F105 Thunderchief (alpha)

[sjsharksfan]

was wondering if since all the best settings and functions are always pushed over to UberKernel eventually, is the UK going to be able to scale above 800MHz eventually?

SJ

[Sabrecat21]

Originally Posted by Spader

Holy Carp!

What is the absolute Max freq the Pre Proccesor can thoretically run at?

like most CPUs, you can SET them to run at just about any speed. The question is will they actually WORK at a given speed

[rwhitby]

Originally Posted by sjsharksfan

was wondering if since all the best settings and functions are always pushed over to UberKernel eventually, is the UK going to be able to scale above 800MHz eventually?

Well, since we have half the people in this thread saying that F105 has dramatically increased their battery life, and the other half saying that it has dramatically reduced their battery life (and both groups pointing the finger at the kernel as being the cause), you tell me whether we can consider this to be an effective alpha test of the F105 kernel (which is the same piece of code delivered to both groups of people).

Nothing gets into UberKernel in the public feed until is has reliable, repeatable results from effective alpha testing by a large number of people.

So, if you want > 800MHz speeds to get into UberKernel, then the alpha testers here in this thread need to do a much better job of performing controlled experiments to prove or disprove the stability and efficacy of the chosen voltage and DLL settings in the F105 kernel for OPPs greater than 800MHz.

-- Rod

[clintwinstead]

So far, F105 is INSANE fast and smooth. No hiccups whatsoever. I cannot believe how smooth this thing is. Will be testing out the battery life over the next day or so.
This is what the Pre should have been from the beginning.

[jamhandman]

I believe the battery drain is higher, but the phone is sooo fast. Web browser in particular is really smooth and functional.

Nice work. I'll try to get more detailed data soon.

[maxima2k53]

this kernal is awesome, but i decided to remove it mainly bc my swap memory got filled up to 90 mb according to govnah and my phone slowed down to a crawl. this has happened to me with compcache kernals b4 so yeah it sucks that i'm removing a 1 ghz kernal, which the speed was great but sucks that the swap memory gets filled up pretty quickly and never goes down unless i reboot

[brennan7]

Originally Posted by rwhitby

Well, since we have half the people in this thread saying that F105 has dramatically increased their battery life, and the other half saying that it has dramatically reduced their battery life (and both groups pointing the finger at the kernel as being the cause), you tell me whether we can consider this to be an effective alpha test of the F105 kernel (which is the same piece of code delivered to both groups of people).

Nothing gets into UberKernel in the public feed until is has reliable, repeatable results from effective alpha testing by a large number of people.

So, if you want > 800MHz speeds to get into UberKernel, then the alpha testers here in this thread need to do a much better job of performing controlled experiments to prove or disprove the stability and efficacy of the chosen voltage and DLL settings in the F105 kernel for OPPs greater than 800MHz.

-- Rod

Why don't we lay out a groundwork of what apps should be running, how often they should be called upon and used, and then base a comparison from lets say the f102 kernel (because it defaults from 500) to this kernel (which also defaults to 500) and give this man what he wants.

Whaddaya say alpha testers?

[pip smith]

Originally Posted by brennan7

Why don't we lay out a groundwork of what apps should be running, how often they should be called upon and used, and then base a comparison from lets say the f104 kernel (because it defaults from 500) to this kernel (which also defaults to 500) and give this man what he wants.

Whaddaya say alpha testers?

If you set a detailed plan for a reliable test then I will install the f105 and help test out.

[rwhitby]

Originally Posted by pip smith

If you set a detailed plan for a reliable test then I will install the f105 and help test out.

Please do comparisons using the F102A, F104A, F105, Warthog and UberKernel packages in the testing feed.

Please don't use previous kernels (e.g. F102 without the A or F104 without the A) unless you absolutely are not able to get the results you want with the latest versions of the kernels in the Preware testing feeds, cause there are issues when people make mistakes by incorrectly uninstalling those kernels.

-- Rod

[rwhitby]

Originally Posted by maxima2k53

this kernal is awesome, but i decided to remove it mainly bc my swap memory got filled up to 90 mb according to govnah and my phone slowed down to a crawl. this has happened to me with compcache kernals b4 so yeah it sucks that i'm removing a 1 ghz kernal, which the speed was great but sucks that the swap memory gets filled up pretty quickly and never goes down unless i reboot

So it seems you're throwing the baby out with the bathwater at the moment ...

Assuming you've closed all your open apps, and have enough free memory to hold what is currently in swap space, then you could just disable compcache to clear the swap. You could then choose to leave it disabled, or enable it again to start with a fresh swap space.

