Don't know what that is, haven't had the time to read up on all the things in CSSU.

And on a side note, have been using the 512Hz for 20 hours now and battery is the same as before, well it seems better since im not seeing those voltage spikes like before according to batterygraph, it is to early to claim better or worse battery life, BUT will say one thing, lower spikes will give better battery life, since a higher C rate always drains the battery faster, meaning the end capacity will be lower.

And Estel it is not a placebo, things do load faster no doubt at this point.

EDIT: And according to batterygraph CPU load seems much more calmer (lower)

full day has passed, no more battery used then normal.

so the 512Hz does not mean more battery usage, i would rather say a little less...., but not enough on the graph to say true on that statement.

Things are still very snappy using 512Hz.

it's funny, with better OC, 512Hz and stripped transitions my N900 is the fastest it has ever been, and thanks to the 720p (576p) has gained a nice bump up the tech ladder.

For me now the future replacement for my N900 has to be even better then previously thought

Rebuilt last kernel-power with a patch to 512Hz.
Visible changes have not noticed, I'll check the battery consumption.

Same here, running for few days. And seems video recording is unaffected too

So far running with 512Hz mod on my both N900 and no issues with battery consumption or any other areas. Other device has been on idle (still in connection to network) last 4 days and there is still 27% battery left. Pretty OK I would say.

What comes to usability. I run tests with 'pi_css5-vfp3' (pii calculation test sw) and results between 128 and 512 were identical. So no power increment gained, but overall responsiveness is a bit improved. Programs starts a little bit faster. I did comparison between my N900's, one running with 128Hz and other with 512Hz. 512Hz was always faster (applications up and running). Both devices are using identical sw settings and images (back up from one to other). So I have to disagree with comments about that no visible changes observed.

Edit: Performance delta between 128 vs 512 is not very big, so difference is not so easy to observe. Comparing when devices are side by side is giving this evidence that 512 is a bit more snappier.

Anyway, increasing CONFIG_HZ to 512 was a proposed countermeasure against video stutter. And seems it does not help. That "increased responsiveness" reminds me of "crappatch" users reports, i.e. beauty is in the eye of the beholder .

On the other hand upstream gst-dsp has a lot of changes for the last month, one of them ensuring DSP buffers alignment at 128 byte boundary (DSP cache line size). That should affect performance (hopefully). Unfortunately some of the other changes break almost all DSP codecs on n900 and it will take me a while until I do the backport.

EDIT:
@Gusse - is it possible to shot a video with enough samples of devices side by side, so we can make a definite conclusion whether there is a speedup?

Also, both devices are not only clones of same backupmenu images, but same revision, etc?

Also, the...

...doesn't sound very tempting. Is it worth at all, to make CPU report to kernel 512 times per second (instead of 128), even if it respect sleep state? Increased battery usage may be also hard to observe, yet bigger than "mystical" performance boost

/Estel

Im just going to say it again, mine is faster and more snappy now, not like 2x but i can defiantly see it, something as simple as opening calculator took longer at 128 then now, every thing else opens just as fast, when i press "Contacts" it opens up as fast as i can press it and have all contacts displayed in the blink of an eye, the same with sms/messaging and here i would like to point out the there are LOTS of texts in there.

In the camera sense, more frames in viewfinder in 576p & 720p, this is not important but shows an extra performance.

In recording 576p is still full on, no dropped frames, 720p still suffers from dropped frames but is FAR LESS then before, i have not OC DSP on mine, if i did i could probably record 720p, but looking at the recordings on big screen there seems to be very little gain anyway.
There is how ever a much bigger gain from normal recording (high 16:9) to the DVD16:9 (576p)

I really find it strange that you can't see the performance boost, can't just be me and Gusse that can see it, and he actually have 2x N900 to test it with.

After lots of testing under different loading conditions I got pretty much even results between 128 vs 512. I cannot say why time to time one device seems to be much faster than other and wise versa. Even under very stable conditions (offline, long timeouts between loadings, etc...) produced very different results what comes to application loading times and overall usage feelings (transition smoothness, responsiveness, etc...). Actually, now I don't even know what is really stable test condiction ...

So, earlier observations based on too short time frame and application coverage, meaning that results were too subjective.

As a conclusion, this mod is not improving overall performance, but neither it had negative impacts. At least I didn't see any drawbacks or drawbacks were so small that it was not easy to observe those.

I know, it sounded too good to be true, but time to time I felt that way

man!!!
freemangordon has a real psychological when it comes to performance increase