PC Watch: PS3 Article

["Nerve-Damage"]

PC Watch: Full article can be found here.

Need a better translator...

*BabelFish Snip*

XDR DRAM finally in the design of the 3rd generation in shipment

The circumstance that it could spend time on verification by the fact that the schedule of PS3 slips is the same concerning the tip/chip around. For example, PS3 is completely new memory, "XDR DRAM" is used in main memory. Rambus develops XDR DRAM, Toshiba, 3 corporations of エルピーダ and Samsung Electronics supply. Because it is the technology whose also XDR DRAM completely is new, it was one of the risk primary factor.

But, with favor of the gap of schedule the DRAM vendor, the development of XDR DRAM, from PS3 came to the point of finishing rather before. Although the tip/chip itself has been completed, it is the case that you have stood by not to insert in real mass production to PS3 production. Depending upon the DRAM vendor, without entering into mass production there is also a place where it moves to the tip/chip design of the 3rd generation.

As for PS3, first it was the expectation which adopts XDR DRAM of the 512Mbit item. At this point in time, it is presumed that of memory sharing architecture was thought, but SCE reconsidered the system, XDR DRAM modified specification to the 256Mbit item. Because of that, the DRAM vendor advanced preceding the design of the 256Mbit item. However, the time where the design of the 256Mbit item completes, requirements specification of SCE side returned to the 512Mbit item again, (around 2004 November). Then, the DRAM vendor deferring the production of the 256Mbit tip/chip, did the design of the 512Mbit item, arranged mass production system. However, because to being in the stage which even 1 year from before is inserted in mass production, the production of PS3 itself does not start, so far as for XDR DRAM you slept.

For example, when エルピーダ is taken as example, the B die/di of 100nm the sample was done from 2005, but after entering into this year, it has moved to C die/di of 90nm. So when it does, 256Mbit 1 generation, as for XDR DRAM which 512Mbit mass-produces in 2 generations becomes the design of 3rd generation. By the fact that process advances, the speed yield which feels concern with XDR DRAM (yield rate) problem, it is the expectation which is rather cancelled in principle.

The fact that schedule is late vis-a-vis Cell and XDR DRAM which precede, is GPU "RSX of the NVIDIA design which development starts from after (Reality Synthesizer)". But, RSX based architecture G70 which NV40 (GeForce 6800) from is really succeeded (GeForce 7800 GTX) is type architecture. Because also 2 years it continued it is the architecture which withers, based architecture itself the risk primary factor is little. Funny perhaps, seeing it is thin, but, certainty is high selection. Of course, if there is a problem at stage of the physical design of the tip/chip, story is another, but either the story such as that is not audible.

As for being visible from such circumstance, the imbalance of of the hard of PS3 and development of the software is. Like the laser of BD, SCE control if the element which cannot be divided was excluded, the individual component of hard advanced at pace of that appearance. Although that it is, software side has not been attached. Although more the resource and approach run period should be provided in development of software layer, as the treatment is late, it is visible. This, is shallow probably is the weak point of SCE whose experience as a software platform company still.

 

[London Geezer]

*headache*

one to the rescue please!

 

[pipo]

...so far as for XDR DRAM you slept.

It won't happen again. I promise.

 

[Shifty Geezer]

*headache*

As far as translations go, I thought this was the best one ever! Is it just saying that the XDR has slowed the process up do to changes in spec? What's the difference between 1st, 2nd and 3rd generation XDR 256 Mb and 512 Mb?

 

[Angelcurio]

God, that translation was worse than my broken english

 

[London Geezer]

This, is shallow probably is the weak point of SCE whose experience as a software platform company still.

Poetry. The "This," at the beginning is charged with emotion. Amazing.

 

[NERO]

My goodness, now we know everything about the PS3 all over again.

Funny perhaps, seeing it is thin, but, certainty is high selection. Of course, if there is a problem at stage of the physical design of the tip/chip, story is another, but either the story such as that is not audible.

Gadzuks, does my brain hurt.

 

[Shifty Geezer]

Anyone hear know what I'd mean if I said 'Star Control II - Orz'?
Any *Happy* *Campers* on this forum?

 

[Bobbler]

Anyone hear know what I'd mean if I said 'Star Control II - Orz'?
Any *Happy* *Campers* on this forum?

Orz is generally an unhappy emoticon from what I've seen.