If you don't have enough memory to clear the swap space in this way with no apps open, then you have a serious memory leak somewhere (and note that kernels or compcache kernel modules don't affect application memory leaks).

-- Rod

[pip smith]

Originally Posted by rwhitby

Please do comparisons using the F102A, F104A, F105, Warthog and UberKernel packages in the testing feed.

Please don't use previous kernels (e.g. F102 without the A or F104 without the A) unless you absolutely are not able to get the results you want with the latest versions of the kernels in the Preware testing feeds, cause there are issues when people make mistakes by incorrectly uninstalling those kernels.

-- Rod

If a controlled test can be agreed on then I will follow the same procedure for all those kernals.
Would you like me to test the kernals in a particular order?

[jamhandman]

testing with a plan? Crazy I say.

had device on touchstone and temps stayed around 34 C. Nice.

[jamhandman]

just noticed that setting brightness to zero (brightness unlinked) sets the CPU to 500 (min freq)

[Moekind]

I'm on Bell and I installed the kernel from 1.4.1.1 testing feed is that ok?

[maxima2k53]

Originally Posted by rwhitby

So it seems you're throwing the baby out with the bathwater at the moment ...

Assuming you've closed all your open apps, and have enough free memory to hold what is currently in swap space, then you could just disable compcache to clear the swap. You could then choose to leave it disabled, or enable it again to start with a fresh swap space.

If you don't have enough memory to clear the swap space in this way with no apps open, then you have a serious memory leak somewhere (and note that kernels or compcache kernel modules don't affect application memory leaks).

-- Rod

so ur saying that the whole compcache, i have to disable it and then enable it for it to start from fresh? so it doesn't do it by itself?

[rwhitby]

Originally Posted by maxima2k53

so ur saying that the whole compcache, i have to disable it and then enable it for it to start from fresh? so it doesn't do it by itself?

Why would it do it by itself?

The normal operation is that processes use memory, and then release it. If the amount of used memory exceeds the RAM, then some least recently used pages are moved to swap to make space. This process fails when the swap is full.

If both your RAM and swap are full, then you have too many things using too much memory. Close all the things. If the memory (ram plus swap) is still being used, then you have a memory leak. Find the app which is causing the memory leak and stop using it. If the app is actually LunaSysMgr or BrowserServer, then report the bug to Palm.

-- Rod

[maxima2k53]

Originally Posted by rwhitby

Why would it do it by itself?

The normal operation is that processes use memory, and then release it. If the amount of used memory exceeds the RAM, then some least recently used pages are moved to swap to make space. This process fails when the swap is full.

If both your RAM and swap are full, then you have too many things using too much memory. Close all the things. If the memory (ram plus swap) is still being used, then you have a memory leak. Find the app which is causing the memory leak and stop using it. If the app is actually LunaSysMgr or BrowserServer, then report the bug to Palm.

-- Rod

wuts the best way to find the app that has a memory leak, and what if it is the lunasysmgr or browserserver? i guess im screwed?

Update: anyways the apps that i normally leave open on my phone all day are messaging, email, facebook, and i just started using twee

[illadelphatic]

I am running some perfomance tests running as many apps as I can. I have tried this with several alpha releases of UK and am currently using thunderchief (obviously). With UK running at 800mhz and with 24mb compcache, I would be able to run 34 apps at most before the phone became unresponsive. Right now I am running F105, I have 41 apps open (pretty much every kind of app except 3d games); my temp is at 42C and my memory is at 234/127...and the phone is still very responsive. I am going to try to push it further...47 apps, still good...50! phone slightly sluggish but memory at 234/127 still, temp 48C...app 52 was wikiHow, phone really slowing down...53: foursqure, phone crawling now, screen more or less unresponsive. I'd like to get back to the govnah card, but cant.

Edit: I'll run battery life tests in the next couple of days of normal usage.

[doctorx]

Following up to my previous posts, battery consumption vs UK 800 MHz is higher. This is with moderate browsing, multitasking, having the screen on constantly, and in an average signal area, on EVDO, no WiFi.

[SirataXero]

So,

I just came back from the same gym and put my phone in the same locker. Except this time I did not have Battery Monitor (at 15 sec interval) and/or Govnah (at 1 s inteval) running.

I came in at around 40% battery and left in 2 hours at around 20%. This sounds a lot more like normal operation.

Thus, as I stated earlier, I chose to blame the kernel... never said I'm sure that's what it is. Anywho, I figured it definitely was Govnah/Battery Monitor combo and i'm pretty sure now that it is. I can't say the kernel is off the hook, cuz I definitely have poorer battery life, but not drastically poorer, just a little bit.