By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Gaming - Xenon vs Cell Which one really is better ??

NJ5 on 09 July 2008
NNN2004 said:
NJ5 said:


My genuinely bogus crap which is even proven by an accurate reading of the quotes you post? I still laugh every time I remember your claims about dynamic branch hints. This thread was quite enlightening about your modus operandi:

http://www.vgchartz.com/forum/thread.php?start=100&id=29781

This, post in particular, will tell any programmer all they need to know about your knowledge:

http://www.vgchartz.com/forum/post.php?id=872071

Especially when followed by our exchange about your misunderstanding of the quote in that post. You seriously thought dynamic branch hints are the be all end all solution to flushed pipelines due to branches.

 

calm down guys we are here to discuss no to fight.

 


I'm quite calm even when I don't seem so ;) For once, I just want MikeB to admit that he's wrong and doesn't know that much about programming.

Anyone experienced with assembly programming on a few different CPUs (MikeB doesn't program in assembly by his own admission; not even one CPU) will be shaking their heads after reading the thread I linked to. His lack of knowledge is very obvious there.

 



My Mario Kart Wii friend code: 2707-1866-0957

Around the Network
Fishie on 09 July 2008
MikeB said:

PS3 Cell of course. Much much better (8 independent processors vs 3 cores at same clockspeed, the SPEs can do a much better job than the PPE or a 360 can at well designed tasks for them to accomplish) and not only that, the way it's implemented in the PS3 architecture counts also. The Xenon has to share its L2 cache amongst all three cores and access to main memory provides far less bandwidth as it has to share the bus with the GPU and the PS3's XDR Ram provides much lower latencies.


Here you go again talking shit about stuff you dont understand. The SPE`s are NOT 8 independant processors, they are instead part of the processor and furthermore there might be 8 of em but only 7 of em work and only 6 are available for the games.

NNN2004 on 09 July 2008
Fishie said:

 

 The SPE`s are NOT 8 independant processors, they are instead part of the processor and furthermore there might be 8 of em but only 7 of em work and only 6 are available for the games.

thats mean Cell have a total of 138 TFLOPS.

 



Fishie on 09 July 2008
MikeB said:
bdbdbd said:
@Mike B: Whether they are GPP:s or not, the number of transistors per SPE limits their ability to work as GPP, which makes them work as specialized (where name implies too, if i recall) work more efficiently. Actually, if they wouldn't need to work as specialized, you wouldn't need the PPE to control the SPE:s.

 

The place the PPE takes within the Cell, the PPE is needed to do this management. IBM could build a full system with only one SPE, acting as its CPU.

To quote a friend:

"There has been a lot of debate about how the Cell will perform on general purpose code with many saying it will not do well as it is a “specialised processor”.  This is not correct, the Cell was designed as a general purpose processor, but optimised for high compute tasks.  The PPE is a conventional processor and will act like one.  The big difference will be in the SPEs as they were designed to accelerate specific types of code and will be notably better in some areas than others, however even the SPEs are general purpose.  The Cell has in essence traded running everything at moderate speed for the ability to run certain types of code at high speed."

Read his articles, they are informative:

http://www.blachford.info/computer/Cell/Cell4_v2.html

You crack me up man, CELL is a general purpose processor but urm it cant run general purpose code well without programing to the SPE`s strengths?

So urm it is NOT a general purpose processor then as you have to write SPE specific code to get the performance you want from it, nice contradiction of what you claim within the same sentance there.

 



Marty8370 on 09 July 2008

Fishie - SPE's are independent cores in there own right. Try doing some research, before spouting crap.



Around the Network
Retrasado on 09 July 2008
Fishie said:
MikeB said:

PS3 Cell of course. Much much better (8 independent processors vs 3 cores at same clockspeed, the SPEs can do a much better job than the PPE or a 360 can at well designed tasks for them to accomplish) and not only that, the way it's implemented in the PS3 architecture counts also. The Xenon has to share its L2 cache amongst all three cores and access to main memory provides far less bandwidth as it has to share the bus with the GPU and the PS3's XDR Ram provides much lower latencies.

 

Here you go again talking shit about stuff you dont understand. The SPE`s are NOT 8 independant processors, they are instead part of the processor and furthermore there might be 8 of em but only 7 of em work and only 6 are available for the games.

Mike. he's right. Although the Cell has 8 SPEs, for the PS3, they manufacture the chips with one SPE designed to not work. (the reason for this is if one of the SPEs doesn't work, they don't have to manufacture an entire new chip.) And one of the other SPEs is reserved for the OS at all times.

Also, check out the following from Wikipedia: (quoted because I really can't say it much better in my own words.)

