Cpu For Pandora 2 ?

[Blue Ion]

replacing a bike tire vs replacing a boeing 747 tire....

on your own alone, with the same tools you would use for the car...

 

[CPUnltd]

last I remember reading about PS2 emulation, somebody said the reason it was so hard is because the clock of the processor was hard to emulate/multiply... SNES runs at 3.58MHz... something not all that hard to emulate either by exact speed or by multiplied speed (ex. 7.16Mhz @ 2x, 10.74 @ 3x, etc)... with PSX, the speed is 33.8688MHz, but some people have been able to pull that off with either exact speeds or multiplied speeds, hence why so many PSX emulators run a lot of games, but different varieties of games (some games require even more odd-ball clock speeds than others and not every PSX emulator is able to match the clock speed for every game in question. Emulator-Zone lists the CPU for PS2 as such:

CPU: Emotion Engine 300MHz, 128-bit INT, 128-bit FP, 24KB L1, 16KB Scratch, 8KB VU0, 32KB VU1, 450 MIPS, 6.2 GFLOPS, 66M Vertices/Sec, 2.4 GB/s Internal, 1.2 GB/s Graphics, 3.2 GB/s Memory

...which, from my understanding, is not all that easy to emulate at exact speeds or multiplied speeds... that is the argument that was brought to the table (that I'm fleshing out a bit more due to my own personal research) in another thread revolving around PS2 emulation... if it was just a raw 300MHz processor setup, we would be able to at least do basic games and homebrew on PS2 emulators by now, but they are VERY few, far between and far from fully functional (try saying that 3x fast )... Since I wasn't the one to state this as the real issue behind PS2 emulation, I can't wholeheartedly back the claim... but since it was the argument I saw cast elsewhere for this subject, I figured I'd drop what information I know about it and found on my own time... I'd love to be wrong if it means actual/better PS2 emulation... but that's what discussions are for... figuring out what the real problem is that keeps us from doing what we wish to do with our devices.

 

[Blue Ion]

:/

So... we just underclock our pandoras to 300 Mhz, and bam, perfect, flawless and fullspeed ps2 emulation. Right?
Or overclock to 729 and have perfect Wii emulation, or 733 and have perfect xbox emulation.

 

[Drack]

I once replaced the CPU of a desktop computer. How much harder would doing so on the Pandora be?

It's not similar in the least.

Desktop CPUs fit into ZIF sockets specifically designed to be swappable. Just remove the heatsink and pop a lever and out it goes.

The Pandora's CPU is physically connected to the board itself. What you're suggesting would require you to cut all the pins connecting it to the board, then solder each individual leg from the new CPU to where the old pins were, making solid joints without bridging any connections. And that's assuming the new CPU is pin-compatible, which I bet it's not for anything that's not a Cortex-A8.

Upgrading the CPU on a Pandora board is not going to happen, period.

 

[typs lik dis]

Couldn't I put all the blobs of solder on the new CPU, then melt them all at once with some sort of flat soldering iron and apply it to the board? That seems not inanalogous to putting paste on a CPU heatsink. And couldn't a pin adapter be created?

 

[gruso]

It's much, much smaller than you think it is.

 

[Exophase]

Couldn't I put all the blobs of solder on the new CPU, then melt them all at once with some sort of flat soldering iron and apply it to the board? That seems not inanalogous to putting paste on a CPU heatsink. And couldn't a pin adapter be created?

I am 100% positive that anything that isn't an OMAP3x won't be pin compatible and therefore won't be something you can drop in a Pandora motherboard. I would be pretty surprised if even anything other than an OMAP3530 is. The newer SoCs probably won't even be the same package.

 

[typs lik dis]

Thx.

 

[urjaman]

I would be pretty surprised if even anything other than an OMAP3530 is.

I had thought that 3430, 3530 and 3630 would be pin-compatible (if the same package), but I havent actually checked (I dont know if TI even has the info available for others than the 3530).

 

[ari64]

last I remember reading about PS2 emulation, somebody said the reason it was so hard is because the clock of the processor was hard to emulate/multiply... SNES runs at 3.58MHz... something not all that hard to emulate either by exact speed or by multiplied speed (ex. 7.16Mhz @ 2x, 10.74 @ 3x, etc)... with PSX, the speed is 33.8688MHz, but some people have been able to pull that off with either exact speeds or multiplied speeds, hence why so many PSX emulators run a lot of games, but different varieties of games (some games require even more odd-ball clock speeds than others and not every PSX emulator is able to match the clock speed for every game in question.

