If PS3 can really do 1Tflops

[Paul]

I agree, 1Tflop.. I don't think anyone here really understands the true power of that.

Not to mention the GS3 will be vastly better than the PS2's GS. Back in 99 Sony had a whole well kinda like a roadmap of what was to come.

Phase 1 was the PS2 which was 10X that of workstation cards at the time.

Phase 2 Never really came to light but Im thinking it was the GScube.

Phase 3 They said that there would be a 1000X leap over Phase one, which was the PS2. They said at this stage PS3 would be taking form.


The Phase 3 workstation, scheduled to appear around 2005-2006, will have the Emotion Engine 3 and Graphic Synthesizer 3, which will have drastically changed architectures. As a result, the workstation will have 1,000 times performance of the development tool workstation, and will handle 4,000 x 2,000-pixel pictures at 24-to-120p. At this stage, Playstation3 will be taking a shape, Sony said.

http://216.239.37.100/search?q=cach...+Sony+Phase+"emotion+engine+2"&hl=en&ie=UTF-8
[/i]

 

[Panajev2001a]

that roadmap was quite scrapped and re-done when Cell entered the picture...

 

[Paul]

No, Cell IS Emotion Engine 3.

Cell isn't a chip, it's a design and architecture, it won't be named Cell when it comes out. It will either be named Emotion Engine 3 or Broadband Engine.

 

[Vince]

Actually, as Panajev knows his stuff. Also, he knows, as recently as 2001 has Sony's plans been changed in a positive manner thanks to increased technological progress. Just a bit diffrent.

 

[Paul]

Sony's phase 3 STILL stands. They touted 1000X over PS3, and they still do now.

 

[Vince]

Sony's phase 3 STILL stands. They touted 1000X over PS3, and they still do now.

Howmuch do you wanta wager that they're plans have changed from that chart (circa 1999) to today, 2002. As recently 1.5 years ago they made the transition in design rule from 100nm to 65nm or smaller - thereby affecting the chips capabilities and capacities.

 

[Paul]

So your saying PS3 will be more than a 1000X leap over the PS2?

1000X the power is what Sony has been claiming all along, and I have no question they will get it.

 

[Paul]

They claimed 1000X in 99 and they still claim it today, end of story.

 

[Vince]

They claimed 1000X in 99 and they still claim it today, end of story.

Hehe, I love this mentality. The architecture has changed, believe me, lets not play linguistic games.

 

[Paul]

Who cares if the architecture has changed, the claim on performance hasn't and that is WHAT matters.

The 'Cell' Won't be called that I Hope you know, when it comes out. It's just a design name.

 

[Vince]

Who cares if the architecture has changed, the claim on performance hasn't and that is WHAT matters.

Haha... I think you haven't visited the 3D Technology forum lately

Where's Joe DeFuria and Hellbinder when you need him?

 

[Paul]

Im not denouncing the PS3's powers belive me, infact I even have a PS3 wallpaper I made Im about the biggest PS3 geek you'll find.

However, Sony has claimed 1000X the power all along, and It's obvious a architecture change didn't effect the performance of the thing.

Although the internernal workings DID change, i'll give you that. But the projected outlook on the power didn't.

 

[Vince]

Hehe, I know. When I saw your comment about the inner working not being a factor, just preformance - I had to say something tongue-deeply-in-cheek inlight of Joe's crusade against the exact statement you just said.

 

[zidane1strife]

those CGI movies are made with thousands of CPUs running for months or years in non-realtime. even if you had a thousand CPUs, you still couldnt render those movies in realtime. because it takes them so much time (non-realtime) to begin with. and even the more efficient and hardwired GPUs couldnt do it in realtime either. although they could help speed up the non-realtime creation of those movies.

Don't those farms have tons of latency, and processors not dedicated primarily to realtime gphx?

Do the cpus used in those farms have the pixel functionality of modern gpus? I dunno, but if they don't...

Anyways effects don't have to be the same, hacks and tricks have always been around, and equally good looking effects can be achieved through trickery...

I believe it will be done, just look at today's games, and imagine an inmense jump in texture quality, geometry, shader effects, and their precision... 100 fold GFX... and that fairy did look close to the FFX movie at least to the average eye it does, models like that one or beyond will be possible next gen, with artistry, and technical wizardry, they will look pretty close to cgi, most consumers won't be able to tell the difference, and that's what matters.

Anyways, the ps2 made quite some jumps in it's last stages of dev., It exceeds the announced 100xpsone perf. if the figures about the psone perf. I've read here are true... so we can't entirely dismiss the possibily of perf that far exceeds what has been announced... for example say sony decides to delay till 2006, or something like that...

