Overclocking

[jockyw2001]

Inspired by GSLEON3 I have started to investigate overclocking options of the ARM11 (runs OS) and ARM9 (runs radio). I disovered a few things which may be of interest, but so far no spectacular results. There is more research needed and I hope people like anton tomov, cmonex, olipro and pof can help out.

Here is a summary of my findings for the ARM11:
- it normally runs on 384MHz which is equal to a masterclock of 768MHz divided by 2
- the ARM11 clockspeed is 768/N MHz where N is a frequency divider (N=1..16)
- i haven't found an option for clock stepping. (If anyone has a MSM7x00 datasheet please share it with me so this can be investigated)
- i found ARM11 clock control code both in oemsbl and spl. The latter is great as it allows to control clock from WM :)
- there are two commands in a mfg spl to set ARM11 clock speed and to test it: "freq n" (n=0..6) and "test"
- "freq 6" sets the fastest speed: 384MHz and testing with "test" results in 436ms on my kaiser (MSM7200). Would be nice if someone tests this on a MSM7201 based kaiser
- i reverse engineered the "freq" command and found a table at 0x8c00e230 with 5 entries for each speed setting. For "freq 6" the settings are:
384000 ; equal to clockspeed/1000
1 ; determines which master clock to use (value is either 0 or 1)
1 ; master clock divider-1 (increasing it leads to slower test speeds)
96000 ; some other clock derived from from master clock
3 ; divider-1 for deriving the 2nd clock
- i hot patched these values with the "mw" command (e.g. mw 8C00E2A8 1 6D600) but could not establish any better result with "test" :(
- setting the divider to 0 (actually means divide 768MHz by 1) caused the device to freeze
- i didn't do any benchmarking in WM
- i haven't investigated ARM9

Okay, enough for now. I hope this triggers more interest and research for this topic.

[kwickone]

i am sure there is interest. I have pinged both Wizcode (anton tomov) and Immiersoft over the past few months. Both have been slow to respond, and both seem to be having issues finding any sort of documentation on the processor.

Here's hoping you make more progress :D

[GSLEON3]

Jocky,

did you get any indication as to how they are handling the default clock?

Have you found any reference correlating to:
AYGSHELL.dll
OLEAUT32.dll
COREDLL.dll

I would think that the system would use them to report current status so that the current clock rate settings can be changed appropriately.

Most of the current scaling apps out there import their data from these three libraries, so I wonder if ther is within them some way to walk back to the actual processes.

Of course my hacking abilities are pretty much non-existent, but I do see that unfortunately a lot of the critical areas in the popular overclocking apps are encrypted. Have to see who can help me with that. ;)

[pingels]

I might be a noob...but here's my thought:
1)The diamond runs on 528mhz
2)That same table/coding should be in the diamond ROM, except it should have more "steps" right?
Anyway, that's a noob's thoughts at 11pm. (past my bedtime for sure : )

[wingman_214]

i figure that if the diamond can run at 528mhz(on a 900mah battery)with the same chipset as the kaiser why cant the kaiser run at 528mhz with little affect on battery life.
should be a matter of setting the clockspeed in the bios or the kaiser equivalent. it should be in the radio bootloader(i believe it runs first on kaisers) ipl, spl or something or other.

however my bubble was burst when i read that the diamonds processor was made in a 45nm die and the kaisers processor was made in a 60nm die.

but the only thing i know is that i dont know shit.

[NuShrike]

The acpu table you're talking about could be found in msm_clk.dll and in the Android GIT repo:
http://git.android.com/?p=linux-msm.git;a=blob;f=arch/arm/mach-msm/clock-7x00a.c;h=43f66c5b4e4da45b9e2ed1104444555eecd49171 ;hb=HEAD

Note, the Android ACPU table is slightly different from the one for the Kaiser, so I assume it's the difference between either the 7200 and the 7201 or the 90nm and the 45nm variations. The Android one is also missing one or two of the lower frequencies.

Btw, for msm_clk.dll, it's 6-column table, and the Android one is 7-columns.

To get and set the acpu clock, I think there's a few calls in either clkregim.dll or TrsTai_extension.dll. I'm sure you can find it by looking at the EXPORTS table.

