Support NeoGAF

[Blaster1X]

OLD....?


IBM recently released a white paper showing real results on what the cell processor can really put out:

http://www-128.ibm.com/developerworks/powe...llperf/?ca=drs-

At this year's E3 (or thereabouts) Sony proclaimed that their processor could achieve 200GFLOPS! However, according to IBM's white paper, only 155.5 GFLOPS was actually achieved (Table 4). BUT, IBM's tests used all 8 SPEs. The PS3 will only use 7 SPE's, due to manufacturing yield issues.

The efficiency of the Cell is 75.9% (Table 4), with of a theoretical peak of 201GFLOPs (Figure 5)--running 8 SPEs at 25.12GFLOPS apiece (Table 2). Similarly, the theoretical peak for the PS3's processor will be 176GFLOPS, using 7 SPEs at 25.12GFLOPS apiece. Assuming the same 75.9% effieciency, we could easily interpolate the PS3's Cell to be capable of 133.6GFLOPS.

The take home message is that with the PS3 being cabable of 133.6 GFLOPS and the Xbox 360 being capable of 115.2 GFLOPS, the PS3 is not nearly as far ahead of the Xbox 360 as we were lead to believe. All in all, we should expect relatively similar power coming from both consoles, processor power and ease of programming all considered.

my take: now since that implies that the Cell isn't nearly as powerful as Sont Claimed, and they're operating with 1 less SPE, well I guess you could say that this means the PS3 won't be as powerful they claimed (typical), that and it has a weaker GPU than the 360 (The PS3 RSX GPU is basically a 7800GTX shrunk to a 90nm process, so nothing new features or architecture-wise).

 

[BlueTsunami]

This is old

 

[Timen]

With cell being developed during the next few years, is it possible Sony revises the processor with each new ... revision ... of the playstation 3? in other words, what are the odds the PS3 will have 8 SPE's and a higher efficiëncy after being on the market for say, two years?

edit: aawww

 

[D3VI0US]

New to me, the plot thickens.

 

[BlueTsunami]

http://www.ga-forum.com/showthread.php?t=73783&highlight=CELL

 

[VictimOfGrief]

This is old

QFOLDness.

 

[Tech Nine]

Actually,i've never seen this. Yeah it's new to me

 

[PedYup]

So actual efficiency of Cell vs theorical peak of XBox360 CPU. Very fair comparison then.

 

[Elios83]

So actual efficiency of Cell vs theorical peak of XBox360 CPU. Very fair comparison then.

LOL :lol

 

[gofreak]

The take home message is that with the PS3 being cabable of 133.6 GFLOPS and the Xbox 360 being capable of 115.2 GFLOPS, the PS3 is not nearly as far ahead of the Xbox 360 as we were lead to believe.

:lol

Compare peaks to peaks, real world numbers to real world numbers. Also, include the PPE. Also, there is no one figure for "efficiency", the numbers above are cherrypicked from certain tasks.

The difference suggested on paper certainly won't be betrayed in the real world - if anything the difference between Cell and Xenon as far as floating point go will actually be larger than the paper figures suggest, IMO.

that and it has a weaker GPU than the 360 (The PS3 RSX GPU is basically a 7800GTX shrunk to a 90nm process, so nothing new features or architecture-wise).

Wait and see, but how architecturally "new" it is has nothing to do with power.

 

[BlueTsunami]

:lol

Compare peaks to peaks, real world numbers to real world numbers. Also, include the PPE.

The PPE isn't being included in the comparison?

 

[gofreak]

The PPE isn't being included in the comparison?

No. Just the 7 SPEs.

 

[BlueTsunami]

No. Just the 7 SPEs.

Then BlasterX1, you should really read the article again. I didn't know what he was refereing to but also comparing actual to theortical isn't a really good comparsion

Since operations in each data block are independent from those in other blocks, the matrix multiplication algorithm is easily parallelized to all eight SPUs. Figure 5 shows that the matrix multiplication performance increases almost linearly with the number of SPUs, especially with large matrix sizes. Using eight SPUs, the parallel version of matrix multiplication achieves 201GFLOPS, very close to the theoretical maximum of 204.8GFLOPS.

What does that say Blaster?

