Release CPU stress test


Mine has been testing for the last 24h and has been at 954MHz for the last 12h so I think that's stable. Is there a way to edit the traybar menu?


EDIT: Also, is it possible to test memory overclocking stability in a similar way?
 
Last edited by a moderator:
Fahrstuhl: 12h should be more than enough - if it doesn't find any error at a given clock speed in 1 or 2 hours I would already consider it stable.


onpon4: Ah, that explains why you got that low number. I would be interested to see the OPP5 stable top speed too...


Updating the aggregated info (all @OPP5 unless mentioned otherwise):


- original (CC) Pandora: 778@OPP3 (onpon4), 928 (TrashyMG), < 936 (Linux-SWAT), 946 (Googer), 954 (Fahrstuhl), 956 (PokeParadox), 967 (Takedown), 969 (ZXDunny), ~978 (BaDToaD), ~979 (FaeMinx), 1006 (Ian J)


- Rebirth (GC) Pandora: 956 (apersonthinks), 985 or 989 (TrashyMG), 999 (_wb_), ~1000 (lemondragon)


- 1GHz Pandora: 1157@OPP4 (Lafazar), 1175, 1178 (foft), 1214 (kamui_kun)
 
I only ever used OPP5 once to test how high I could overclock, and I think the most I could overclock with that without crashing was around 900 MHz. In contrast, I think I can overclock up to 850 or so without crashing at OPP3. So I really don't expect overvolting to make the number much higher.
 
Ok, I have news for everyone who thinks 1 hour is a safe bet for a stable overclock.


I was just about to post that after an hour and 10min my Pandora was stable @ 975Mhz


... however, it just dropped down to 974Mhz a couple of minutes later.


So my info needs to be adjusted down to 974Mhz for now... and I'm going to wait at least 2 hours before I confirm that this is my top clock speed @ OPP 5.


Edit: So yeah... after about 1 hour 25min @ 974Mhz it dropped down to 973Mhz... Doesn't look like I'll be using my Pandora for anything else today... At this rate, who knows what reliable amount of time is required before a stable overclock can be determined? 12 hours? ... Personaly, I'm now looking to maintain 973Mhz for at least 2 and a half hours before I'm satisfied...
 
Last edited by a moderator:
But doesn't that mean also that if you use a normal program/game with neon at that speed you are likely to use it for more than an hour before it might crash? For a game that might suffice.
 
But doesn't that mean also that if you use a normal program/game with neon at that speed you are likely to use it for more than an hour before it might crash? For a game that might suffice.

Is it worth risking it over a few MHz?
 
Ok well... I thought I was there.


I had successfully gone over 2 and a half hours @ 972Mhz


I decided to leave it and go and have supper.... and what do you know, somewhere around the 3 hour mark it had dropped to 971Mhz


So... it's been sitting at 971Mhz for roughly 1 and a half hours now... I'm just going to leave it overnight.


I strongly suspect, based on these ever increasing gaps, that many of the top clock speeds already posted are infact incorrect.


Am I able to do anything else with my Pandora while the stress test is running? I.e. use the Free e-book reader? I'm right at a juicy part of the book and I've been *dying* to get back to it.


Of course I don't want to mess up the test so would rather go without my book for an evening if there is any risk.
 
Am I able to do anything else with my Pandora while the stress test is running? I.e. use the Free e-book reader? I'm right at a juicy part of the book and I've been *dying* to get back to it.
Of course I don't want to mess up the test so would rather go without my book for an evening if there is any risk.

I'd say any time the Pandora is running something else is time when it isn't stress testing. But reading a book probably isn't going to steal that many cycles from it. I don't think you'll mess up the test..
 
Mine is an original (CC) Pandora: 973 at OPP5 :)
 
Last edited by a moderator:
My GC will do 1017 at opp5 stable for about an hour or so. interestingly i was running off USB and the charge light dropped off as soon as i started this up and it used more juice than USB could provide. So it ran out of battery before i could do much longer
 
Last edited by a moderator:
Uploaded an update, with these changes:

  • it's now possible to set upper limit (for those who get system crashes)
  • the script is now aware about overclock limits in CPU settings
  • prints a timestamp so you can see when last error happened
  • runs a bit longer every 10 steps to increase chance to detect errors earlier and avoid crashes


And finally the results on my own units, after > 30 hours of testing:

  • CC 256M pandora: 917 @ OPP5
  • 1GHz: 1249 @ OPP4


The CC one stepped down after more than 10h of runtime at 918, so there is no proper time limit for this test I guess..


BTW for those wondering, DM3730 only has 4 OPPs, which are named OPP50, OPP100, OPP130 and OPP1G, for nominal clocks of 300, 600, 800 MHz and 1GHz. For pandora this was left as just numbers 1-4 to compatibility with OMAP3 based software.
 
Last edited by a moderator:
Alright. You can *finally* put me down at 968Mhz @ OPP 5


Now I'm going to download the new version of the stress test and begin work on OPP 3 (Which I'm actually far more interested in).


Notaz - Any thoughts on making a stress test to determine non-neon maximum clock speed? Or is there already a pnd for that?


P.s. Are you getting us to do these tests because the new revision of PCSX ReARMed is going to require a significant overclock?
 
Notaz - P.s. Are you getting us to do these tests because the new revision of PCSX ReARMed is going to require a significant overclock?

Double resolution rendering (excited!) will require overclocking on non-1ghz Pandoras
 
Back
Top