PS3 & Cell Stations no local HDD storage

[David_South#1]

I've said a couple of things before about Cell.
Most of them I still believe to be true.
Some are confirmed in early news about the 5 solid state papers.

Confirmed
90nm (I stated it would start in the most familiar process 90nm, not 65nm.)
Each Processor Element is a seperate chip.
(I said BE is a MCM. But others disagreed.)


Here is something we never discussed but I wanted to mention.

PS3 and Xbox2 are not expected to include a hardrive in the box.
Back when I posted my history on how Xbox2 developement was mixed with the Apple G5's that were shipped to Microsoft, I did a little reading about Xbox2 stuff, rumors etc. One of the things that caught me was the mentioning of Isreal, with USB Flash keys (I believe these will be used so that gamers can carry their ID's & accounts to any station supporting a connection to the LIVE Network. Inluding Networked Arcades that have Thumb scanners on the machine.) Another was mention that Microsoft was also trying to drop the HDD.

Paraphrased,
Microsoft said yes we want to get rid of the HDD.
Microsoft also said it was the biggest requirement of its surveyed users.

In other interviews with Sony I saw much the same thing. Minus the fact that they didn't include an HDD to start with or successfully intergrate one later. (I loved my PS2 despite that and some sorry toe stomping.)

So I visited IBM and Linux @ IBM looking for stuff and followed the 5 part GRID gaming and commerce stuff at IBM. They were at Part 4 at this time when I noticed how central everything was to the servers and not the devices. How cell phones user and other devices used the network to suppliment its storage. "Cell used in Cell-Phones" is a quote that rattled and left a spark.

So I did some searching to learn more and found something called iBoot.

I'm just posting this to predict an iBoot like model will probably be used by both companies. It aids in stopping people from using a Local storage device to pirate games and defends against Linux zealots trying to prove even toasters can run that OS. (Though they may still figure out how to run using a live image.)

I have a few other ideas on how network storage is in a lot of ways smaller and cheaper and more flexible to maintain. But mostly how it is safer for the provider nd network.

At the University of Florida (where I live) they already use a model a lot like this. At each boot the computer receives a download from the users storage space on the network. The PC we use just seem to have some sort of live image or ROM setup that our data is overlayed onto, but does not perminently change. When we are done with a station and log out, all our info is saved to some central network storage place and the PC is completely clean of history or any crumbs for the next user to find. If they have an account with storage space then their stuff is overlayed or loaded into the local station. It's too quick for and build to be completely installed each time. So I figure it is OS an Image of some sort with personal stuff plugged in and removed by each logon.

Anyways I hope I've helped give you an idea of the concept and shared an idea, iBoot, that doesn't seem to be something future guesser's are aware of.

 

[PC-Engine]

Confirmed
90nm (I stated it would start in the most familiar process 90nm, not 65nm.)
Each Processor Element is a seperate chip.
(I said BE is a MCM. But others disagreed.)

I haven't read any of the papers, but were those disclosed in said papers?

 

[Guden Oden]

Confirmed
90nm (I stated it would start in the most familiar process 90nm, not 65nm.)
Each Processor Element is a seperate chip.
(I said BE is a MCM. But others disagreed.)

You seem to forget these cells aren't what will end up in the PS3, so your "predictions" are still UNcomfirmed at best...

 

[one]

Each Processor Element is a seperate chip.
(I said BE is a MCM. But others disagreed.)

At the University of Florida (where I live) they already use a model a lot like this. At each boot the computer receives a download from the users storage space on the network. The PC we use just seem to have some sort of live image or ROM setup that our data is overlayed onto, but does not perminently change. When we are done with a station and log out, all our info is saved to some central network storage place and the PC is completely clean of history or any crumbs for the next user to find.

Seems like you've never heard of Unix and X Window System server since 1984 and all the concept of workstation

 

[David_South#1]

PC-Engine,
I'm just online Sunday evenings, but I'll scout on the news I'm reffering to.

iBoot at IBM is often refered to as part of SCSI.
Another play on this idea, since i'm just playing on light details in memory is using you PC or some other storage memium not within the console.

Option A. Storage on far away server. (Some of that already happens.)
Option B. Storage on local PC. (Possible prone to hacking.)

The notion that neither company wants to include an HDD seems apparent.
So I'm just exploring possible alternatives by starting this thread.

One,
I've heard of a more than a few other examples including BSD, Linux. The point isn't that the method was new, it was that these gaming providers may REALLY be seeking a means of not including an HDD.

 

[David_South#1]

Confirmed
90nm (I stated it would start in the most familiar process 90nm, not 65nm.)
Each Processor Element is a seperate chip.
(I said BE is a MCM. But others disagreed.)

I haven't read any of the papers, but were those disclosed in said papers?

Here is the first Hint.
This one confirms the 90nm comment.
http://games.kikizo.com/news/200411/091.asp

I'll look to confirm the that each PE is a seperate chip.

But reading the quote you will see they say that Cell comprises only ONE 64bit unit to a chip. The PU is a 64bit core, the APU are 32bit. 1 PU and a few APU equal one chip.

So on this point you will probably find me right again.

 

[David_South#1]

OK, here it is the original news on 90nm and PE = Cell.

http://forum.pcvsconsole.com/viewthread.php?tid=13702&page=1
http://www.eet.com/semi/news/showArticle.jhtml?articleId=54200580

"At the top level, the architecture appears to be a pool of "cells," or clusters of perhaps four identical processing elements. All of the cells in a system — or for that matter, a network of systems — are apparently peers. According to one of the ISSCC papers on the Cell design, a single chip implements a single processing element. The initial chips are being built in 90-nm SOI technology, with 65-nm devices reportedly sampling."

There it is one, So laugh at yourself. 8)

I've felt a bit bitter towards this forum or at least some of the members for how hard it has been to convince you of the exact things now being stated.

Also I suggested on chip system security features, by mentioning the coincidental bit size of Public Key encryption and how it could be processed in a single cycle by a PE. At the time people blasted me or just mocked such comments and down played any idea of my suggestion.
http://www.psinext.com/itex.php?iid=407

Though we still don't know what sort of features and support Cell will provide it is stated that the security support is on chip.

 

[nAo]

Emh...I believe almost everyone here was aware that this first CELL implementation has just one PE per chip (after STI press releases..)
We still don't know if future implementations (65 nm?) will have more than one PE per chip.

 

[one]

David_South#1, read the comment by Guden Oden... Besides CMP and MCM are not mutually exclusive, if Cell has MCM capability like POWER4, so your conclusion doesn't make sense. I mean your comment only restates everyone knows about the 1st-gen Cell Processor already and is out of context

 

[PC-Engine]

So when DeadmeatGA said that CELL in PS3 would be only one PE or two PEs at most he was right. I guess that 1 TFLOPS figure is science fiction afterall.

 

[Ty]

Where did that 1TFlops figure come from anyhow? Sony itself? Or just some people speculating?

 

[Phil]

it's quite ironic that the most vocal people about such figures are usually always those with an anti-Sony bias - and then later, blame the company for hype they're partly responsible for. :?

 

[wco81]

Sony will probably push some form of Memory Stick storage for saves. Too lucrative.

Plus they can't assume every console will be connected to the Internet. For all the PR releases about XBL and SOCOM II, the fact remains that only a small percentage of people who own the Xbox and the PS2 get online with their consoles.

So the idea of online storage as the only form of storage is a non-starter.

It would be interesting if they were able to make hybrid Blu-Ray discs with ROM data and a small portion reserved for rewriteability. Supposedly, the cost delta between a read-only blue-laser diode and a read/write blue laser diode isn't that great.

On the other hand, HD-DVD, owing to its much thicker surface laser, requires a much more powerful blue-laser diode for write operations.

But initial BD-RE media is expected to be pretty pricey. So even putting say 5 GB of rewriteable media (out of a 25 GB layer), might still be too costly.

 

[David_South#1]

nOa,
I've argued that because 90nm was my expected start scale, only one maxed PE per chip.
65nm will follow, I've said that back in my /seventh/ post to this fourm.
But would you honestly NOW guess that something like the Broadband Engine is what the PS3 might use??

Since before STI, but after the patents I've been arguing that PS3 would receive one or two PE Cells and no more (by two it is meant, one VS & one PE). Everyone and their brother (so it seems) on this forum has argued this is wrong, that Fig 6 the 4PE = BE is the correct example of what people should call Cell. I have also argued that Cell is a modular design philosophy and because PS3 is using it does not mean it is pictured in the patents.

Why have so many at this forum been locked on Fig. 6?

I recall discussing with you the debate on eDRAM. Many here were at that time stuck on 64Megabytes per PE.
Do you and others still believe 64MB of eDRAM?
I argued that only SRAM was on the PE because eDRAM would slow it down.
That the only eDRAM is 32MB on the VS as image cache.
Updated, I still see VS as an overview for the GPU. That the Pixel Engine is nVidia's Design.

one,
Guden ~ The specific processors of PS3 are unconfirmed. But to you what else does that mean>?
I'm using MCM as an adjective. Multi Chip Module. When manufacuring permits, they may be combined.
What is CMP & why must cell compete with Power4? Single precision VS double? You make no sense?
If my comments only restate what everyone knows, why in the past have so many replies disagreed?

PC-Engine, and before him I said the same.
But I do not believe "a one Teraflop cell" is science fiction. Just science future. 10Ghz in 2010 to be precise.
This is the way I originally digested the news way back from the Nekkei Ken talks Cell story.
Before that a BE has the same /potential/ and may be why so many here think of it.

Ty,
IBM and Sony, but Ken Kutaragi was first to says Cell will do 1TFlOp/s.
The majority of those I encounter see Cell as an actual chip. Cell is a modular design architecture.
LEGO or Fish is the easiest example I can offer.
No matter how many of them or in what sort of sentence, it is still Lego, Fish, or Cell.

wco81,
"So the idea of online storage as the only form of storage is a non-starter."
Your point is absolutely correct. Option (A) is aimed /exclusively/ at online users.
But don't ignore option (B).
Over the years it has increasingly become normal that a console owner has a PC in the house too.
This would also overcome most of the online bandwidth problems and resemble my user experience.

The Hybrid ROM/WRITE disc seems to me /way/ expensive.
ROM is a Physically pressed or relief layer. RW is a chemical spray.
Trying to apply both on a single layer is beyond IMO reasonable consideration.
Having two seperate layer, though innovative IMO requires a process turnover slowing production.
Then there is as you say, consumer cost.