Overclocking And Overvolting

[quadomatic]

My N900 overclocking results seem to be very good in comparison to other people's results on the N900 forums, but the N900 overclocking thread is very full of "noise" from people just copy-pasting other people's settings and FUD, only a handful of people are actually setting and testing the voltages themselves so we don't have a very good sample of how the yields really are for overclocking and what voltages they may require.

Very informative post, thanks!

I'm betting we'll get a better sample of average overclocking potential from the Pandora community.

 

[ste_167]

875 here, I get freezes in the 880s. Battery life and stability are most important to me but it's good to see that everyone is getting good speeds. I'd have been a bit gutted if 550Mhz was my limit!

 

[GizmoTheGreen]

800 here, and the pandora crashed while running psx4pandora...? :/

 

[mvickers03]

800 here, and the pandora crashed while running psx4pandora...? :/

I get a stable OC @ 880 with Mupen64 although with PSX4Pandora I can only goto 810, seems a strange but there is a big difference.

 

[juniorm33]

800 here, and the pandora crashed while running psx4pandora...? :/

I get a stable OC @ 880 with Mupen64 although with PSX4Pandora I can only goto 810, seems a strange but there is a big difference.

Different demands, so not strange really.
Once we can change the voltage, may become stable at 880 for both (if not higher)

Be nice to find out at what speeds the multiplier changes at etc.
As there are reports that PSX4Pandora sometimes runs faster at say 600 than 650. Been speculated it could be the bus speed.

 

[GizmoTheGreen]

800 here, and the pandora crashed while running psx4pandora...? :/

I get a stable OC @ 880 with Mupen64 although with PSX4Pandora I can only goto 810, seems a strange but there is a big difference.

ah, i ran at 800 with mupen64+ as well.
however psx4pandora crashes alot when im over say, 750?

 

[mali]

I didn't have a look at the Mupen64 and PSX threads yet, but it might be possible that PSX uses some more NEON calculations and NEON is rather sensitive to overclocking and crashes early. I read about it on Beagleboard IRC ages ago.

 

[juniorm33]

It seems the memory is asynchronous going by what Exo says, so that doesn't seem like it would explain why some people are getting faster speed in PSX4Pandora at a lower frequency.

Mali, it seems that PSX4Pandora is more demanding on the system (or part of it) than Mupen64, so the overclocks people were getting may not be stable with PSX4Pandora.

At the moment PSX4Pandora seems like the app to stress overclocks at the moment.
Although I'd be happy to clock at different rates depending on stability of the software I'm using.

 

[darionapole]

why don't we just use superpi to test the stability? and maybe a second one that would use the neon instructions? that way we could have X MHz stable for general purpouse, and Y MHz stable for apps using cpu specific instruction like neon.

 

[Yod4z]

laharl we need more than superpi because inside the Soc we have Cortex-A8, DPS and SGX and RAM then to test stability we need to test all of this. Maybe PSX4all use more the memory or more the SGX than Mupen64.

 

[darionapole]

but wasn't the SGX, DPS, and RAM clock independent from the core clock? as far as i read, the RAM's clock is linked to the SGX's clock, i remember reading that if you overclock the ram it also overclocks de SGX, but that it wasn't stable, and that it would be better to use the OMAP #number-following-pandora's-omap, firmware or driver (don't remember) that was 200MHz instead of 110......maybe i misread everithing...

 

[mvickers03]

Once we can change the voltage, may become stable at 880 for both (if not higher)

Be nice to find out at what speeds the multiplier changes at etc.
As there are reports that PSX4Pandora sometimes runs faster at say 600 than 650. Been speculated it could be the bus speed.

I hope so the more the better. I didn't notice the difference TBH 810 is considerably faster than 750 as is 750-700, 700-650 and 650-600

ah, i ran at 800 with mupen64+ as well.
however psx4pandora crashes alot when im over say, 750?

I must be lucky, I just finished resident evil 1 @810MHZ and it was running for 6 hours in one sitting. cant wait to see what a OV can do

 

[Vitel]

Has anybody already compiled the overvolting kernel? I wanna try it. I have manually fixed my broken case and swapped the nubs, so my warranty is definitely void now.

 

[mali]

^ urjaman did it, maybe you should send him a PM.

 

[Vitel]

Will do, thanks.

 

[MonkeyChops]

please keep us updated on how far you get, I'm very interested in this.

 

[urjaman]

Will do, thanks.

And I'm here to note that Vitel should have email (i sent him my kernel binary).
I'm personally interested in how fast his pandora will go
And just a note to you people building your own kernels: this voltage patch he is using is in the bugtracker already (cant bother to point to it now, sorry.).

 

[mali]

^ Does your kernel allow selection of all available voltage values? Which Vsel did you use for your 1050MHz?

edit:
Some info from Droid/Milestone

                     [max_vsel]
[frequ.] [stable] [possibly unstable] [probably unstable]
            >=                                <=
   550      56 (original setup)      
   800      48          46 .. 44              42
   1000     60          58 .. 56              54
   1100     64          62 .. 60              58
*  1200     76          74 .. 70              68
*  1330                   84

*: Always unstable in some CPU's! Will damage CPU on prolonged use.
According to OMAP3430 datasheet, max_vsel up to 66 should be acceptable.
Above 80 will certainly severly damage the CPU on long-term scale.
The most appropriate and stable settings seem to be
either 800 MHz or 1000 MHz at max_vsel of 56 to 60.

Vcore=Vsel*12.5mV+600mV

 

[Vitel]

I seems to be quite stable at 1Ghz (OPP5). Currently I'm not going to try any further.

 

[mali]

^ Which voltage is it at?