Pass Prime but not Super PI?

[Daddyjaxx]

I'm not giving up on this 4400. How can a CPU fail SuperPi in three minutes and run Prime for 3 hours? I know 3 hours is nothing, but I upped the CPU 22MHz from my stable settings and SuperPi fails in the 3rd iteration. I'm going to let this thing run, but if the dang thing passes Prime, is that more important than SuperPi?

[lowboy]

You are running dual prime?

[Guest GripS]

Sounds backwards. I can pass a 32m Super Pi and fail 2 instances of prime95 within 5 minutes at the same settings. You got me on that one.:confused:

[sailindawg]

How are you running SuperPi ? Are you running 2 instances from 2 seperate folders and setting the affinity for each instance ? What type of failure notification are you getting from SuperPi ?

[Daddyjaxx]

Aaagggh. It doesn't matter. I woke up this morning to a PC that was on, but I had a black screen. It was still running at 5 a.m and died sometime after that. That means it ran Prime for over eight hours before it went kaput. Well....back down to 2706 for me....

[sailindawg]

That means it ran Prime for over eight hours before it went kaput. Well....back down to 2706 for me....

 

Don't I know that feeling. My 4400 is 8 hour P95 stable, 3D, SuperPi, etc....at 11x245. Tried 10x280, one core passes 11+ hours, the second core failed after 6.5. The pc was SuperPi 32M stable, 3D stable and game stable. I could rip a video with no errors. Go figure. No problem booting either. These 4400's are tough to overclock and need to be slowly massaged to get anything out of them. I took my TTBT off and reapplied AS5. Now, at 11x245, temps seemed more balanced under load when watching temps with XS Core Temp program. I have been seriously considering water cooling with this 4400. But you know the crazy thing, I had my SD at 9x324, sheer speed, nice. However, even at 11x245, I am appreciating this 4400.

 

The other day I was downloading the Company of Heroes demo, downloading from Steam and surfing the net with no lag and the simultaneous downloads never slowed each other down or slowed down a bit. With my SD, I could only download one item at a time, otherwise the downloads would lag. I thought I was bandwidth limited by Comcast. Nope. I was limited by cpu power. These dual cores are interesting compared to single cores. Now if I can only get 10x280 8 hours P95 stable on both cores with my current air cooling. At the least, I can load off CMOS Reloaded and game away at 10x280. Now if only EVGA would ship my Step-up, I can get back to playing.

[Kingfisher]

Yeah, it's frustrating to see but the botom line is your PC has to pass all tests to be deemed stable.

Sure you can run your PC any way you see fit.

 

But ,

There is nothing like playing a game or working on a project that you invested your precious time in and find a blue screen or frozen screen pop up on you. :confused:

 

Get it stable with good temperatures and enjoy your PC, if it can't clock up to what others with the same hardware have, big deal.

[FTCold]

In answer to your original question though:

 

How can a CPU fail SuperPi in three minutes and run Prime for 3 hours?

 

We dig into the actual processor itself. While this may not be the real reason in your case, it is a real reason none the less. When we run these benchmarks we have no idea (unless GNU or open source) of what the source code is. It's really not that big of deal anyway unless you want to know how much of your processor you are really testing. Take Prime95 for example. I haven't checked it's source (if it even has free source?) but I'm willing to wager that it only tests your general purpose registers. Also note that since we are all using socket 754/939/940 and are using 32-bit windows or linux half of the register isn't being checked. So long story short, there are different instructions that use different parts of the chips and different special or general purpose registers that are tested for each different benchmark. The benchmarks here such as prime95 and super pi do a great job of testing the most used parts of your cpu, but who knows, you might actually be unstable using SSE2 instructions :confused:

[sailindawg]

the botom line is your PC has to pass all tests to be deemed stable.

 

I agree wholeheartedly !

[h_Us3r]

In answer to your original question though:

 

 

 

We dig into the actual processor itself. While this may not be the real reason in your case, it is a real reason none the less. When we run these benchmarks we have no idea (unless GNU or open source) of what the source code is. It's really not that big of deal anyway unless you want to know how much of your processor you are really testing. Take Prime95 for example. I haven't checked it's source (if it even has free source?) but I'm willing to wager that it only tests your general purpose registers. Also note that since we are all using socket 754/939/940 and are using 32-bit windows or linux half of the register isn't being checked. So long story short, there are different instructions that use different parts of the chips and different special or general purpose registers that are tested for each different benchmark. The benchmarks here such as prime95 and super pi do a great job of testing the most used parts of your cpu, but who knows, you might actually be unstable using SSE2 instructions :confused:

 

This is a very solid perspective.