New SP2004 version coming soon!

[LuckMan212]

I just got word from Johnny Lee, author of SP2004 that a new version is coming soon which will incorporate the latest prime95 code as well as fix a few things such as remembering the affinity settings when testing dual cores using 2x SP2004s. Not sure if he will be renaming it to SP2006 or not. :D

If you all have any other bugs or suggestions, now might be a good time to post here or write in, so he might work on them while he's tinkering with the code........ :)

edit:
new beta is out...
I don't think he got to a bunch of things still, right off the bat I notice FSB /HTT/multi is still reporting wrong...
but anyway.......

http://www.dblsys.ca/sp2004/beta/beta.htm

please test and report bugs to Johnny!!!
his preferred email is: typo.pl AT gmail.com

[Jaco]

FSB and multiplier are always wrong , if he could fix that, that would be great.

Look at the picture. I'm running my CPU 260x10 , but SP2004 is showing 194x13,5 (wish i had a 13 or 14 multi , LOL)

[vapb400]

I dont know how to describe it, because I dont have it on this comp, but when i minimized, i rarely could get the program back up. The options in the systray did nothing

[LuckMan212]

ok I have directed JL to this thread, so hopefully he will check in on it and take a look at some of these issues. Not sure if he's a member here (but he should be!) :p:

[K.I.T.T.]

Theres also a 6 second bug (only withlarge FFTs) encountered by me and some other members with dual core opterons. Sometimes, the first core started will randomly fail at 6 seconds, and the 2nd core started will keep on going. It doesnt depend which order, eg if you start core 0 1st and core 1 second, core 0 will fail in 6 seconds. If you start it the other way round, core 1 will fail in 6 seconds. I've run it for 8 hours straight, error free in SP2004, but this error keeps on randomly popping up. i the latest version of "original" prime, this error doesn't exist. it would be nice if he could fix it :)

[LuckMan212]

hmm! interesting! I think I may have been "hit" by that bug myself and not even known it (probably wrote it off as hardware failure). Good to know, hope it can be fixed.

[geoff2k]

Is he going to release the source for the next version? I appreciate that he licenses it under CC, but its hard to make derivative works without the source. ;-)

[soundx98]

love to see two versions with the latest Prime95 versions
32bit with P95v2414 and x64 with P64v2414

[high5]

any updates?

[freecableguy]

needs to be SMP aware so that one instance handles all cores.... :rolleyes:

[LuckMan212]

edit--moved to top post

[Clint]

Theres also a 6 second bug (only withlarge FFTs) encountered by me and some other members with dual core opterons. Sometimes, the first core started will randomly fail at 6 seconds, and the 2nd core started will keep on going. It doesnt depend which order, eg if you start core 0 1st and core 1 second, core 0 will fail in 6 seconds. If you start it the other way round, core 1 will fail in 6 seconds. I've run it for 8 hours straight, error free in SP2004, but this error keeps on randomly popping up. i the latest version of "original" prime, this error doesn't exist. it would be nice if he could fix it :)

Aha!
I knew it wasnt my CPU that failed me...thats nice to hear!:)

[K.I.T.T.]

Aha!
I knew it wasnt my CPU that failed me...thats nice to hear!:)
:toast:

testing new version now :fact:

[K.I.T.T.]

So far it seems ok :)

No 6 second bug encounterd yet. One possible feature would be to enable to record the CPU speed dynamically, rather than through windows. ie. if you boot at 290 x 10 and use clockgen to go to 300 x 10 and THEN start to prime, it will still show 2900 MHz. The FSB/multi thing needs sorting out as well. Using something like the code used in CPU-Z and adding an option that can tell the viewer whether the CPU speed had been changed during priming (cheating) would make it one hell of a program :)

Not saying theres anything wrong with it. Just an idea i had :)

[guess2098]

just wonder, large is harder to run, if CPU is not stable and large FFT will fail very quick, but why small FFT call CPU test? i can easy to run small FFT with 1.3v FX60 @ 3G but i know that is totally not stable at all

[dogsx2]

just wonder, large is harder to run, if CPU is not stable and large FFT will fail very quick, but why small FFT call CPU test? i can easy to run small FFT with 1.3v FX60 @ 3G but i know that is totally not stable at all

I've always thought that large is the harder of the two.

