Xenos vs. Cell

[danteye]

I understand! but you are saying that xbox, which will be in our homes 6 months before then ps3, will be almost powerfull as ps3?
anyway....thanks a lot for your answer!

Thanks as well!

But nobody around here knows at this time which one is going to be the best one. That's where all the threads are about: speculation.

The most important part isn't going to be the actual hardware either way, but it will be the people who make the games for them. If those developers do well and make great games, the console that runs them will be popular.

do you think that ps3 will be able to move the graphics of killzone 2 in real-time?

and i have heard some rumors about 10 mb of edram in ps3?
are only rumors?

 

[Laa-Yosh]

I haven't been keeping up with the latest FLOPS tally, but if what you say is accurate, then, obviously, Cell is the greater engineering achievement since Cell can use its 218GFLOPS to render graphics, or to simulate a nuclear explosion, or to repaginate your 300-page novel. Let me see a GPU do all that as well. It's easier to achieve greater FLOPS when you sacrifice flexibility.

CELL's 218 GFLOPS are only valid if you're using its 32-bit, non-IEEE floating point format. Once you want to use the 64 bit IEEE format, CELL's performance will drop to about 25-30 GFLOPS, which isn't that much more than today's dual 64-bit Intel and AMD systems. And we're still talking about peak, ie. theoretical performance. Thus I have some doubts about CELL and professional/scientific applications...

 

[Carl B]

CELL's 218 GFLOPS are only valid if you're using its 32-bit, non-IEEE floating point format. Once you want to use the 64 bit IEEE format, CELL's performance will drop to about 25-30 GFLOPS, which isn't that much more than today's dual 64-bit Intel and AMD systems. And we're still talking about peak, ie. theoretical performance. Thus I have some doubts about CELL and professional/scientific applications...

When in coding for a game though would you want to do that? And believe me when I say I'm interested in learning the 360's DP performance as soon as those numbers are released - I don't exactly see it smashing Cell.

 

[Titanio]

When in coding for a game though would you want to do that?

No

AFAIK, DP is of marginal use in games. Maybe a few calculations you'd want to do with double precision, but the vast majority are done in SP.

 

[Carl B]

When in coding for a game though would you want to do that?

No

AFAIK, DP is of marginal use in games. Maybe a few calculations you'd want to do with double precision, but the vast majority are done in SP.

Absolutely my points. 8)

But I AM curious still on the DP performance regardless just for the sake of knowing.

 

[Laa-Yosh]

When in coding for a game though would you want to do that? And believe me when I say I'm interested in learning the 360's DP performance as soon as those numbers are released - I don't exactly see it smashing Cell.

Phat has been talking about using Cell based computers for offline rendering, scientific simulations and so on. These all require DP at least.

And no need to bring up X360's DP speed, because noone mentioned how cool it'd be to use it for the above mentioned applications. Only Cell is hyped as the supercomputer on a chip.

 

[Carl B]

Phat has been talking about using Cell based computers for offline rendering, scientific simulations and so on. These all require DP at least.

And no need to bring up X360's DP speed, because noone mentioned how cool it'd be to use it for the above mentioned applications. Only Cell is hyped as the supercomputer on a chip.

Well, if we're not talking about games performance, and just going on it's possible supercomputing applications, the Cell is still fairly strong as a DP chip, especially for the costs I imagine the economies of scale will bring it down to relative to it's competitors. Major drawbacks would be dearth of software at the moment and the additional chips needed for communication between more than two Cells.

 

[Frank]

do you think that ps3 will be able to move the graphics of killzone 2 in real-time?

I really don't know. A lot depends on the GPU and the developers. It is too early to say.

and i have heard some rumors about 10 mb of edram in ps3?
are only rumors?

Yes, rumors mostly, I think. I don't think it will have any edram, as it will have it's own GDDR memory. But it might after all.

 

[Laa-Yosh]

Well, if we're not talking about games performance, and just going on it's possible supercomputing applications, the Cell is still fairly strong as a DP chip, especially for the costs I imagine the economies of scale will bring it down to relative to it's competitors. Major drawbacks would be dearth of software at the moment and the additional chips needed for communication between more than two Cells.

Think about the costs again... The PS3 will probably cost more than $300, and will still be sold at a loss. What about RAM, can it use 'cheap' SDRAM? Or do you have to buy the necessary 2-4 GB or more RAM in XDR memory for your workstation? You'll probably end up with a price-to-performance ratio very similar to Intel/AMD systems... There's a reason for the VFX studios' transition to the cheap Linux PC as their renderfarm.

 

[danteye]

do you think that ps3 will be able to move the graphics of killzone 2 in real-time?

I really don't know. A lot depends on the GPU and the developers. It is too early to say.

and i have heard some rumors about 10 mb of edram in ps3?
are only rumors?

Yes, rumors mostly, I think. I don't think it will have any edram, as it will have it's own GDDR memory. But it might after all.

thanks again for your answers !

 

[Carl B]

Think about the costs again... The PS3 will probably cost more than $300, and will still be sold at a loss. What about RAM, can it use 'cheap' SDRAM? Or do you have to buy the necessary 2-4 GB or more RAM in XDR memory for your workstation? You'll probably end up with a price-to-performance ratio very similar to Intel/AMD systems... There's a reason for the VFX studios' transition to the cheap Linux PC as their renderfarm.

Cell will be produced in volumes the niche chips never will, so I would expect a good price-performance ratio there; but good point on the 'support structure' such as RAM, that must very well knock it back in the other direction somewhat. It's all a bunch of unkowns anyway though until something commercial comes on the market. IBM was talking about selling two Cell-based servers earlier inthe year, but I haven't heard anything new on those plans lately. Too bad, because their pricing could have given some indication.

 

[Laa-Yosh]

Cell's volumes of a maximum of 100 million in 5-6 years isn't exactly that much when compared to the "niche" chips from Intel and AMD. Yep, a Xenon costs a lot - but there'll be dirt cheap dualcore 64-bit processors on the market within a year from both manufacturers. Cell isn't even out before next year.

 

[Carl B]

Cell's volumes of a maximum of 100 million in 5-6 years isn't exactly that much when compared to the "niche" chips from Intel and AMD. Yep, a Xenon costs a lot - but there'll be dirt cheap dualcore 64-bit processors on the market within a year from both manufacturers. Cell isn't even out before next year.

Well, I'm not talking about AMD or Intel (non-Itanium) chips, as those aren't the ones traditionally used by the scientific community. But I agree, that's been changing over the last couple of years, and in that context I don't know what to tell you. Except that I'm a big AMD fan as well.

 

[jvd]

aye we should be able to get pretty fast dual core cpus from both companys by summer time next year . I would wonder if you could get a dual core athlon 64 for like 200ish