OVERCLOCKING SPV....

[Guest ClintEastman]

I should imagine it's the bus speed that's killing some other component, and it's not the CPU falling over, seeing as the CPU is designed to do higher speeds. If you think of it it can't be a heat prob because it was never designed to be cooled.

It could be a voltage problem, EG. running a lower voltage to save the battery life.... IMHO

The bottom line he told me was that it was running at 84MHz, cos that is the only speed it is stable.

And that's a very strange comment from redmond.....

That's saying that HTC didn't hit the spec on it.

[edit]Just checked on ti site, it's the slowest OMAP in production[/edit]

[Guest Hax]

I would have thought that it'd be more likely that EMC requirements haven't been met and slowing the processor down gives the SPV a better chance of passing the EMC tests. Could also be that the phone's been so badly designed that there is internal EMC interference killing the address/data lines, etc... ("Crosstalk" is the term - IIRC :)

Hax

[Guest DJHope]

The TI OMAP processor, is the processor contained within our beautiful machine and is designed to run at 132mhz. This processor simply containes an ARM 920t core and extra GSM and radio processing modules which have been generated by Texus Instruments, i cant see why itd be unstable at this speed (since thats what its designed for) but i supose it could kill the battery in 30mins or something hence the lower clock speed!

DJ Hope

[Guest slowphones]

The problem being, everything runs off the clocks that need to be changed to make it run faster, this will cause major problems with devices - SD card would be the first one that would fail. It's possible to run the processor above 100Mhz and for it to stay stable but it would require a change in more than just the clock speed... I think we will have to wait for a firmware and OS update... someone please prove me wrong

[Guest spacemonkey]

It runs internally similar to real life pcs. There are 2 clock speeds in the phone, the bus clock and the core clock. Here after BCLOCK and CCLOCK.

BCLOCK is normally slow, it's the speed memory and peripherals talk to the chip at.

The control you can do in processor from DjHope's manual is:

set it to Fast Bus speed. This actually underclocks the processor to BCLOCK speed. This is intended to make the phone super stable and predictable for things like startup etc. It's fully synchronous in this mode.

set it to Synchronous. This says that CCLOCK will be an exact multiple of BCLOCK. This means, if the BCLOCK is say 21Mhz and CCLOCK is 84Mhz (making up those numbers) then the processor is running 4 times faster than the bus, but it can always rely on the fact that the bus will be aligned with a clock cycle. As in clock cycles 1,2,3 the cpu cycles but can't use the bus, cycle 4 the cpu cycles and reads from the bus, etc.

set it to ASynchronous. This says that CCLOCK can be any speed greater than BCLOCK and doesn't need to be a multiple. So the CPU ticks along at whatever speed it likes with no regard to the bus. When the CPU needs data from the bus (like a memory read then it will delay it's current clock cycle to synchronise with the bus (a maximum loss of 1 clock cycle) read the memory and then continue on it's merry way.

So, to overclock, we need to in the phone increase CCLOCK to make the processor run faster. Additionally we may need to make a call to the processor to change it's mode from synch to asynch (or vice versa). Other components should be unaffected because BCLOCK remains unchanged. Pauls guy experienced instability, this could be either because (A) the chip just can't handle it (needs a voltage change?) or some other freaky thing is happening, or (:) he has pushed up CCLOCK without setting the synch/asycnh correctly so the cpu can't talk to the phone reliably or © some other problem.

So the primary thing to address is if there is a way from software to communicate to the phone BIOS and adjust the CCLOCK speed. After that it's a case of experimentation and seeing what we can do.

That's my understanding of the hardware at the moment anyway.

[Guest Eran]

I'm seeing you got into the hardware and kinda understood parts of it.

Could you possibly think of tryin' to do it?

I'm sure many people would try it out and give you feedback to try and make it stable and good as it has to do with a software tweak nothing inside the hardware itself

Eran

[Guest spacemonkey]

The problem is I don't actualy know that much. What I've posted is what I can infer from the processor documentation.

To move forwards I'd need some documentation relating to the rest of the hardware and how to instruct the bios. This documentation is probably only shared within MS and HTC would be my guess.

My outline of a software solution would be something like:

Set processor to fast bus mode (BCLOCK)

Adjust CCLOCK speed

Set processor to synchronous/asynchronous mode as appropriate

I have no clue as to the commands or how to adjust CCLOCK speed.

[Guest ClintEastman]

I get the funny feeling that the SPV was NEVER ment to go at full speed, it's the slowest OMAP so they couldn't use a slower chip because there isn't one, so they used the 710 and under clocked it to the speed the phone was specd to. :cry:

Thats not to say there isn't a way round it, seems to me that Spacemonkeys idea is the best we have so far.

I wonder what jimmysoft's approch is?

[Guest madu]

Research well done Spacemonkey (and DJHope o'course).. Now we just need someone clever enough in this area to put this together in some kinda usable tool.. But we all know that bla bla. Just my regards to you guys (most of you actually).

[Guest stevesco]

Does FLORIN_M have a comment on this?? :twisted:

[Guest slowphones]

The problem is within your explanation, there are problems with the TI chip that mean it must run in the current configuration ( at least it had to last year..). The other area of problems is that to speed the clocks up you also need to change the memory timings which are set both in the bootloader and then in the MS startup code, it is all possible but we need someone at MS to either let me do it... or get the manufacturers to get employee's that can do it!