[tzitzos]

thx for the beta version!!

will test it later ;)

[[XC] moddolicous]

I've always thought that large is the harder of the two.
Yea, thats why he says the the large should be the CPU test. Another thing, can you make a linux version? I'm sure there are some people that would like a linux version to test their OC's.

[masterofpuppets]

Yea, thats why he says the the large should be the CPU test. Another thing, can you make a linux version? I'm sure there are some people that would like a linux version to test their OC's.
I would also like to see a Linux version.. there is a Linux equivalent of Prime95 (mprime), and a simple GTK or Qt frontend to that wouldn't be hard at all.

[marius7]

LuckMan212 , can u ask Johnny Lee to keep command line switches from Prime95 ? For example -t is like starts tourture test. For the rest of them see Prime95 -> Help -> Advanced Usage -> Undoc.txt.

[LuckMan212]

LuckMan212 , can u ask Johnny Lee to keep command line switches from Prime95 ? For example -t is like starts tourture test. For the rest of them see Prime95 -> Help -> Advanced Usage -> Undoc.txt.
hey guys I am not connected with Johnny I just emailed him initially to ask about new versions of SP2004 and he happened to write back and seemed to start working on it. If you check the top post, you will see I have given his direct email so please send him your bugs & requests because I have already nagged him enough I think :rolleyes: Besides, the more people he hears from the better I think since It may encourage him to finish the next version! :)

[[XC] moddolicous]

I'm surprised he doesnt have an account here.

[LuckMan212]

I'm surprised he doesnt have an account here.

me too... I would guess 70-80% of the downloads of SP2004 are from XS users.

[krampak]

but when i minimized, i rarely could get the program back up. The options in the systray did nothing
That happens to me too ! Normally after some hours of testing (when I wake up) and the program is still running at 100% CPU.

(that's on SP2004 0.3).

[eva2000]

just wonder, large is harder to run, if CPU is not stable and large FFT will fail very quick, but why small FFT call CPU test? i can easy to run small FFT with 1.3v FX60 @ 3G but i know that is totally not stable at all
cause small ffts test cpu only while large and blend test ram too...maybe mem controller too ? maybe for you the mem controller on cpu fails with large ffts ???

i always find small ffts better for cpu stability for me...

but i use blend for total system

Idle & Blend with 875MB memory allocated to each instance of stressprime
http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/stressprime_idle_tn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/stressprime_idle.jpg) http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/stressprime_blend_custom_875MB_tn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/stressprime_blend_custom_875MB.jpg)

Dual blend load running
http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/stressprime_blend_1_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/stressprime_blend_1.jpg)

[no_n|ck]

There is a bug with running Large FFT's. It will fail in 1-4min on one core. I did a 12h of custom with 880mb for each instance, 12h of small fft's + i did 12h of Large FFT's on prime95 - so my system seams stable. It only fails in sp2004beta with large fft's (sometimes in a few seconds, or max at 4min). Can anyone with duale confirm this?

EDIT; system restart seams to help :D (passing 30min mark)

[eva2000]

large ffts no probs with FX-60 + 2GB OCZ 4000EB :)


Dual stressprime large FFTs with 890MB allocated to each instance

Idle and Load
http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/11x/275-275-3320-7-14-2213_1.52-1.4-1.5-2.64_1552_ds8dds2_7F5_trrd0_dc0/stressprime_largeFFTs_idle_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/11x/275-275-3320-7-14-2213_1.52-1.4-1.5-2.64_1552_ds8dds2_7F5_trrd0_dc0/stressprime_largeFFTs_idle.jpg) http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/11x/275-275-3320-7-14-2213_1.52-1.4-1.5-2.64_1552_ds8dds2_7F5_trrd0_dc0/stressprime_largeFFTs_1_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/11x/275-275-3320-7-14-2213_1.52-1.4-1.5-2.64_1552_ds8dds2_7F5_trrd0_dc0/stressprime_largeFFTs_1.jpg)

[LuckMan212]

i just found this weird bug....... I was stress testing last night and Daylight savings time kicked in....... these 2 SP2004's crashed simultaneously. I knew it was too much of a coincidence so I checked the time in my screenshots, the "before" and "after" were 1:50am and 3:00am, respectively. (I have a cron job that takes screenshots of my desktop at intervals) These were only 10 mins apart so I knew daylight savings had kicked in.

