I think I missed somethibg there. So that means that is 32MB is 256000000 bits and transistors count is the same?
Questions about PS2
- Thread starter Liandry
- Start date
Yes. A byte is 8 bits. For the PS2's main RAM, 32 megabytes is 32*1024*1024*8 = 268,435,456 bits large. With DRAM, at one transistor per bit, the transistor count would be about 268M. For the PS2 GPU's EDRAM, at 4 megabytes, the transistor count would be 4*1024*1024*8 ~= 33.5M.
It was the "8 transistors for one bit" that stood out to me, but it looks like the transistor count numbers in your post were fine.
It was the "8 transistors for one bit" that stood out to me, but it looks like the transistor count numbers in your post were fine.
53.5 M transistors total as I understand it.
After looking at pictures of various models of motherboard, I can't find any evidence of a PS2 where CPU+GPU+RAM was combined into a single chip. It looks like they had combined CPU+RAM with separate GPU at the end.
davis.anthony
Veteran
Sony did combine the EE+GS+RAM in to a single chip.
Slim consoles released in 2007 with the SKU number SCPH-7900x were the first PS2's to use this combined chip.
The SCPH-7900x is used for PS2 mainboard trimming due to it using this combined chip.
Slim consoles released in 2007 with the SKU number SCPH-7900x were the first PS2's to use this combined chip.
The SCPH-7900x is used for PS2 mainboard trimming due to it using this combined chip.
I can't find any hard evidence for that, besides a bunch of people repeating something someone else said. I'm looking for either actual pictures of a motherboard, or well researched documentation, like from homebrewers or official Sony docs.
From what I can find, the motherboard used for the SCPH-7900X (790XX is technically more correct, since it ranges from 79001 to 79012) was separated into a CPU+RAM chip and a GPU chip (plus extra chips like the disc drive controller). There is a picture of it's motherboard on the PSXDEV wiki, and the largest chip near the bottom-left is the CPU+RAM (CXD2976GB), and the chip above it is the GPU (CXD2980BGB). I can find no pictures of a motherboard with combined CPU+GPU+RAM, or any reliable docmentation of a chip like that existing. The later SCPH-900XX seems to use the same pair of chips that the SCPH-790XX boards used, and is just a rearranged version of the SCPH-790XX motherboard.
From what I can find, the motherboard used for the SCPH-7900X (790XX is technically more correct, since it ranges from 79001 to 79012) was separated into a CPU+RAM chip and a GPU chip (plus extra chips like the disc drive controller). There is a picture of it's motherboard on the PSXDEV wiki, and the largest chip near the bottom-left is the CPU+RAM (CXD2976GB), and the chip above it is the GPU (CXD2980BGB). I can find no pictures of a motherboard with combined CPU+GPU+RAM, or any reliable docmentation of a chip like that existing. The later SCPH-900XX seems to use the same pair of chips that the SCPH-790XX boards used, and is just a rearranged version of the SCPH-790XX motherboard.
Oh man, I just had a SCPH-9000x torn apart at work last week and I sold it yesterday. I just did a quick search and there aren't any really good pics I could find online, but that board is tiny. I wish I would have know and I could have taken some pics. Here's a thread from another forum that has some, though. I don't know how the board or components differ from the 7900X's, but the 9000X are the slims with the internal power supply.
Panino Manino
Newcomer
I had forgotten that the Super Slim model existes, but looking that wikipedia link, when we get to 2007 the two RAM chips disappear.
If the RAM isn't on the board it can only be on one place.
It might be on the back of the board, though.
VU1 was more like a mesh shader , but it's big limitation was the lack of direct memory fetches. To simulate a fragment shader deferred lighting you could pull the colour/z buffer back into main ram and then pass it through VU1 as a full screen grid of particles or even an image upload.
I've got an OG PS2 and a slimline PS2 here in my collection. If taking them apart and photographing the MB will help I can try and put some time aside to do that?
I red your post again and only now fully understand that. Thank you so much. You actually may be right. GS again is separate from EE, EE now combined with IOP + RAM and SPU2 + RAM, and they are on same platform with RAM chip. When I will get PS2 Super Slim next time I will check what's inside.
Now I have new question. In Picture attached there is VU1 diagram. I know data in PS2's EE goes to GS by 3 Paths. So my question is why there is Path 2 from VIF inside VU1, if data can be moved through Path 3 outside VU1 and through Path 1 inside VU1. Why to move data inside VU1 and without any reach to VU1 core move it already through Path 2? There is also possible to use only 1 Path at a time.
Attachments
davis.anthony
Veteran
Because it saves bandwidth and allows developers (with the proper syncing of busses) to maximise utilisation of the hardware.
davis.anthony
Veteran
You save bandwidth on Path 3 by using the other paths, so there's more texture bandwidth.
Using all 3 paths to send data to GS was the key to getting good results on PS2.
Just chucking everything over one or two of the paths limits bandwidth, spread it across all 3 and you have more bandwidth.
For example, you could send texture and vertex data over path 3.
Or you send the vertex data over path 2 which would then free up bandwidth on path 3 for more textures.