Japanese article about PS3 backwards compatablilty

xbdestroya said:
Some English-speaking media seems to be going a little further into the article and indicating that PS3 has an ample supply of Flash-memory on board to keep the OS and all updates off the hard drive, in order to allow for their vision of third-party hard drive compatability.
The author of the article in the UltraOne magazine which IGN is talking about is, Munechika Nishida, and he clarifies some points of his article in his own blog in the OP of this thread. The flash RAM is for the OS and its updates, while the HDD is for individual game patches.
 
Last edited by a moderator:
INKster said:
I thought the Xbox 360 also had a 16MB Flash ROM chip onboard to store the OS...

Truthfully I forget, but doubtless that's the case. Afterall, it has to be able to boot without a hard drive installed. I'm not saying this is unique to PS3... it's not. But it's another peice of information for the PS3 info-tapestry.
 
one said:
The author of the article in the UltraOne magazine which IGN is talking about is, Munechika Nishida, and he clarifies some points of his article in his own blog in the OP of this thread. The flash ROM is for the OS and its updates, while the HDD is for individual game patches.

Well, I didn't think to the contrary at all - obviously game patches would eventually overwhelm any flash solution - but just wanted to know if the flash was confirmed.

It might be frustrating to upgrade your harddrive though and sit through an our or two of patch downloads. I guess that's how it is now anyway, for FFXI and the like even on PS2.
 
Last edited by a moderator:
I guess you'd only download patches on games as you use them. And if they're massive patches, I'll be furious. I don't want a flamin' PC for gaming, and if that's what Sony have in mind for 'PC functionality' (buggy games released before being finished and then patched with multi-megabyte fixes) they can keep it!

As for flash for OS, I suppose that's no different to PSP, but will it mean the entirety of Linux will be on Flash? Or will Linux be HDD based (talk was of Linux being on the HDD) and the OS in flash will be the X-media bar interface and basic media functions including web browser?
 
sonyps35 said:
They really need to not spend money on this.

I really dont know what Sony is doing sometimes these days.

BC is nice, if it were snap your fingers I'd rather have it than not, but I wouldn't invest any significant resources whatsoever in it, if I was running Sony or any videogame company.

Just, wow.

Considering the PS3 costs $600, they better give users a reason to stick with the playstation brand. If playing their old games required them to keep their old systems out, they could just as easily switch to Xbox 360.

EE+GS is roughly the size of an Athlon64 chip at 90nm; so it probably costs them about ~$50 to fab it.

1. No way an Athlon 64 costs $50 to fab, at least I've never heard of microprocessor costs being that high.
2. With such simple chips, EE+GS would be a good candidate for 65nm. Also, since the tech is so old, yield rates could be near 100%.
 
Fox5 said:
1. No way an Athlon 64 costs $50 to fab, at least I've never heard of microprocessor costs being that high.
2. With such simple chips, EE+GS would be a good candidate for 65nm. Also, since the tech is so old, yield rates could be near 100%.

Hey whatever, I'm just going off some random Intel chip cost estimates that came out a year or two ago and adjusting them for AMD. :cool:

If you know better, I'm more than happy to default to your knowledge on the subject. I certainly don't profess to know the internal costs for either AMD and/or Sony.

I don't know exactly what you mean by never heard of microprocessor costs beign that high though; do you mean for AMDs K8, or...? Because there are definitaly chips out there that exceed $50 to fab.

But yeah - they're about the same size on 90nm.
 
I recently posted some pricing info on Intel and AMD chips in regards to manufacturing costs in my PS3 marketing thread. There are a number of sources, so they seem pretty accurate.

In 2002 AMD was at $21 a CPU and Intel at $22. Last year, in late 2005, Intel's desktop chips were at $40. CPU makers get the benefit of 1 or 2 designs across the board and segregating the chips on frequency so that does impact average cost because they are not throwing away a lot of chips that don't quite make it up to spec.
 
xbdestroya said:
Hey whatever, I'm just going off some random Intel chip cost estimates that came out a year or two ago and adjusting them for AMD. :cool:

If you know better, I'm more than happy to default to your knowledge on the subject. I certainly don't profess to know the internal costs for either AMD and/or Sony.

I don't know exactly what you mean by never heard of microprocessor costs beign that high though; do you mean for AMDs K8, or...? Because there are definitaly chips out there that exceed $50 to fab.

But yeah - they're about the same size on 90nm.

Strange how intel can fab a chip that is ~440 mm^2 at $140 per chip (a version of Itanium 2) then ~86 mm^2 chips cost them $50+ to fab (assuming yields for each are good)...

Probably EE+GS@90 nm by this fall will probably cost them $15-20.

I cannot believe that the margins would be so thin: retailers online sell 90 nm Athlon 64 3000+ CPU's (bottom of the line, but hey they are new... they are 90 nm :D) for about $100 (I do not think AMD sells it to them at $99 per chip at 1,000xshipment quantities) and to AMD each of the cost $50+ just to manufacture each chip excluding taxes, transport, retailers margins, etc... ?
 
Last edited by a moderator:
Panajev2001a said:
Strange how intel can fab a chip that is ~440 mm^2 at $140 per chip (a version of Itanium 2) then ~86 mm^2 chips cost them $50+ to fab (assuming yields for each are good)...

Probably EE+GS@90 nm by this fall will probably cost them $15-20.

