The "Freeola Customer Forum" forum, which includes Retro Game Reviews, has been archived and is now read-only. You cannot post here or create a new thread or review on this forum.
[URL]http://www.megagames.com/news/html/hardware/cellvsintelround1.shtml[/URL]
I made a post a while back stating that "is it worth doing a new computer revamp with the cell chip on the horizon. I was simple told go ahead cause it wont be compatible with PC’s because of the architecture difference. I have just found this webpage and it seems very interesting and very worrying for most PC owners.
Here are some of the main points.
According to the Cell consortium, the new chips will be made up of 64-bit processors (each with 2.5 MB of memory) and should debut at 4GHz speeds.
Cell will also carry a greater number of transistors compared to current Intel chips with the triumvirate announcing 234 million compared to the 125 million carried by the latest Pentium 4 CPUs.
According to its creators the main advantage of Cell architecture is its flexibility. Jim Kahle, an IBM fellow said, We support many operating systems with our virtualization technology so we can run multiple operating systems at the same time, doing different jobs on the system.
IBM declined to confirm which Operating Systems, other than Linux, have been tested on Cell.
If Playstation3 ends up having 4 Cell processors, as has been rumoured, the console will end up being officially classified, with today's standards, as a super computer.
Now if I have not misunderstood the cell may be used in computers, and if the ps3 has 4 cell chips, should that make it around 16ghz?
I made a post a while back stating that "is it worth doing a new computer revamp with the cell chip on the horizon. I was simple told go ahead cause it wont be compatible with PC’s because of the architecture difference. I have just found this webpage and it seems very interesting and very worrying for most PC owners.
Here are some of the main points.
According to the Cell consortium, the new chips will be made up of 64-bit processors (each with 2.5 MB of memory) and should debut at 4GHz speeds.
Cell will also carry a greater number of transistors compared to current Intel chips with the triumvirate announcing 234 million compared to the 125 million carried by the latest Pentium 4 CPUs.
According to its creators the main advantage of Cell architecture is its flexibility. Jim Kahle, an IBM fellow said, We support many operating systems with our virtualization technology so we can run multiple operating systems at the same time, doing different jobs on the system.
IBM declined to confirm which Operating Systems, other than Linux, have been tested on Cell.
If Playstation3 ends up having 4 Cell processors, as has been rumoured, the console will end up being officially classified, with today's standards, as a super computer.
Now if I have not misunderstood the cell may be used in computers, and if the ps3 has 4 cell chips, should that make it around 16ghz?
Page:
Garin wrote:
> Your original assertion
>
> Miserableman wrote:
> * Software designed to run on CELL architecture will need a complete
> rewrite compared to software running on standard single-processor
> architecture. The software design techniques to develop
> multi-processor systems haven't even been invented yet. It is like
> being all excited over being able to build tall buildings because
> someone invented the brick, yet nobody's invented scaffolding. Read
> this for more information:
>
> Your current assertion
>
> Miserableman wrote:
> Just because something won't crash and die immediately on a
> multiprocessor architecture doesn't mean it's making efficient use
> of
> the available hardware. The Playstation 3 is supposed to have four
> processing units: how exactly do you construct a videogame that
> makes
> efficient use of all the processing power when it is aligned in this
> way? "With extreme difficulty" is the answer, you have to
> have four tasks on the go at all times. I work in a videogames
> company and a lot of the programmers here are dreading working with
> multiprocessor architecture.
>
> At least we've made progress, we've gone from all software will
> require rewrites and we dont have the software design techniques , to
> software will still work and we can write multiprocessor software
> "With extreme difficulty".
>
> If you want to make a point, at least stick with it. If you wanted
> to say multiprocessor architecture can be more challenging to use
> efficiently then you wouldn't get any arguments from me. But for me
> thats light years away from proclaiming we dont know how to do it in
> the first place.
I'm not saying you couldn't port software developed today to a CELL processor. I never said that, and if I didn't make that clear then I'm sorry. You could run x86 or PPC code on an abacus and a piece of paper if you want to.
What I'm saying is that the last major overhaul in the way software is developed, Object Orientation, took years to develop and even longer to get to the point where it were widely practised in the industry. For concurrent programming to permeate into mainstream programming is another OO-magnitude overhaul, it's going to be years and years before the programming profession is able to mass-produce multiprocessor code that offers significant benefits over single processor code. Until that point all this hoopla about the PS3 having FOUR PROCESSORS OMG is bunk, as the software we are currently developing will struggle to use more than half of them at once. Doom III or Half-Life 2 ported to a CELL wouldn't run any faster than on x86 without a big rewrite and a few debugging suicides.
