Go on the record & predict CELL performance

[Brimstone]

I predict 256 GOPS and 128 GFLOPS.

 

[Urian]

I predict 192GFLOPS and 192 GOPS.

 

[Phil]

CELL is just an architecture...

I would have to think that after months and months of speculation, this should be finally clear to everyone?

 

[Fafalada]

You mean the one at PS3 (in which case, you should clarify) or the one that will be shown at ISSCC?

Anyway, I am feeling nostalgic today, so I will go with 294.9 GFlops/Gops (for broadband engine).
Needless to say, "size" is not why I chose this number. 8) (and it IS a guess, just in case anyone thinks I'm cheating).

 

[version]

256 Gflops, 1 Tops

 

[archie4oz]

Anyway, I am feeling nostalgic today, so I will go with 294.9 GFlops/Gops (for broadband engine).

How about 294.912GFLOPS to be more precise... And a 147.456 *useful* GFLOPs for the visualizer, 'cause NV will be forced to incorporate some boneheaded Sony clipper design that robs the developer of the rest of the calculation resources...

 

[cthellis42]

Now the REAL challenge is to name how many Sony MktgFLOPS and NVFLOPS total the PS3 will achieve.

If science can actually calculate a figure that large.

 

[AlNom]

Pi

 

[Paul]

I predict that Broadband Engine will be a TFLOPS....

If I'm wrong, oh well who cares. Do I think it's likely? No... But I want to be an optimist.

 

[cthellis42]

You mean it will be so blazingly fast that it will be able to FULLY CALCULATE pi?

Wow!

 

[Guden Oden]

42 hey baboreeba and 47 bop-a-lop-abop-bam-boom.

 

[Brimstone]

You mean the one at PS3 (in which case, you should clarify) or the one that will be shown at ISSCC?

Anyway, I am feeling nostalgic today, so I will go with 294.9 GFlops/Gops (for broadband engine).
Needless to say, "size" is not why I chose this number. 8) (and it IS a guess, just in case anyone thinks I'm cheating).

The one that will be shown at ISSCC in a few days.

 

[Shifty Geezer]

You mean a PE - 1 PU and 7/8/9 APUs, the silicon being described at ISSCC.

Um, 256 GOPs, of some sort or other. For PS3 I'd hope they get several PEs on a chip for 512-1024 GOPS.

If they don't, I'm gonna sue Ken Kutaragi for selling me a PS3 on false pretences

 

[Inane_Dork]

1 Megaflop






Get it? Just kidding.

 

[Alejux]

Around 512 GFLOPS/GOPS. I'm basing this on the "knowledge" that the IBM Cell workstation will be able to achieve 16 TFlops, with 32 chips.

Of course, I bet that will be the peak performance. I imagine it will be around half of this speed sustainable.

 

[Titanio]

At the ISSCC: 1 PE, 8 APU, 256GFlops.

Probably the same in PS3. That'd an awful lot of power by anyone's measure, and probably enough to make PS3's CPU the most powerful of the next gen systems (so why any more?).

 

[London Geezer]

1 Megaflop






Get it? Just kidding.

That would win the award for Most Stupid Joke That Made You Laugh Out Loud.

 

[Brimstone]

I predict that Broadband Engine will be a TFLOPS....

If I'm wrong, oh well who cares. Do I think it's likely? No... But I want to be an optimist.

Once the die size of 1 PE is revealed the real chances of a 4 PE Broadband CPU for the PS3 should become a lot clearer. It could turn out that STI has more streamlined design than what people here have estimated. Sadly though, this information might not be revealed in the next few days.

 

[Mythos]

@ISSCC 2005 The Cell, 1:8/90NM configuration.

256Gflops-294Gflops.

 

[jvd]

it will offer 1 flop performance .


Yup 1 flop