March 31 Cell thread

[cthellis42]

I'm wondering, if this thread is going to make it 'til the 31st. Guess it will be locked before any details emerge. Current signal:noise ratio, 1:99

The topic is all noise anyway. We can't talk about anything substantial until something substantial comes out, and the topic asks if people think we might get something substantial by the end of the month.

Just where in tarnation is THAT supposed to go? Heh...

Hardly matters anyway. The moment any information comes out--even one-line mentions in completely unrelated articles--we get one side or the other making a post, quoting subjectively, and making whatever assumptions they want.

 

[Paul]

*cough* 1 TFLOPS *cough*

It's no rumour, that's the official target of the Broadband Engine. Whether they HIT IT OR NOT is a different story though, we will see, My estimate has been 256GFLOPS(said in a thread months ago), though whatever happens happens.

 

[Wunderchu]

I believe that there actually is quite a reasonable probability that CELL will have 1 TFLOP performance .. especially if PlayStation 3 is released in late 2006 , as most people believe will be the case......

.. also, remember what DeanoC's response to Deadmeat was, in this thread:

http://www.beyond3d.com/forum/viewtopic.php?t=10627&postdays=0&postorder=asc&start=0

Of course SCEI could pull a teraflop in a box size of a dishwasher right now, I am talking about a $399 console launching in 2005. That was the promise, PSX3=1 TFLOPS, right???

Looks like I will be the one to laugh at you...

What if a non Sony next-gen console was released with a reasonable fraction of a TFLOP in 2005? Would you change your mind for a Sony Console release a year later?

 

[jvd]

I still wouldn't believe a 1 tflop cell chip.


Because really a fraction of a tflop has already been released. He 1 flop is a fraction of a tflop.

But even if a non sony chip comes out in 2005 that offters 256gflops i doubt that sony would be able to double that performance let along quadrouple (sp?) that performance in less than a year .

 

[cthellis42]

...and yet it STILL doesn't matter, as whatever "marketing XXXX" whether "XXXX" is FLOPS, <someoneelse'sdefined>FLOPS, "polygon performance," "sustainted poly performance," "particle performance," or whatever-the-fuq a company uses, we still will have no idea how it correlates to the end products WE are concerned with. PS2's numbers don't line up to their games; Xbox's numbers don't line up to their games; Gamecube's doesn't seem to either, but at least has a closer correlation than the others...

Perhaps certain simplified applications will be able to take advantage of those "marketing numbers" more efficiently, but they're pretty meanless to the vast majority of those actively wanting those types of machines. And we won't find out what games will be made of by seeing the console tech-specs.

Personally I guess it depends if I become a part of a scientific research team needing great floating-point-performance-per-dollar in a few directions on custom-built code I'll be looking at other things as well, but I don't consider those chances too great.

 

[V3]

Can't wait. I really wanted to know, if they'll go the MCM route or not.

But even if a non sony chip comes out in 2005 that offters 256gflops i doubt that sony would be able to double that performance let along quadrouple (sp?) that performance in less than a year.

Cell is aim for 256 GFLOPS. Its the Broadband Engine that's meant to be rated at 1 TFLOPS. And that's done by putting 4 Cells in either a chip or MCM.

I hope they'll talk about STI cell, and not more of their Blue Gene.

I also wanted to know if, they really manage to do realtime processing over local network.

 

[PC-Engine]

1 TFLOPS using *MCM*, sure. 1 TFLOPS *total* for PS3, sure. 1 TFLOPS *single chip cpu with at least 16MB of eDRAM*, I think not.

 

[London Geezer]

I still think that some way or another, the number "1TFLOP" will make it into the final product. As cthellis said, it doesn't have to mean anything, just like the "theoretical 66 million polygons" on ps2 (and other numbers for other consoles) doesn't mean much. If Nvidia can say they are pushing hundreds of GFLOPS today in their GPUs, then anything could happen...

 

[V3]

From NV GDC slides

The Future
• Increasing power
– More vertex & fragment processors
– GFLOPs, GFLOPs, GFLOPs!

