Support NeoGAF

[Kleegamefan]

How many Gflops can a Pentium 4 EE and a FX 55 do these days?

 

[FeedMeAStrayCat]

a half of one

 

[Panajev2001a]

It depends on the clock speed: they (the Pentium 4 and the Athlon 64) both can do only 4 FP ops/cycle.

So, at 4 GHz they would push a theoretical peak of 16 GFLOPS.

 

[DonasaurusRex]

guess it would depend on the application

I dont really think the cell can be touched in media applications if the real world performance adds up to what Sony/IBM has said. Eventually with multiple cores AMD and INtel will catch or surpass the CELL but that will depend if they can work out heat issues to get the clock speed up to CELL levels, and have better logic gates and switches, right now its a stall in that area for both companies, hence the switch for multi core platforms.

 

[doncale]

Eventually with multiple cores AMD and INtel will catch or surpass the CELL

yep, but they'll have a tough time catching even the first generation CELL processors. by the time they approach CELL performance, STI will have leap-frogged Intel & AMD again with 3rd or 4th generation CELL-'1' with more SPUs and/or be moving on to a true next-generation CELL-'2'

 

[DonasaurusRex]

yep, but they'll have a tough time catching even the first generation CELL processors. by the time they approach CELL performance, STI will have leap-frogged Intel & AMD again with 3rd or 4th generation CELL-'1' with more SPUs and/or be moving on to a true next-generation CELL-'2'

Most likely yes but that depends on what happens with the K10 and next Gen Intel design whatever its gonna be called. In the server market if the Cell takes off i dont expect it to be caught if IBM gets market penetration because then they will be updating it as much as possible. Then again you never know competition does some crazy things.

 

[ourumov]

Unless the new double cores appear with a new set of 3D instructions as a substitute of the SSE/MMX, I doubt it.

 

[DonasaurusRex]

Unless they get pressure in the server market dual cores will probably just be 2 cpu's on one die. Until the new generations appear that is.

 

[TTP]

It depends on the clock speed: they (the Pentium 4 and the Athlon 64) both can do only 4 FP ops/cycle.

So, at 4 GHz they would push a theoretical peak of 16 GFLOPS.

This compared to the 256 GFLOPS the Cell can push? OMG, am I reading this right?

*man, I suck so much at tech stuff. Can someone extend the comparison to GC, PS2 and Xbox CPUs too? Yeah, I know it's not an indication of the overall systems power. Just show me the numbers. Thanks ^_^.

 

[gofreak]

This compared to the 256 GFLOPS the Cell can push? OMG, am I reading this right?

*man, I suck so much at tech stuff. Can someone extend the comparison to GC, PS2 and Xbox CPUs too? Yeah, I know it's not an indication of the overall systems power. Just show me the numbers. Thanks ^_^.

PS2's CPU had a peak of 6.2GFlop.

It all sounds very impressive, but remember that Intel/AMD etc. could, if they wanted to, engineer similar floating point monsters. But floating point performance isn't as useful for applications generally as it is for certain types of app (games, multimedia, simulation etc.) - they have to accomodate a much broader spectrum of program.

edit - ourumov has the GC details

 

[ourumov]

GC CPU peaks at 1.9 GFLOPS: It has a FPU split in two that can do 4 FP/cycle. It's the Flipper integrated T&L that peaks at 10 GFLOP...

 

[TTP]

well, that's pretty impressive then. thanks for the numbers

 

[Kleegamefan]

Wow....well that really puts the 256Gflops of CELL and the 90(ish) Gflop of the Xenon CPU into perspective...thanks

 

[Fafalada]

*man, I suck so much at tech stuff. Can someone extend the comparison to GC, PS2 and Xbox CPUs too?

XBox CPU was 2.9GFlops peak.
Add another ~3.73GFlops for the vertexshaders in its graphic chip.