New PS3 Cell Tech Specs.

[Seb]

Although not confirmed by Sony, some of the technical aspects of the design are emerging in excerpts from papers to be released today for the 2005 International Solid-State Circuits Conference as well as in patent filings.

See: PS3 Cell News

Key points:

> The processor leverages a multicore 64-bit Power architecture with an embedded streaming processor, high-speed I/O, SRAM and dynamic multiplier.

> The Cell architecture rests on two concepts: the "apulet," a bundle comprising a data object and the code necessary to perform an action upon it; and the "processing element," a hierarchical bundle of control and streaming processor resources that can execute any apulet at any time. The apulets appear to be completely portable among the processing elements in a system, so that tasks can be doled out dynamically by assigning a waiting apulet to an available processing element.

> Giving scale to the performance targets for the project, one of the ISSCC papers puts the performance of the streaming-processor SRAM at 4.8 GHz. This suggests the data transfer rate for 128-bit words across the local bus within the processing element.

> Each processing element comprises a Power-architecture 64-bit RISC CPU, a highly sophisticated direct-memory access controller and up to eight identical streaming processors.

[Munky]

Sounds interesting. Hopefully it performs as well as they say.

[LegendaryArcturus]

Or better.

[PeanutButterMunky]

February 2005...I can't wait!! =)

[knowitman]

What is in February 2005?

[PeanutButterMunky]

It's the official release of the PS3 specs, along with various other information.

[FmanFORPS3]

seriously.....its pretty close

[watashi]

I can't wait...there will be alot of people going crazy...although ALOT of Americans that have a PS2 say they regret ever getting one...so maybe it'll be the opposite when they get an Xbox while we get the best system again...that's saying as if everyone is gonna get a PS3...

[gaming ultima]

here a nice pic from one of the presentations

oh and considering that the prototype cell workstation calc 2TFLOPS and by the time its into its final product stages it will be able to calc 16TFLOPS, which makes the ps3 to be around 1TFLOP seem reasonable

[Blake]

although ALOT of Americans that have a PS2 say they regret ever getting one...

I have no clue what people you are hearing say this but Im pretty sure most people who own a PS2 are satisfied with they're purchase. Same would go with Xbox and Gamecube though.

[Seb]

oh and considering that the prototype cell workstation calc 2TFLOPS and by the time its into its final product stages it will be able to calc 16TFLOPS, which makes the ps3 to be around 1TFLOP seem reasonable

According Paul Zimmons, a PhD graduate in computer science from the University of North Carolina;

"I believe that while theoretically having a large number of transistors enables teraflops-class performance, the PS3 [Playstation 3] will not be able to deliver this kind of power to the consumer," he wrote in response to an e-mail query from EE Times. "The PS3 memory is rumored to be able to transfer around 100 Gbytes/second, which would mean it could process new data at roughly 25 Gflops (at 32 bits) — far from the 1-Tflops number."

....interesting comments.

[john kordic]

According Paul Zimmons, a PhD graduate in computer science from the University of North Carolina;

"I believe that while theoretically having a large number of transistors enables teraflops-class performance, the PS3 [Playstation 3] will not be able to deliver this kind of power to the consumer," he wrote in response to an e-mail query from EE Times. "The PS3 memory is rumored to be able to transfer around 100 Gbytes/second, which would mean it could process new data at roughly 25 Gflops (at 32 bits) — far from the 1-Tflops number."

....interesting comments

Hi! there my first post on these boards, long time lurker first time user

I'd just like to know how many Gflops those current high end PC's can achieve in comparison to the 25 Gflops estimated for the Ps3.

[siren]

If memory serves me right the nVidia 6800's do about 20 - 25gflops.

Now keep in mind that this is the CPU. Theoretically the GPU will be quite a bit faster.

The biggest question is whether it will be all raw power, or if they will include some advanced functionality and features that will make it easier to develop for.

[PeanutButterMunky]

And we still have to wait for the official specs. I know Paul Zimmons is a very intelligent individual. But this new CELL architecture is like nothing anybody has seen before. And Sony has been quite secretive up until this point, and will be completely OPEN in February =)

oh and considering that the prototype cell workstation calc 2TFLOPS and by the time its into its final product stages it will be able to calc 16TFLOPS, which makes the ps3 to be around 1TFLOP seem reasonable

Yes, but the 16 Teraflops comes from a "rack" of CELLs, and they didn't specify how many. So there is no way to tell at this point. All in all...I'm very excited about the next generation. The next Microsoft console included =)

[watashi]

no, on the actual SCEA forums many of them, I read, said they should have never bought the PS2...most of them say this due to the lack of support for the HDD in America. Also some of them are moving the Xbox because of the DOA...

[PeanutButterMunky]

no, on the actual SCEA forums many of them, I read, said they should have never bought the PS2...most of them say this due to the lack of support for the HDD in America. Also some of them are moving the Xbox because of the DOA...

What did that have to do with ANYTHING?

EDIT: Oh...you're replying to Blake. Well I disagree with you as well. Look at PS2's sales figures. Everyone I know is satisfied with their PS2. And most of my friends still only have a PS2, or have the most games for the system, and hardly any for the others.

There's no way I would move to the Xbox because of DoA. That's a ridiculous reason and has no basis for anything. And by many do you mean very few in comparison to the people who aren't annoying little whiners? Nothing is perfect.

[knowitman]

I only have ONE friend getting an XBox for DOA, but he still has a PS2. I know no single person that was dissatisfied with their PS2. Most people join things to complain not praise. If everyone that liked their PS2 came to the forums and said that it would FAR out number the people that were dissatisfied.

