How was Ps2 thought to work ???

ShinHoshi

Newcomer
Concepts like "software rendering", "synthesis programation", "streaming" appear a lot when PS2 development is on talk but I just don't understand what was Ken Kutaragi idea...

You cannot trust on the EE to make a R-T engine and make all textures from equations because there is not enough computational power but on the other hand PS2 doesn't have a powerful T&L like DC, XBOX and GC have...

So what was the idea ??? Was it only the first step ???
 
I used a bad term, yes. But I think that the POWERVR on DC can do more things than the GS on PS2 (built in). Correct me if I am wrong since I am not a huge specialist.
 
PS2s Vectorunits is its T&L. They are by far the most powerful and flexible vertice transformation processors of any console.
If you mean hardwired T&L, well then you right, PS2 doesn't have any of that.
 
Dreamcast had a lot of effects that the PS2 couldn't do without doing it slowly in hardware. Things like anisotropic filtering and unused bump mapping. DC has no T&L unit accept for the SH4, which had to do T&L, AI, physics, game code, and all without the use of extra vertex units like the PS2 has.

PS2 has some very nice tech in it. The EE is a very nice processor coming from where I stand. It might be hard to code for, but the results gotten from it are excellent when things are done right. The only bad thign about the PS2 is the GS, not because it has to do many things in software, but because it has a puny 4 MB of very fast eDRAM. If they stuck it with 8 MN instead I'd think we would see PS2 comparing next to Xbox and Gamecube much better than it does in general.
 
Then tell me how GC gets by with only 1Mb for textures, and xbox with probably under a quarter of that?
 
I used a bad term, yes. But I think that the POWERVR on DC can do more things than the GS on PS2 (built in). Correct me if I am wrong since I am not a huge specialist.


You are right in that the PowerVR2DC / CLX2 in Dreamcast can do more things than the GS in PS2. it's just that the GS is a highly parallel design with many (16) pipelines plus on-chip memory/eDRAM. the DC's PowerVR2 only has a single pipeline and no eDRAM, but it has some features not found on the GS, and access to more VRAM than the GS has if the GS doesnt tap PS2's 32 MB of main memory. that 8 MB for PowerVR2DC combined with texture compression, is part of what gives DC better & more textures. the features of PowerVR2 also. (that Sonic mentioned)

however, the DC has no T&L on board the PowerVR2 graphics chip, as it relies on the SH-4 CPU's massive computational power (for its day) to provide transform & lighting data, or geometry & lighting, or however you want to say it. both the GS and PowerVR2 graphics chips in PS2 and Dreamcast respectively depend on their host CPUs
(EE and SH4 respectively) for T&L. only the Xbox's XGPU and GameCube's Flipper graphics processors have geometry/lighting engines. or what is now called Vertex Shader in Xbox XGPU's case.

Only the arcade NAOMI 2 has a powerful T&L engine, as one of its graphics processors, which provides T&L to twin PowerVR2DC / CLX2 rasterizers (same as the one in DC just that there is two of em)


oh and btw, i am not a huge specialist either :)
 
both the GS and PowerVR2 graphics chips in PS2 and Dreamcast respectively depend on their host CPUs
(EE and SH4 respectively) for T&L. only the Xbox's XGPU and GameCube's Flipper graphics processors have geometry/lighting engines. or what is now called Vertex Shader in Xbox XGPU's case.

Er um... The GS relies on VU1 (and VU0 to a lesser extent) not the "CPU"

Only the arcade NAOMI 2 has a powerful T&L engine,

Well technically VU1 is the same thing (just programmable)...
 
Megadrive1988 said:
You are right in that the PowerVR2DC / CLX2 in Dreamcast can do more things than the GS in PS2. it's just that the GS is a highly parallel design with many (16) pipelines plus on-chip memory/eDRAM. the DC's PowerVR2 only has a single pipeline and no eDRAM, but it has some features not found on the GS, and access to more VRAM than the GS has if the GS doesnt tap PS2's 32 MB of main memory. that 8 MB for PowerVR2DC combined with texture compression, is part of what gives DC better & more textures. the features of PowerVR2 also. (that Sonic mentioned)

The PVR2 must have something equivalent to the eDRAM, for its on-die tile buffer.
Something to bear in mind though, is that DCs VRAM is linked to the PVR2 with a bus only two thirds the bandwidth of the EE to GS bus, and it's over that same bus that framebuffer has to be read and refreshed.

True, it does have texture compression, but PS2 can do multitexturing better than DC, which in turn means that by double texturing it can get even better compression ratio than DCs vector quantization (or S3TC).
 
True, it does have texture compression, but PS2 can do multitexturing better than DC, which in turn means that by double texturing it can get even better compression ratio than DCs vector quantization (or S3TC).

Hmmm...That's a thing I really never will understand. Sure that PS2 has more fillrate and bandwith...which in a rough way means you "paint faster" and change the "colours on your hand faster" (hand=4MB VRAM). But that is always considering that you never surpass the 4MB limit, right ?
If we needed to use a 5MB texture (just suppose that we use a very bad compressed texture), then DC could use it while PS2 couldn't, right ?
 
I don’t think any game has ever used textures larger than a few hundred kilobytes. But in the interest of this hypothetical example, the PS2 could split the texture into smaller bites, and then but them together when rendering (indeed, if you don’t want to impair fillrate seriously, you have to keep to textures under certain size, otherwise the 8kb texture buffer will "overflow" and have to reload all the time).
 
Yes, I know that a 5MB textures is almost an impossible thing...But, let's suppose another case:
Just imagine a frame of an scene in which we need tons of textures. The total size of the textures on the scene reaches 8MB. Is the bandwith/fillrate enough fast to paint half of them, then update th 4Mb buffer and paint the unfilled polygons ?
In other words, how many updates of the framebuffer can you do in 1/60 s (game running at 60 fps) ?
 
Megadrive1988 said:
Only the arcade NAOMI 2 has a powerful T&L engine, as one of its graphics processors, which provides T&L to twin PowerVR2DC / CLX2 rasterizers (same as the one in DC just that there is two of em)

To be specific, Videologic Elan at 100MHz.
 
Reznor007 said:
Megadrive1988 said:
Only the arcade NAOMI 2 has a powerful T&L engine, as one of its graphics processors, which provides T&L to twin PowerVR2DC / CLX2 rasterizers (same as the one in DC just that there is two of em)

To be specific, Videologic Elan at 100MHz.

Also 10 Mpolys/sec with 6 complex lights at only 100MHz ;) The Elan is friggin lighting monster. Imgtec guys are brilliant :oops:

I wanna see a 500 MHz Elan 2 damit!! :devilish:

Oh btw the PVRDC doesn't require a lot of memory bandwidth because it's a TBDR. Comparing it's bandwidth to the GS's is pointless ;)

In other words a motorcycle doesn't require a V8 with 500 hp to do 0-60 in 3 secs...a Viper does ;)
 
Dreamcast had a lot of effects that the PS2 couldn't do without doing it slowly in hardware. Things like anisotropic filtering and unused bump mapping.
DOT3 bump mapping on PS2 (4 passes) vs DOT3 on DC (2 passes) would again see PS2 performing it much faster though, due to sheer difference ein fillrate.
 
Back
Top