PS3 Visualizer

[London Geezer]

eh, thats kind of silly, 1000x. would you say PS2 is 1000x better than PS1? and would that mean, that PS3 will be 1,000,000x better than PS1!? and PS4 will be a billion times better than PS1??? and PS5 will be a trillion times better???

PS2 is maybe 1000x better than monochrome Pong

well ps2 is not 1000x better than ps1 and NO ONE EVER SAID IT IS....

 

[qwerty2000]

the ps2 is 300x more stronger than the ps1. But the ps3 is excpeted to be 1000x maybe more powerful than the ps2

 

[megadrive0088]

but we've already been over the likelihood that PS3 will not be 1000x PS2 in raw performance. meaning PS3 is not going to do 6.2 TFLOPs which is 1000x PS2's 6.2 GFLOPs.

 

[megadrive0088]

do u even read other people's posts?

my WONDERFUL ( so said megadrive ) post made it clear that whatever the numbers, the result will be the same, something around 1000x better than ps2. even if the polygon throughput is not 1000x more than ps2, all things together will make it that fast.

just read my WONDERFUL post will u...

yeah my thoughts are pretty much the same or similar. I dont think qwerty is fully understanding / reading what is being said here. that's not a flame. but one needs to read a bit more carefully.



qwerty this is what is being said, simply:

1. Sony claimed PS3 will be 1000x the performance of PS2 back in 1999

2. PS3 will probably NOT have 1000x the raw performance of PS2.

3. PS3 will LIKELY have 100~300x the raw performance of PS2

here's the important part
4. PS3's increased efficiency per GFLOP (usable sustained power) compared to PS2 plus much much better integer power, eDRAM/caches, programable shaders, gfx features, high res textures, bump-mapping, FSAA, new rendering methods (procedural stuff) screen resolution and other things that PS2 didnt have, all combined with the 100~300x raw performance increase will probably give something like a 1000x improvement over PS2.

 

[Paul]

PS4 CPU! Over 100TFLOPS power!

Of course this assumes the leap of CPU power between the EE and the supposed Cell 1tflops remains the same between the leap of ps3 to ps4.

 

[qwerty2000]

do u even read other people's posts?

my WONDERFUL ( so said megadrive ) post made it clear that whatever the numbers, the result will be the same, something around 1000x better than ps2. even if the polygon throughput is not 1000x more than ps2, all things together will make it that fast.

just read my WONDERFUL post will u...

yeah my thoughts are pretty much the same or similar. I dont think qwerty is fully understanding / reading what is being said here. that's not a flame. but one needs to read a bit more carefully.



qwerty this is what is being said, simply:

1. Sony claimed PS3 will be 1000x the performance of PS2 back in 1999

2. PS3 will probably NOT have 1000x the raw performance of PS2.

3. PS3 will LIKELY have 100~300x the raw performance of PS2

here's the important part
4. PS3's increased efficiency per GFLOP (usable sustained power) compared to PS2 plus much much better integer power, eDRAM/caches, programable shaders, gfx features, high res textures, bump-mapping, FSAA, new rendering methods (procedural stuff) screen resolution and other things that PS2 didnt have, all combined with the 100~300x raw performance increase will probably give something like a 1000x improvement over PS2.

i know 2-3 the i dont expect the ps3 to have 1000x times raw perfomce of the ps2 but it's possible to have the 300x time raw polygon power than the ps2

 

[megadrive0088]

more or less, yeah, I suppose so.

 

[Gunhead]

Each 128 SIMD instruction could produce 8 FP ops/clock instead of the 4 FP ops/clock they have in the article...

I don't understand, if they are operating on 8 32bit chunks of data, that's 256 bits, not 128.

They are still operating on 4 32bit chunks of data (a 128bit vector), but doing two ops (multiply + add) to each chunk in the vector.

FMAC/FADD units are funny that way -- they do two ops per cycle to the same data

 

[V3]

They are still operating on 4 32bit chunks of data (a 128bit vector), but doing two ops (multiply + add) to each chunk in the vector.

FMAC/FADD units are funny that way -- they do two ops per cycle to the same data

 

[megadrive0088]

Regarding the Visualizer's Pixel Engines and Nvidia - I was thinking that maybe each Pixel Engine will be a complete NV40 or NV45 rasterizer part/pixel pipelines w/pixel shaders. without the vertex shaders of course since the APUs/PUs will be doing that part.

If using today's Nvidia GPUs, each Pixel Engine of Visulizer would be Nvidia's NV35 minus the vertex shader/vertex engine. Each Pixel Engine has a 4:2 or 8:1 configuration, in NV35's case.

Obviously the Pixel Engines of Visulizer will be somewhat newer than NV35 technology if Nvidia is indeed doing the rasterizing part of PS3's Visulizer.

So whatever the NV40 or NV45 rasterizer/pixel pipeline config is, would be each Pixel Engine of Visulizer. Understand what I am trying to suggest here? that each Pixel Engine is not one pipeline, but the full (customised of course) rasterizing portion of an Nvidia NV4X GPU. or posssibly even NV5X depending on PS3's release. so i'm not saying 4 Nvidia GPUs in PS3, but 4 rasterizers in Visualizer. still a Sony GPU.

If PS3's GPU has 4 Pixel Engines, and Nvidia's contribution is based on NV4X/NV5X. if NV40, say 8:2 (8 pipes, 2 TMUs per pipe) we would have

8:2 - 8:2 - 8:2 - 8:2 <--- PS3 Visualizer's 4 Pixel Engines
or
16:1 - 16:1 - 16:1 - 16:1
or some other configuation


hey I realize it probably won't be quite like this. just a thought I had the other day.

 

[megadrive0088]

*RRV annoucer's voice*

Waal, what do ya think?