check stability. Rinse n' repeat!

Whilst doing all this, keep your eyes on the CPU Temp reported by Motherboard Monitor 5 (http://mbm.livewiredev.com) and ensure that the CPU does not go above 50 deg C. If it starts heading towards the 50s then stop and settle with the fastest you can get out of it below 50 deg C.

Personally.... don't use Prime95 as a stability test. What do you DO with your PC?? Benchmarking, Gaming, Office work? None of these involve calculating Prime. If you're primarily a gamer, then your stability test should be the game the most graphically intensive game you own that you tend to play the most, or for the longest periods of time. THAT is the true stability test for your system. As long as your rig can do what you ask it to do on a regular basis, it is stable for your purposes.

Prime95 stability is just for willywavers and purists. My rig fails Prime everytime, but is stable for everything else I ask it to do... 3DMark, F@H, DF, Photoshop, Cubase, Battlefield Vietnam... so what difference should it make to me that it bombs prime...? None at all.

Other than that.... upgrade your heatsink to a Thermalright SP9x series with a decent fan on top to get the CPU temps down a bit to allow a higher overclock.

And to really show prime is not all thats needed....

Nope. F@H won't submit incorrect results. It detects the corruption in the result and terminates the work unit. However.... why is it my rig is F@H stable but NOT Prime stable? Obviously mathematically the CPU is functioning correctly to be able to run F@H without corruption of the work unit... as well as everything else I do that relies on maths (muxing, demuxing, 64 channel layered mixdowns in realtime etc etc)

Hence... balls to Prime.

prime is not the bee all and end all

[Rave]

Oh yeah? How exactly is F@H going to know the WU is corrupt? Unless the processor has such a serious stability error that it actually crashes the WU (this was actually quite common with the Gromacs core), there's no way that F@H can know that it has generated an erroneos result. The only way it could know that there was an error would be to compare it to the results generated for the same WU by another computer. This is what the author of Prime95 has to say on the subject (cribbed from ocforums):

Originally posted by Kendan
F@H especially with the Gromacs core will give you a test of stability (you will lock up or bluescreen if you are not stable) and a load temp.

No, no, no! As author of prime95 please allow me to comment on some misconceptions in this thread.

1) Prime95 is a brutal test of your memory, reading and writing well over a hundred megabytes every second. The latest version can be told how much memory to use.

2) Prime95 does not test SSE, but does test SSE2.

3) Do not confuse lock up and bluescreen with stability. Aggressive overclocking causes instructions to very occasionally return the wrong result occasionally (e.g. 1+1=3). If this is in a critical part of the OS you will crash. More commonly, this occurs in a non-critical part of the program you are running, causing incorrect results to be returned to F@H. You will probably see torture test errors in prime95 long before you see crashing stability problems.

4) Running several stability test programs at the same time is only useful if they all run at the same priority. Running several different programs is a very good idea but you are probably better off running them one after the other. Prime95 is a great test for memory and CPU (especially FPU) and a lousy stress test for your video system and peripherals.

5) Heat is not the most common cause of errors. Users that contact me more often suffer from over aggressive memory settings, too little voltage, and weak power supplies.

F@H, Seti and others are all worthy projects. You are only helping them if your computer is returning accurate results. It is critically important to run prime95 (for just an hour or two) as well a couple of other stability test programs before committing your computer to these projects full-time. Oh, and please you the prime95 Torture Test rather than starting and abandoning a work-unit.

I have no problem with your recruiting people to run F@H full-time as opposed to prime95 or seti or whatever. There is plenty of idle cpu cycles to go around.

Here's google's cache, since the site itself appears to be down

http://64.233.183.104/search?q=cache...+results&hl=en

Rich :¬)

[Firelord]

Ok then if a PC is stable at prime but craps out at 3D, does then that mean that the rig is stable? no it doesn't.

You can't reley on prime to prove a stable system.

ok then I build a pc, its a standard pc using default BIOS settings for CPU and memory but its not prime stable at that but can loop memtest, 3d mark, sandra, video encoding, f@h and user interaction. Does that mean my stock machine is unstable? nope it doesn't, it means that maybe prime doesn't like the rig itself.

My point is is that if all this guy does is game or surf the web and it does that with no crashes, restarts or errors then why should he have to worry if his rig can pass 12hrs of prime?

[Rave]

I'm not really arguing with that particularly. It really depends whether your computer does stuff where accuracy is critical or not. For gaming, as long as the computer doesn't actually crash, then you're fine. For word processing etc., any instability is likely to be obvious. For video encoding, you'll probably be fine albeit that you might get the odd glitch here and there.

For stuff like Finite Element Analysis, and medical DC like F@H, UD or FAD, accuracy is critical, and you shouldn't be running stuff like that unless your computer is 100% stable. Since (by the author's own admission) Prime95 doesn't test SSE it's not a great test for Athlon XPs; I'd recommend BurnK7 as an additional test.

It is possible that there are systems that Prime95 doesn't like, e.g.systems running at stock that should definately be fine. I'd try reinstalling Prime95 in the first instance. At the end of the day though, if Prime95 is generating an error, then something is wrong, even if by rights it shouldn't be. Who's to say that other programs like F@H aren't going to suffer from similar errors?

Rich :¬)