Is this for real? Can anybody confirm this info (CELL info)

[Josh378]

http://shplorb.ods.org/


http://games.slashdot.org/comments.pl?sid=131807&cid=11006251

Actually, CELL is based around the 970. Expect about 80-90% performance compared to an equivalently clocked 970. Where it goes nuts is that there's a number of vector units attached that are basically "VMX on steroids" to quote one of the main guys at IBM behind this. The vector units (or Data Plane Processors as they're calling them) can also communicate between each other as well as with the central core. The workstations are actually headless server blades, each of which will have 2 CELL's on them and they'll be running Linux.

This stuff isn't bullshit, it was all disclosed Thursday at the Australian Game Developers Conference. I didn't sign a NDA so it's all good. I also fondled a PSP =]

http://arstechnica.com/cpu/02q2/ppc970/ppc970-1.html

-Josh378

[Trawler]

Looks like the PS3 and Xenon will be a pretty even match after all...

[Qroach]

Looks like the PS3 and Xenon will be a pretty even match after all...

That wouldn't surprise me if it were true. At the very least I don't think you will see much of a difference in their graphical capabilities.

[Josh378]

NM, this info is true, but I think it's talking about the PU's of the CELL hardware. This info was released WEEKs before the Sony Nvidia info was released....

-Josh378

[shplorb]

Hi, I'm the guy who posted that stuff to Slashdot. Yes, it is all true. If you don't believe me then that's your problem.

A lot of you also obviously appear to have no idea what you're talking about.

[Mulciber]

Hi, I'm the guy who posted that stuff to Slashdot. Yes, it is all true. If you don't believe me then that's your problem.

A lot of you also obviously appear to have no idea what you're talking about.

a lot of who? :roll:

[one]

Hi, I'm the guy who posted that stuff to Slashdot. Yes, it is all true. If you don't believe me then that's your problem.

A lot of you also obviously appear to have no idea what you're talking about.

a lot of who? :roll:

Perhaps those who ignore the effect of "VMX on steroids" and the targeted Cell clockspeed? :wink:

Looks like the PS3 and Xenon will be a pretty even match after all...

That wouldn't surprise me if it were true. At the very least I don't think you will see much of a difference in their graphical capabilities.

[Ty]

Looks like the PS3 and Xenon will be a pretty even match after all...

How do you come to that conclusion? We aren't even sure what the PS3 is capable of (QRoach - you mentioned that developers haven't even been given specs which I can confirm as well for our company and even portions of SCEA) nor exactly what the Xenon is going to be about either.

So it seems to me that you're comparing incomplete data of the PS3 with incomplete data of the Xenon and saying they're similar. Why? Because both sets of data are incomplete? I'm not saying either is going to be more powerful than the other but I don't understand how either of you can even begin to make that call at this point in time. Not trying to argue so if you have reasons why you feel this way please share them.

[Mulciber]

Hi, I'm the guy who posted that stuff to Slashdot. Yes, it is all true. If you don't believe me then that's your problem.

A lot of you also obviously appear to have no idea what you're talking about.

a lot of who? :roll:

Perhaps those who ignore the effect of "VMX on steroids" and the targeted Cell clockspeed? :wink:

Looks like the PS3 and Xenon will be a pretty even match after all...

That wouldn't surprise me if it were true. At the very least I don't think you will see much of a difference in their graphical capabilities.

wouldn't that be....a couple?

[V3]

Hi, I'm the guy who posted that stuff to Slashdot. Yes, it is all true. If you don't believe me then that's your problem.

A lot of you also obviously appear to have no idea what you're talking about.

I belive you, but can you explain your slashdot comment more in depth. You use rather vague analogy in your comment, can you disclose other details that you picked up ?

[Megadrive1988]

this is nothing really new. maybe more detail, but nothing surprising. like others have said, this would be about the PU portion of Cell Processing Elements. the PU is the CPU core. one per Processing Element. PU is based on POWER or PowerPC.

[Trawler]

I think it's a fair bet to say that there are more than a couple of posters in this forum who don't know what they're talking about. :wink:

[Megadrive1988]

remember, the APUs aka SPUs are going to be the workhorse of Cell, and thus, of PS3. along with the Nvidia components. the processing capabilities of the PUs (be it PowerPC 970 or otherwise) is only a small part of story.

[j^aws]

Hi, I'm the guy who posted that stuff to Slashdot. Yes, it is all true. If you don't believe me then that's your problem.

A lot of you also obviously appear to have no idea what you're talking about.

We discussed it here when you posted it at slashdot,

http://www.beyond3d.com/forum/viewtopic.php?p=424665#424665

So you were at the Australian dev conference?

[MrFloopy]

shplorb wrote:
A lot of you also obviously appear to have no idea what you're talking about.

Now don't be impolite on your first visit. Don't make me swivel my swiss ball on monday and give you a slap :)

loved the day off, you? Great time to be sick!

Yes there are a lot of people on this board that are misinformed, but some of them (one in particular, although V has been quiet lately) seem to take great pleasure in their own magnificance so don't burst their bubble.

[Panajev2001a]

No please, burst our bubble.

After-all, I have always been interested in knowing how things really stand, understand what that implies and have learned from the process something new :).

[MrFloopy]

No please, burst our bubble.

Sorry for teasing cause really can't say too much more, but can say having been in this industry for 14 years, one thing I have learnt is that noone wins by default. In my experience both Sony and Microsoft are new comers to this game. To think that one company has control over this industry is foolish and simple minded.

Also:

1) There has always been and always will be fan boys/girls, no matter how many big words they use in their posts.

2) In the past 20 years there has been nothing significantly new in computer graphics techniques. (Although some applications of such have been interesting).