The PPE, which is capable of running a conventional operating system, has control over the SPEs and can start, stop, interrupt, and schedule processes running on the SPEs. To this end the PPE has additional instructions relating to control of the SPEs. Unlike SPEs, the PPE can read and write the main memory and the local memories of SPEs through the standard load/store instructions. Despite having Turing complete architectures, the SPEs are not fully autonomous and require the PPE to prime them before they can do any useful work.

Thus, while the SPEs do give the Cell a powerful computing capability, the increase is not nearly as great as it would be if each SPE was a complete processor in the way a dual or quad-core processor has two or four complete processors.



Not trying to be a fanboy. Of course, it's hard when you own the best console eve... dang it

Fishie on 09 July 2008
MikeB said:

@ Deneidez

(Plus PS3 has only 7 SPUs and its has lower clocks.)


Same clock and PPE/VMX wasn't taken into account for this test.

8 SPEs x 25.6 GFlops = 204.8 Glops. The PS3 Cell as a whole is ~218 GFlops.

In any case the article is accurate, only your conclusions are wrong. The main thing to take notice of in that article is really that the Cell is able to achieve near its full potential, unlike most other CPUs.

Nope, one SPE is killed of for higher yields and one at all times is reserved by the OS so you only have six to use.

 



Fishie on 09 July 2008
Staude said:
Katilian said:

Staude said:
The Cell.
It has UNLIMITED powaz !

No seriously. It's all about opinion anyways, but the cell was designed with super computers in mind and is between 8-40 times better than a pc cpu.

And how exactly do you quantify it being better than a PC (i assume you mean x86 based) CPU? My GPU can perform certain tasks significantly better than my CPU, but it would suck as a general purpose processor.

 

I read it somewhere. In either case MikeB is pretty spot on with his layout of the cell. (most of it anyhow)

I've read a lot about the cell to dig in why it has so much hype, but it's because like with the PS2, the PS3's design is so that you can extract more and more power from it the more familiar you are with the arcitecture. Sure it does mean bad ports at first, but once it sets off, it's miles ahead of anything else.

I could easily imagine the PS4 (if using a new version of cell) not having a GPU and simply running everything on the Cell.

Sony dropped out of their CELL partnership so now it is owned exclusively by Toshiba and IBM.

You wont see a CELL processor in the next Playstation, Sony learned a very expensive lesson with CELL.

 



Fei-Hung on 09 July 2008
Fishie said:
MikeB said:

PS3 Cell of course. Much much better (8 independent processors vs 3 cores at same clockspeed, the SPEs can do a much better job than the PPE or a 360 can at well designed tasks for them to accomplish) and not only that, the way it's implemented in the PS3 architecture counts also. The Xenon has to share its L2 cache amongst all three cores and access to main memory provides far less bandwidth as it has to share the bus with the GPU and the PS3's XDR Ram provides much lower latencies.

 

Here you go again talking shit about stuff you dont understand. The SPE`s are NOT 8 independant processors, they are instead part of the processor and furthermore there might be 8 of em but only 7 of em work and only 6 are available for the games.

 

@ Fishie actually all 8 of them work but 6 are designated for games. In fact the only detail he got wrong there is 8 independent processors  (although many people have phrased it the same throughout various articles about both the 360 and the ps3 saying they have 3 processors working as one or 6 processors working as one). Even then, all it was is just wrongly phrased rather then to be talking 'shit' as you would say. does this mean you talk shit to? c'mon people calm down for gods sake. Or do you guys have nothing better to do then to make mountain out of mole hills?



NJ5 on 09 July 2008
Fishie said:

You crack me up man, CELL is a general purpose processor but urm it cant run general purpose code well without programing to the SPE`s strengths?

So urm it is NOT a general purpose processor then as you have to write SPE specific code to get the performance you want from it, nice contradiction of what you claim within the same sentance there.

 

I think what MikeB meant (and he's right here) is that the SPEs are Turing-complete, unlike (at least some) GPUs. The SPEs can effectively run any code you throw at them, much like any PC CPU, or a well-trained monkey :P

However, to get good performance out of the PS3's SPEs, you have to vectorize your code and use the best data types for it. For example, double-precision arithmetic is very weak there (1 GFLOP/s in each SPE, very low for a 3.2 GHz processor). Another problem is that the SPEs are strongest when you use them simultaneously, in order to make up for their deficiencies. This is not a problem with some kinds of algorithms (like the ones used in folding@home and other scientific applications) which are easily parallelizable.

However, on a complex application like a game engine, which processes many different workloads, often in an inherently serial fashion, you will eventually run into non-parallelizable code, which will prevent you from using the totality of the Cell's raw computing power. Of course, future engines will use more of the Cell's power, but nowhere near as much as some developer quotes make it seem (e.g. "game X uses only 30% of the Cell!!").

 



My Mario Kart Wii friend code: 2707-1866-0957