How powerful are you expecting Cell to be

[GwymWeepa]

And what do you hope games will accomplish with all that power?

 

[Paul]

If Broadband Engine is not a Teraflops peak theo performance, throw it back at Sony.

 

[pcostabel]

It will probably reach 1TFLOP with some fuzzy math.
Actual performance will be half that. I expect it to be able to do software rendering on selected parts of the scene, but still use a traditional rasterizer for the bulk of the geometry. I am intrigued by the possibility of connecting multiple ps3 for performance boost, and I can't wait for the workstation version.

 

[Wunderchu]

If Broadband Engine is not a Teraflops peak theo performance, throw it back at Sony.

indeed


(as I posted in this thread: http://www.beyond3d.com/forum/viewtopic.php?t=11142 ):


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?

 

[GwymWeepa]

If Broadband Engine is not a Teraflops peak theo performance, throw it back at Sony.

indeed

(as I posted in this thread: http://www.beyond3d.com/forum/viewtopic.php?t=11142 ):


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?

Quick question since I'm not as tech savvy as many of you, how much better than the cpu's of the ps2 and xbox is 1 tflop?

 

[cthellis42]

I will go so far as to say it will have a LOT more floating-point performance than a Dorito! Maybe not a Sun Chip, though, as they're damn tasty. Especially the French Onion.

 

[GwymWeepa]

I will go so far as to say it will have a LOT more floating-point performance than a Dorito! Maybe not a Sun Chip, though, as they're damn tasty. Especially the French Onion.

Sun chips are awesome, I love the cheddar, my fav of all time though are ruffles sour cream and onion.

 

[Megadrive1988]

I don't even care how powerful the Cell-based CPU in PlayStation 3 is. I would hope that whatever its peak performance is, it is very effiecient and has a high percentage of sustained, real-world performance, in actual games.

 

[j^aws]

My Casio fx82 Calculator may finally be owned by this thing! It will be able to calculate pi a trillion times faster...we finally might get circular tyres in games!

 

[Wunderchu]

Quick question since I'm not as tech savvy as many of you, how much better than the cpu's of the ps2 and xbox is 1 tflop?

according to Sony, the PS2 can do 6.2 GFLOPS:

http://www.us.playstation.com/consoles.aspx?id=2/info/415007657.html


.... 1 TFLOPS is about 161x more than 6.2 GFLOPS ........

 

[GwymWeepa]

Quick question since I'm not as tech savvy as many of you, how much better than the cpu's of the ps2 and xbox is 1 tflop?

according to Sony, the PS2 can do 6.2 GFLOPS:

http://www.us.playstation.com/consoles.aspx?id=2/info/415007657.html


.... 1 TFLOPS is about 161x more than 6.2 GFLOPS ........

Dear god, that be quick.

 

[Psychogenics]

I expect some really good gaming.

 

[GwymWeepa]

I expect some really good gaming.

Me too, but what sorts of things will the Cell be capable of doing that'll enhance that? I expect massive games, I would love a kessen game where every little troop is shown, and they fight the enemy with superb AI, I want them to feel real.

 

[jvd]

under a 125 gflops sustained.

xbox 2 will be around 60-70 gflops sustained

gamecube 2 400 tflops sustained

Just kiding


Really i don't expect much from the cell chip in the ps3 . Under a 125 gflops sustained. Mabye twice that in theroy never going to be seen

 

[Fafalada]

Quick question since I'm not as tech savvy as many of you, how much better than the cpu's of the ps2 and xbox is 1 tflop?

XB CPU is rated at 2.8GFlop.
XB2 CPU is (currently) rated at 84GFlop.
PS2 was covered already.

Really i don't expect much from the cell chip in the ps3 . Under a 125 gflops sustained. Mabye twice that in theroy never going to be seen

If you're one of those who actually believe SH4 sustained 900MFlops in your average games, then you should believe Cell chip will be 80-90% effective too.
So 125GFlops would come from a BE variant with 150GFlop max

 

[GwymWeepa]

Quick question since I'm not as tech savvy as many of you, how much better than the cpu's of the ps2 and xbox is 1 tflop?

XB CPU is rated at 2.8GFlop.
XB2 CPU is (currently) rated at 84GFlop.
PS2 was covered already.

Really i don't expect much from the cell chip in the ps3 . Under a 125 gflops sustained. Mabye twice that in theroy never going to be seen

If you're one of those who actually believe SH4 sustained 900MFlops in your average games, then you should believe Cell chip will be 80-90% effective too.
So 125GFlops would come from a BE variant with 150GFlop max

Ahh, didn't realize the Xbox2 numbers were known...how does that compare to modern x86 chips?

 

[DeanoC]

If we are talking PS3:
I'm expecting about 0.5 TFlops across the entire system, with prehaps 0.3 TFlops in a Cell CPU.

How efficient in real world situations, is a complete unknown at this point. The XB2 is looking fairly good (has quite a few features to help sustain real world performance) but we don't have any information on cache sizes, register sets, ISA etc. for Cell yet.

 

[Jov]

If we are talking PS3:
I'm expecting about 0.5 TFlops across the entire system, with prehaps 0.3 TFlops in a Cell CPU.

How efficient in real world situations, is a complete unknown at this point. The XB2 is looking fairly good (has quite a few features to help sustain real world performance) but we don't have any information on cache sizes, register sets, ISA etc. for Cell yet.

Wound anyone count the multimedia chip in all this? If so how many xFlops can we expected based on what it will do?

As for the XB2 features, are you stating this based on the ATI R6xx and IBM CPUs? Or are there more we don't know about?

 

[jvd]

If you're one of those who actually believe SH4 sustained 900MFlops in your average games, then you should believe Cell chip will be 80-90% effective too.
So 125GFlops would come from a BE variant with 150GFlop max

Na i would say a chp is anywhere from 40-70% effective depending on design .

 

[ERP]

If you're one of those who actually believe SH4 sustained 900MFlops in your average games, then you should believe Cell chip will be 80-90% effective too.
So 125GFlops would come from a BE variant with 150GFlop max

Na i would say a chp is anywhere from 40-70% effective depending on design .

Do you really think that you're average game codebase contains anything like that density of fp instructions?

Even in vertex transform code it's unlikely you can sustain even that. Lets take DC as an example. It's possible to do a simple vertex transform on DC in 12->13 CPU cycles, unfortunately the TA takes at least 26 CPU cycles to accept a simple tri. So thats 4 dot products a reciprocal and a 4 component multiply 33flops/vertex at a rate of either 13 cycles (if you don't actually write the result anywhere) or 26 if you do. Peak would be 8 flops/cycle so in this real world (although extreme) case your getting more like 1->3 flops per cycle.

So even in an extreme case your only seeing 12%->38% CPU FPU utilisation. And contarary to what some cell fanatics here would have you believe the majority of game code is nothing like that flop intensive.