IBM backstabbing sony and neglected Apple ? (long read)

[ADEX]

I did an updated article on the Cell a while back and I got an IBM guy (one of the Cell architects) to read over it for accuracy.

I didn't have to change much but this subject is an area I had to correct, he said it was very much a joint project.

If you even look at some of the presentations you can see it was designed for multiple purposes, requirements are quite different in different areas and IBM isn't an expert in all of them.

BTW You may also note the original patent credits a pair of Sony guys as the inventors.

 

[blakjedi]

Well, I'll do my own bolding now for a moment.

In fact, they almost did go it alone...
Well Toshiba wanted to do just that - so the superiority wasn't clear, IMO.

My friend, this is easy. If we stop right here, we realize something. "Almost" means they could have but realized that they "shouldn't". They realized that they shouldnt cause there was something better. What was better wasnt THEIR design obviously.

Also being companies that specialize in 'DSPs' as such does not inherently make them poor contributors to the project.

True. Texas Instruments is one of, if not the best DSP maker on the planet and their contribution to any semiconductor project should be taken seriously. The same for Toshiba.

Let me ask you, do you consider the SPE's elevated DSP's or dumbed down complex cores? There's a middle ground here somewhere, and it's a middle ground where I could definitely see Toshiba and Sony's expertise in their own respective areas of chip design as having contributed to the overall design in very significant ways.

You may be right in that I do not have enough knowledge of IBM's in-house DSP designs from whence the SPE design may have derived. This is in direct opposition to what we KNOW about PPC/PPE cores.

Also you mention chipset - but since it's really Rambus tech which is the key kynchpin of the chipset (if indeed chipset is what you meant), I don't see how it's anything beyond what Sony and Toshiba would normally be capable of - both very familiar with Rambus afterall.

No I mean the key ICs.

Most on this board probably know my disdain for analysts at this point; I unfotunately can't make an exception for your friends.

That is CLEAR. LOL. Knowing analysts and having worked in one of the big financial house for a time and also being involved in the dot come/ VC era... analysts make plenty of mistakes...as do treasury secretaries ( or chancellors of the exchequer in some places), CFOs, CPA and everyone else. However in truth analyst information is much greater, more confidential and sooner than all of our information so take that for what you will.

Dave B from this site has the same type of access to information that analysts do except that he doesnt try to predict markets or marketshare with his information. But his information is typically accurate and ahead of our knowledge curve. Same with them.

But its no skin off my teeth...

 

[therealskywolf]

Titanio you are correct. Now tell me, in what area do you think Cell is worse than Xenon or an AMD processor or Intel PRocessor.

IF you think it ain't "perfect", then you know that there's something not up to snuff in it.

Well give me your thoughts, i realize my response was more of a rant than anything and i accept that it was wrong.

Actually lol F it. Don't answer it, i don't need it. I already know the answer.

 

[Carl B]

My friend, this is easy. If we stop right here, we realize something. "Almost" means they could have but realized that they "shouldn't". They realized that they shouldnt cause there was something better. What was better wasnt THEIR design obviously.

They got talked down, for one reason or another. Doesn't mean they couldn't have been right! (but of course, I don't care much either way)

That is CLEAR. LOL. Knowing analysts and having worked in one of the big financial house for a time and also being involved in the dot come/ VC era... analysts make plenty of mistakes...as do treasury secretaries ( or chancellors of the exchequer in some places), CFOs, CPA and everyone else. However in truth analyst information is much greater, more confidential and sooner than all of our information so take that for what you will.

Oh well, aren't you so cool?

Hey I was all up in the dot-com/VC era as well my friend, and have several friends in analyst positions also (though they don't cover tech, so maybe yours are cooler). Honestly my own cynical take on analysts comes from the fact that although they do in fact have greater *access* to information than the rest of us, they get into these ruts where they in fact put forth minimal effort, since they know people aren't really going to hold their feet to the fire. When you work long-ass Wall Street hours, you tend to get a little quick and a little sloppy with your analysis is what I've seen. Granted not everyoen fits this bill I'm writing.

I mean whatever, there's a half-full, half-empty way to view market/idustry analysis; but it goes without saying that a lot of us on this board have a lot more passion for the analysis than do some of the analysts that, for lack of a better word, are 'stuck' with it.

We'll agree to agree on the one thing I think we both have no qualms on saying: that regardless, IBM played a very central role.

 

[PC-Engine]

Here's some history for you. Toshiba wanted CELL to be comprised of SPEs only with no main core. Now tell me this, if this was their original plan why did they go to IBM?? How do you think this orginal concept would work if it had gone through? How good would the chip be?? How would developers feel about having ONLY SPEs???

 

[Carl B]

PC-E everyone who wants to, knows that history - afterall I linked to it in post #27 of this thread. I'm all for discussing what you're positing, but frankly I think it's a different thread topic altogether. We'd have to know what those cores were like firstly, as it stands to reason that they probably would have been different than the SPE's to one extent or another.

Plus, did they ever exclude a main core (albeit less robust) from the design-philosophy?

Going back to the past:

Later 3 companies had meetings to discuss the architecture of CELL. The target performance of the project was 1 TFLOPS. Toshiba proposed Force System that has many simple RISC cores and a main core as the controller.

And this is what YOU wrote PC-E!

I read that the original plan was to use a MIPS core as the main core with SPEs dangling off of it.

 

[PC-Engine]

Yes that's what I wrote but then ONE corrected me saying the orginial plan was no main core. IBM responded by saying there should be a main core based on a PPC when their proposal of an all PPE core CELL was rejected since IBM got pissed off. They were saying it had to have a PPC core in it otherwise there would be no point in them (IBM) joining the group.

Actually, Toshiba's configuration is heterogeneous where it has a main core ("master processor" in the patent) and other cores ("processor elements"), but somehow one embodiment may lack a main core.

Regardless, even if it had a main core why would they need IBM if they were uber CPU architects???

 

[Carl B]

Well if he corrected you, it must have been in a different thread, because that first quote is his own translation afterall.

EDIT: Well, I see you're linking to something (or One is) but that link doesn't give me anything useful. Can you direct to the thread? That seems to be a different thread so might as well go re-read that one as well.

(@below: yeah yeah yeah PC, give it time for edits or we'll have nothing but one liner posts - and a lot of them!)

 

[PC-Engine]

Well if he corrected you, it must have been in a different thread, because that first quote is his own translation afterall.

Read above. It's in the SAME thread a few posts down.

 

[Carl B]

Read above.

Ok well now, read my edit above.

 

[Shifty Geezer]

Now tell me this..., if this was their original plan why did they go to IBM?? How do you think this orginal concept would work if it had gone through? How good would the chip be?? How would developers feel about having ONLY SPEs???

How's about before I answer your question, you answer mine...

If PPC is serviceable and well known then go with it. What logical reason is there not to do it that way?

See post number 56 for context. No-one's denying IBM's invaluable contribution. It's your idea that they did all the work and Sony+Toshiba did absolutely nothing which everyone of sense is having trouble with.

 

[Carl B]

What the hell PC-E, your post came *after* his! And no wonder why, as here's his full post:

Buy it every 2 weeks, thx


Actually, Toshiba's configuration is heterogeneous where it has a main core ("master processor" in the patent) and other cores ("processor elements"), but somehow one embodiment may lack a main core.

Link

Good god man, talk about selective screening to 'prove' a point.

Plus the whole 'why would they need IBM thing' is explained at the beginning of that thread as well: Sony simply wanted to bring them on for their expertise, not over some sort of inherent roadblock on 'main core, sub-core' implementation.

 

[blakjedi]

but it goes without saying that a lot of us on this board have a lot more passion for the analysis than do some of the analysts that, for lack of a better word, are 'stuck' with it.

Dude you are like the great swami or something! That exact point was what I left out my last post. I forgot to mention that while they have information its just that to them... numbers, analyses, projections blah we as gamers have the pulse of the industry...

For example they may project one console or the other will "win" based on x blahs and y blehs... but we all know that Sony is in a very good position to go for the real triple crown of gaming i.e. three generations of hardware and fifteen years of dominance. This feat is unlike any other in nearly any other industry...

 

[Carl B]

Blakjedi you and I are on the same wavelength or something.

 

[blakjedi]

Blakjedi you and I are on the same wavelength or something.

Heh some days I argue nuance... others i argue big picture... today is big picture day!

I would really like it if everyone stoped arguing about "better" and started discussing capability. Like based on locking caches and directly streaming data, what will that allow xenon/xenos to do that we have never seen before? There is a reason why IBM and MS designed a system that could do that, knowing full well the float math dominance that cell would have... to what benefit?

Conversely, the design choice for SPE local stores has been controversial; when is local store better than a cache and what did Sony/IBM expect to be able to with that configuration?

I personally think its 6 in one hand...well... you know the rest, but both systems will deliver amazing if not unique games to the industry based on technologies that we might never have seen in the PC space otherwise.

Our first taste will be on november 22nd 2005 and EVERYONE should be excited for it. At the very least even if you dont like or are particularly thrilled with the lineup you should excited by the fact that by the end of its lifetime you may see things on x360 which are technically unlikely (as opposed to impossible) on any other system due to US.

As B3d ers thats thrilling in and of itself.

Horespower wise, PS3 will give a glimpse into the most finely tuned rocket processor (blast processing anyone? ) that software has ever been placed on... a milestone in and of itself.

Hopefully everyone will enjoy the technology more than the debate cause in the end thats what the tech is for...

 

[Edge]

One Power core to manage many spe's. The spe's are the main technical advantage to CELL, so the loss of a PowerPC core to the competition is rather meaningless, and so no reason to blame things on Ken. They could have used any generic processor to manage the SPE's, and MS could have taken any processor, and asked for a multi-core product.

Sorry, but this is just nonsense.

 

[SedentaryJourney]

Heck look at the EE, it's just a MIPS core with VU tacked on.

I know! Anyone with absolutely no engineering talent whatsoever could totally do that in like 3...maybe 7 seconds man!

 

[The_Standard]

What are you talking about? SONY/Toshiba had almost NOTHING to bring to the table except some eDRAM tech, some process tech, and some vague ideas about using many small simple math cores in Force Project or some such. In fact SONY has no CPU design experience whatsoever that's why they hired a former NEC supercomputer processor architect. Toshiba only licenses MIPS cores and fabs them. They've not designed any CPU based on their own design. Everybody knows IBM did the bulk of the *actual* design for the CELL chip while SONY/Toshiba came up with some concepts that had not even been proven. Heck look at the EE, it's just a MIPS core with VU tacked on. It's measely cache size proves they didn't know what the hell they were doing. Now you're saying IBM needed their awesome CPU design skills???



The truth? Says who? Where's the proof?

PC-Engine:

You have no idea what you are talking about. The twin "ring" EIB design, the very backbone of the Cell, came from KK-san's own head.

 

[Fafalada]

Heck look at the EE, it's just a MIPS core with VU tacked on.

At least the VUs had a decent instruction set.
Thanks to MS and Sony going with IBM we're now all stuck using crapped "real man's" SIMD for the next 5 years.
People bitch about lack of OOOe in the new CPUs - yet that was never really there in the first place. Meanwhile SIMDs have actually taken a massive step backwards from previous generations, and that's somehow supposed to be alright.

 

[blakjedi]

At least the VUs had a decent instruction set.
Thanks to MS and Sony going with IBM we're now all stuck using crapped "real man's" SIMD for the next 5 years.
People bitch about lack of OOOe in the new CPUs - yet that was never really there in the first place. Meanwhile SIMDs have actually taken a massive step backwards from previous generations, and that's somehow supposed to be alright.

Faf - Im interested in learning more. How are SIMDs a step backward? Are you referring more to the architecture of 360 or cell here? Arent massive arrays of SIMDs the norm anyway with graphics processors?

Just questions to which i do not have answers.

Thanks friend!