Does Cell Have Any Other Advantages Over XCPU Other Than FLOPS?

[Guilty Bystander]

Uhm excuse me?? What entire pipe? We're not talking GPU's here.

Pfff... I don't mean Pixel Pipeline.
That's a term we use that the entire hardware is using one algorithm which in this case is Floating point (previously Integer).

What are the chances Xenon and Cell will get anywhere near peak performance? G5 cpus have much higher theoretical flops than Opterons, but their measured performance (in synthetics) is within range of what the Opterons do.
Besides, isn't the primary use for FLOPs graphics, which video cards already handle?

Well I dunno about the Xenon but the Cell has been tested and it can hit 201GFlop/s in realtime.
This means the Cell is 92% efficiënt.

Also you do realise that with the benchmarks used the G5's and other Apple Power-PC based CPU's have to emulate x86 whereas the Opteron doesn't.
If the benchmarks were programmed for Power-PC from the ground up results would be a lot different.

Nope complex physics, A.I. and everything else that makes a game of application tick can be done with Floating point it just can be done better with Floating point.

Ok to kinda get back on topic, XeCPU has 3 PPEs yet its die is smaller than PS3CELL which has 1 PPE and 7+1 SPEs. I can definitely see an advantage for XeCPU for uses beyond a gaming console. Imagine a CPU with 3 PPEs and 6 VMX units at 3.2GHz at 90nm being used as a cheap buidling block for supercomputers. CELL OTOH doesn't seem to be very useful outside of consoles and CT/MRI radar functions.

Uhmm... Cell excells any CPU out there or currently in development in anything (being bandwidth, Floating point, Integer etc.)

 

[scificube]

No problem - it's just annoying when people post bold statements - it can be forgiven if there is no public reference material ( apart from presentations, which are often easy to misrepresent )

This is the architecture book for the cell broadband engine. In here it talks about SPU interupts - both from external causes ( via signal notification registers ) and internal ( by decrementor counter )

http://www-306.ibm.com/chips/techlib/techlib.nsf/techdocs/1AEEE1270EA2776387257060006E61BA

Thank you. I'll definitely try to read this since you bothered to dig it up for me.

 

[darkblu]

Well I dunno about the Xenon but the Cell has been tested and it can hit 201GFlop/s in realtime.
This means the Cell is 92% efficiënt.

i'm pretty sure it can hit its theoretical maximum in realtime - question is in what algorithms. IOW, it's a matter of the problem domain and not reaching full potential per say.

Also you do realise that with the benchmarks used the G5's and other Apple Power-PC based CPU's have to emulate x86 whereas the Opteron doesn't.
If the benchmarks were programmed for Power-PC from the ground up results would be a lot different.

sorry, would you clarify the above - what does the ppc have to emulate and why?

Uhmm... Cell excells any CPU out there or currently in development in anything (being bandwidth, Floating point, Integer etc.)

actually no, there are capabilities of current desktop/server cpu's that cell does not have the blocks for at all - like self-teaching branch predictors, which can be very useful for certain situations.

 

[ihamoitc2005]

Weighted sum

I'd say they're mostly bound by memory bandwidth as you're essentially only computing weighted sums...

Floating point precision is good gain as well as high concentration of single-chip floating point processing from CELL to make floating-point neural net practical for more applications.

 

[aaronspink]

Once again you confuse yourself my friend and you forgot origin of the comparison. It is you who said Xenos is better coprocessor for assistance of PPE in tasks PPE is not good for because you said Xenos is more specialized and therefore superior than SPE for such tasks.

Apparently you enjoy being a tard.

If you meant certain parts of Xenos only, you should be more specific. Also, picking only parts of Xenos for your design means you do not really like Xenos for such use but rather prefer new kind of chip using some parts of Xenos but leaving out other parts no?

I was pretty specific. If I'd known you where incapabable of being intelligent I would of exactly delineated my suggestion to the computational complex of the Xenos design, but that should have been pretty f*cking obvious to anyone with half a brain. But I forgot myself and didn't realize I was conversing with the mentally challenged.

Also, you are incorrect that single SPE has no performance. Single SPE has 25.6 Gflops performance when used as companion of PPE. It is a individual processor but many can be added to PPE as with STI CELL which has 8 SPE added to 1 PPE.

A single SPE has no method of getting either data or instruction into or out of the SPE. In fact, it has no way of allowing any control what so ever. The SPE relies of a large collection of logic on die in order to move data around. Without this large collection of logic the SPE is useless. You seem to want to make comparisons where you only include what is labelled as an SPE while leaving out the large amount of support logic that is required for the SPE to function.

As you can see, even after adding all other components of entire CELL including bus and extra "dead" SPE, Xenos is not superior. Activate 8th SPE and Xenos is suddenly 15% inferior.

We might as well add the 4th SIMD array from the Xenos design then.

So you see my friend Xenos is not effective as companion to PPE.

And one again you have chosen to do a flawed comparison. Apparently you brain is incapable of understanding the issues. If you really want to make this comparison in the fashion you are trying, why don't you use as your area estimate 1/2 of the Xenos Die area, because that IS the area for the flops that are in question, that is the computational core. And that computational core also includes 33% redundancy.