> Your original assertion
>
> Miserableman wrote:
> * Software designed to run on CELL architecture will need a complete
> rewrite compared to software running on standard single-processor
> architecture. The software design techniques to develop
> multi-processor systems haven't even been invented yet. It is like
> being all excited over being able to build tall buildings because
> someone invented the brick, yet nobody's invented scaffolding. Read
> this for more information:
>
> Your current assertion
>
> Miserableman wrote:
> Just because something won't crash and die immediately on a
> multiprocessor architecture doesn't mean it's making efficient use
> of
> the available hardware. The Playstation 3 is supposed to have four
> processing units: how exactly do you construct a videogame that
> makes
> efficient use of all the processing power when it is aligned in this
> way? "With extreme difficulty" is the answer, you have to
> have four tasks on the go at all times. I work in a videogames
> company and a lot of the programmers here are dreading working with
> multiprocessor architecture.
>
> At least we've made progress, we've gone from all software will
> require rewrites and we dont have the software design techniques , to
> software will still work and we can write multiprocessor software
> "With extreme difficulty".
>
> If you want to make a point, at least stick with it. If you wanted
> to say multiprocessor architecture can be more challenging to use
> efficiently then you wouldn't get any arguments from me. But for me
> thats light years away from proclaiming we dont know how to do it in
> the first place.
I'm not saying you couldn't port software developed today to a CELL processor. I never said that, and if I didn't make that clear then I'm sorry. You could run x86 or PPC code on an abacus and a piece of paper if you want to.
What I'm saying is that the last major overhaul in the way software is developed, Object Orientation, took years to develop and even longer to get to the point where it were widely practised in the industry. For concurrent programming to permeate into mainstream programming is another OO-magnitude overhaul, it's going to be years and years before the programming profession is able to mass-produce multiprocessor code that offers significant benefits over single processor code. Until that point all this hoopla about the PS3 having FOUR PROCESSORS OMG is bunk, as the software we are currently developing will struggle to use more than half of them at once. Doom III or Half-Life 2 ported to a CELL wouldn't run any faster than on x86 without a big rewrite and a few debugging suicides.
Mr. Bill Gates has probably got a team working on an operating system that will support it already. If not a lot of money will be made by other companies doing what Microsoft missed out on. Will it be the fall of Microsoft?
Ignore the transistor count. It's an outdated method of measuring performance. Pentium 4 Northwood cores at 3 GHz were faster than Prescott cores at 3 GHz, despite having 70 million less transistors. Prescott did scale up well though (provided it was cool enough) because of the bigger cache and long pipeline. The first dual core Intel chips are to contain over a billion transistors.
Speed is also semi irrelevant nowadays. For instance, a 2 GHz Athlon 64 Clawhammer will outperform a 3 GHz Pentium 4 Prescott.
And these differences are just between x86 processors. The CELL is a vector processor. We are in fact comparing apples to peas.
Windows is unlikely to ever run on the CELL processors. At least, not in a proper retail manner. It would require a substantial re-write to port to an altogether different architecture. If the CELL can do code emulation, it'll be painfully slow. Even after converting x86 binaries to Alpha binaries, FX!51 struggled to get a 500 MHz Alpha chip to run as fast as a 200 MHz Pentium Pro. Nor is there really demand for Windows on anything other than x86.
The only use for CELL outside video games will be specialised workstations. The unit price of the processor is likely to be high. And it's also going to be rather warm. At 85 degrees I'd be worrying about stability. The thermal limits for most processors are less than 100 degrees.
I'm not a fan of x86. It's a 64 bit hack of a 32 bit hack of a 16 bit hack of an 8 bit architecture. But they have made decent progress with it. Dual core in the near future will allow proper multiple operating systems. They are getting faster too. Yesterday saw a Winchester core Athlon 64 running on air cooling at 3 GHz. It wasn't stable admittedly, but it did indeed post at that speed.
CELL has a lot of hype. But most of it is just that, and what isn't just hype is unlikely to catch on, at least in PCs. I don't ever expect to see Time or Dell selling CELL based PCs. You haven't made the wrong choice.
Speed is also semi irrelevant nowadays. For instance, a 2 GHz Athlon 64 Clawhammer will outperform a 3 GHz Pentium 4 Prescott.
And these differences are just between x86 processors. The CELL is a vector processor. We are in fact comparing apples to peas.
Windows is unlikely to ever run on the CELL processors. At least, not in a proper retail manner. It would require a substantial re-write to port to an altogether different architecture. If the CELL can do code emulation, it'll be painfully slow. Even after converting x86 binaries to Alpha binaries, FX!51 struggled to get a 500 MHz Alpha chip to run as fast as a 200 MHz Pentium Pro. Nor is there really demand for Windows on anything other than x86.