Slightly unfortunately, a lot of the meat behind these calls are now hidden behind ONC RPC calls in the Diamond dlls.

[jockyw2001]

Someone with a 7201, please enter bootloader and type "test" and post the number in ms shown.

Thx for the hint about diamond spl, i'm checking that out. I can nicely test the diamond spl with the upcoming "FrankenKaiser SPL Loader".

Ah thx for the link NuShrike. Yes that is indeed the clockspeed table:
26 /*
27 * ACPU speed table
28 */
29 struct clkctl_acpu_speed acpu_freq_tbl[] = {
30 { 19200, ACPU_PLL_TCXO, 0, 0, 19200, 0, 0},
31 { 61440, ACPU_PLL_0, 4, 3, 61440, 0, 0}, /* VDD assumed */
32 { 81920, ACPU_PLL_0, 4, 2, 40960, 1, 0}, /* VDD assumed */
33 { 96000, ACPU_PLL_1, 1, 7, 48000, 1, 0}, /* VDD assumed */
34 { 122880, ACPU_PLL_0, 4, 1, 61440, 1, 3},
35 { 128000, ACPU_PLL_1, 1, 5, 64000, 1, 3}, /* VDD assumed */
36 { 176000, ACPU_PLL_2, 2, 5, 88000, 1, 3}, /* VDD assumed */
37 { 192000, ACPU_PLL_1, 1, 3, 64000, 2, 3}, /* VDD assumed */
38 { 245760, ACPU_PLL_0, 4, 0, 81920, 2, 4},
39 { 256000, ACPU_PLL_1, 1, 2, 85333, 2, 5},
40 { 264000, ACPU_PLL_2, 2, 3, 88000, 2, 5}, /* VDD assumed */
41 { 352000, ACPU_PLL_2, 2, 2, 88000, 3, 5}, /* VDD assumed */
42 { 384000, ACPU_PLL_1, 1, 1, 128000, 2, 6},
43 { 528000, ACPU_PLL_2, 2, 1, 132000, 3, 6},
44 { 0, 0, 0, 0, 0, 0, 0},
45 };

108 struct clkctl_acpu_speed
109 {
110 unsigned int a11clk_khz;
111 int pll;
112 unsigned int a11clk_src_sel;
113 unsigned int a11clk_src_div;
114 unsigned int ahbclk_khz;
115 unsigned int ahbclk_div;
116 int vdd;
117 };


The main difference is the additional PLL clock source of 1056MHz called "ACPU_PLL_2"
I don't think we have that PLL in a standard Kaiser, but I will try to verify it with some experiments.

NuShrike: do you know whch clockspeed is set on a stock Kaiser? Can u patch Android and try to force 528000 ?

[eXceed]

We're talking about two completely different CPUs here, the Kaiser CPU is built on a 90nm node process whereas the Diamond is built on a 65nm node process, the difference in power consumption and heat generation between the two is most likely the reason that all Diamonds can run at 528Mhz without fault... Tbh I think you're being optimistic to assume the Kaiser CPU can withstand that speed for any length of period, and also how are you going to test for stability? As far as I know Stress Prime is a windows only app! SOrry to put a dampener on the idea but I don't condone this...

[jockyw2001]

We're talking about two completely different CPUs here, the Kaiser CPU is built on a 90nm node process whereas the Diamond is built on a 65nm node process, the difference in power consumption and heat generation between the two is most likely the reason that all Diamonds can run at 528Mhz without fault... Tbh I think you're being optimistic to assume the Kaiser CPU can withstand that speed for any length of period, and also how are you going to test for stability? As far as I know Stress Prime is a windows only app! SOrry to put a dampener on the idea but I don't condone this...
Of course, but you see we're doing research and experiments and trying to find out details on how to control clockspeeds. There are very likely other PLLs in a stock Kaiser, not necessary running on 1056MHz, but perhaps 800 or 900MHz. To obtain useful results requires people with experience and persistance ...

[eXceed]

Yes of course, sorry I didn't mean to completely force the idea out the window as such, but I don't want people to expect good results!

