ps3 cell

dobby

Distinguished
May 24, 2006
1,026
0
19,280
hi there, just wondering, there has been aload of talk about sony's PS3, and more specifically the cell. it is suposed be extremely powerful, but caan anyone actually tell me how powerful, (say if comparison to a e6300 or something) because if it is as or more power full than the latest processor series then maybe it is worth the heavy price tag.
 

shinigamiX

Distinguished
Jan 8, 2006
1,107
0
19,280
Humm, and oft-asked question... the CELL is a budget server chip developed by Sony, Toshiba and IBM. The price tag is not as heavy as you might think, especially when one considers that an Intel Core 2 Extreme X6800 costs $999.
While it's hard to compare the CELL to the chips used in our desktop (the old apple-oranges story), the CELL is really not as powerful as Sony advertises it to be. Some will say "but it's an 8-core 3.2GHz Supercomputer!" The 8 cores are not cores in the traditional sense, rather SPEs. In the PS3, one is disabled to enable better yields and one is reserved for the OS, leaving 6 SPEs to run the other stuff. The problem lies in the fact that each SPE can only do one thing. As for the clockspeed, well, I think all of us here know just how irrelevant clockspeeds are to a CPU's performance. If somehow someone did manage to let the CELL function as a desktop CPU (which isn't really possible) I doubt it would do all that well. It's more specialized, unlike CPUs, which take a jack-of-all-trades approach. In fact, the CELL is intended to bridge the gap betweeen CPUs and highly-specialised processors like GPUs.
That being said, however, the CELL does really stand out in certain areas. One being multimedia decoding; a CELL has been demoed decogin 48 HD signals simultaneously, which is why there has been talk of placing CELLs in HDTV. Another thing the CELL excels at is Scientific calculations.

I hope this answers some, if not all, of your questions. :)
 

gman01

Distinguished
Jun 25, 2006
272
0
18,780
I know people put linux on the ps3 and make it a desktop computer....

I am sure if you google 'linux' 'ps3' 'benchmarks' you will find speed tests comparing it to various intel/amd processors.... If you do please report back to us where it sits....
 

shinigamiX

Distinguished
Jan 8, 2006
1,107
0
19,280
http://www.geekpatrol.ca/2006/11/playstation-3-performance/
The Geek Patrol benchmarked some apps using Linux on PS3 and state that the CELL, when running those applications, is roughly equivalent to the low-end 1.6GHz PowerPC G5. Take those with a grain of salt though, as the G5 had 5 times the RAM. Being the kid that I am, I'm not sure if RAM is relevant to those benchmarks...
 

Dante_Jose_Cuervo

Distinguished
May 9, 2006
867
0
18,990
This is very true. Now... of course with a grain of salt you have to consider the intended function of a CELL. That... and I'm pretty sure that the RAM had something to do with it. Then again... a PS3 is pretty decent for gaming so who cares about those benchies as long as I can kill things...
 

Pippero

Distinguished
May 26, 2006
594
0
18,980
As far as i know, Linux on the PS3 does not really make any use of the SPEs of Cell.. so you're just running the general purpose core, which in fact is just comparable to a Power PC.
I think Cell is a great CPU for some specific applications (especially embedded multimedia devices), but not as a general purpose CPU.
 

gman01

Distinguished
Jun 25, 2006
272
0
18,780
http://www.geekpatrol.ca/2006/11/playstation-3-performance/
The Geek Patrol benchmarked some apps using Linux on PS3 and state that the CELL, when running those applications, is roughly equivalent to the low-end 1.6GHz PowerPC G5. Take those with a grain of salt though, as the G5 had 5 times the RAM. Being the kid that I am, I'm not sure if RAM is relevant to those benchmarks...

check this complete ps3/linux article out dated 3 days ago:
http://www-128.ibm.com/developerworks/power/library/pa-linuxps3-1/

according to this you can't get gfx acceleration with linux - so who knows what it could really do if sony unlocks the video acceleration
 

mpjesse

Splendid
that's an interesting article. but there's a LOT of things not considered. like you said earlier, the Cell has 8 cores that can be programmed for seperate tasks. if a program or OS was created to take advantage of this, it would certainly smash anything out there. since superclusters are usually highly customized, the potential for cell is huge, especially with IBM's linux.

but there's one other thing to say: IBM's yields with the Cell sucks goat balls. The majority of the chips only have 4-6 working cores... the other ones are broke. (i think the PS3 only really needs 3 cores to work properly, so it's not a big deal right now)

but if IBM intends to use this thing in a supercluster (i.e., a Blue Gene), they'd better get the chip yields to improve.
 

shinigamiX

Distinguished
Jan 8, 2006
1,107
0
19,280
that's an interesting article. but there's a LOT of things not considered. like you said earlier, the Cell has 8 cores that can be programmed for seperate tasks. if a program or OS was created to take advantage of this, it would certainly smash anything out there. since superclusters are usually highly customized, the potential for cell is huge, especially with IBM's linux.