EDITED:

PS: I remember reading in that next gen magazine how they expected the ps2 would push around 5Mpolys ingame peak... no one expected what happened at the celebration of the 50th million psone sold... it appears many were conservatives saying they won't achieve it, this will just push a few M polys

 

[Sonic]

Paul, the chip in the PS3 will be called whatever IBM wants it to be called.

 

[randycat99]

I'll call it Marklar.

 

[Panajev2001a]

Sony plans did change... We were supposed to have seen at least two phases, one of them was supposed to see the EE2 ( never came out )...

What we saw instead was the GScube with GS I-32 and the GScube with GS I-64 ( archie mentioned GScube-64, but I think he meant a GScube with 64 MB of e-DRAM for the GS... using 64 "PS2 blades" instead of 16 would have been impressive as well though )

SCE's R&D proceeded from the first plans using the two GScubes, but we have not really seen even the real Phase 1 Workstation
( unless you call the GScube projects phase 1 ) let alone phase 2...

No, all that R&D hasn't gone to waste... all that research in parallel rendering was used and is still being used in the Cell project... ( including the GS 1.5 or GS I-32 with 32 MB of e-DRAM and 4x as wide busses )

But the Cell project kinda changed the plans and so have the new research on state of the art manufacturing processes...

A manufacturing process like .10um was thought to be used for the EE3, the alliance with IBM and the well proceeding of the research with Toshiba did change Sony's plans positively... they didn't do a 180, they were already looking in that direction...

Actual estimations for the so called Broadband Engine state 65 nm or lower ( 65 nm initially and 45-50 nm should follow for more "ideal" Broadband Engine and Visualizer implementation [yelds, heat, etc...] ) this shift of gears have pushed Sony's plans into more ambitious territory ( also thanks to the help of IBM and RAMBUS technologies, now both partners in the PlayStation 3 project )...

1,000x the "performance" was the goal all along... that will mean not necessarily 1,000x the RAW specs... as with 1,000x the RAW specs ( 6.2 TFLOPS ) we would be able to pass 1,000x the "rendering speed" of PS2...

The technology to realize it has... that was my point... the Broadband Engine is NOT the EE3 as we had no real EE2... Still Sony has said, when they originally first mentioned the roadmap, that EE3 would have had an original architecture... A bit like calling the next IPF processor Pentium VI Still they might even do that, heck they moved into RISC-ier territory since the days of the Pentium Pro ( actually they started with the 486 and a bit earlier, but the Pentium Pro was such a huge step in that direction... )...

You ended up agreeing on this issue as you were not really making a different point... we didn't comment on the "target" performance, but the way ( in this case more reliable ) of achieving it

 

[jvd]

1000x faster at cooking eggs on its heatsink ...

I think all the cell fans will be disapointed. Cell could end up never passing the 1ghz clock. It may never hit 1 tflop. looking 6 months down the road at hardware is a hard thing to do ... look at nvidia. They can tell you all about it . Lets wait till a year before the system is set to come out to talk about it .


Btw 1tflop is alot of speed and power. But 1tflop does not equal 100s of pcs working on a scene for weeks at a time . Mabye just mabye it can do the first toy story with cheats. But there is no way it can do monsters inc . In a few short months the big movie companys will start to use even faster chips . Athlon 64s , new p4s , new ithiums and new video cards that can start to help with the load , r300 and nv30s . Then one day they will be able to use cell chips and use 100s of them for days at time to do one scene . Every day the target gets further and although realtime makes huge leaps , prerendered gets to use all those leaps to stay ahead.

 

[randycat99]

Btw 1tflop is alot of speed and power. But 1tflop does not equal 100s of pcs...

Actually, 1 TFLOP would be about 1 hundred 3 Ghz P4's. If one P4 can run the hell outta Doom 3, I don't think a lot of people will be complaining about a game that utilizes 100x the resources of Doom III will look like. So what is there to be disappointed over?

Mind you (if you hadn't read the first post in this topic), this topic isn't about whether or not you think the PS3 will deliver 1 TFLOPs. It is about what can we expect assuming it does.

 

[jvd]

Btw 1tflop is alot of speed and power. But 1tflop does not equal 100s of pcs...

Actually, 1 TFLOP would be about 1 hundred 3 Ghz P4's. If one P4 can run the hell outta Doom 3, I don't think a lot of people will be complaining about how a game that requires 100x the resources of Doom III will look like. So what is there to be disappointed over?

1 p4 would not run the hell outta doom 3. Carmack didn't write a software path for it. 1p4 and a 9700pro run doom 3 okay. But with no extra eye candy ie aniso and fssa .