[austinsnyc]

this needs to be made a sticky so that all can help with testing any way that they can. It would be amazing to have our Kaisers running faster from a simple change to the SPL and some .dll's (Well I am sure its not to easy) I will do what ever needs done for testing. Should be easy with the linux image compared to the WM6.1 OS. Am I correct on that?

[wingman_214]

this needs to be made a sticky so that all can help with testing any way that they can. It would be amazing to have our Kaisers running faster from a simple change to the SPL and some .dll's (Well I am sure its not to easy) I will do what ever needs done for testing. Should be easy with the linux image compared to the WM6.1 OS. Am I correct on that?

agreed, need to make this a sticky, the msm 72xx/75xx platform looks like it is going to be around for a while. overclocking may be whats needed to get that last bit of performance out of our kaisers.

[wingman_214]

in any which case jocky, i am willing to flash a test spls' and or roms with modified dlls and what not.

[mikeeey]

have you tried playing around with "HTC Peformance"? aparently it overclocks the Titan/6800 and actually works. But i tried it on my kaiser and didnt notice any difference. ill upload it if your interested.

[wingman_214]

have you tried playing around with "HTC Peformance"? aparently it overclocks the Titan/6800 and actually works. But i tried it on my kaiser and didnt notice any difference. ill upload it if your interested.

that cab has been proven to do absolutely nothing.

[austinsnyc]

We Should All Start Posting What We Have...

MSM7200
Has had FrankenKaiser.exe run on it before.
Kaiser got it like the week they came to nyc.

[wingman_214]

i have an at&t tilt.
MSM7200
security unlocked
Spl 1.0.olipof
Display driver 1.60.00.00

[Isaygarcia]

unbranded TyTN II
Spl 1.0.OliPof
with some mixing of drivers that partially work (but crashes every time i try something new)

[jockyw2001]

Guys, please enter SPL tricolor, connect with MTTY and type "test" followed by return. Post the value it returns and also whether you have a MSM7200 or 7201. Test only works if you have HardSPL installed.

For reference: my msm7200 kaiser returns 436ms

[aonanodad]

AT&T Tilt
MSM7201
SPL-1.0.OliPof
test returns 436ms

[Dredd67]

HTC TyTN II
MSM7200
SPL-1.0.OliPof
test returns 436ms

[cushcalc]

AT&T Tilt
MSM7200
KAIS1*0 MFG
SPL-1.0.OliPof
CPLD-8
Security Unlocked

test returns 436ms

[spaanplaat]

T-Mobile MDA Vario III
MSM7200
SPL-1.0.OliPof
test 436ms

[pen-pen]

Vodafone V1615
MSM7200
SPL-1.0.OliPof
436ms

[jockyw2001]

Okay, so far results are all 436ms. From now on only post a value if the result is lower than 436ms.

[jockyw2001]

AT&T Tilt
MSM7201
SPL-1.0.OliPof
test returns 436ms
Nearly overlooked your reply. Thx!
Shows that MSM7201 uses the same primary PLL clock frequency.

The challenge will be to find a faster PLL clock and configure MSM to use it for clocking the ARM11.

[fisha21]

Not sure if you've seen the following post?

http://forum.xda-developers.com/showpost.php?p=2232922&postcount=532

Haven't tried it at all, and don't know if there is any basis to it. Just trying to facilitate the exchange of information!

[johanromijn]

Not sure if you've seen the following post?

http://forum.xda-developers.com/showpost.php?p=2232922&postcount=532

Haven't tried it at all, and don't know if there is any basis to it. Just trying to facilitate the exchange of information!


Strange, values keep changing between 1352, 1382 and 1411.
Sometimes it's faster with HTC Performance enabled, sometimes with HTC Performance disabled. I think this needs an SPB Benchmark test.


-------------------
I heard a lot of people about this HTC Performance.
Some said it worked, some said it's impossible it will work.

So, I tried it myself with gxmark(faster clock speed, better graphics isn't it);

The results for GXmark were lower with HTC Performance installed.

With HTC Performance enabled: 1st try: 1382, 2nd try: 1352
With HTC Performance disabled: 1st try: 1411, 2nd try: 1411

BUT: It makes a difference.