3) No matter how many times people promise revolution in computing, it remains evolution because in computing, revolution is the norm. if you don't understand that then think a little longer.

Anyway, my advice, just enjoy what you have. Yes next gen will be better but not by what you think.

If your not happy with games today, you wont be happy with next generation games either.

If you are happy with todays games then there is a chance that you will still be happy with the next generation.

[Panajev2001a]

No please, burst our bubble.

Sorry for teasing cause really can't say too much more, but can say having been in this industry for 14 years, one thing I have learnt is that noone wins by default. In my experience both Sony and Microsoft are new comers to this game. To think that one company has control over this industry is foolish and simple minded.




True, no company wins by default and at the start of a generation all makers start from 0 units basically. Id o nto think we can deny the fact that brand-name does count in today's market. Brand-name power can drive market-share up which in turns can crank developers' suppmarket.


Also:

1) There has always been and always will be fan boys/girls, no matter how many big words they use in their posts.

2) In the past 20 years there has been nothing significantly new in computer graphics techniques. (Although some applications of such have been interesting).

It depends how you look at it. A lot of work has been done perfecting the models and then speeding them-up and so on: I think that, for example, 20 years ago people knew how light interacted with surfaces physically, but that does not mean that you had ready and efficient (computationally wise) ways of achieving that result on screen.

I could say that nothing significantly was done in the last 20 years in regards to cars' engine technology if I only cared about going from point A to point B and back to point A with no regards even to how quickly I can travel.

Then you have to add the fact that it is still an excitement when techniques only used in off-line CG rendering become available to be used in real-time.

3) No matter how many times people promise revolution in computing, it remains evolution because in computing, revolution is the norm. if you don't understand that then think a little longer.

It is evolution, we all agree that CELL is definately based on tons of great concepts which due to technology limitations could only stay at the theoretical/lab environment kind of stage.

I was not looking to CELL as a revolution: parallel computing is not a new idea and any kind of optimization of the concept is and can only be defined as evolution.

I look at CELL as an architecture developed with the idea that it might be time we push, for the Desktop PC space and for CE products, parallel computing to give the needed jump in media processing's performance even though some efficiency might go a way to run your single instance of Web Browser/Office Application etc...

[london-boy]

Looks like the PS3 and Xenon will be a pretty even match after all...

That wouldn't surprise me if it were true. At the very least I don't think you will see much of a difference in their graphical capabilities.

I still have to understand where the 1Tflop will come from. If this is true and X2 and PS3 are so similar, then why would one target performances of ~60Gflops and the other 1Tflop... :?

[MrFloopy]

True, no company wins by default and at the start of a generation all makers start from 0 units basically. Id o nto think we can deny the fact that brand-name does count in today's market. Brand-name power can drive market-share up which in turns can crank developers' suppmarket.

Of course marketing and brand name counts, thats why sega and nintendo are still top of the tree...... oh wait! Your right it's not an insignificant issue, however consumers are fickle masters.

I could say that nothing significantly was done in the last 20 years in regards to cars' engine technology if I only cared about going from point A to point B and back to point A with no regards even to how quickly I can travel.

Yes I remember when cars sudenly went 1000 times faster or same speed with 1/1000th of the fuel. I understand the analogy though :)

Yes you are correct, however mine was an anolgy too that I though was closer to home. Don't suddenly expect that anyone has suddenly cracked some magical ideal not though of before.

It's not a matter of progress, thats implied, but a matter of rate. There is a historical precedence of improvement. This is more likely to slow down than accelerate.

I look at CELL as an architecture developed with the idea that it might be time we push, for the Desktop PC space and for CE products, parallel computing to give the needed jump in media processing's performance even though some efficiency might go a way to run your single instance of Web Browser/Office Application etc...

It certainly looks like that parallelism is the way forward, and that is not a bad thing, but be aware that even if someone suddenly developed a system that was a 1000 times faster than what is available today that it would not be released today. Simple economics dictates that you don't over stretch expectations.

Your on the ball though.

[nAo]

Even if CELL PUs are 970 based there's nothing in that comment we didn't expect. We already know A|SPUs are SIMD stream processors along the lines of PS2 vector units.We already know they can 'talk' each other and with the central core..so..what's the big deal?

[MfA]

Whether this is legit or not Cells cores will almost certainly use a VMX like instruction set, same could be said about EE's VUs, but calling it VMX on steroids is slightly misleading ... VMX still depends on the thread of execution supplied by a huge superscalar core, the vector unit can only ever make up a small part of said core. Giving the vector units their own threads of execution lets you put more power in less area than you could by just using multiple superscalar cores with VMX, that is a pretty big difference.

[london-boy]

Ok so let me get this straight, the Xbox2 CPU will have 3 Double core Powerwhatever cores, which are the same cores as the Cell ones.

So, if i'm not getting things horribly wrong, unless the BE has a lot more cores than the XCPU, at whatever cost they will be, the 2 will be quite identical...? :?

[Titanio]

Ok so let me get this straight, the Xbox2 CPU will have 3 Double core Powerwhatever cores, which are the same cores as the Cell ones.

So, if i'm not getting things horribly wrong, unless the BE has a lot more cores than the XCPU, at whatever cost they will be, the 2 will be quite identical...? :?

No. Cell's power seems to lie in its APUs, not the PUs (the 970s).

All those flop estimations you've seen are based off the APUs alone and don't count the PU.

[london-boy]

Oh ok, so i did miss something.

[nAo]

Giving the vector units their own threads of execution lets you put more power in less area than you could by just using multiple superscalar cores with VMX, that is a pretty big difference.
Actually a lot of people can't grasp this idea..I don't know why.