I emailed Johnny and he says its already fixed in the next beta. but I thought id post so in case others crashed last night they might know why...:)
http://img77.imageshack.us/img77/8827/2006040201504wg.th.png (http://img77.imageshack.us/my.php?image=2006040201504wg.png) http://img87.imageshack.us/img87/4347/2006040203001hy.th.png (http://img87.imageshack.us/my.php?image=2006040203001hy.png)

[typo_pl]

Apr 05 2006 - v0.40 released - Wow. Has it really been over a year since the last release?
- switch to Prime95 24.14 numeric routines
- fix bug when restoring from minimized/tray icon state
- during stress/torture test, change text selection to show progress
- restore CPU Affinity setting on startup, save CPU Affinity and Priority on exit
- miscellaneous tweaks to layout and dialog text
- show when page faults occur during stress test, N.B. if using MBM5 in Blend or Custom Test, paging may occur often
- show number of rounds in dialog
- warn if system CPU affinity mask different from process CPU affinity mask

I'm having trouble updating one of the mirror sites, so only one mirror site is available right now.

Check http://www.geocities.com/stressprime2004/ for the up-to-date list.

Let me know if you have any comments, suggestions, or bug reports.

Thx,
J

[lawrywild]

downloaded and will be testing tonight :)

[high5]

Apr 05 2006 - v0.40 released - Wow. Has it really been over a year since the last release?
- switch to Prime95 24.14 numeric routines
- fix bug when restoring from minimized/tray icon state
- during stress/torture test, change text selection to show progress
- restore CPU Affinity setting on startup, save CPU Affinity and Priority on exit
- miscellaneous tweaks to layout and dialog text
- show when page faults occur during stress test, N.B. if using MBM5 in Blend or Custom Test, paging may occur often
- show number of rounds in dialog
- warn if system CPU affinity mask different from process CPU affinity mask

I'm having trouble updating one of the mirror sites, so only one mirror site is available right now.

Check http://www.geocities.com/stressprime2004/ for the up-to-date list.

Let me know if you have any comments, suggestions, or bug reports.

Thx,
J
thanks for the update Johnny! :up:

keep up the good work! :toast:

[LuckMan212]

been using this version for a few days... ran 2x overnight for about 14 hrs no errors, using blend. Thank god it remembers its affinity settings now, a big time saver. It seems that determining CPU multi / HTT was too much of a hassle so now it just shows the raw CPU speed, which is OK, there are other tools for FSB/HTT.

Nice job Johnny, thanks mate :)

[Delirious]

I seem to have issues with it, stress prime always gives me errors after 10min - 1 hour on an overclock that i know is 12+ hours of prime95 stable.

Anyone else haveing issues?

[ST4]

Apr 05 2006 - v0.40 released - Wow. Has it really been over a year since the last release?
- switch to Prime95 24.14 numeric routines
- fix bug when restoring from minimized/tray icon state
- during stress/torture test, change text selection to show progress
- restore CPU Affinity setting on startup, save CPU Affinity and Priority on exit
- miscellaneous tweaks to layout and dialog text
- show when page faults occur during stress test, N.B. if using MBM5 in Blend or Custom Test, paging may occur often
- show number of rounds in dialog
- warn if system CPU affinity mask different from process CPU affinity mask

I'm having trouble updating one of the mirror sites, so only one mirror site is available right now.

Check http://www.geocities.com/stressprime2004/ for the up-to-date list.

Let me know if you have any comments, suggestions, or bug reports.

Thx,
J

I don't know if this has been addressed with this new version (will try tonight), but one alarming ethical issue i saw was the ability to change affinity and cpu cores on the fly. Why ethical? Seeing folks post left and right especially on ebay about "dual" priming stable cpu's while one could easily circumvent the program to fake it for a screenshot. A simpel solution to this is to force the user to STOP before making any changes once the Prime run has started (just gray out the options)...just a thought and keep up the good work! I could never go back to Prime 95 anymore because of SP2004!

[ST4]

I don't know if this has been addressed with this new version (will try tonight), but one alarming ethical issue i saw was the ability to change affinity and cpu cores on the fly. Why ethical? Seeing folks post left and right especially on ebay about "dual" priming stable cpu's while one could easily circumvent the program to fake it for a screenshot. A simpel solution to this is to force the user to STOP before making any changes once the Prime run has started (just gray out the options)...just a thought and keep up the good work! I could never go back to Prime 95 anymore because of SP2004!

awww just tried out the new version and its ben rectifed! :toast: