![]() |
![]() |
#1 |
Jun 2004
22×3×5 Posts |
![]()
I'm running Prime95 (24.14) on a Celeron D 3G machine (256M mem).
I'm pretty sure that P95 is the only program running on that machine, no other hidden programs stealing CPU cycles But, very strange to me, the time per iteration is not a constant within the period of one exponent. After I boot up the system, fresh start up Prime95, the iteration is often 0.100 sec for a 37M level exponent which is slower than it should be. I have to pause the program, and run some other softwares such as IE surfing internet or/and WMplayer watching a video clip, then resume Prime95, if I'm lucky, the iteration time can drop to 0.084 which is the expected performance for Celeron D 3G, but sometime I have to repeat this several times to get the speed. The difference is HUGE, 20-30% in speed! My other machine (P4s) have similar problems, but the difference is small, about 1-2%. So I don't worry too much about them. Any thought? Last fiddled with by nngs on 2007-04-13 at 21:53 |
![]() |
![]() |
![]() |
#2 |
"6800 descendent"
Feb 2005
Colorado
2×359 Posts |
![]()
I've noticed video cards can steal a lot of CPU cycles. That could explain why you have to run a graphics intensive program in order to get the video card into the right "mode". You might try updating the video card driver, or if it is a Windows machine, try starting Windows in safe mode.
|
![]() |
![]() |
![]() |
#3 | |
Jun 2004
22×3×5 Posts |
![]() Quote:
to the lowest one if using integrated video card. But my video card is a regular AGP card, and I lowered the resolution to 800X600 (the lowest I could get), but no help. I ran under Safe mode too, which gave a iteration time at ~ 0.095 sec, and my tricks to run other programs to get better iteration time did not work well under the safe mode. I thought this could be due to the position of the Prime95 code/data in the system memery or in L2 cache? some positions give best performance? |
|
![]() |
![]() |
![]() |
#4 |
"6800 descendent"
Feb 2005
Colorado
2×359 Posts |
![]()
If you have an old PCI VGA card, I would try it. You might also try telling Windows to shut the monitor off after a couple of minutes (power save mode, not screen saver mode) to see if that affects the iteration times.
I like to make sure peripherals like USB ports, serial ports, and hard drives are set so that Windows does not turn them off to save power. Also any settings that can slow down the CPU to save power should be disabled. I am still suspecting the video card and/or driver more so than the position in system memory or cache. I have seen indications that memory position can affect performance in Linux, but I have never seen it in Windows. Of course, YMMV.. |
![]() |
![]() |
![]() |
#5 |
"Jason Goatcher"
Mar 2005
DB316 Posts |
![]()
If I may ask an impertinent question. I've never owned a power saving cpu that I know of. Assuming one is the type to buy computers just to crunch, is it possible that the powersaving feature, while slowing down a computer, may save enough energy that buying and crunching on another box with the savings makes using the powersaving feature best-bang-for-buck? I'm not saying it's true, I'm just wondering if anyone has considered this possibility.
|
![]() |
![]() |
![]() |
Thread Tools | |
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
strange thing happening in factordb | firejuggler | FactorDB | 96 | 2019-12-06 15:32 |
ecm thing | 3.14159 | Miscellaneous Math | 3 | 2016-12-16 23:58 |
my thing | firejuggler | Aliquot Sequences | 1 | 2010-05-31 06:57 |
Hm... strange thing... | Yxine | Factoring | 1 | 2006-08-10 13:48 |
Prime95/LLR on dualcore CPU - strange thing? | Cruelty | Software | 1 | 2006-07-27 04:53 |