but there's one other thing to say: IBM's yields with the Cell sucks goat balls. The majority of the chips only have 4-6 working cores... the other ones are broke. (i think the PS3 only really needs 3 cores to work properly, so it's not a big deal right now)

but if IBM intends to use this thing in a supercluster (i.e., a Blue Gene), they'd better get the chip yields to improve.
Right. But if I recall correctly, one SPE is basically only capable of a single operation at a time. For example, the Sony OS takes up one SPE entirely. There's a lot of wasted power right there. Another example is Resistance: Fall of Man. In the game, 2 SPEs are assigned to collision detection. In an intense scene, the two SPEs are working like mad. But when there isn't anything happening they just sit there. Alan Wake on the XBox 360, by contrast, only uses one core for collision detection. The hard part is utilizing all of the CELL's power to the max.
 

Pippero

Distinguished
May 26, 2006
594
0
18,980
but there's one other thing to say: IBM's yields with the Cell sucks goat balls. The majority of the chips only have 4-6 working cores... the other ones are broke. (i think the PS3 only really needs 3 cores to work properly, so it's not a big deal right now)
Waitamin.
If a chip doesn't have 7 working SPEs, then it can't be used for a PS3.
Regardless of the fact that today the games might not be using the SPEs at all.
 

sviola

Distinguished
Mar 9, 2006
313
0
18,780
but there's one other thing to say: IBM's yields with the Cell sucks goat balls. The majority of the chips only have 4-6 working cores... the other ones are broke. (i think the PS3 only really needs 3 cores to work properly, so it's not a big deal right now)

but if IBM intends to use this thing in a supercluster (i.e., a Blue Gene), they'd better get the chip yields to improve.

That's not exactly what's happening. Altough yelds are terrible (only 30% of all Cell have the 8 cores working), the PS3 Cell uses 7 of the 8 SPEs. Of these 7, only 6 are available for developers to use, the other one is only for Sony to use. The 8-working-core Cells are separated to use by IBM in Supercomputing, while the ones that have 6 or less working SPEs are being used to other electronic uses (seems they intend on using it in electronic home appliances).
 

morfinx

Distinguished
Jan 8, 2007
11
0
18,510
I recently attended a Cell talk by IBM. It turned out it was by the chief scientist of the Cell project, Dr. Peter Hofstee, who was giving the presentation. So he is "THE MAN" to talk to about Cell. Let's start with some pics:

celltalk01.jpg

celltalk02.jpg

celltalk03.jpg

celltalk04.jpg


Cell is of course a Power based architecture, with 1 PowerPC core and 8 SPEs (1 SPE disable in PS3 to improve yield). Each SPE has 256KB cache-like SRAM local storage, a DMA (direct memory access) unit, which is VERY important in this architecture.

To use an analogy, you are working on a plumbing project. Traditional processors are akin to you get a pipe from Home Depot, run back, install it, and figures out you need a fitting, goes back to Home Depot, buys it, run back, install it, and figures out you need an elbow. Very inefficient. Traditional techniques is akin to having a long supply line from Home Depot to your house, so ideally you reach out the window, and grab what part you need. With cell, the SPE creates a "shopping list", and the DMA fetches from memory and puts it in the local storage. Memory latency is the biggest slow down to performance, and that's what Cell is good at.

There are 128 registers, because that's the number of bits (7) available in the op codes to address registers. There are no reorder buffers, branch predictors, or reservation stations. Loop unrolling is used to speed things up, taking advantage of the large number of registers. Since each core is in-order, a lot of complexity associated with out-of-order execution is eliminated.

Every program has branches, and some are quite branch heavy. So Cell uses branch hint instructions that the compiler generates to guide it, as well as select instruction to calculate two outcomes before committing to a path. Cell is not the best at every type of application, but it is great for many types of apps. For example, variable length decode of video codecs has a lot of branches, which a lot of people predicted will run poorly on the Cell, but it actually runs great.

Some more highlights:
- 300 GB/s peak, 200 GB/s sustained transfer on the internal bus
- 26 GB/s bandwidth to memory, which is utilized much more efficiently by the Cell architecture than traditional ones. It's not uncommon to see 90% sustained utilization during a program execution.
- Rambus Flex I/O - configurable I/O, could be used as coherency links between multiple Cells
- Very power efficient, runs 3.2Ghz@1V.
- Highly scalable, up to 6Ghz achieved in lab


celltalk05.jpg

celltalk06.jpg


By 2008, an enhanced Cell will be released featuring more double precision float point units for each SPE. Notice by 2010, there will be a cell with a total of 34 cores!


