JF_Aidan_Pryde
Regular
Would anyone be playing PS1 games a decade after its launch? I'm not sure you need to emulate it, although it certainly wouldn't be difficult or anything..
PlayStation III Architecture
- Thread starter alexsok
- Start date
Panajev2001a said:Vince, I have one thing to say buddy...
man you take LONG showers
Been hearing this alot lately. Actually, I saw Chicago with a friend, which surprisingly wasn't as painful as I anticipated and will have to recommend it. Felt kinda awkward seeing it, but I figured if she's willing to put up with me...
Overall I think you're on a good path, I think you and V3 are underestimating what even the present cell-based design is capable of and are a bit 'off' in your ideas concerning lithography and preformance.
For example, Intel will scale Prescott with it's 100M transistors and 1Mbyte of embedded RAM to around 5-6Ghz. Sony wants to design a processor with 64Mbtyes of embedded RAM, perhaps approaching 90M gates and clock it at 4 Ghz. Just a bit more ambitious IMHO.
I mean, there's a fine line between being ambitious and suicidal. Your previous thouhgts of the preformance Kutaragi was speaking of [1 TFlop] was based on a 100nm process and was blatently impossible. You instead convinced yourself that it was possible and are now like, "woah, they're not pushing hard at all" on 65nm. Perspective gentlement, perspective.
Panajev2001a
Veteran
well, I wasn't going onl by will power... there was an IBM paper talking about cellular computing ( wasn't the "nair" paper... uhm... ) manufacturing technology, e-DRAM densty per cm^2 and that didn't seem at the time too unfeasible...
I guess I was grossly underestimating the size and complexity of CELL...
The thing is also that I wasn't expecting to have KB and KB of registers alone, 4 MB of SRAM ( Local Storage, going to take up a nice chunk of die-area ) + 64 MB of e-DRAM...
going by SRAM alone this CELL chip we seen in the patent has 4x the amount of Prescott ( which is on a 90 nm process ) and plus we have the 64 MB of e-DRAM ( 1 Kbits bus )...
The way I conviced myself of the previous "possible" CELL architecture was to envision it working like memory chips ( then more e-DRAM would have been possible maybe )... basically it would not be CPU integrating memory, but memory chips integrating processing logic ( a tiny bit related to 3DRAM maybe ? ... )...
The reason that I moved the thought to 65 nm is that I found your argument, and even more the links you posted making enough evidence there was quite a consistent PUSH for sub 90 nm technology... and my min accepted it moving to the next problem... which for me was even more interesting... I guess I never really stopped after the new patent was released to see how more complex than Prescott this chip was and that 100 nm would have been a worse process from the one Prescott was going to use... 65 nm seems quite a good jump from that and gives me hope...
More interesting it is starting to appear the fact of "common ISA for the APUs and each APU can have a variable number of Execution Units while still running all code designed for the common APU ISA.."
Does that interest you at all ? I'm sure it does, but burying me for the 100 nm mistake is more fun isn't it ?
J/K, I am actually glad to e prooven wrong... my head was in the sky and now is approaching earth again... holpe I can stop eating Helium for breakfast
Panajev2001a
Veteran
<Ladies' man voice [SNL]>: "It's a lady!"
So you liked Chicago ? I am still not very likely to go see it... I am very old school with musicals... I might try it though...
I still think that Rocky Horror Picture Show, Little Shop of Horrors and Jesus Christ Superstar are the "where it's at" in terms of musicals
JF_Aidan_Pryde
Regular
Vince said:Panajev2001a said:Vince, I have one thing to say buddy...
man you take LONG showers
Been hearing this alot lately. Actually, I saw Chicago with a friend, which surprisingly wasn't as painful as I anticipated and will have to recommend it. Felt kinda awkward seeing it, but I figured if she's willing to put up with me...
You got asked out on a date!? 8)
She digs hot tempered, right wing 'Cell' conservatives or what?
j/kThats definatly a cool aspect. The whole idea behind using a cellular architecture was an excellent choice IMHO. I've talked about it before and people just shruged me off, but starting in 2004/2005 Sony and perhaps even Toshiba and Matsushita, will begin making products based around Cell and the era of truely pervasive computing will begin.
One of my favorite books is Visions by the string physicist Michio Kaku, and was taken back at his predictions of the future and just how prevelent computing will be. Fast forward 6 years and I'm sitting here, seeing the beginning of this forming - and at a pace thats far outstripped what he thought was even possible.
