-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
The CELL Hype Begins: Supercomputer on a Chip
- Thread starter monkeymagic
- Start date
automagnus
Member
The hype began about a year ago
inpHilltr8r
Member
Begins? Where have you been?
Chittagong
Gold Member
ROFL
Begins?
Begins?
Shogmaster
Member
automagnus said:
A year ago? Did I get transported back in time to 2004?
radioheadrule83
Banned
If it doesn't deliver will anyone call them on it? If anyone does, will the rabid horde listen? By the hair of Solid Snake's silver tash, I doubt it. So why not. Jack me in Sony!
VictimOfGrief
Banned
"Begins" must be a relitively "old" beginning.
Kleegamefan
K. LEE GAIDEN
acidviper said:
shh.....don't let the haters hear you.....
Chittagong
Gold Member
Syb said:
The Double Dual Apple Power Mac G5s are performing so slow in graphics because they have that NVIDIA 7800 card that doesn't have unified shaders... oh wait.
It's my understanding that in the current definition of supercomputer, the CELL would not even come close to ranking. Doesn't the CELL top out at well below a teraflop? Here's an article on the 2005 top 500 rankings.
http://www.informationweek.com/story/showArticle.jhtml?articleID=173602325
http://www.informationweek.com/story/showArticle.jhtml?articleID=173602325
Deg said:Its a CPU comparision.
Is there any pc more powerful than G5? I am not aware of quad processors for Pc
http://www.sun.com/servers/entry/v40z/index.jsp
8 fucking cores if you want.
choplifter
Banned
acidviper said:
It has been known that Cell's R&D budget was $400 million since before 9-11 happened - Cell was announced in March 2001 with that budget.
http://www-03.ibm.com/chips/news/2001/0312_sony-toshiba.html
the PlayStation3 is more than just Cell - overall, Sony has spent billions on the entire PS3 effort: Cell development, GPU development (dont forgot the GPU(s) Sony worked on before turning to Nvidia for help) Blu-Ray, Software / Tools / APIs / Chip-Manufacturing / Games Etc
sonycowboy
Member
choplifter said:
Billions is a hell of alot of money. The ONLY way to spend billions is when you actually buy the hardware and sell it below cost. Otherwise, you don't even get fucking close at all.
But thanks for the speculation.
Chittagong
Gold Member
littlewig said:
No.
Flachmatuch
Member
Hehe you can get cheap (around $3000) 4 CPU (2x2 core) Opteron blades. For (a lot) more money you can go up to 32-64 etc processors for PC stuff even...so not a really big deal unless I misunderstand something.
I don't really understand the Cell money hype, $400m is not that much for research, and even though IBM and Toshiba are in it too, I don't think this is nearly as much as Intel's spending. It's not Sony who realized that the way ahead was massive parallelism, the whole pc thing has been heading that way for quite a while now: current pc/videocard setups are actually assymetric multiprocessing (not even talking about separate I/O processors); pipelining is actually parallelizing procedures - even current cpus can run more instructions than their clockrates in a lot of cases. Thing is, this requires a completely different approach to software design, and in general purpose computing you had to make every step of the way ahead work better than the previous one, so you had to have incremental development of compilers, higher level cpu functions (OOO execution for example allows for better usage of hardware resources), even "smart" emulation (like what Transmeta tried and failed at.) Maybe such a huge jump could work if all the architecture, from execution resources to high level software development was designed in one step, I'm not sure, but this seems to be what the Cell was aimed at. Right now, it doesn't look general purpose though, it looks more like it isn't ready for that yet, but for well-understood stuff like simulation and graphics, where you understand the nature of the software very well and there's no need for completely new architectures (the graphic pipeline architecture for polygon-based modelling/rendering is pretty sophisticated now), it could work pretty well. I'm not sure the Cell is that great for general-purpose stuff, but I think it will be ace for games
I don't really understand the Cell money hype, $400m is not that much for research, and even though IBM and Toshiba are in it too, I don't think this is nearly as much as Intel's spending. It's not Sony who realized that the way ahead was massive parallelism, the whole pc thing has been heading that way for quite a while now: current pc/videocard setups are actually assymetric multiprocessing (not even talking about separate I/O processors); pipelining is actually parallelizing procedures - even current cpus can run more instructions than their clockrates in a lot of cases. Thing is, this requires a completely different approach to software design, and in general purpose computing you had to make every step of the way ahead work better than the previous one, so you had to have incremental development of compilers, higher level cpu functions (OOO execution for example allows for better usage of hardware resources), even "smart" emulation (like what Transmeta tried and failed at.) Maybe such a huge jump could work if all the architecture, from execution resources to high level software development was designed in one step, I'm not sure, but this seems to be what the Cell was aimed at. Right now, it doesn't look general purpose though, it looks more like it isn't ready for that yet, but for well-understood stuff like simulation and graphics, where you understand the nature of the software very well and there's no need for completely new architectures (the graphic pipeline architecture for polygon-based modelling/rendering is pretty sophisticated now), it could work pretty well. I'm not sure the Cell is that great for general-purpose stuff, but I think it will be ace for games
CarlosX360
Banned
QFT.automagnus said:
I think he meant, the REAL hype begins? :lol
Sony invests in East Fishkill fab...
