Super Grafx Arcade Card attempt?

[soop]

This is one of my favorite threads - gives me goosebumps

The other two are the 68k/6280 thread and the bubble bobble thread.  Not that anyone cares.

[Bernie]

China Warrior sucks ass. 

[esteban]

China Warrior sucks ass. 

Absolutely. It sucks the cutest, juiciest, sweetest ass in the cosmos. I WANT SOME. 

[touko]

A good SGX tech demo, will be the first level (not entirely) of final fight ..

What's in Final Fight that can't be done as a regular ACD?

I had a discussion with stef (the gens's dev, and other geny tech demos) on the feasibility of FF  ..
And it's very difficult, i think, to make a better version than geny one,even with AC ..
Why ???
PCE or SGX don't have a good dma to put data in vram quickly in vblank, all should be done with CPU .
If you keep the originals arcade sprites, each sprite update is about 3ko of data .
But the size of all sprites sheets must be compressed to fit in AC, and depacked on the fly.

I think is too much for 6280 alone ..

But i don't know how geny version is close to arcade ..

[CrackTiger]

A good SGX tech demo, will be the first level (not entirely) of final fight ..

What's in Final Fight that can't be done as a regular ACD?

I had a discussion with stef (the gens's dev, and other geny tech demos) on the feasibility of FF  ..
And it's very difficult, i think, to make a better version than geny one,even with AC ..
Why ???
PCE or SGX don't have a good dma to put data in vram quickly in vblank, all should be done with CPU .
If you keep the originals arcade sprites, each sprite update is about 3ko of data .
But the size of all sprites sheets must be compressed to fit in AC, and depacked on the fly.

I think is too much for 6280 alone ..

But i don't know how geny version is close to arcade ..

Stef is very talented and knowledgeable when it comes to Mega Drive, but always seems to not be very familar with the PC Engine and always has the same conclusion/answer: MD>PCE. What it sounds like he's describing is what the PCE is supposed to be superior at. If you try to do things the wrong way for PCE, then of course it won't turn out as good as the hardware a particular method is designed for.

We already have Sapphire out performing any other 16-bit console game for number of sprite frames. It's a no-brainer.

The Sega-CD port of Final Fight isn't very good overall for the hardware and is very far off from the arcade. It just happened to be the best console port until the Capcom Collections.

[OldMan]

PCE or SGX don't have a good dma to put data in vram quickly in vblank, all should be done with CPU .

No, but if the data is in a spare area of VRAM, it can be dma'd to where it needs to be quickly. Which allows you to use the rest of the frame to refill the transfer area....

If you keep the originals arcade sprites, each sprite update is about 3ko of data .
But the size of all sprites sheets must be compressed to fit in AC, and depacked on the fly.

Why must it be compressed?  The Arcade card has more memory than the system 3.0 card (iirc). In any case, can't it be loaded into adpcm memory and sent from there? (64K adpcm memory + direct to vram writing from adpcm.... maybe not dma speed, but very close, I'd bet.)

Honestly, pushing 3K of data per frame sounds doable to me....

[touko]

@blacktiger: yes you're right, i did some tests, and i can refresh 5 sprites (arcade sized) without any optimisations, and with a sprites animation @30fps,and it uses 50% of CPU cycles .
But if MD has all the arcade patterns (however i doubt), it's impossible to fit 500 ko/caracters without any very good compression,even with AC ..

The Sega-CD port of Final Fight isn't very good overall for the hardware and is very far off from the arcade. It just happened to be the best console port until the Capcom Collections.

And i doubt heavily that capcom  was able to fit all arcade animations in 768 ko of cache .

No, but if the data is in a spare area of VRAM, it can be dma'd to where it needs to be quickly. Which allows you to use the rest of the frame to refill the transfer area....

There are too many datas, and of course you need to have a cache in vram of each caracter.
in my tests 50% is for a transfer of half size of data/ frame for a 30fps anim.

Why must it be compressed?  The Arcade card has more memory than the system 3.0 card (iirc). In any case, can't it be loaded into adpcm memory and sent from there? (64K adpcm memory + direct to vram writing from adpcm.... maybe not dma speed, but very close, I'd bet.)

Honestly, pushing 3K of data per frame sounds doable to me....

In fact the problem is not transfer, is doable,it's all sprites patterns .
For exemple,the sprite sheet of axel, his size in arcade version is about 500 ko ..
For main caracters (cody,guy haggar), is 3x larger, stef told me that md version has all the coin op animations ...