Thats what I see as the true revolution, that a single ISA/Core, ect will flawlessly be the backbone of a networking faric that connects our PDAs, PCs, TVs, Cell Phones, and can distribute digital content flawlessly. The limit for this type of thing is only the human imagination... and Cell is barely a first step, but it's a start.
IMHO, thats whats impressive. I'd much rather buy a Sony PS3 thats less powerful than a Xbox Next or Cube2, but allows me to seemlessly interact with all my other little gadgets and get digital content (movies, music, ect).
JF_Aidan_Pryde said:
Actually, there is a longstanding joke in my family about that: Basically whenever I'd go anyplace with Lauren, someone would say, "She's still around? I guess she hasn't gotten to know you yet" <shrug>
But, as I always say, don't judge people based on the 'net. That is the fallacy of the Digital Age - then again, I still hate talking to people on anything but in person. lol.
Besides, find me an Italian thats not hot-tempered!
Panajev2001a
Veteran
that would be me... uhm... ok I can be wicked hot tempered too
it must be in the blood... I have never read that book and if you give me the title I can try to look for it at my college/library... it sounds interesting...
If I understand well enough your concept of pervasise computing would be seeing computers operating all around us, in most aspects of our life...
Even before Cell we were seeing this happening... TVs with more and more functionalities, Cellular phones able to work like a PC ( Smartphones, etc... ), controllers of micro-wave ovens getting more and more sophisticated...
Till now we saw each component around our house getting more and more sophisticated, but we haven't seen them interoperating much...
too many different ISAs, lack of a modular enough architecture DESIGNED to be this modular and flexible to be adapted in both ovens, Stereos, TVs, consoles and servers...
Cell is not a new concept, it is the realization of an idea that has been around for a long time while the necessary elements for making it possible were becoming available...
That is the advantage Sony, IBM and Toshiba have... they're not changing the current state of IA-32 or IA-64 processors to fit those needs, they have embraced a new different way of looking at the computing problem when they started thinkering with the Cellular Architecture idea...
it is a good step towards the vision you are thinking about...
Ethernet at the first mile and your whole house wired with Ethernet plus power over Ethernet ( possible, possible ) will bring the rest...
Cell is the first step towards having in each "intelligent" device nice processors that are built from the ground up to work with each other, so that devices can share data in a way we're not accustomed as of yet... Ethernet at home and power over Ethernet will help this happen as they will deliver the paths over which these intelligent devices can talk...
If eletrical and safety problems could be solved with power over ethernet and your house were all Ethernet ready ( wired all over
) you could do quite nice things: your console could take power from an ethernet port ( the console could have efficient batteries that would recharge during idle times and when the console is off ) and use the same port to deliver the frame to the TV and use that port to communicate with the other appliances in your house... programming your ovens, your shower, your TV and VCR... all from a Cell based device...Sony and Toshiba's products ( if Matsushita joins in... well well, that would be quite "something" ) have a very good diffusion in the consumer electronics market and a product like Cell could fit in each and every single of their products...
The programming model would be uniform for all your devices/appliances ( basically ) as each software Cell could be processed by ANY of the Cell processors ( APUs ) on the same network as the ISA is constant andcare has been put into potential compatibility issues ( and as Intel has taught us, backwards compatibility is a powerful tool )...
Once Cell reaches a certain market penetration, the "critical mass", it should help sustain itself and justify more expansion under the basis of benefiting the current userbase and providing a common platform for as many users possible...
So much to talk ( I do enjoy talking about this Vince ), so little time ( my fiancee has already given the usual three "honey !" yells to go to sleep and shut the computer off, this might get ugly
... did I mention her mother is half Italian, now the idea shold be clearer )Have a good night,
Panajev
P.S.: My question was about implementation and as I said how can we have same ISA on all APUs if each APU is a SIMD device ( basically two VUs, one for FP and one for Integer operations ) that can have a variable amount of execution units ?
Think if we are trying to execute a 4-way parallel FP MADD on a APU which has 1 or 6 FP Units... my only answer would be micro-code them and have the micro-code change if we change the number of execution units, what would be yours ?
Panajev2001a
Veteran
if PS3 seems to be as programmable as it seems and versatile and Blu-Ray is in it... raw performance crown will be meaningless in my eyes... and Sony would take advantage of each bit of any ace under their sleeves to outsmart MS... and their plan seems to show again a good dose of foresight
Don't forget that cell has at least an extra 4MB of eSRAM on top of that. and 1024 bit buses for each PE. Anyway modular design and eDRAM is an advantage.
Using 65nm they'll surely able to put more eDRAM. Anyway the final spec and the process it's going to be use is still up in the air, though I guess from this we have pretty good idea.
Though GPUs on sub 0.1 tech is going to be interesting as well.
Cell will be extremly important for Sony. Read this article: http://msnbc.com/news/861338.asp?0dm=N248T
The Sony DNA.
Fredi[/url]
The Sony DNA.
Fredi[/url]
The last posts reminded me of a bit old news from a few month ago:
http://www.zdnet.com.au/newstech/enterprise/story/0,2000025001,20268030,00.htm
Next step, connect those home-grids via broadband to other grids and use those clusters to host multiplayer-games and save a lot cash on server-farmes and bandwidth-costs? Something like p2p for multi-playergames?
Considering the millions of future online-players this would make some sense.
http://www.zdnet.com.au/newstech/enterprise/story/0,2000025001,20268030,00.htm
Guess this verifies Sony's plan of a home-grid.
Next step, connect those home-grids via broadband to other grids and use those clusters to host multiplayer-games and save a lot cash on server-farmes and bandwidth-costs? Something like p2p for multi-playergames?
Considering the millions of future online-players this would make some sense.
Hm ... good question. Possible solution would be to assign priorities to those data/program-packages. Priority #1 home-grid or single PS3, priority #2 wan-grid ... if the single PS3 would idle (pausing, chatting with other online-players,etc.), it should be possible to use the available resources to process wan-grid-packages.PC-Engine said:What if the only SONY product one owns is a PS3?
Anyway, I am not an expert ... just some crazy ideas.
JF_Aidan_Pryde
Regular
PC-Engine said:What if the only SONY product one owns is a PS3?
Then it's time to buy more!
Isn't that the whole strategy anyway? It's kind of like Microsoft, whose products lock you with their software. But here, it's Sony locking you in with their hardware (through CELL chips and OS).But haven't Sony said it would sell "the Cell" architecture to anyone who want's it, even to MS and Nintendo for use in their next gen concoles.
This can mean that many home electronic products from a lot of other manufacturers can be compatible with the rest of Sony products...
simplicity
Newcomer
Oh man, am I the only one on this thread who thinks this whole Cell Grid computing strategy is unlikely to succeed?
I think Cell might be a reasonable architecture for creating a cheap-and-powerful-but-hard-to-program video game console, but all this other stuff just doesn't make any sense.
First, if you go back to the launch of the PS2 you'll see that Kutigari/Sony/Toshiba were saying pretty much the same thing for the PS2 (we're more than just a game console, we'll be the hub of your home, we'll sell the chipset into all sorts of other products) that they're saying now about the PS3. And we all know how that turned out: No other commercially viable consumer product needed the raw performance of the video game console chipset. Even PVR products could get away with less performance, using a normal CPU and a dedicated MPEG2 encoder.
What's fundamentally different this time around?
Second, this whole "Cell will succeed because it establishes an ABI that can be used to bring Grid computing to the masses" idea doesn't make sense. There's already many existing ABIs that could be used to implement Grid computing: (the Win32 ABI, the Linux ABI, the Java ABI and the .NET ABI, the KaZaa ABI, whatever the distributed-key-cracking guys cobbled together, whatever the SETI-at-home guys have done, plus thousands of other research and startup company ABIs)
In order for Cell to popularize grid computing, two things must be true:
1) Grid computing must prove to be extremely useful.
2) For some reason you can't use the existing Win32/Linux/Mac/Internet infrustructure to implement it.
It's hard to believe that both of these statements could be true at the same time. If Grid computing were really that useful, someone would have hacked an imperfect version on top of the hardware that we already have. (In fact, thousands of people have done so, many times over the past forty years. The reason we don't have Grid computing today has much more to do with the limitations of Grid computing than with the lack of standard ABIs for implementing it.)
Cell's Grid Computing play is another in a long line of "let's replace the existing world order with a different world order that I control" architecture. I can see what's in it for IBM and Sony, but what's in it for me? Why can't I do the same thing with a WinCE box? Or a stripped down Linux box?
Sure, we've had cluster computing for heavy duty number crunching for years, but why would I need it in my home?
I think Cell might be a reasonable architecture for creating a cheap-and-powerful-but-hard-to-program video game console, but all this other stuff just doesn't make any sense.
First, if you go back to the launch of the PS2 you'll see that Kutigari/Sony/Toshiba were saying pretty much the same thing for the PS2 (we're more than just a game console, we'll be the hub of your home, we'll sell the chipset into all sorts of other products) that they're saying now about the PS3. And we all know how that turned out: No other commercially viable consumer product needed the raw performance of the video game console chipset. Even PVR products could get away with less performance, using a normal CPU and a dedicated MPEG2 encoder.
What's fundamentally different this time around?
Second, this whole "Cell will succeed because it establishes an ABI that can be used to bring Grid computing to the masses" idea doesn't make sense. There's already many existing ABIs that could be used to implement Grid computing: (the Win32 ABI, the Linux ABI, the Java ABI and the .NET ABI, the KaZaa ABI, whatever the distributed-key-cracking guys cobbled together, whatever the SETI-at-home guys have done, plus thousands of other research and startup company ABIs)
In order for Cell to popularize grid computing, two things must be true:
1) Grid computing must prove to be extremely useful.
2) For some reason you can't use the existing Win32/Linux/Mac/Internet infrustructure to implement it.
It's hard to believe that both of these statements could be true at the same time. If Grid computing were really that useful, someone would have hacked an imperfect version on top of the hardware that we already have. (In fact, thousands of people have done so, many times over the past forty years. The reason we don't have Grid computing today has much more to do with the limitations of Grid computing than with the lack of standard ABIs for implementing it.)
Cell's Grid Computing play is another in a long line of "let's replace the existing world order with a different world order that I control" architecture. I can see what's in it for IBM and Sony, but what's in it for me? Why can't I do the same thing with a WinCE box? Or a stripped down Linux box?
Sure, we've had cluster computing for heavy duty number crunching for years, but why would I need it in my home?
Panajev2001a
Veteran
The Cell architecture offers uniform ISA together with great scaling & modularity, scaling and modularity... repeat 100x and maybe you will see this in a more bright light...
simplicity said:
I think your missing the idea here and instead trying to find problems where there really aren't. Attempt to clear this up:
First off: GRID computing was mentioned by Okomoto at GDC '02 as an example of how their utilizing the R&D of major universities, public projects like Globus (which is the standard in the type of Grid your speaking of), ect to help them tackle problems with distributed computing. Not that they intend to utilize it for massive cluster -like computing, it could be as simple as inter-device communications, sharing of data, storage, ect. All of these fall under the guise of Grid based computing
You're just taking it to an extreme - which is wrong. I hate to use the word ignorant, but you need to realise that Sony isn't stupid. They know that developers can't feasibly program to a cluster of x Cells [where x is the number of cell based products in home] and get tangible results while making a game that works across the 50M PS3's that be in homes within 2.5 years. Thats just idiotic at this point.
Second: Cell is elegent in that as just one computational core and it's associated ISA, it spans the entire spectrum of computing needs. While, obviously, not everything will be based off of this architecture - much can be and can be economically.
This is important because I beleive, as Kutaragi has stated and Kaku has written about, that the future will be one of pervasive computing. Where almost every item we use will in someway be networked together and contain low-cost processing elements. It seems alien today, but all truely forward looking staements do. The problem, as Kutaragi talked about in that one interview, is that the internet of today is more like a bunch of stand alone islands with their own ISA, OS, procesing elements, ect. Cell, atleast for Sony and anyone who adopts this, will eliminate those burdens - without software thats costly in it's incompatability problems and/or preformance sacrifices.
So, I don't see how Cell can fail - atleast in it's role in PS3, the console.. lol (like we have the 'console' and the 'vision' and to some people the 'fantasy' aswell). It appears to be quite powerful, and extremely flexible - which could be great for developers if Sony and the gang work out the kinks experienced with PS2 and providing extraction from the hardware at first.
Panajev2001a said:The Cell architecture offers uniform ISA together with great scaling & modularity, scaling and modularity... repeat 100x and maybe you will see this in a more bright light...
Damnit, you beat me! And your explination has a higher code density than mine - you do in 3 lines what I did in 50. Shit, I caught what Lazy8 has.
Similar threads