The only use for CELL outside video games will be specialised workstations. The unit price of the processor is likely to be high. And it's also going to be rather warm. At 85 degrees I'd be worrying about stability. The thermal limits for most processors are less than 100 degrees.
I'm not a fan of x86. It's a 64 bit hack of a 32 bit hack of a 16 bit hack of an 8 bit architecture. But they have made decent progress with it. Dual core in the near future will allow proper multiple operating systems. They are getting faster too. Yesterday saw a Winchester core Athlon 64 running on air cooling at 3 GHz. It wasn't stable admittedly, but it did indeed post at that speed.
CELL has a lot of hype. But most of it is just that, and what isn't just hype is unlikely to catch on, at least in PCs. I don't ever expect to see Time or Dell selling CELL based PCs. You haven't made the wrong choice.
Your original assertion
Miserableman wrote:
> * Software designed to run on CELL architecture will need a complete
> rewrite compared to software running on standard single-processor
> architecture. The software design techniques to develop
> multi-processor systems haven't even been invented yet. It is like
> being all excited over being able to build tall buildings because
> someone invented the brick, yet nobody's invented scaffolding. Read
> this for more information:
Your current assertion
Miserableman wrote:
> Just because something won't crash and die immediately on a
> multiprocessor architecture doesn't mean it's making efficient use of
> the available hardware. The Playstation 3 is supposed to have four
> processing units: how exactly do you construct a videogame that makes
> efficient use of all the processing power when it is aligned in this
> way? "With extreme difficulty" is the answer, you have to
> have four tasks on the go at all times. I work in a videogames
> company and a lot of the programmers here are dreading working with
> multiprocessor architecture.
At least we've made progress, we've gone from all software will require rewrites and we dont have the software design techniques , to software will still work and we can write multiprocessor software "With extreme difficulty".
If you want to make a point, at least stick with it. If you wanted to say multiprocessor architecture can be more challenging to use efficiently then you wouldn't get any arguments from me. But for me thats light years away from proclaiming we dont know how to do it in the first place.
Miserableman wrote:
> * Software designed to run on CELL architecture will need a complete
> rewrite compared to software running on standard single-processor
> architecture. The software design techniques to develop
> multi-processor systems haven't even been invented yet. It is like
> being all excited over being able to build tall buildings because
> someone invented the brick, yet nobody's invented scaffolding. Read
> this for more information:
Your current assertion
Miserableman wrote:
> Just because something won't crash and die immediately on a
> multiprocessor architecture doesn't mean it's making efficient use of
> the available hardware. The Playstation 3 is supposed to have four
> processing units: how exactly do you construct a videogame that makes
> efficient use of all the processing power when it is aligned in this
> way? "With extreme difficulty" is the answer, you have to
> have four tasks on the go at all times. I work in a videogames
> company and a lot of the programmers here are dreading working with
> multiprocessor architecture.
At least we've made progress, we've gone from all software will require rewrites and we dont have the software design techniques , to software will still work and we can write multiprocessor software "With extreme difficulty".
If you want to make a point, at least stick with it. If you wanted to say multiprocessor architecture can be more challenging to use efficiently then you wouldn't get any arguments from me. But for me thats light years away from proclaiming we dont know how to do it in the first place.
Garin wrote:
> Miserableman wrote:
>
> No-one's ever had to write mainstream software concurrently. It's
> going to take a very long time for software gurus to develop
> techniques to create things like videogames that fully utilise
> multiple processing units safely.
>
> Quake 3
> UT 2003
> Both support dual processor machines albeit in crude manners, even
> windows itself will support and utilize dual processors. Just
> because the processors are on the same chip, its not going to
> confound the programming world and invalidate techniques they've been
> using for past 25 years.
>
> And if what you say its true, then I guess the PS3 is screwed
> already.
Just because something won't crash and die immediately on a multiprocessor architecture doesn't mean it's making efficient use of the available hardware. The Playstation 3 is supposed to have four processing units: how exactly do you construct a videogame that makes efficient use of all the processing power when it is aligned in this way? "With extreme difficulty" is the answer, you have to have four tasks on the go at all times. I work in a videogames company and a lot of the programmers here are dreading working with multiprocessor architecture. To quote one of our senior programmers from a discussion on our internal newsgroup:
"Concurrent programming is b*stard hard. During the 90s when I couldn't avoid it, I always delegated it. I've known three people who were capable - two Israelis and a Russian guy (actually he was Ukrainian) (no, he wasn't Artem). It drove me up the wall. Not happy about this. You thought OO was hard to get the hang of..."
Sure people will be able to create games that work on the PS3, that use multiple processors at once, but it's going to be a long long time before the majority of software companies can reliably churn out software that runs significantly faster on multiprocessor architecture than single-processor architecture.
