SNES Overclocking: Good speed increase?

I remember years ago seeing scematics for overclocking the SNES. Found them thanks to the wayback machine! I also managed to find the crystals for it on ebay for a relativly cheap price, 99p, a few years ago they were about £50, and there's still a seller with an xtal at that price, pah.

So yeh, I've got both PAL and NTSC SNES's, and have a game doctor on my PAL one (really not designed for a NTSC SNES; the ridge in the middle is the only bit holding it up and it can move backwards and forwards ). So they're both region-free and 50/60 modded, one based on a 505-timer, the other a reset-button PIC that I coded.

So my question is this; as the homebrew super mario world games can go so slow during some parts where there's lots of sprites on the screen, will upping the CPU speed to 4.16Mhz help with these, or will it just make everything run way too fast like the gameboy advance CPU speed hacks do?

 

it'll be too fast and i think it would mess sound too...

try and let us know

 

If you did overclock you wouldn't overclock the PPUs or APU, so no, everything wouldn't run faster, a frame of game logic would just complete faster. Programs however could lose sync. Raster effects could be broken. DMA could exceed RAM speed. Because the SNES uses waitstates and contains an internal divider you'll have to overclock the MASTER CLOCK to the CPU (to 25 MHz, 25 / 6 = 4.166. 25 MHz is one of the MOST COMMON frequencies), while still feeding the PPU 21.477273 MHz.

The slow SMW hack wouldn't be any faster in an emulator so perhaps the author should just better optimize their code. If the slowdown has nothing to do with the game logic but the (lack of) time in which to update VRAM, overclocking won't help because games are hardcoded to transfer only so much data per Vblank.

 

I'm not sure what causes the slow-downs with them, but on a screen with quite a few jumping koopas (changed sprites) and things happening, it goes slow.

Yep I got a 25Mhz xtal, also got a 38Mhz one for my dreamcast to see how that improves!

 

This was a dumb idea... Lost the wire I had on the original pin pad, so had to order a 21.606Mhz crystal (closest I could find) and decided to get a 23Mhz one too. Well, at the moment all I have now is 25Mhz (4.16Mhz CPU speed) which is erm... Well, pilotwings the title has some annoying dodgy sprites all over it, suprisinly the actual game isn't all that bad.
Then I tried a mario game. Through my GD7... Wow, flashing sprites, corrupt sprites, you name it. Could be due to the GD7 using slow chips, not too sure. Ah well, shall replace the crystal with a 23Mhz and add a 21Mhz to fix my big mistake and retry!

 

Yeah, overclocking is rarely smart XD

Remember you're only supposed to overclock the CPU, not PPU. So you shouldn't have replaced the crystal, but instead bought a 25 MHz complete oscillator in it's own package, and solder that to the CPU clock pin, lifted.

It shouldn't be surprising it won't work with copiers, copiers have their own oscillators and state machines designed around strict DRAM constraints.

 

That's what I did do, it only connected to the power and CPU pin.

 

OK, even a SLIGHT overclocking (Had to get a 21.676Mhz crystal to replace the screwed up pad) to using a 23.296Mhz crystal (3.882Mhz core CPU clock speed) gives severe GFX glitches on everything.
Even on an overclocked DSP game.

I made a video in case anyone is interested; http://www.youtube.com/watch?v=prs66pabTTg
A shame really. What's weird though is that in gunforce (I didn't record this), it gives loads of graphic glitches, but actually does overall decrease the lag of having many things on the screen at once.