 

[D3VI0US]

The thickening of the plot continues. :lol

 

[gofreak]

Then BlasterX1, you should really read the article again. I didn't know what he was refereing to but also comparing actual to theortical isn't a really good comparsion

That's not even the main problem here. You can't just compare actual figures attained in a real world task to Xenon's theoretical peak. What would its efficiency be like in this task? Less than Cell, I'd guess (and that'll generally be the case with FP heavy stuff - as I say, the paper figures understate the likely realworld difference between their averages, imo).

This is absolutely pointless as we do not have Xenon's performance in the quoted tasks, though I'm willing to bet that in many its efficiency would be worse.

 

[Elios83]

Actually it would be interesting to know the efficiency of the Xbox360 CPU.
Many developers said the shared L2 cache between the three cores had an important impact on performance (the code was running better on the betakit with the separated PowerPCs than on the final CPU).

 

[Blaster1X]

Then BlasterX1, you should really read the article again. I didn't know what he was refereing to but also comparing actual to theortical isn't a really good comparsion



What does that say Blaster?

eight SPUs... = 201GFLOPS. note PS3 has 7 SPUs.

 

[BlueTsunami]

Also, Blaster is cherry picking numbers from the charts. Some tasks CELL is better at than others. It says that very clearly in the article itself

 

[Blaster1X]

Also, Blaster is cherry picking numbers from the charts. Some tasks CELL is better at than others. It says that very clearly in the article itself

I want the truth.. you cant get the truth from Sony. So this IBM article does show us real numbers.

 

[Elios83]

eight SPUs... = 201GFLOPS. note PS3 has 7 SPUs.

1 SPE= 8 FP intructions per cycle.
1 PPE= 12 FP intructions per cycle.

PS3 Cell is 1 PPE+ 7 SPE @ 3.2GHZ = 217.6 Gigaflops.

 

[Blaster1X]

Also, BlueTsunami you seem very concern over this news?

 

[teiresias]

Blaster1X - trolling PS3 - certainly NOT!!!

There is a pretty massive thread all about this article over at B3D and there was already lengthy discussion of this article here as well, this topic should just be locked actually.

 

[BlueTsunami]

Also you say its easy to interpolate that the PS3 will have Cell that can only achieve 133Gflops? Where does that number come from?

Also, BlueTsunami you seem very concern over this news?

Just concerned at someone who clearly doesn't understand what their reading and relaying it to other users as fact

 

[gofreak]

I want the truth.. you cant get the truth from Sony. So this IBM article does show us real numbers.

Please. We have NO real numbers for Xenon. How do you think it'd have faired in these tasks? You are expecting people to swallow that Xenon is 100% efficient, generally, and Cell 75% efficient, generally, in order for your post to make sense. That's simply ludicrous.

Xenon will have a harder time reaching its FP peak than Cell, I'll say that much.

This is like taking, say, polygon draw rates for an actual game on a GPU, and comparing that to the peak polygon draw rate of a weaker card, and saying they're as powerful. Simply stupid.

 

[BlueTsunami]

Merry Christmas Blaster

http://www.beyond3d.com/forum/showthread.php?t=26161

 

[Kung Fu Jedi]

Basically, this boils down to what we've been hearing from developers for some time, PS3 and Xbox 360 are very similar in power.

 

[Raven.]

Basically, this boils down to what we've been hearing from developers for some time, PS3 and Xbox 360 are very similar in power.

Heh, from some pc dev.s that hate going multi-thread. What we've been hearing from some console/ps2 dev.s is quite the contrary, not to mention what we've been seeing.(Things like, models so detailed as to've individual eye lashes, and eyebrow hairs at 60fps on the first and significantly weaker ps3 kits.).

Btw, Xenon's been said to've cache issues, and not sure if I heard right but some time ago here on gaf it was said a good chunk of the cache ran at half the speed.(not sure how that discussion ended). On the cell Faf hinted at the spus beating the ppe in Flt tasks, what do you will happen when you compare them with cores similar to the ppe but that will be having to contend with sharing the cache between themselves?

 

[Blaster1X]

Blaster1X - trolling PS3 - certainly NOT!!!

Checking my past posts i'm no more a trolll then you are.

 

[Blaster1X]

1 SPE= 8 FP intructions per cycle.
1 PPE= 12 FP intructions per cycle.

PS3 Cell is 1 PPE+ 7 SPE @ 3.2GHZ = 217.6 Gigaflops.

Wrong again.

Either Sony or IBM got that math wrong? who do you believe?