• Fast approaching TFLOPs!

• Supercomputer on a chip

– Start planning ways to use it!

• Massive multi-GPU Supercomputers?

This is like I expected. TFLOPS will be the norm, next gen. Xbox2 with ATI GPU, will surely able to push TFLOPS.

 

[ChryZ]

I am quite amused, that the "1TFLOP" is still subject of discussion. It's clearly a marketing tool and if required, then the numbers will be massaged to fit.

IMHO it will be more interesting to learn about the CELL architecture, the technology behind it, the PS3 implementation and how well it will work

 

[PC-Engine]

From NV GDC slides

The Future
• Increasing power
– More vertex & fragment processors
– GFLOPs, GFLOPs, GFLOPs!

• Fast approaching TFLOPs!

• Supercomputer on a chip

– Start planning ways to use it!

• Massive multi-GPU Supercomputers?

This is like I expected. TFLOPS will be the norm, next gen. Xbox2 with ATI GPU, will surely able to push TFLOPS.

LOL ever heard of NVFLOPS...or NV PR BS? Where can I sign up to become NV zombiized??? Do they teach SONY math at universities?

 

[London Geezer]

LOL ever heard of NVFLOPS...or NV PR BS? Where can I sign up to become NV Zombiized??? Do they teach SONY math at universities?

Chill matey... have a cup of tea...

 

[jvd]

I remember when the comment of 1tflop had just come out. How the tune has changed . How the tune has changed.

It will be interesting rereading the old threads and comparing them with when the official info is out .


Anyway if we can continue to keep all the noise to this thread that will be great.

I should make an official noise thread to let you all bicker in and then just hang anyone that bickers in other threads.

 

[Deadmeat]

...

It will be interesting rereading the old threads and comparing them with when the official info is out .

Looks like I will be the one to have the last laugh....

 

[V3]

LOL ever heard of NVFLOPS...or NV PR BS? Where can I sign up to become NV zombiized??? Do they teach SONY math at universities?

I see, you're still living in stone age

 

[Megadrive1988]

even I was saying stuff like '1 TFLOP is not enough because it doesn't give us 1000 times PS2's 6.2 GFLOP performance'

now we would concider it AMAZING if the Broadband Engine / PS3 CPU had 1 TFLOP.


I do see how expectations have been lowered dramatically. most here seem to expect 128 ~ 512 GFLOPs for the BE and somewhat more for the entire PS3 when counting the GPU.

then there is the problem of memory. even if you had 1 TFLOP, do you think 256 MB of main memory would be enough to utilize that power? no I don't. it's PS2's 32 MB main memory all over again. maybe even worse.( i think DM would agree with me here)

but then there *is* the insanely high-bandwidth eDRAM. so I don't know.

256 GFLOPs CPU / 64-128 GFLOP GPU / 256 MB main memory /
64 MB ~ 96 MB eDRAM might be a good bet. maybe lower, maybe higher.

speculation won't end until we know.....

 

[jvd]

well for them to have a vpu to make up the diffrence of a tflop from a 256 gflop cpu would be a huge departure from their current ideas .

 

[Megadrive1988]

If the PS3 CPU is 256 GFLOPs there is no way the GPU/VPU would make up the difference of a TFLOP. PS3's GPU/VPU will most likely have *less* floating point performance (within the APUs of the GPU/VPU) than that of the CPU. or at most, equal to that of the CPU. not counting the pixel engines. but probably less.

Anyway, it will really be about how everything is implimented, what can be achived on-screen and in-game, rathar than some vauge and near-useless FP numbers which can be twisted so easily (like Nvidia does).

 

[nobie]

Right, isn't the XGPU supposed to be capable of 120 gigaflops? I wouldn't be surprised if MS says XGPU2 does 120 TERAflops. Take all the NVflops and Sonyflops numbers with a grain of salt.

 

[cthellis42]

I wouldn't waste 42 million zarniwhoops thinking about it.