> Miserableman wrote:
>
> No-one's ever had to write mainstream software concurrently. It's
> going to take a very long time for software gurus to develop
> techniques to create things like videogames that fully utilise
> multiple processing units safely.
>
> Quake 3
> UT 2003
> Both support dual processor machines albeit in crude manners, even
> windows itself will support and utilize dual processors. Just
> because the processors are on the same chip, its not going to
> confound the programming world and invalidate techniques they've been
> using for past 25 years.
>
> And if what you say its true, then I guess the PS3 is screwed
> already.
Just because something won't crash and die immediately on a multiprocessor architecture doesn't mean it's making efficient use of the available hardware. The Playstation 3 is supposed to have four processing units: how exactly do you construct a videogame that makes efficient use of all the processing power when it is aligned in this way? "With extreme difficulty" is the answer, you have to have four tasks on the go at all times. I work in a videogames company and a lot of the programmers here are dreading working with multiprocessor architecture. To quote one of our senior programmers from a discussion on our internal newsgroup:
"Concurrent programming is b*stard hard. During the 90s when I couldn't avoid it, I always delegated it. I've known three people who were capable - two Israelis and a Russian guy (actually he was Ukrainian) (no, he wasn't Artem). It drove me up the wall. Not happy about this. You thought OO was hard to get the hang of..."
Sure people will be able to create games that work on the PS3, that use multiple processors at once, but it's going to be a long long time before the majority of software companies can reliably churn out software that runs significantly faster on multiprocessor architecture than single-processor architecture.
Miserableman wrote:
> No-one's ever had to write mainstream software concurrently. It's
> going to take a very long time for software gurus to develop
> techniques to create things like videogames that fully utilise
> multiple processing units safely.
Quake 3
UT 2003
Both support dual processor machines albeit in crude manners, even windows itself will support and utilize dual processors. Just because the processors are on the same chip, its not going to confound the programming world and invalidate techniques they've been using for past 25 years.
And if what you say its true, then I guess the PS3 is screwed already.
> No-one's ever had to write mainstream software concurrently. It's
> going to take a very long time for software gurus to develop
> techniques to create things like videogames that fully utilise
> multiple processing units safely.
Quake 3
UT 2003
Both support dual processor machines albeit in crude manners, even windows itself will support and utilize dual processors. Just because the processors are on the same chip, its not going to confound the programming world and invalidate techniques they've been using for past 25 years.
And if what you say its true, then I guess the PS3 is screwed already.
gamesfreak wrote:
> I'm a little overwhelmed with this.
>
> Will this 'cell' archtecture have a totally different instruction set
> to say, the normal x86 that we have now?
Yes. I'd be suprised if you see Windows released for CELL before 2010, if at all.
> I'm a little overwhelmed with this.
>
> Will this 'cell' archtecture have a totally different instruction set
> to say, the normal x86 that we have now?
Yes. I'd be suprised if you see Windows released for CELL before 2010, if at all.
gamesfreak wrote:
> Yeah PPC isn't it?
PPC = PowerPC which is the name all Mac processors have had since IBM took over creating them. the G4/G5 uses IBM's 64-bit POWER series
[URL]http://www.apple.com/g5processor/executioncore.html[/URL]
Fwiw, the XBOX2 will be using PowerPC processors instead of x86/intel which is quite an interesting snippet as effectively, the XBOX2 will be using "dumbed" down G5/G4 processing/architecture!;)
> Yeah PPC isn't it?
PPC = PowerPC which is the name all Mac processors have had since IBM took over creating them. the G4/G5 uses IBM's 64-bit POWER series
[URL]http://www.apple.com/g5processor/executioncore.html[/URL]
Fwiw, the XBOX2 will be using PowerPC processors instead of x86/intel which is quite an interesting snippet as effectively, the XBOX2 will be using "dumbed" down G5/G4 processing/architecture!;)
Tyla wrote:
> Dunno, as Mac's don't use x86 and haven't done for a long time!;)
Yeah PPC isn't it?
> Dunno, as Mac's don't use x86 and haven't done for a long time!;)
Yeah PPC isn't it?
gamesfreak wrote:
> I'm a little overwhelmed with this.
>
> Will this 'cell' archtecture have a totally different instruction set
> to say, the normal x86 that we have now?
Dunno, as Mac's don't use x86 and haven't done for a long time!;)
Cell is pretty sexy though:
[URL]http://images.spong.com/news/i/n/internetsp158258l.jpg[/URL]
> I'm a little overwhelmed with this.
>
> Will this 'cell' archtecture have a totally different instruction set
> to say, the normal x86 that we have now?
Dunno, as Mac's don't use x86 and haven't done for a long time!;)
Cell is pretty sexy though:
[URL]http://images.spong.com/news/i/n/internetsp158258l.jpg[/URL]
Page: