How Do You Achieve High Clock Speeds?


I know I can't wait to hang a compressor off the back of my '2x. ;)

In all seriousness, ramping up voltage to the MMSP2 is likely to help, but given the high integration level of the chip, it's equally likely you'll create instability among all the various coprocessing units like the video decoder or the blitter. It's something I do intend to try, though. I'll publish anything I learn. Were the ARM920T a distinct IC it would be far more feasible to achieve substantial overclocks, but if you ask me, achieving 317 MHz on a 200 MHz-rated processor like mine is more than I could ever have hoped for; and the performance gains are STAGGERING.
 
I remember at one point you were talking about ways to get 350MHz out of it (with some hardware modding). I was kind of hoping that it was a precursor to turning that into an actual project.
 
Epicenter posted on May 12 2006 at 04:17 PM said:
I know I can't wait to hang a compressor off the back of my '2x. ;)

In all seriousness, ramping up voltage to the MMSP2 is likely to help, but given the high integration level of the chip, it's equally likely you'll create instability among all the various coprocessing units like the video decoder or the blitter. It's something I do intend to try, though. I'll publish anything I learn. Were the ARM920T a distinct IC it would be far more feasible to achieve substantial overclocks, but if you ask me, achieving 317 MHz on a 200 MHz-rated processor like mine is more than I could ever have hoped for; and the performance gains are STAGGERING.

I was meaning compressor cool it like those guys do in the netherlands to SLI 7900GTX and an overclock x2 AMD to get 40,000 score on 3dMark 2006. They don't run it that way, just for running becnhmarks.

I meant as far as voltage, ensuring that the load was taken off the power supply circuit doing the upconverting of the voltage by using 3 AA batteries (3.6v is better than 2.4v to start with, the NiMH batteries just go down from there under high load.)

I assume you mean to modify the power supply circuit, any idea if a pencil-trick mod like with the GP32 could service surface? (I meant surface, but service worked pretty well too :))

What do you mean by Staggering performance gains?

Do you mean that the performance goes up more than you would have thought by the clock speed, or that it goes up enough to make a certain program scream along at an insane pace?
 
Last edited by a moderator:
I've heard from more than one guy "in the know" that - basically - we're talking about stuff we don't know. Like Epicenter said - ARM chips work in a way us PC guys aren't used to. I don't understand enough about electricity, but suffice it to say, heat isn't the issue.

Epicenter, how'd you get 317mhz out of your GP2X? I'll be doing pagan dances over the weekend to see how mine comes out. I called upon dark and terrible forces. No doubt, after the weekend, a horrific cloud will slowly work its way across the Atlantic, cutting a swatch of sunken ships and dashed lives lost at sea.

But I had to ask myself - "Do I know any of these people?" ;)
 
David Beoulve posted on May 12 2006 at 07:13 PM said:
I've heard from more than one guy "in the know" that - basically - we're talking about stuff we don't know. Like Epicenter said - ARM chips work in a way us PC guys aren't used to. I don't understand enough about electricity, but suffice it to say, heat isn't the issue.

Epicenter, how'd you get 317mhz out of your GP2X? I'll be doing pagan dances over the weekend to see how mine comes out. I called upon dark and terrible forces. No doubt, after the weekend, a horrific cloud will slowly work its way across the Atlantic, cutting a swatch of sunken ships and dashed lives lost at sea.

But I had to ask myself - "Do I know any of these people?" ;)

LOL, but seriously heat is not the issue yet.

Heat usually becomes an issue when you need to overvolt a chip to get it running overclocked where you want it, then sometimes it starts overheating (go ahead ask me how I know :)).

I like playing with overclocking stuff and I needed to raise the core voltage on my Duron 800 (my god, has it been that many years ago) to get it to clock to 1ghz, but then it wound up overheating, so I underclocked it again and left the pencil lead on the overvolt mod (yeah, that's right, the original thunderbird chips used a pencil-mod), it ran better than without the volt addition.

I wonder if the MMSP2 is getting properly filtered voltage that stays high enough under full load, that is why I would check out the 3 AA and voltage regulator adjustments first.

Of course if the inefficiency of the voltage regulator and batteries are making the MMSP2 needlessly hot, that is another matter, and maybe one we should look into.
 