[The_One]

oh and considering that the prototype cell workstation calc 2TFLOPS and by the time its into its final product stages it will be able to calc 16TFLOPS, which makes the ps3 to be around 1TFLOP seem reasonable

According Paul Zimmons, a PhD graduate in computer science from the University of North Carolina;

"I believe that while theoretically having a large number of transistors enables teraflops-class performance, the PS3 [Playstation 3] will not be able to deliver this kind of power to the consumer," he wrote in response to an e-mail query from EE Times. "The PS3 memory is rumored to be able to transfer around 100 Gbytes/second, which would mean it could process new data at roughly 25 Gflops (at 32 bits) — far from the 1-Tflops number."

....interesting comments.

I don't want to bad mouth him... but unfortunately, his calculations were more then a "little" off.
Reasons:
1. Memory bandwidth is NOT a big factor in FLOp/s performance, unless you're getting chocked... (e.g.: Trying to buffer a 256MB file into a mere 128MB of RAM)

2. Theoretically speaking, the multi-threading capbilities should more then make up for it's poor single thread FLOp/s performance. It should handle up to 10-15 threads like a breeze. Just for comparison, a single P4 with SMT can only handle up to 4 threads, after that, you're pretty much just pushing it.

3. Just for the sake of argument, the PS3 doesn't have enough RAM, and it chokes. The eDRAM should make up for this. You can buffer loads of stuff into the eDRAM for fast access. eDRAM can pretty much be used as a cheap cache. Thus, his memory bandwidth argument is overthrown.

4. Taking the the formula from number 2, let's just SAY that some wacked out dev is trying to make a single threaded game to run on the CELL. Now, according to the Register, 8PE's at 4Ghz were pulling 2TFLOp/s (the whole RACK crunches 16TFLOp/s, but a single rackmounted BOX process only 2 TFLOp/s). Now, we divide 2 TFLOp/s by 15 to get the individual performance PER thread... what do we get? 66.66...GFLOp/s. There you have it folks, THE lowest possible FLOp/s rating (with 8 PE's at 4Ghz) even processing at single thread would be ~67GFLOp/s... that's approx 2.5 times more then his mere 25GFLOp/s "guesstimate".

5. On the other hand... if ONE single CELL has only 4PE's (Which is more then likely), then I'd have to say he's pretty close, since half of ~66GFLOp/s is ~33GFLOp/s, still 8 GFLOp/s more... But you get the idea .

6. Then again, we HAVE to take into account CELL's multi-threading capabilities, NO ONE is stupid enough to NOT use it's multi-threading capabilities to it's fullest extent. Thus, the advantage given by multi-threading should make up for it's poor single thread performance...

Have fun with these numbers, Cheers!

-Your friendly neighbor, The_One .

Oh yes... if anyone notices ANY error with my calculation(s), please do tell me :P.

[john kordic]

2. Theoretically speaking, the multi-threading capbilities should more then make up for it's poor single thread FLOp/s performance. It should handle up to 10-15 threads like a breeze. Just for comparison, a single P4 with SMT can only handle up to 4 threads, after that, you're pretty much just pushing it.

3. Just for the sake of argument, the PS3 doesn't have enough RAM, and it chokes. The eDRAM should make up for this. You can buffer loads of stuff into the eDRAM for fast access. eDRAM can pretty much be used as a cheap cache. Thus, his memory bandwidth argument is overthrown.

4. Taking the the formula from number 2, let's just SAY that some wacked out dev is trying to make a single threaded game to run on the CELL. Now, according to the Register, 8PE's at 4Ghz were pulling 2TFLOp/s (the whole RACK crunches 16TFLOp/s, but a single rackmounted BOX process only 2 TFLOp/s). Now, we divide 2 TFLOp/s by 15 to get the individual performance PER thread... what do we get? 66.66...GFLOp/s. There you have it folks, THE lowest possible FLOp/s rating (with 8 PE's at 4Ghz) even processing at single thread would be ~67GFLOp/s... that's approx 2.5 times more then his mere 25GFLOp/s "guesstimate".

Ok! but say you use it's muli-threading capabilities (10 or more) how many GFLOP/s are attainable. His there a way to calculate the overall power of this machine. Cause now it looks like a high-end pc with an amazing video card from the future. Is it just gonna be like the last generation console: in 2-3 years Pc's will have reach the graphic capabilities of the ps3 (and xbox2) ?

[The_One]

Ok! but say you use it's muli-threading capabilities (10 or more) how many GFLOP/s are attainable. His there a way to calculate the overall power of this machine. Cause now it looks like a high-end pc with an amazing video card from the future. Is it just gonna be like the last generation console: in 2-3 years Pc's will have reach the graphic capabilities of the ps3 (and xbox2) ?

At FULL potential, according the the patent with 4 PE's, 1 TFLOp/s. However, this is the theoretical max performance, NO processor can ever attain 100% efficiency (as with anything else in life ).

As for the PC reaching the PS3... It should happen within 8 months, since every 8 months, a new GPU core is produced (this is an estimate, it's not always 8 months). If the first core doesn't beat the PS3's GPU, then the second one should... which would mean at the most, a 16 month lead.

[watashi]

Well if you think about it...if you don't have the dis-satisfied customers, what makes you think an xbox is? It's the fact that the HDD players do not like there PS2 anymore *they spent 100 on the dang thing for 3 socom 10 dollar each map* I'd understand them...

get it?

[The_One]

No... I don't get it... You're going totally off topic .

[PeanutButterMunky]

^^^^^^

He's been doing that for a while. He keeps bringing up that same point over and over again. A point that has nothing to do with this thread, or any other topic for that matter.

And did you come from the PSI Next Forums, The_One? :P It's good to see more intelligent posters in here =)

[Blake]

You have to give watashi an "A" for effort though, he really tries guys...nothing wrong in that.

[PeanutButterMunky]

I know...I'll give him that :P