What the heck are you talking about?

 

[darfgarf]

Couldn't I put all the blobs of solder on the new CPU, then melt them all at once with some sort of flat soldering iron and apply it to the board? That seems not inanalogous to putting paste on a CPU heatsink. And couldn't a pin adapter be created?

in regards to the soldering, paste is applied, then the whole thing is cooked in a reflow oven, you can do single chip soldering with a heatgun, but it is extremely hard to even get a cpu off a board, let alone clean it all up, re-paste and put a new one in (all that with them being compatible of course

 

[Letalis Sonus]

in regards to the soldering, paste is applied, then the whole thing is cooked in a reflow oven, you can do single chip soldering with a heatgun, but it is extremely hard to even get a cpu off a board, let alone clean it all up, re-paste and put a new one in (all that with them being compatible of course

Don't forget that it's double-packed, trying to do so will likely kill the other BGA, which means one needs super-human BGA soldering abilities!

 

[CPUnltd]

last I remember reading about PS2 emulation, somebody said the reason it was so hard is because the clock of the processor was hard to emulate/multiply... SNES runs at 3.58MHz... something not all that hard to emulate either by exact speed or by multiplied speed (ex. 7.16Mhz @ 2x, 10.74 @ 3x, etc)... with PSX, the speed is 33.8688MHz, but some people have been able to pull that off with either exact speeds or multiplied speeds, hence why so many PSX emulators run a lot of games, but different varieties of games (some games require even more odd-ball clock speeds than others and not every PSX emulator is able to match the clock speed for every game in question.

What the heck are you talking about?

You know what... nevermind what I'm talking about... when I find the thread to back up the information, I'll post it... till then, you won't hear from me on this topic again... period...

 

[Pocak]

I had thought that 3430, 3530 and 3630 would be pin-compatible (if the same package), but I havent actually checked (I dont know if TI even has the info available for others than the 3530).

The DM3730 is meant to be mostly pin-compatible. But, even if it can work with the Pandora's PCB at all (at first sight, MMC1 might be a problem), it would require changing some resistors and upgrading the OS.

 

[Aninhumer]

You know what... nevermind what I'm talking about... when I find the thread to back up the information, I'll post it... till then, you won't hear from me on this topic again... period...

I don't know what you've read, but multipliers have absolutely nothing to do with emulation difficulty.

 

[urjaman]

at first sight, MMC1 might be a problem

IIRC (from reading the pinmux map in u-boot sources) we only use the 4-bit interface in the MMC/SD slots. So a minor board component revision and a new kernel would be needed to support a DM3730 instead of the 3530. Hadnt even heard of that before btw

 

[Exophase]

DM3730 is on Beagleboard-XM. Part of me is hoping that Ari64 gets one and compares the performance of his recompiler vs on the Beagleboard, with CPU clock speed kept the same. I want to see what effect the move to 32/32KB L1 cache has, as well as the move to a faster external memory interface (would also be good to test this at the same speed and see if they've otherwise improved latencies)

 

[ari64]

DM3730 is on Beagleboard-XM. Part of me is hoping that Ari64 gets one and compares the performance of his recompiler vs on the Beagleboard, with CPU clock speed kept the same. I want to see what effect the move to 32/32KB L1 cache has, as well as the move to a faster external memory interface (would also be good to test this at the same speed and see if they've otherwise improved latencies)

My beagleboard is a rev C3 - It has a 3530 like the Pandora. But if you want to send me a Beagleboard-XM, I'll benchmark it.

 

[Exophase]

I don't have one to send, unfortunately. It's "only" $180 at Digi-key; do you think it'd be useful for you to have one? We could probably pretty easily raise that kind of money if you do. But I doubt we can just to satisfy my curiosity, unless I give you the whole amount

 

[darkblu]

I don't have one to send, unfortunately. It's "only" $180 at Digi-key; do you think it'd be useful for you to have one? We could probably pretty easily raise that kind of money if you do. But I doubt we can just to satisfy my curiosity, unless I give you the whole amount

I have a 32KB L1D/L1I-equipped A8, if something needs to be benchmarked.