Last edited by a moderator:
I have a GP2X First Edition,Duracell 2500 mAh and fantastic DaveC cup. Mame run perfectly at 295 Mhz and the screen clearly :p . Lucky GP2X. :D Great machine. ;)
 
Epicenter posted on May 13 2006 at 12:17 AM said:
I know I can't wait to hang a compressor off the back of my '2x. ;)

In all seriousness, ramping up voltage to the MMSP2 is likely to help, but given the high integration level of the chip, it's equally likely you'll create instability among all the various coprocessing units like the video decoder or the blitter. It's something I do intend to try, though. I'll publish anything I learn. Were the ARM920T a distinct IC it would be far more feasible to achieve substantial overclocks, but if you ask me, achieving 317 MHz on a 200 MHz-rated processor like mine is more than I could ever have hoped for; and the performance gains are STAGGERING.
I envy you!

MattB posted on May 13 2006 at 10:36 AM said:
>300Mhz ?? Wow! The best i can achieve on my GP2X is 255mhz
Ditto!
 
Last edited by a moderator:
Heat usually becomes an issue when you need to overvolt a chip to get it running overclocked where you want it, then sometimes it starts overheating (go ahead ask me how I know smile.gif).
Makes good sense. I'll talk with my friend, Jock. He overclocks every PC he gets. Sometimes I wonder if he likes playing games or overclocking things. You and he are probably a bit alike in that regard ;)

So - batteries might change the overclockability, and more importantly, the connections and equipment used to pull power from the batteries? That sounds... moddable?
 
When I say the performance is staggering-- I mean in terms of utilizing actual applications. The framerate in GNGeo2X is blistering, Quake2X runs about as fast as it does on my laptop in software mode at any reasonable resolution, which for a fixed-point ONLY processor is pretty damned incredible. My own game engine runs astronomically faster, even the menus are more snappy and responsive.

Granted running at 317 MHz is by no means fully stable, it's quite crash-prone. Usually I can sit comfortably at 307 or 308 MHz without incident though. Very few things will crash at that level, Quake being one of the things that does after a bit. This is why I'm looking into ways to improve stability.

Let's face it, 2 ARM9s, an LCD controller, LCD, backlight, video hardware, audio hardware, USB hardware .. this is a lot to power from 2 AAs! It's no wonder PocketPCs utilize high-current level Lithium Ion cells. Installing a better power source-- a Lithium Ion pack or a higher-current AC supply than the default GPH one (maybe 2 amperes+?)

I'm looking into raising core voltage to the MMSP2, it is worth an attempt. If there's a dirty power stream coming into the MMSP2 that could probably be corrected with some capacitors to smooth it out, and/or some filter capacitors tied to ground (IIRC very high value capacitance, they should act like open circuits and react only to high frequency interference that should NOT be found on low-frequency power lines).
 
Epicenter posted on May 13 2006 at 10:05 AM said:
Let's face it, 2 ARM9s, an LCD controller, LCD, backlight, video hardware, audio hardware, USB hardware .. this is a lot to power from 2 AAs! It's no wonder PocketPCs utilize high-current level Lithium Ion cells. Installing a better power source-- a Lithium Ion pack or a higher-current AC supply than the default GPH one (maybe 2 amperes+?)

I'm looking into raising core voltage to the MMSP2, it is worth an attempt. If there's a dirty power stream coming into the MMSP2 that could probably be corrected with some capacitors to smooth it out, and/or some filter capacitors tied to ground (IIRC very high value capacitance, they should act like open circuits and react only to high frequency interference that should NOT be found on low-frequency power lines).

If the power is what needs to be more, how about using 9000 mah D-cell, or less rated C-cell NiMH batteries?

I like where you are going with the power thought, it is the same one I had. Do you have a DSO, or a regular scope that you can see the power going into the MMSP2?

I hope you have a good macro camera and good schematic skills (or a friend who can help you out), because I am sure many will want to try this kind of stuff, I mean me of course ;).

So everything screams, that is very interesting, I wonder how the PSX emulator would take to a nice fat overclock.
 
Last edited by a moderator:
Because I'd like to be able to close my GP2X up when I am done. ;) As for ramping up vcore to the ARM920T, I've looked into it and the ARM920T has its own DISTINCT voltage line. It is pin VDDA920 on the MMSP2 and can be found in the datasheet. However, due to the lovely FPBGA package that prevents any of the pins from coming through the PCB, tracing down the line to that pin will be a lot of trial and error, or a process that would necessitate removing the MMSP2 from the machine, after which point I'd not really expect it to go back on very cooperatively..

At any rate, the ARM920T seems to take 1.8vdc by default, and its maximum rating is 1.95v. At 1.9v or 1.95v, I'd expect much more stability past 300 MHz .. Maybe my '2x would hit 320 with stability. Right now it only stands it for several seconds, I cannot even get through one run of 2xMark.
 
Back
Top