SuperPi 32M - rought estiminate of time need?

[trodas]

After twaking a bit with my Duron 750 (now x5 multi), I set it slowest as it could go in bios and let it run SuperPi 32M overnight.

 

Yep, booting to WinXP took about 15min alone, but I was patient. Now in the morning, I saw this:

 

 

So it took 1h 11min 44.375sec for the Duron to come up with just the "initial value."

 

What will be the total time of the test?

 

Could I roughly estiminate, that when doing on same board SuperPi 1M test, then it took with AXP 47.765 sec ( http://hwbot.org/submission/2462459_ ), while 0.625 sec took the initial value...

 

So that is 76.424x longer for the vhole result, witch could push the time to 5502.528 minutes = 91 hours = 3.8 days?

 

Could that be somewhat roughly accurate?

[TerraRaptor]

First 2 loops are a good point to estimate. Set maxmem low to make run longer:)

[trodas]

Thanks for the tip. I run some super-low Aquamark scores and then set it up on another bench table, where it can sit for 4+ days, lol.

 

Maxmem low? How to set it? Where? I never thought about it. I was just trying to take a look, what scores will HWbot accept as these super-low ones (Special Achievements, Submit 5 SuperPI 32M runs which took over 12 hours to finish.) and I had a good way to produce that result, based on very special downclock of Duron.

Yes, one can just use original Pentiums... but when I stumbled upon a bug, that produce about 30MHz FSB on one MSI board:

http://valid.canardpc.com/evdcpe

...then I was like - that is the interesting way to do it

 

Sadly I never managed to run the x5 Duron yet into the mainboard, witch is kinda sad. But I will keep pushing and trying to figure out what went wrong... And even 30x7.5 was too fast and Duron made these SuperPi 32M in around 8 hours or so... There it is, 8h 32min: http://hwbot.org/submission/2435812_ So w/o working multiplier x5 I cannot manage that low scores, even having very very low FSB 30MHz

[TerraRaptor]

Add a switch to boot.ini /maxmem=200 mb - that will make superpi run slower as it won't be able to allocate 250mb of RAM.

[trodas]

Well, I settled for simple yet effective disabling of L1/L2 caches. The result is, that it seems to be stuck at loop 6 for like 3 days or so... making me mad. I wonder, if that happend for someone before and if there is a chance that it will resume operation any time soon... The aim was "longer that 12h", not longer that 12 days!