Hey you know what, awesome. :cool:

I was answering the die size question, and taking a stab at the pricing. Like I said I'm not coming at these costs with any inside information, just have to work with whatever docs leak from whatever year they leak. It's an inexact science! But I give myself credit for trying. ;)

But I mean, where does that Itanium fab cost come from itself?
 
xbdestroya said:
Hey you know what, awesome. :cool:

I was answering the die size question, and taking a stab at the pricing. Like I said I'm not coming at these costs with any inside information, just have to work with whatever docs leak from whatever year they leak. It's an inexact science! But I give myself credit for trying. ;)

But I mean, where does that Itanium fab cost come from itself?

It was from an article by the Micro-Processor Report magazine (MPR) and I have first found it quoted by Paul DeMone over at RWT and actually according to what I remember reading over at RWT's (realworldtech.com) forums (DeMone has friends at Intel and some Intel folks regularly post over there) the real cost was a bit nearer $130 per device and not $140 IIRC (if I remember wrong add $10 to each cost because I remember the cost delta between the two).
 
Panajev2001a said:
It was from an article by the Micro-Processor Report magazine (MPR) and I have first found it quoted by Paul DeMone over at RWT and actually according to what I remember reading over at RWT's (realworldtech.com) forums (DeMone has friends at Intel and some Intel folks regularly post over there) the real cost was a bit nearer $130 per device and not $140 IIRC (if I remember wrong add $10 to each cost because I remember the cost delta between the two).

Panajev that is a great point of reference then, because there has been a lot of Cell and RSX fab-cost guessing over the course of the last several months; and even though Itanium benefits from cache-madness on the die, I think at their (Cell/RSX) reduced sizes and redundant quad/SPE, this can help us get a handle on what a realistic price estimate for these chips might be.

Granted there are several other factors involved, but...

Anyway thanks for that info, and thanks for the revised estimate on the EE+GS.
 
xbdestroya said:
Panajev that is a great point of reference then, because there has been a lot of Cell and RSX fab-cost guessing over the course of the last several months, and even though Itanium benefits from cache-madness on the die, I think at their reduced sizes and redundant quads/SPEs, this can help us get a handle on what a realistic price estimate for these chips might be.

Sure, Itanium 2 benefits a lot of the cheap redundancy that having LOTS of SRAM on chip brings and IMHO it is one of the good ideas that the IPF line of processor has going: moving the problem out of expensive yield and power consumption unfriendly blocks and focus on keeping latency down by integrating some world class cache hierarchy on chip.
 
450mm wafers sooner than later?

Gubbi said:
That article is from 2002. Intel has moved to 300mm wafers since then, with AMD following, so cost is likely to be slightly lower for Intel. Larger die, more complex process, but more dies/wafer.

In semi-related news:

Geek-dot-com said:
Intel is not holding back on letting the public know about significant future technologies, and--whereas many Intel employees are at a loss to understand how it is possible that they're able to talk about things they could not discuss just a few weeks ago--it's all good stuff for us.

In 2007, Intel is planning to be ramping to 45 nm production in at least two fabrication facilities, but what's even more interesting is that at least one of them may soon convert from 300 mm (11.8") wafers to 450 mm wafers (17.7") in 2008. These notably larger wafers increase literal surface area by 225%, with the potential to obtain over twice the number of chips per wafer. Add in that Intel will be at 45 nm, meaning each chip will be smaller still, and that's a tremendous number of potential product from each wafer. If the die size were to shrink down to 50 mm[SIZE=-2]2[/SIZE], that would equate to roughly 3,100 chips compared to 700 available today using 100 mm[SIZE=-2]2[/SIZE] on 300 mm wafers.

Those numbers represent a 400%+ increase in the number of potential products available from the processing of a single 450 mm wafer. That will be a significant advantage for Intel.

Some sensationalism there, but the big news is that it seems Intel may be moving some fabs over to 450mm wafers in 2008. I thought the timeline for 400mm was more in the 2012 range, so I am not sure how accurate as it is a blurb and not news. An interesting comment is that only 25% of semiconductors have moved to 300mm.

Anyhow, carry one :D
 
I'm not very knowledgeable about producing CPUs, but don't you have to plan on having an EE+GS in the system to make it work well? If so, Sony has been planning on this for some time.
 
Yeah I think that EE+GS has been planned for a while, and this emulator project has likely been running concurrent.

Still though there must be a hardware consideration built into the roadmap other than just EE+GS--->emulator, right? Because once that's gone, emulator or no, I just don't see how GS bandwidth could be emulated; will there be a replacement chip for the EE+GS, hardware considerations on RSX, or indeed would total emulation be a feasible achievement?
 
I think the solution already is a hybrid of sorts of software and hardware-supported emulation, with simply a heavy dependency on the presence of certain hardware for now. Over time they'll reduce that dependency..maybe later the EE will be removed altogether, for example. But I have my doubts if they'll ever remove all the hardware and solely rely on the core PS3 hardware.
 
Well it's pretty clear the EE is out of there at some point; but the indication seems to be they're looking to toss the GS out as well (since they hint at the entire EE+GS being removed), and I'm just wondering what the implied asterisk is next to the GS removal.
 
Shifty Geezer said:
As for flash for OS, I suppose that's no different to PSP, but will it mean the entirety of Linux will be on Flash? Or will Linux be HDD based (talk was of Linux being on the HDD) and the OS in flash will be the X-media bar interface and basic media functions including web browser?
Linux has been always associated with HDD by Kutaragi. (Linux is the driver of HDD, Linux is pre-installed in HDD) The main reatime OS resides in the flash RAM. Probably Linux will be just one of applications on PS3 along with games but have less access privileges.
 
Back
Top