Official PS3 Thread

[Paul]

Mamoru Maekawa, Bunsan Operating System (Distributed Operating System), Tokyo: Kyoritsu Shuppan Co., Ltd., Dec. 25, 1991, pp. 175-182, ISBN 4-320-02570-9.
IBM System/370 Kakucho Architecture (SA22-7085-0 IBM System/370 Extended Architecture Principles of Operation, IBM Japan, Apr. 1984, pp. 3-8 to 3-10.
"IBM Wins Playstation 3 Contract," BBC News, Mar. 12, 2001

So... What does this mean? Especially the part in bold. Not the part where IBM wins the contract, but why it's in that patent in general.

 

[Panajev2001a]

Thread pollution ? I fail to understand why you would consider it thread pollution and I'd appreciate if you could point out why you do think so.

I was just pointing out how I failed to recognize that patent to be talking about R&D work totally unrelated to the Cell joint-venture...

 

[Panajev2001a]

Mfa,

the last link I posted ( from which that last quote comes from ) is related to the other "famous" Cell patent... just go look at the Images section and you can see it

 

[Sonic]

This thread is being locked for obvious reasons.

Thread has been reopened.

 

[Panajev2001a]

JVD,

That's the author making a speculation. Sony denied Cell in PS3 previously as well, the same way they denied about Emotion Engine being in PS2.

I still remember the IEEE event in which Sony Toshiba presented what will be soon known as Emotion Engine and refused to link it with the PlayStation 2...

Of course engineers at that meeting started laughing, but maybe not everybody remembers that event

 

[MfA]

Well a block quote with three quarters of a screen of numerical gibberish is a bit much, especially since it was sparked by your interpreting something I said as something I didnt say It is tantamount to Paul quoting your message in its entirety, instead of just the relevant bit like he did.

 

[bbot]

I've read the patent and I recall that there is nothing in it that says Fig. 6 is for PS3 (the picture Paul posted in this thread). Therefore you can't say for sure whether the Cell chip for PS3 will have a performance of 1tflop.

 

[bbot]

Paul,

How did you come to the conclusion that Figure 6 in the patent specifically refers to PS3. My guess is you took the rumor that PS3 will have a performance of 1tflop and noticed that particular cell chip has a performance of 1tflop and made the connection. Which is, of course, unwarranted.

 

[Panajev2001a]

Well a block quote with three quarters of a screen of numerical gibberish is a bit much, especially since it was sparked by your interpreting something I said as something I didnt say

And I have apologized for it... I will edit the message, but I posted that much as to sort of proove I was not making that link up myself

 

[Paul]

"How did you come to the conclusion that Figure 6 in the patent specifically refers to PS3. My guess is you took the rumor that PS3 will have a performance of 1tflop and noticed that particular cell chip has a performance of 1tflop and made the connection. Which is, of course, unwarranted."

We've been through this whole damn thing before. Read it here.

http://www.beyond3d.com/forum/viewtopic.php?t=3793

 

[Panajev2001a]

Paul,

How did you come to the conclusion that Figure 6 in the patent specifically refers to PS3. My guess is you took the rumor that PS3 will have a performance of 1tflop and noticed that particular cell chip has a performance of 1tflop and made the connection. Which is, of course, unwarranted.

That chip in the patent does have 1 TFLOPS if and only if we have 8 APUs per PU and 4 PUs in the Broadband Engine.

Also the relation between Cell, that patent and PlayStation 3 is the topic of this discussion and there have been quite a bit of posts in the last three pages regarding the issue... please read the whole thread ( and that thread linked by Paul ) as it will help you and us to discuss the topic

We can repeat those arguments of course the issue is not closed...

I also found a bit strange we have in one of those Cell patents something like "IBM wins PlayStation 3's contact"...

 

[Paul]

Haha yes Panajev, that for me is proof enough. They obviously slipped up putting that in there, or maybe they didn't think anyone would find it?

 

[bbot]

If the chip in Figure 6 is for PS3, then why didn't it SPECIFICALLY state so? And Kutaragi has stated that PS3 will be based on Cell, but he didn't give the specifications for the Cell chip for the PS3, so Pana, your point is unwarranted.

Let's see if jvd agrees with this point.

 

[Panajev2001a]

I understand and that is also why we have all those Cell related patent qith Suzuoki Masakzu in them

Also, let's look at that quote again:

United States Patent 6,526,491
Suzuoki , et al. February 25, 2003

[...]

Other References

Mamoru Maekawa, Bunsan Operating System (Distributed Operating System), Tokyo: Kyoritsu Shuppan Co., Ltd., Dec. 25, 1991, pp. 175-182, ISBN 4-320-02570-9.
IBM System/370 Kakucho Architecture (SA22-7085-0 IBM System/370 Extended Architecture Principles of Operation, IBM Japan, Apr. 1984, pp. 3-8 to 3-10.
"IBM Wins Playstation 3 Contract," BBC News, Mar. 12, 2001.

