is the vdp1 32bits
- Thread starter Coolgame
- Start date
Chilly Willy
Established Member
It makes no sense to talk about how many "bits" a video processor is. These are not CPUs. You normally talk about the fill rate under various circumstances - how many pixels can be processed and stored in a given time. VDPs may have really wide buses to the vram to give a high fill rate. The max fill rate for VDP1 is specified as 400,000 pixels per frame (NTSC), or 24,000,000 pixels per second. That's under ideal conditions where you're just filling the frame buffer. The fill rate drops for various things like lighting and shading.
Chilly Willy said:
i know what the fill rate means and the theoretical fill rate of the vdp 1 (400,000 per 1/60sec).
reading the documents is what got me asking the questings in the first place.
but the thought put (how many bits the processor process at a time) can put the gpu at just as much of an advantage as it would the cpu.
one example is, if the vdp1 runs at 32bits that could put the pixel fill rate at 48 mega pixels per sec. get me?
Chilly Willy
Established Member
Coolgame said:
You're thinking of GPUs like blitters - if only you double the width of the blit, you increase the fill rate...
It doesn't work that way for GPUs. If you look at modern GPUs, they didn't increase the "bits" for processing, they added more pipes. Each pipe handles pixel processing (or geometry, but we're just concerned with the pixels here). Again, there's no bitness, merely a fill rate for the enabled features. Then to give better results, you improve the efficiency of the pipe, or add more pipes in parallel. Modern video cards have done both to a great extent - pipes are faster and more efficient, and you have a ridiculous number of pipes running in parallel.
The other significant area of speed in a video card is vram bandwidth - all those pipes require moving a LOT of data as fast as possible, so cards will often increase the width of data fetched/stored from/to vram. An el-cheapo card may be 64 bits wide, while a high-end card will be at least 256 bits wide, and maybe as much as 1024 bits wide. The memory itself is pushed as far as it can go, DDR, DDR2, DDR3...
Chilly Willy said:
almost all of what you said is true.
true GPUs today is depended on pipelines, because pipelines is a efficient
way to process video/image data.
true the bits method went the way of the dinosaurs and it doesnt matter now as it did with the era of older console like sega genesis, super nintendo ect...(just about everything as i know is 64bits),but with the sega saturn it's a different story as it's such a primitive console, 32bits for the VDP1 could give it an edge, im not thinking higher bits architecture will make it a graphical monster in today standards, but just think, as we all know
color depth and pixel fill rate is affected by the bits architecture that's what it did for the VDP2.
the VDP2 display 16bits color data two at a time, because it's 32bits, if that's so with the VDP1, then with it's two famebuffers, which the VDP2 can process both at the same time because it's a 32bit chip, its can render to both famebuffer at the same time, at 28mhz and give 4mhz for polygon calculation that should be 48mega pixels or 800,000 in 1/60sec.
get me?
any way I just want to know is there any info on it being a 32bit chip? the VDP1 that is.
Chilly Willy
Established Member
Coolgame said:
Okay, to simplify it as much as possible almost to the point of irrelevance, VDP1 is 16 bit. All data passing through it is a max of 16 bits wide. I could make an argument that it's really a triple five-bit processor due to how each color channel is processed by its own ALU and all hardware ops are therefore only 5 bits, but you probably don't want to hear that.
Chilly Willy said:Okay, to simplify it as much as possible almost to the point of irrelevance, VDP1 is 16 bit. All data passing through it is a max of 16 bits wide. I could make an argument that it's really a triple five-bit processor due to how each color channel is processed by its own ALU and all hardware ops are therefore only 5 bits, but you probably don't want to hear that./>/>
But don't forget, the draw commands are 32bit!
Chilly Willy
Established Member
rorirub said:
True, but irrelevant. It's one of the factors defining a RISC processor - a fixed width opcode. IIRC, the x86 opcode can be up to 96 BYTES in length, but that doesn't make it a 768-bit CPU. The SH2 opcodes are 16 bits wide, but it's a 32-bit RISC chip.
Given the fixed width of the VDP1 commands, the best you can take from that is it's derived from a RISC design on the processing.
