Anyone care to explain this superpi result (I appear to have beaten 8packs WR)

[Frozennova]

Been benching a little SuperPi 32M tonight. For reference 3570k 49x101, Mushkin 1600MHZ CAS6 @ 2222mhz 9-10-8-24-30-1T, went to make a cuppa while it ran and came back to the following time. System had locked by the time I got back so I could only take a photo. The checksum validates @ http://www.techpowerup.com/superpi/

Spoiler

 

[Avenged7Fold]

Looks to me like it carried out only 7 loops then it locked up. I am not familiar with superPI but it probably glitched and will validate your 7th loop time rather than the full test.

 

[Frozennova]

Yea pc had hardlocked when I came back to it. Just awfully strange I've never seen it do that before.

 

[mmj_uk]

This is the kind of thing that happens when system are not fully stable, software goes haywire and inexplicable things happen.

 

[Frozennova]

I knew it was likely to be stability, the main reason I find it strange is due to the checksum being valid. By rights if I could have got the screenshot with the relevant cpuz information it'd be a valid on hwbot.

 

[Frozennova]

I knew it was likely to be stability, the main reason I find it strange is due to the checksum being valid. By rights if I could have got the screenshot with the relevant cpuz information it'd be a valid on hwbot.

 

[PCZ]

I knew it was likely to be stability, the main reason I find it strange is due to the checksum being valid. By rights if I could have got the screenshot with the relevant cpuz information it'd be a valid on hwbot.

Even if the automated systems accepted it a mod would remove it straight away.
Why don't you try submitting it.

No don't you would most likely be banned.

 

[mmj_uk]

The checksum is calculated after the tests have run based on the results, it's just to stop people from photoshopping different test times in afterwards.

If the times had been more realistic you might have got away with submitting it, what are the odds of it glitching and that happening though?

 

[Frozennova]

Even if the automated systems accepted it a mod would remove it straight away.
Why don't you try submitting it.

No don't you would most likely be banned.

I had no intention on submitting it as I know that it wouldn't end well. Guessing nobody else has ever seen this happen then

 

[PCZ]

Yes we have seen it happen.
Not just with SuperPi but other benches as well.

All sorts of crazy can happen when you try to bench on extremely unstable systems.
If you scroll through the original PD thread you will find where nkata and myself were trying to get 100 Gflops on IBT.
Something we were successful at doing but every result was a non number (invalid) and yet IBT declared a successful pass
100.png

Not just when benching.
I was running LLR tests on KN pairs to test for prime numbers.
Normally you would expect one pair in around ten thousand to be prime.
The PC I was using a BD 8150 had a brain fart and decided to declare every test it had done that day to be prime (100's of pairs)
That was a major embarrassment for me as the results are emailed out to other members of the project.
Not to mention having to sort through the logs and redo the work.
I decided then that 100% duty cycle running FPU intensive software was not something BD was ideally suited for

 

[Frozennova]

Interesting.

Glad to see its not just me seen things like this.

I got my XP install working properly in the end 6m46 or so. Still need to tune secondaries and tertiaries but the current time puts me 5th. I think i will be able to push the cpu frequency with better cooling as I'm needing to use silly volts. 1.5v bios around 1.48 cpuz load.