Primary Examiner: Nguyen; Hiep T.
Attorney, Agent or Firm: Lerner, David, Littenberg, Krumholz & Mentlik, LLP

March 12, 2001... uhm, I think I saw that date on "some other document" before...


edit: I correct myself... the patent was filed March 22, 2001... it was filed AFTER the joint-venture became OFFICIAL and known to the general public...

 

[Panajev2001a]

If the chip in Figure 6 is for PS3, then why didn't it SPECIFICALLY state so? And Kutaragi has stated that PS3 will be based on Cell, but he didn't give the specifications for the Cell chip for the PS3, so Pana, your point is unwarranted.

Let's see if jvd agrees with this point.

Why should they do such a thing ? The chip in Figure 6 might not be for PlayStation 3 as the Broadband Engine and Visualizer might be even better than that

Seriously, that patent doesn't have to cover the final and exact IMPLEMENTATION of the chips in PlayStation 3 as they can see modifications over the example that the patent was providing...

 

[zidane1strife]

Well no micron process is being used yet. Since .65 doesn't exist yet we don't know the yeild and clock speed it will be able to hit .

Well, I've heard intel plans to hit 30Ghz using that manufacturing tech(65nm)... although I might've heard wrong... but if not... other parties should easily hit 4-9+Ghz...

1 tflop and everyone assumes the ps3 will have a 1tflop

Not everyone, I believe far more can be achieved.... they originally thought it feasible to achieve 6Tflops at .1 micron(EE3, original idea had it at 500M transistors, and in the minds of the designers it would've been enough to achieve 1000x... that was yrs ago, prior to cell)... now that they are at 65nm... I tend to believe far more will be achieved, and many will be surprised...

Why because like the first hints that the ps2 could do 80 million polygons per second (or whatever the exact number was) and then we come to find out that that is peak numbers with nothing happening in it but sony never promised those specs it was just what the press was saying sony got out of it with out looking bad .

Ps2 was said to do 75M poly raw for the GS, and 66M poly raw for the EE(vu1). It clearly does what was said to be done ingame and then some.... some have gotten 30M+ ingame, which is FAR more than the 15-20M polys sony said would be feasible ingame...

Prior to the announcement many said 5M polys ingame would be the real deal... but those like me believed FAR more would be achieved.... Today many believe 1Tflops.... but many like me... again believe... FAR MORE will be achieved...

The chip in Figure 6 might not be for PlayStation 3 as the Broadband Engine and Visualizer might be even better than that

Indeed, in less than 12months the PS3 will likely be announced and all will be revealed...

 

[Evil_Cloud]

Don't expect the hardware to be THAT good, because I heard sony PR talking before... Remember the earliest PS2 game material? They said it was in-game but actually it was CG.

 

[Panajev2001a]

Ps2 was said to do 75M poly raw for the GS, and 66M poly raw for the EE(vu1).

Sorry for the anal correction and my language

66 MVertices/s RAW for the EE was considerign both VUs... but in a way you are right... if we consider a simple Transform on VU1 can take only 5 cycles ( on VU0 it would still take 7 cycles as we do not have an additional FMAC and FDIV ) you could achieve that RAW number using the VU1 alone

 

[vers]

Ps2 was said to do 75M poly raw for the GS, and 66M poly raw for the EE(vu1).

Sorry for the anal correction and my language

66 MVertices/s RAW for the EE was considerign both VUs... but in a way you are right... if we consider a simple Transform on VU1 can take only 5 cycles ( on VU0 it would still take 7 cycles as we do not have an additional FMAC and FDIV ) you could achieve that RAW number using the VU1 alone

it is a simple transform on VU0 or VU1:

;----------------------------------------------------
;trans for 4 vertex
;----------------------------------------------------
mulax.xyzw acc,vf4,vf29x nop
madday.xyzw acc,vf5,vf29y nop
maddz.xyzw vf1,vf6,vf29z nop
mulax.xyzw acc,vf4,vf30x nop
madday.xyzw acc,vf5,vf30y nop
maddz.xyzw vf2,vf6,vf30z nop
mulax.xyzw acc,vf4,vf31x nop
madday.xyzw acc,vf5,vf31y nop
maddz.xyzw vf3,vf6,vf31z nop

its 9 cpu cycles for 4 vertex
1 vertex=9/4=2.25 cycles !!!

vu0+vu1 = 266 millions vertex transform /sec
cpu core about 50 millions transform/sec

= about 300 millions vertex transform/sec !!!!!

 

[Panajev2001a]

Sorry vers, but the lowest transform ( which uses perspective divide IIRC ) cannot go down lower than 5 cycles on VU1 and that is using the help of the FDIV in the EFU...