Hehe, more like $120 million in just the East Fishkill fab. $1.1 billion overall in semiconductors.
Flachmatuch
Member
Just to give some sort of perspective: AMD's investment in just their Dresden fabs will exceed $4.7b by 2007, intel investment for Fab28 will be around $4b, fab12 retool cost $2b - in fact, intel has announced a $4b investment in various fabs around just the US in only 3 months at the end of this year.
"G5 almost audibly groans under the burden, though it's no slouch. In fact, it's currently the top of the line for PCs."
As a programmer, I feel I must clear something up:
G5 is by NO MEANS "top of the line" and a MAC is by NO MEANS a PC.
For a PC, AMD is king, and has been for a while.
Intel may get ahead next year though, but they have a tendancy to screw up their launches.
In the server market, there's tons of options. 8-way Opteron systems are beefy beasts, as are a bunch of stuff from SUN or IBM. The Xeons are aging, but still compete.
The CELL will certainly be the most power chip most people will have ever seen.
But it is NOT by any means anything new or amazing.
Hype train, money hats, and consumer ignorance ftw.
As a programmer, I feel I must clear something up:
G5 is by NO MEANS "top of the line" and a MAC is by NO MEANS a PC.
For a PC, AMD is king, and has been for a while.
Intel may get ahead next year though, but they have a tendancy to screw up their launches.
In the server market, there's tons of options. 8-way Opteron systems are beefy beasts, as are a bunch of stuff from SUN or IBM. The Xeons are aging, but still compete.
The CELL will certainly be the most power chip most people will have ever seen.
But it is NOT by any means anything new or amazing.
Hype train, money hats, and consumer ignorance ftw.
Flachmatuch
Member
conker said:
Mac is not a PC only if PC means a Microsoft OS, but otherwise it definitely is a "personal computer." It's not "IBM PC compatible" (well...it is if you really want it to be) but it is a PC. Also, G5 is "top of the line" the PPC lineup :-D...and you're forgetting the Pentium-M which is actually leading a couple of SPEC benchmarks.
"Most power on a chip" is not something that can be defined very easily either, for a couple of algorithms it seems to be very good, but definitely not general purpose enough for this claim to be so undeniable.
Shompola said:
Only 8?
How about 64 cores?
ftp://ftp.software.ibm.com/common/ssi/rep_sp/n/XSD00994USEN/XSD00994USEN.PDF
Flachmatuch
Member
I guess they mean the single PPE (powerpc processing element) and 8 SPEs (Synergistic Processing Elements - vector units basically.)Oblivion said:
Flachmatuch
Member
aaaaa0 said:Only 8?
How about 32 cores?
ftp://ftp.software.ibm.com/common/ssi/rep_sp/n/XSD00994USEN/XSD00994USEN.PDF
That's 32 processors, 4 processors in a blade, scalable to 8 blades. 2 core on each Xeon MP - so the chips have only 2 cores. The Cell is a supercomputer on a chip because the specialized SPE's can do specific tasks at very high speeds, so in a few cases (a lot of which occur in gaming) it can perform really well - and there's 8 of these specialized cores on a single chip.
Flachmatuch
Member
aaaaa0 said:
Yep but that has nothing to do with it - the chips are still 2 core Xeon MPs. You can configure old IBM mainframes to see however many "cores" you want, but that's irrelevant wrt the semiconductor part.
Flachmatuch said:
Uh, the OS sees a single machine image, with 64 identical real cores to run tasks on. How is that not the very definition of a multi-processor machine? It's a 64-way multiprocessor, and that's what the OS sees. The CPUs just happen to be in seperate boxes, with big fat busses connecting them.
Flachmatuch
Member
aaaaa0 said:
The claim was that the cell is a supercomputer on a chip. If you consider 8 3U blades a "chip", you're pretty strong
Edit: actually they're 4 processor machines, so it's just 4 blades, my mistake.
Flachmatuch said:The claim was that the cell is a supercomputer on a chip. If you consider 8 3U blades a "chip", you're pretty strong
Well then the 8-way AMD that someone just posted above doesn't count either, since it's 4 dual-core AMD64s.
The question was:
Is there any pc more powerful than G5? I am not aware of quad processors for Pc
And yes, there are most definitely more powerful PCs than a G5. By a lot.
Lord Error
Insane For Sony
Nothing is amazing if you compare it to some crazy expensive end of the line solutions or the top of the line supercomputers. Cell chip however will be sold in a device that's going to be very cheap, even compared to PCs, and is generally aimed towards mass market products. That's why it makes sense comparing it to dual G5, as they are featured in a consumer computer you can buy in any store, and is close to the top of the line in that market. For the rendering demo they showcased, Cell does hugely better job than those dual G5s, IBM had a comparision chart that displayed that, but the demo was of course suited to Cells architecture.conker said:
Flachmatuch
Member
Similar threads
- Poll