i really have my doubts about CELL!

[Shifty Geezer]

In the end the EE was owned by a 733 Celeron.

In what way 'owned'? If you set the EE against a 733 Celeron doing a flop intensive task, fully optimized for both, AFAIK EE would win. The Celeron was paired with a more powerful GPU which is why XB outpowers PS2 (on some, not all) tasks.

Considering this IS sony we're talking about, and there have been no true benchmarks of this processor

Except the ISSCC showing.

here's another great quote from 2000:

The best laid schemes o' mice and men gang aft aglay. They hoped for that, and it didn't work. So they're trying again. That's KK's vision and he's pushing it.

I can't see any reason to complain about Cell's performance, with solid performance figures the same as any other chip (a peak GFlop is a peak GFlop after all), based on the lack of uses Sony hoped to implement of a totally different processor.

 

[Shifty Geezer]

To unleash the power of cell programmers will have to write their software pretty close to the cell hardware.

...it is obviously not a good choice for Macs or PCs at the moment. (The point here is that there is no good abstraction (OS and Programming) of the cell architecture that lets everyday programmers use its power for everyday tasks.)

Although I can imagine that Intel or AMD may pick up an idea or two from cell in the future.

Intel are working towards a similar design. Multicore, multithreading is the future. Cell's just first there. Eventually programmers will have to write for this still computer.

Up to now, the investment on technology has been in the hardware, to make the programming easier. The future will see the onus move onto programmers, on all platforms, and the software development tools will become the centre of advancement, IMO

 

[BlueTsunami]

I think it's a valid opinion, just because someone doesn't imediately bow-down in awe to the power of cell doesn't make them a troll.

Does anyone remember the CNN article on the "super-computer" like processing power of the emotion engine? it was a fun read, I can't find it, but I found some others.

"The Emotion Engine can perform 6.2 gigaflops (6.2 billion operations) per second, which is about twice as many as a state-of-the-art personal computer like Apple's top-of-the line G4. "
- CNN
http://archives.cnn.com/2000/TECH/computing/10/21/playstation.techno/index.html

In the end the EE was owned by a 733 Celeron.

Considering this IS sony we're talking about, and there have been no true benchmarks of this processor, and all the info has been released by Sony or IBM, which are in effect salesmen trying to sell a product, I see every reason to be skeptical of the "power" of CELL.

here's another great quote from 2000:
"SCEI offered no more details on the plan, disclosed at a Tokyo press conference on Thursday, except to say that it plans to sell the PS2 chipset -- the brain of the PS2 machine -- externally in the future. The plan could see a range of consumer electronics products, including televisions and set-top boxes, doubling as PS2 consoles in the future."
-CNN
http://archives.cnn.com/2000/TECH/computing/06/05/ps2.chips.idg/index.html

lol.....history repeating itself?

The XCPU didn't trump the EE in any way.

The XCPU+NV2A however was a more powerful configuration than the EE+GS, and it bloddy well should have been, after 18 months.

So, now that MS are actually releasing earlier than Sony, history will not repeat itself.

Now go back to your cave.

Once I saw the name Scooby_Dooby.....and started reading the post....I stoped around I think it's a valid opinion. I don't know why Scooby keeps posting about SCE stuff when we all already know where his loyalties lie.

About the expectation of CELL, the CELL hasn't even really been implimented in anything yet. IBM is going to release CELL Based Servers...but what you have to realize (if my understanding of CELL is right), the more CELLs that are in use...the more powerful it becomes. So basically wait and see...

 

[scooby_dooby]

The XCPU+NV2A however was a more powerful configuration (in most things) than the EE+GS, and it bloddy well should have been, after 18 months. Oh and double the memory with double the main bandwidth helped too.

So you just ignore the fact that in 2001 the Celeron was 2 year old technology?

The XBOX had many advantages to PS2 other than the COU, but the point is the EE wasn't everything they hypoed it upo to be, and it certainly wasn't more powerful than the top of the line G4's like they claimed.

Now, go back to your high horse.

 

[Kalin]

Erm, no, it's not Sony. It's Sony, Toshiba and IBM. Can't overlook that.

Besides, as we all know, they're hard at work on bringing CELL outside of PS3 a full year before the console is out (talking about the CELL workstation to be demoed on LinuxTag2005).

There's really no comparison between this and EE, so please let's hold judgment for now and see how it turns out.

 

[scooby_dooby]

Except the ISSCC showing.

you mean the paper released by IBM?

i've read the ISCC article, a few times, IBM "claims" to be able to proccess 250 G-Flops at 4.0Ghz...in an imaginary scenario with 8 SPE's running at 100%

I'm skeptical, it's for good reason, if you look at PS2 now and compared to what sony said about it at the time, it was nowhere near as powerful as they claimed.

Sorry if this time I don't swallow the whole hook, line ands sinker.

 

[London Geezer]

The XCPU+NV2A however was a more powerful configuration (in most things) than the EE+GS, and it bloddy well should have been, after 18 months. Oh and double the memory with double the main bandwidth helped too.

So you just ignore the fact that in 2001 the Celeron was 2 year old technology?

The XBOX had many advantages to PS2 other than the COU, but the point is the EE wasn't everything they hypoed it upo to be, and it certainly wasn't more powerful than the top of the line G4's like they claimed.

Now, go back to your high horse.

No, MS had 18 months to get a machine that could perform better than PS2 cause if they used PC parts released when PS2 was released, they'd have been trumped by Sony's technology.