I don't have / want to install SPB Benchmark.
Anyone with SPB Benchmark installed, who want to try this?

[_Alex_]

...I heard a lot of people about this HTC Performance...

HTC Performance has been around for at least 1+ years and has been proven time and time again on many different HTC devices to be 100% placebo effect and 0% speed increase.

Let's drop discussions about HTC performance and keep focused on creating a real Kaiser overclocking app :D

FWIW, added to the sticky.

*edit* All posts about 'HTC Performance' will be moved into this thread (http://forum.xda-developers.com/showthread.php?p=1902180). Feel free to discuss HTC performance there.

[wingman_214]

jocky, any updates?

maybe olipro and pof have some more ideas.

[austinsnyc]

I just saw this at enadget.com about Andriod but it looks like the same processor as is in the kasier will run at 528MHz

Following the grand reveal of the latest Android build running on fresh HTC-sourced hardware at I/O today, Google took some questions that managed to pull a few extra gems from the rough. First and foremost, key specs of that lovely piece of kit were exposed: 3.6Mbps HSDPA is on board, as is a Qualcomm MSM7201A processor good for 528MHz, 128MB of RAM, 256MB of ROM,

[GSLEON3]

Both MSM7200 & 7201 report 436ms.

[jockyw2001]

... First and foremost, key specs of that lovely piece of kit were exposed: 3.6Mbps HSDPA is on board, as is a Qualcomm MSM7201A processor good for 528MHz, 128MB of RAM, 256MB of ROM,
Good info :)

Someone with a MSM7201 please try this:
- enter spl bootloader and connect with MTTY (USB)
- type "freq 5"
- type "mw 8C00E2B0 1 0"
- type "freq 6"
- if device doesn't freeze, type "test" and report the value
- if it froze then remove and reinsert battery and boot normally

[Vaxon]

Hi,

If you can provide me with info how to type anything in the spl bootloader (after holding camera button and soft resetting I'm not able to type a thing) I can try to do that. Sorry for lame question.

Regards.

[aruppenthal]

Hi,

If you can provide me with info how to type anything in the spl bootloader (after holding camera button and soft resetting I'm not able to type a thing) I can try to do that. Sorry for lame question.

Regards.

Jocky is refering to the use of mtty i believe. If you use mtty it is very important to only use the commands specified by jocky as doing the wrong commands in mtty can have adverse effects. If you search the site for mtty you can find it easily.

[pen-pen]

Got a freeze right after typing the mw 8C00E2B0 1 0 command.

[Vaxon]

I have tried it on MSM7200 but unfortunatly it freezes after freq 6 command |-:

[pen-pen]

Sorry, I just seen that I've typed mw C00E2B0 1 0 instead of mw 8C00E2B0 1 0 :/

So, my device did not freeze after the freq 6, and I've got 436ms on a MSM7200

[jockyw2001]

I have tried it on MSM7200 but unfortunatly it freezes after freq 6 command |-:
Thx, but I already know the result for a MSM7200 device. This test is only for MSM7201 users. Furthermore the instructions have to followed exactly as I wrote them.

[manaaa]

Good info :)

Someone with a MSM7201 please try this:
- enter spl bootloader and connect with MTTY (USB)
- type "freq 5"
- type "mw 8C00E2B0 1 0"
- type "freq 6"
- if device doesn't freeze, type "test" and report the value
- if it froze then remove and reinsert battery and boot normally

hi,

makes it sense to try this on a polaris or nike?


regards

[jockyw2001]

makes it sense to try this on a polaris or nike?

If you have a MFG SPL installed you can try the "freq" and "test" commands.
Just type "test" and see what it says and just type "freq" and it will show the supported clock modes. You can't use the "mw" commands as the clock speed table will be on another address.

[aonanodad]

Good info :)

Someone with a MSM7201 please try this:
- enter spl bootloader and connect with MTTY (USB)
- type "freq 5"
- type "mw 8C00E2B0 1 0"
- type "freq 6"
- if device doesn't freeze, type "test" and report the value
- if it froze then remove and reinsert battery and boot normally

freq 5
Change frequency to level 5 ok

Cmd>mw 8C00E2B0 1 0
8C012480: E8B0FFFF E1A0F00E - EE100F10 E1A0F00E | ................
8C012490: E1A0000D E1A0F00E - E3A00000 EE070F1A | ................
8C0124A0: EE070F9A EE070F90 - E1A0F00E 00000000 | ................
8C0124B0: 00000000 E3A00401 - E1A00000 E1A00000 | ................
8C0124C0: E2500001 E1A00000 - E1A00000 1AFFFFF9 | .P..............
8C0124D0: E1A0F00E 8C01246C - E92D4030 E3520000 | ......$l.-@0.R..
8C0124E0: E1A04001 E1A05000 - 12422001 E3D4C00F | ..@...P..B .....
8C0124F0: 1A000001 E3520000 - 0A000003 E1A01224 | .....R.........$
8C012500: E1A00005 EBFFFFF3 - E204400F E5952000 | ..........@... .
8C012510: E354000A 32843030 - 22843037 E5C23000 | .T..2.00".07..0.

Cmd>freq 6
Change frequency to level 6 ok

Cmd>test
436 ms

Cmd>

[jockyw2001]

Cmd>mw 8C00E2B0 1 0
8C012480: E8B0FFFF E1A0F00E - EE100F10 E1A0F00E | ................
...
Not good, you should type the commands and not use copy & paste

[jockyw2001]

Both MSM7200 & 7201 report 436ms.
GSLEON3 and I tested with his MSM7201 and managed to double the ARM11 clock from 384MHz to 768MHz. George, could you repeat the result?

[johnny13oi]

GSLEON3 and I tested with his MSM7201 and managed to double the ARM11 clock from 384MHz to 768MHz. George, could you repeat the result?

Wouldn't doubling the clock make it a very extreme overclock and could possibly fry the chip or overheat it very quickly?

[aonanodad]

Not good, you should type the commands and not use copy & paste

Sorry.

Here are new results.

-----------------------

test
436 ms

Cmd>freq 5
Change frequency to level 5 ok

Cmd>test
655 ms

Cmd>mw 8C00E2B0 1 0

Cmd>test
655 ms

Cmd>freq 6

-----------------------------

then it froze, but a soft reset revived it fine.

[aonanodad]

OK, I've been doing these tests.

I presume that "mw 8C00E2B0 0 1" modifies the frequency table for the "freq 6" command. Freq 6 causes the device to hang after that command on my 7201 Tilt. Is that the only choice for a faster clock frequency?

Is there a list of the SPL commands anywhere? I have searched the forums and have found some information but not a good list - at least not one that talks about the mw and test commands. Also, is there an annotated source code listing for the ROM that would tell me something about what I am doing when I modify bytes at 8C00E2B0? I know I could probably find more information by digging through these forums, but I figured someone might be nice to me since I volunteered my new Tilt as a gunea pig. ;)

[GSLEON3]

Sorry.

Here are new results.

-----------------------

test
436 ms

Cmd>freq 5
Change frequency to level 5 ok

Cmd>test
655 ms

Cmd>mw 8C00E2B0 1 0

Cmd>test
655 ms

Cmd>freq 6

-----------------------------

then it froze, but a soft reset revived it fine.

No offense, but if you are unsure about what these commands mean, or how to really use MTTY in conjunction with SPL commands, maybe you should leave it alone. I really mean no offense, but there are thing MTTY can do to really wreak havoc on your device...as in kaput, dead...

After all, there are many experienced people working in here & I'd really hate to see your device dead trying to duplicate an experiment already performed. Cut & Paste can get you into bog trouble, & I have a device used solely for some of Jocky's experiments & whims.

[GSLEON3]

OK, I've been doing these tests.

I presume that "mw 8C00E2B0 0 1" modifies the frequency table for the "freq 6" command. Freq 6 causes the device to hang after that command on my 7201 Tilt. Is that the only choice for a faster clock frequency?

Is there a list of the SPL commands anywhere? I have searched the forums and have found some information but not a good list - at least not one that talks about the mw and test commands. Also, is there an annotated source code listing for the ROM that would tell me something about what I am doing when I modify bytes at 8C00E2B0? I know I could probably find more information by digging through these forums, but I figured someone might be nice to me since I volunteered my new Tilt as a gunea pig. ;)

The freezing is caused because there are othe divisors we are still missing. You have factors in scaling for CPU, BUS & RAM speed setting. I also have a feeling that the overclock in the msm series is voltage driven, not frequency.

Jocky, I was thinking, the device has to have an OEM default scale, so even if we patch it if this is controlled somewhere else, or if there is a module we're missing, that could be why it freezes. AI think we need to find a way to disable the OEM clock scaling first & then try applying spped patches.

[aonanodad]

The freezing is caused because there are othe divisors we are still missing. You have factors in scaling for CPU, BUS & RAM speed setting. I also have a feeling that the overclock in the msm series is voltage driven, not frequency.

How do you drive overclocking by voltage? You mean that you would want to overvolt the chip to be able to overclock it?

The Diamond uses a 65nm msm7201 and a lower voltage to get the 528 mhz clock rate without increasing power consumption or heat dissipation.

Other than battery life, heat dissipation will be the biggest problem with overclocking the Kaiser.

[neva dinova]

would it really get that much hotter if we only overclocked it to 450mhz

[no2chem]

hello,

I have a titan and have reverse engineered (actually, rewrote) the msm_clk.dll driver. It has a function msm_set_clk_params, but it only controls the msm_clk_pause function. Interestingly, the QTV driver tries to use the function and sets the clock to 384 MHz.

Anyway, I was wondering how far you got with reversing the freq command - do you know which memory locations are being set (or which registers) control the divider? I haven't looked in the titan spl, but I don't recall the "freq" command.

[jockyw2001]

Anyway, I was wondering how far you got with reversing the freq command - do you know which memory locations are being set (or which registers) control the divider?
Yeah, I got pretty far reversing it.

[NuShrike]

Yeah, I got pretty far reversing it.I've reversed the msm_clk.dll and clkregim.dll for the Kaiser to the point where I get similar or better ASM (which I've posted the full binary in the d3d/opengl/qualcomm thread before), but I don't claim to understand the C that well yet.

Jocky, want to combine with what I have? I am currently too busy to focus on this.

[daedric]

Nevermind...

[austinsnyc]

no check back for updates and dont clutter the thread with questions like yours please or mods delete that crap before this turns the wrong way.

[no2chem]

I've reversed the msm_clk.dll and clkregim.dll for the Kaiser to the point where I get similar or better ASM (which I've posted the full binary in the d3d/opengl/qualcomm thread before), but I don't claim to understand the C that well yet.

Jocky, want to combine with what I have? I am currently too busy to focus on this.

@nushrike and jocky

i'm posting the c code for msm_clk here. feel free to use it, it doesn't seem like it actually does anything really interesting. i also rewrote it to write messages to kitl for debug use. i've also rewrote parts of wce_rex, if you're interested. i haven't looked at the d3d thread, but from what I can understand from the C, it only controls pausing the clock.

[NuShrike]

@nushrike and jocky

i'm posting the c code for msm_clk here. feel free to use it, it doesn't seem like it actually does anything really interesting. i also rewrote it to write messages to kitl for debug use. i've also rewrote parts of wce_rex, if you're interested. i haven't looked at the d3d thread, but from what I can understand from the C, it only controls pausing the clock.Thanks. I do think a lot of the clock control meat is probably in clkregim.dll. Of course, a lot of this is a wrapper for TrsTai_extension.dll, so that's where you start digging into the secret amss qualcomm code.

From what I can tell from wce_rex, it also controls masking off data-abort interrupts during certain kernel operations, as well as allocating and tracking chunks of virtual memory (the KS20, new HTC WM6.1, and Diamond versions).

[jockyw2001]

@nushrike and jocky

i'm posting the c code for msm_clk here. feel free to use it, it doesn't seem like it actually does anything really interesting. i also rewrote it to write messages to kitl for debug use. i've also rewrote parts of wce_rex, if you're interested. i haven't looked at the d3d thread, but from what I can understand from the C, it only controls pausing the clock.
Thanks! I'm sticking to assembler code for the moment. It's very easy to reverse the freq command. You end up with all relevant io regs. I will not publish these yet, but you and NuShrike can send me a PM if you have questions.

[manaaa]

If you have a MFG SPL installed you can try the "freq" and "test" commands.
Just type "test" and see what it says and just type "freq" and it will show the supported clock modes. You can't use the "mw" commands as the clock speed table will be on another address.


hi,

im still with my custom rom and spl. i dunno if olipro' s work here matches the requirements of a MFG SPL. >> http://forum.xda-developers.com/showthread.php?t=351964

he states 'also please note this Hard-SPL is based on an MFG, for those of you who intend to play.'


so, if you need some polaris values, or need to test some files on the polaris for further progress, then count me in, np to install the hardspl and go though the instructions ;)

btw...

i read gsleon and you successfully overclocked the kaiser with double speed. that runs stable?

[GSLEON3]

LOL, fun, but not stable.

[austinsnyc]

Great job guys so it IS possible to overclock the kaiser! Who cares if its not stable at this point!! It shows that it can be done.... Please inform us on how this was done and if any speed increase was seen while using the device when it was overclocked!!
I'm sure LOTS of people would like to hear what you have to say!

[jcespi2005]

Really nice work :)

Congratulations to all for the efforts

Cheers

[GSLEON3]

Nothing worth sharing, really, I swear. While we can change the overall clock setting, the OEM scalling is still controlling Bus & Ram speed, there is too much we are still missing, some of it ties into the AMSS & that I think is an area only Jocky really understands. You'll see some (Very Slight) affect when many apps are open, but you'll deal with much more freezing as well & battery consumption. This is something that took Anton years to do for the px chips.

So far the vast majority of kaisers, all early units, completely freeze when trying to change the clock. We're not there yet.

[djrinc]

All these tests and experiments sound great and all, i hope you guys come up with something specifically for the Tilt...but for those who don't want to wait look for HTCPerformance.cab works fine for me. It's not actually for the Kaiser/Tilt but it's tested and it works. Only 2 speed adjustments but definately noticable increase.

[GSLEON3]

All these tests and experiments sound great and all, i hope you guys come up with something specifically for the Tilt...but for those who don't want to wait look for HTCPerformance.cab works fine for me. It's not actually for the Kaiser/Tilt but it's tested and it works. Only 2 speed adjustments but definately noticable increase.

LOL, dude do you really think you found something that no one else would know about???

C'mon, man. How do you know it works? I happen to know it NEVER has worked on any Pocket PC or Touch Screen device period. I have disassembled it, tested & benchmarked it, & poured thru it with a fine tooth comb using IDA PRO. There have been numerous other endeavors into HTC Performance as well. HTC PERFORMANCE DOES NOT WORK, IF IT WORKS FOR YOU, WELL, CONGRATULATIONS ON BEING SUGGESTABLE. I HAVE A PILL THAT WILL MAKE YOU THE STRONGEST MAN IN THE WORLD & FOR A LIMITED TIME IT'S ONLY $99.99

[thesire]

...I HAVE A PILL THAT WILL MAKE YOU THE STRONGEST MAN IN THE WORLD & FOR A LIMITED TIME IT'S ONLY $99.99[/B]

hahahaha lmao :D can't stop laughing :D:D

[swtaltima]

LOL, dude do you really think you found something that no one else would know about???

C'mon, man. How do you know it works? I happen to know it NEVER has worked on any Pocket PC or Touch Screen device period. I have disassembled it, tested & benchmarked it, & poured thru it with a fine tooth comb using IDA PRO. There have been numerous other endeavors into HTC Performance as well. HTC PERFORMANCE DOES NOT WORK, IF IT WORKS FOR YOU, WELL, CONGRATULATIONS ON BEING SUGGESTABLE. I HAVE A PILL THAT WILL MAKE YOU THE STRONGEST MAN IN THE WORLD & FOR A LIMITED TIME IT'S ONLY $99.99

GSLEON, just curious about something. I had used this program before and it is weird when it activated and you (x) out of it it seems snappier? What I am curious about is why that would be if it does not effect the speed at all. Just curious, but in no way challenging your opinion on the matter.

[GSLEON3]

GSLEON, just curious about something. I had used this program before and it is weird when it activated and you (x) out of it it seems snappier? What I am curious about is why that would be if it does not effect the speed at all. Just curious, but in no way challenging your opinion on the matter.

If that indeed is your belief, let's see if it holds.

Run baench marking utilities before & after use. You can also use SK Tools to see if ti's optimizing threads or something. (Hint: Skip the SK Tools, I've tried that & there was no difference) Also, _Alex_ has done extensive bench marking & if you believe his findings, device performance was actually worse with HTC Performance (Within the margin for error though, so probably no effect)

Also, everyopne that says that it works, always says it works on the lower setting, that makes no sense. Take a look at the registry keys it creates, I don't see how it could do much creating two keys that do nothing othere than reflect what check box you've selected, & accroding to the reg keys, on the lower setting it actually reports as disabled.

Much of the program & API's are deprecated. Just one of those things, like jumping higher in Nike's than you do in Adidas.

[manaaa]

he gsleon,

any convo with anton yet, or further details?

[GSLEON3]

he gsleon,

any convo with anton yet, or further details?

From what I've seen & heard, Anton thinks he's too good for us. I have emailed him, but he thinks it better we share everything with him via email without him putting anything on the table. I've invited him to the forum & given him the user names of all the interested parties. He responded asking for all of our info to be emailed to him, I did that & have yet to hear or see of him yet. To my knowledge neither has anyone else.

My impression is that he'd maybe lurk waiting to find something he can patent & use without having to contribute himself. I've yet to get any further correspondence & it's been over a month now.

[sebastianmc]

please forget HTC Performance!! its like the xcpuscalar on Hermes,,, lol I dont believe theres people that believes in htc performance!!

but I believe a day we ll running our kaisers at 528mhz!! Im waiting for it!!:D:p
regards and sorry for my english:confused:

[dharvey4651]

Has any real progress been made on this? Is it even possible? I know this was a hot topic when the thread was first started but it seems to have died.

I think that even though the ATI drivers are now working that overclocking should still be an active topic of discussion.

Think about it. If you have optimized ATI Imageon video drivers and an overclocked processor then you have the possibility of pushing the phone to its maximum potential and who really knows how high that really is...

[fisha21]

I think that even though the ATI drivers are now working that overclocking should still be an active topic of discussion.


:confused::confused:

Did I miss something? I know the drivers been extracted from the Diamond ROM, but I have not see any proof that they are "working".

[dharvey4651]

:confused::confused:

Did I miss something? I know the drivers been extracted from the Diamond ROM, but I have not see any proof that they are "working".

In the Vogue forum over at PPCGeeks they have made great progress and have the drivers booted and funtioning in a couple of ROMs over there. They're not really optimized to their fullest potential but yes they are working.

[TDO]

That's the understatement of the year.
There is only one function accelereted and you don't even notice this in real live.
So there are no realy functioning drivers available.

TDO

[keethar]

I'm extremely interested in this project. Has any progress been made? I read PHM overclocking would become available for the mogul, would the same method be transferrable to the Polaris of Kaiser.
Thumbs up for the guys working on this!

[arn350z]

Does the cpu have variable adjustments built in?
Why does my tilt show higher than 400 mhz?

[AciDCooL]

Does the cpu have variable adjustments built in?
Why does my tilt show higher than 400 mhz?

So that video player has a overclock function ? :eek:

[DaveShaw]

So that video player has a overclock function ? :eek:

I don't think so. I think the Media player has an accurate (or maybe inaccurate) CPU speed tester.

Ta
Dave

[CrArc]

What exactly are the limitations of the kaiser as things are at the moment? I mean, if these drivers advance to some kind of official release stage (and support 2D acceleration also) will overclocking be necessary? Considering it's all CPU-based right now yet still manages to work at passable speed, offloading the display onto the hardware would make a massive difference, right? Well, whether it will or won't it would be awesome to have support for overclockage, so I pledge to lurk in this thread regularly ;)

[joel2009]

Anyone still working on this? or is the project now dead!

[B0$$$]

ok so there's a lot of talk about overclocking...
anyone actually do this?
and how can it be done?