celltalk07.jpg


Advanced Cell blade servers. IBM is actually building a supercomputer that consists of 2000 processors of AMD + Cell (don't remember if it's 2000 each or 2000 total). They are linked together, and the Cell is there to greatly increase float point performance.

celltalk08.jpg

celltalk09.jpg


If you know anything about computer graphics, you know that ray tracing is EXTREMELY process intensive and hard to accelerate through hardware. So although it has superior quality, no existing graphics card use ray tracing, instead using "hacks" to approximate lighting. So it's insane to see real time ray tracing. The system uses 8 Cell blades.


celltalk10.jpg

celltalk11.jpg

celltalk12.jpg



There is another slide that I forgot to take a picture of. It's Cell doing terrain rendering from color maps and height data and rendering a 3d terrain. A 2.4Ghz Power5 (used in Apple G5) renders about 2/3 fps, which is about twice as fast as an Intel processor at the time. A Cell renders the scene at 30fps.


Lastly, the question y'all have been waiting for. You asked, I deliver :)

Q: What is more powerful, the 3 general purpose PPC cores in the XBox 360, or the Cell in the PS3?

A: He was a bit evasive at first. He says both are different concepts, and IBM isn't biased toward either, and helps all game developers. So I asked, "what about raw performance between the 2?" And Dr. Hofstee answered, "raw peak performance wise, Cell is about 2x as much, but it really depends on the developer to deliver."

So there you have it, straight from the horse's mouth :)
 

derek2006

Distinguished
May 19, 2006
751
0
18,990
So there you have it, straight from the horse's mouth :)
You can't trust everything they say. But then again it was supposed to be a gpu too which have a lot of raw power.

here is another slide that I forgot to take a picture of. It's Cell doing terrain rendering from color maps and height data and rendering a 3d terrain. A 2.4Ghz Power5 (used in Apple G5) renders about 2/3 fps, which is about twice as fast as an Intel processor at the time. A Cell renders the scene at 30fps.

The Cell should be good at terrain rendering because it was original supposed to be a gpu and a cpu. But Sony found out if they did that it will be underpowered. So instead they went with a nvidia 7900 like gpu and used the cell for the cpu.

And the G5 sucks thats why Apple went with Intel. Which has more than 2x the performance according to apple.
 

sunangel

Distinguished
Feb 27, 2006
221
0
18,680
You can't trust everything they say. But then again it was supposed to be a gpu too which have a lot of raw power.

Dude, the guy pwned you. Just accept it. Your losing more face than your gaining.
 

morfinx

Distinguished
Jan 8, 2007
11
0
18,510
You can't trust everything they say. But then again it was supposed to be a gpu too which have a lot of raw power.
He's the chief scientist at IBM, who do you propose we trust?

The Cell should be good at terrain rendering because it was original supposed to be a gpu and a cpu. But Sony found out if they did that it will be underpowered. So instead they went with a nvidia 7900 like gpu and used the cell for the cpu.

And the G5 sucks thats why Apple went with Intel. Which has more than 2x the performance according to apple.
You are missing the point entirely. The actual rendering is trivial, it's the terrain generation that is process intensive.
 

apt403

Distinguished
Oct 14, 2006
2,923
0
20,780
There is another slide that I forgot to take a picture of. It's Cell doing terrain rendering from color maps and height data and rendering a 3d terrain. A 2.4Ghz Power5 (used in Apple G5) renders about 2/3 fps, which is about twice as fast as an Intel processor at the time. A Cell renders the scene at 30fps.

I saw a video of something very similar on youtube a while back, ill try to find it.

found it

if i was the video camera guy i was grab the waffer proped up against the monitor and run.
 

sviola

Distinguished
Mar 9, 2006
313
0
18,780
There is another slide that I forgot to take a picture of. It's Cell doing terrain rendering from color maps and height data and rendering a 3d terrain. A 2.4Ghz Power5 (used in Apple G5) renders about 2/3 fps, which is about twice as fast as an Intel processor at the time. A Cell renders the scene at 30fps.

I saw a video of something very similar on youtube a while back, ill try to find it.

found it

if i was the video camera guy i was grab the waffer proped up against the monitor and run.

That's probably why they didn't invite you there! lol
 

shinigamiX

Distinguished
Jan 8, 2006
1,107
0
19,280
You can't trust everything they say. But then again it was supposed to be a gpu too which have a lot of raw power.
He's the chief scientist at IBM, who do you propose we trust?


Is this the reason why we shouldn't trust him, course he will say his project is 2x as powerful![/quote]
IBM was part of the group that created the PowerPC architecture as well. I think he will know what he is saying.
 

atp777

Distinguished
Jul 21, 2006
279
0
18,780
Agreed.. I wouldn't say 2x as powerful, but it has it's place where it shines. I think when it's finished and developers know how to "work" it, it will be a great solution. We'll just have to see how Intel and AMD go up against it in the future.