The EE was hyped up to be what it was, a math cruncher. It did its job mighty fine, and it's only bloody normal if a platform released 18 months later is more powerful.

Do we have to have this kinds of conversations again after 5 years?

History will not repeat itself simply because this generation is completely different from the last one. The players are using different partners, the time releases are not comparable, so your arguments are useless.

Happy?

 

[nAo]

You guys are trolling cause you keep saying EE can't do that..CELL can't do this..but you can't say what or why.
This is a technical board, have you noticed?
I'll be happy to discuss with you when there will be something to discuss, otherwise just use the big search button.

 

[Fafalada]

So you just ignore the fact that in 2001 the Celeron was 2 year old technology?

So you just ignore the fact that R5900 was technology a lot older then 2 years in 1999?

 

[London Geezer]

Except the ISSCC showing.

you mean the paper released by IBM?

i've read the ISCC article, a few times, IBM "claims" to be able to proccess 250 G-Flops at 4.0Ghz...

I'm skeptical, it's for good reason, if you look at PS2 now and compared to what sony said about it at the time, it was nowhere near as powerful as they claimed.

Sorry if this time I don't swallow the whole hook, line ands sinker.

I smell disappointed fanboi who bought into the INTERNET hype (Not Sony hype).

EE was "hyped" as a processor with a 6.2GFLOP performance, which it had, whose main advantage was having 2 vector units, which it had and did a very good job if programmed properly.
What exactly is the problem there?

 

[N00b]

Intel are working towards a similar design. Multicore, multithreading is the future. Cell's just first there. Eventually programmers will have to write for this still computer.

Up to now, the investment on technology has been in the hardware, to make the programming easier. The future will see the onus move onto programmers, on all platforms, and the software development tools will become the centre of advancement, IMO

The difference is that with Intel's and AMD's multicore cpus each core is a general-purpose core. If you run multiple programs you can transparently profit from a multicore without changing a single line of code. BTW multiprocessor systems are really an ooold technology. Multicore is just an optimisation.

Also, I would not call cell a multicore cpu. IMHO, it is really a single core general purpose cpu with 8 programmable "MMX" units. Please note that I don't mean MMX literal and I am well aware of the differences between a MMX unit and an SPE.

But I agree with you that software programming tools and languages will evole. I don't think that the programmers themselves will move so much. Humans really like to think single-threaded and Joe Programmer, the average VB.Net programmer, is no different.

 

[BlueTsunami]

Except the ISSCC showing.

you mean the paper released by IBM?

i've read the ISCC article, a few times, IBM "claims" to be able to proccess 250 G-Flops at 4.0Ghz...

I'm skeptical, it's for good reason, if you look at PS2 now and compared to what sony said about it at the time, it was nowhere near as powerful as they claimed.

Sorry if this time I don't swallow the whole hook, line ands sinker.

I smell disappointed fonboy who bought into the INTERNET hype (Not Sony hype).

EE was "hyped" as a processor with a 6.2GFLOP performance, which it had, whose main advantage was having 2 vector units, which it had and did a very good job if programmed properly.
What exactly is the problem there?

You hit the nail on the head L-B, yeah..Sony CAN hype certain things..but if your on message boards often (especially ones where they regurgitate information over and over and it turns out like a game of telephone..when the information begins to change into something different).

I personally take the stance most people have with consoles applied to the CELL processor. You'll never know its REAL WORLD performance UNTIL its actually being used in a REAL WORLD environment. Speculating is fine, accusations of SCE-Toshiba-IBM lying about the performance about the CELL is a failed attempt at trying to downplay their processor....especially if your trying to link to something else...

 

[Shifty Geezer]

i've read the ISCC article, a few times, IBM "claims" to be able to proccess 250 G-Flops at 4.0Ghz...in an imaginary scenario with 8 SPE's running at 100%

If you knew what ISSCC was/is, you would understand what those numbers mean. Just as EE matched everything it's ISSCC paper claimed.

As has been pointed out by others here, as a technical forum you need to bring technical reasons why you doubt Cell's performance, not 'Sony are a bunch of liars. They lied about EE. They always lie. Now IBM are lying.' Save they pro- anti- propaganda for other forums.

 

[Shifty Geezer]

The difference is that with Intel's and AMD's multicore cpus each core is a general-purpose core. If you run multiple programs you can transparently profit from a multicore without changing a single line of code. BTW multiprocessor systems are really an ooold technology. Multicore is just an optimisation.

Current designs are dual-core. Future designs include Cell like asymmetric processor and sub-processors AFAIK. I imagine there'll be different approaches of symmetric and asymmetric processing, but for media (home PCs) I'm expecting Cell-like SPE's on one or more general purpose cores from Intel.

Also, I would not call cell a multicore cpu. IMHO, it is really a single core general purpose cpu with 8 programmable "MMX" units. Please note that I don't mean MMX literal and I am well aware of the differences between a MMX unit and an SPE.

Then you are aware SPE's can do everything a 'normal' processor can? It seems people (mostly 'under MS's influence' given their portrayal of SPEs?) see the SPE's as glorified FPUs. They include a full processor API from what I know. They can branch, loop, access memory...the whole shooting match. They are heavily optimized, but they are full processors. This is in contrast to an FPU or MMX unit that only performs a subset of processor functions. It's unfair to think of them as not full processors - they are specialised processors. That makes Cell an Asymmetric Multi-Processor which we'll be seeing more of, whether devs like it or not!