There is nothing to be gained from rudeness my friend. Let us maintain civilized discourse no?

It would be a discourse if you could understand what is actually going on.

Aaron Spink
speaking for myself inc.

 

[nAo]

A single SPE has no method of getting either data or instruction into or out of the SPE.

Quote of the year, period.
btw..my friend Elvis wants to meet your pusher.

 

[typoEDR]

Quote of the year, period.
btw..my friend Elvis wants to meet your pusher.

...tell 'im to stay away from 'em toilets!

 

[PC-Engine]

Okay so now the CELL is only good for consoles and CT/MRI radar functions?

At first you didn't even recognize that the CELL was good for MRI and radar functions. Well as long as you understand it now that is what really counts.

So do you believe that the CELL chip is good for HDTVs or other possible equipment like a Blu-ray player?

Niche market.

We've been over this dude. CELL is a solution looking for a problem. Why do you NEED CELL in CE equipment??? My tv and DVD player works just fine without CELL as do my cellphone and PC..

 

[Mordecaii]

True PC-ENGINE, a TV will work just fine without a Cell in it... but will it work without a processor inside it at all? Nope. Can Cell be used to be replace many of the processors currently in a wide range of consumer electronics? Yes.

 

[PC-Engine]

True PC-ENGINE, a TV will work just fine without a Cell in it... but will it work without a processor inside it at all? Nope. Can Cell be used to be replace many of the processors currently in a wide range of consumer electronics? Yes.

Don't existing tvs already have a processor solution? Why would I need CELL in my tv instead of a traditional microprocessor/DSP?

 

[Mordecaii]

Because you can use Cell instead of having to make different processors for different products, thus reducing costs for the company.

 

[PC-Engine]

Because you can use Cell instead of having to make different processors for different products, thus reducing costs for the company.

I don't own stock in the company...

 

[Mordecaii]

Well, saving money is certainly in the best interest of Sony and consumers since it should allow Sony to continue to make products and hopefully improve on their products. There is also the possibility that Cell will allow new things which the current processors in consumer electronics devices do not allow.

 

[dodo3]

Don't existing tvs already have a processor solution? Why would I need CELL in my tv instead of a traditional microprocessor/DSP?

HDTVs are becoming standard in households, 1 SPE on the Cell processor can decode HDTV @ 77 FPS as suggested in the benchmarks by IBM and Sony. Why would Sony NOT want to use mini-Cell processors for HDTVs?

I think you just don't want Cell to exist at all by replying nonsense.

 

[AlphaWolf]

HDTVs are becoming standard in households, 1 SPE on the Cell processor can decode HDTV @ 77 FPS as suggested in the benchmarks by IBM and Sony. Why would Sony NOT want to use mini-Cell processors for HDTVs?

I think you just don't want Cell to exist at all by replying nonsense.

The problem here is that 1 spe can't do anything, by itself. Cell can't do anything that other processors can do. It all comes down to cost and power and that remains to be seen.

 

[dodo3]

The problem here is that 1 spe can't do anything, by itself. Cell can't do anything that other processors can do. It all comes down to cost and power and that remains to be seen.

Which is why the PPE would be present. If 1 Cell processor was esimated at $150 USD, then what do you think a mini-Cell (being the main core and a single SPE) would cost to manufacture?

I would say depending on the yields of Cell wafers, any Cells with more than 1 bad SPE could be used for an HDTV or some other product.

 

[AlphaWolf]

Which is why the PPE would be present. If 1 Cell processor was esimated at $150 USD, then what do you think a mini-Cell (being the main core and a single SPE) would cost to manufacture?

I would say depending on the yields of Cell wafers, any Cells with more than 1 bad SPE could be used for an HDTV or some other product.

And while that may be useful for sony, how does it benefit you?

 

[dodo3]

And while that may be useful for sony, how does it benefit you?

I am pointing out to PC-Engine of why Sony would want to use Cell for other products, other than the PlayStation 3 console.

Talking about beneifits to consumers, I would imagine since Cell is being fabbed, it would be cheaper for Sony to use Cell as a HDTV decoder rather than making a seperate processor for HDTVs, possibly making it cheaper for consumers to afford 1080p HDTV in the future, IMHO. What do you think?

 

[AlphaWolf]

I am pointing out to PC-Engine of why Sony would want to use Cell for other products, other than the PlayStation 3 console.

Talking about beneifits to consumers, I would imagine since Cell is being fabbed, it would be cheaper for Sony to use Cell as a HDTV decoder rather than making a seperate processor for HDTVs, possibly making it cheaper for consumers to afford 1080p HDTV in the future, IMHO. What do you think?

It's not like we're talking about replacing a future product they would have to R&D, we're talking about replacing a product that is already in production in the millions of units. AFAIK decoders that are in use now don't cost hundreds of dollars so I question how much there is to save.

 

[Edge]

Guys I can say it a million times, but arguing with PC-Engine is a complete waste of time. The guy is an anti-Sony fanatic, and will never change, no matter how right you are. I would just ignore him.

Almost all his post these days are one liners, that contribute nothing to a meaningful discussion.

Kinda of funny for a guy who is such a huge Nintendo fan, can't seem to see it's Microsoft that is slowly but surely killing Nintendo as a hardware player. He should be Sony's biggest fan.