But looking on google I see it's also one of the races in the Start Control game.

 

[Shifty Geezer]

But looking on google I see it's also one of the races in the Start Control game.

You're not a *Happy* *Camper*. You are a *Silly* *Cow* You blow much *Bubbles*

 

[NERO]

Anyone hear know what I'd mean if I said 'Star Control II - Orz'?
Any *Happy* *Campers* on this forum?

I get the reference. I was once happy but never a camper. Is the article translation a parallel of Orz(inese)?

 

[Bobbler]

You're not a *Happy* *Camper*. You are a *Silly* *Cow* You blow much *Bubbles*

I don't know if I should take offense to that or not!

 

[Inane_Dork]

Anyone hear know what I'd mean if I said 'Star Control II - Orz'?
Any *Happy* *Campers* on this forum?

Heh. I remember that game. They liked saying, "squishy."

 

[Shifty Geezer]

Heh. I remember that game. They liked saying, "squishy."

Thank you! That's some *special source* for you!
Although they only said Squishy in one lot of dialogue, but it was whenever you great them IIRC.

For the record (and those confused by my ramblings) the Orz were a race of aliens who's dialogue was so alien that the universal translator couldn't translate properly, so it used 'best guesses' when you spoke to them, producing such insightful remarks as...

Hello to our *house*. Do you feel *better* yet?
If you are *campers* you will enjoy *the change*, but maybe not yet.
It is best if many happy Orz are coming to your *house*.
Let's *spitting* the fun words for several *pieces* and then surprising things!!!

I guess this game is where Google got the idea for translation of Japanese from

 

[dlm]

Poetry. The "This," at the beginning is charged with emotion. Amazing.

That sentence is actually not too bad. Start with this:

This, is shallow probably is the weak point of SCE whose experience as a software platform company still.

Now move the "is shallow" to the end of the sentence and swap the order of "probably is", then move the comma to after "SCE", finally swap "still is" at the end and you get this:

This is probably the weak point of SCE, whose experience as a software platform company is still shallow.

And aside from a bit of passive voice ("weak point of SCE" instead of "SCE's weak point") is perfectly proper English. Just needed to move things around a bit.

 

[one]

Basically this Hiroshige Goto's article says the hardware of the PS3 has got very mature by now. A typical custom CPU for a game console is developed in a very short time and can't take enough time for validation, but in the case of Cell it could secure over a year just like other computer CPUs. This higher product quality is required for the PS3's Cell since they plan to offer an open programming environment. For a closed game development world, you could issue a notice to licensees and compiler vendors, like "this CPU has a bug that it skips the last branch when a loop is entirely on a cache line, you should avoid it by adding NOP". I/O interfaces were like that too. But it's no more allowed for Cell for the wider market and the PS3 as an entertainment computer.

The 90nm SOI yield for the Cell is expected to be very good. For the 65nm process, a subset of LongRun2 technologies is applied to the Cell. 2 weeks ago David R. Ditzel (Transmeta CTO) said the work with SCEI had almost finished.

XDR DRAM is also mature. It was a new untested technology back then but now Elpida's design has already reached the 3rd generation in 90nm. The design of RSX is a safe bet. Overall the hardware side is in a very good shape due to the delay.

This, is shallow probably is the weak point of SCE whose experience as a software platform company still.

 

[nonamer]

If that is all true, then it begs the question: How long will the PS3 stay at 500-600 dollars and how many can they make? Wouldn't it be reasonable to assume that the PS3 could see a $200+ price drop alongside 15-20 million units before the end of next year?

 

[Carl B]

If that is all true, then it begs the question: How long will the PS3 stay at 500-600 dollars and how many can they make? Wouldn't it be reasonable to assume that the PS3 could see a $200+ price drop alongside 15-20 million units before the end of next year?

RSX and Cell will drop in cost rapidly enough... this was never really a question. The dark cloud over it all is the Blu-ray drives. But supposedly the diode replication process has finally been ironed out, so we'll see what happens.

 

[patsu]

Basically this Hiroshige Goto's article says the hardware of the PS3 has got very mature by now. A typical custom CPU for a game console is developed in a very short time and can't take enough time for validation, but in the case of Cell it could secure over a year just like other computer CPUs. This higher product quality is required for the PS3's Cell since they plan to offer an open programming environment. For a closed game development world, you could issue a notice to licensees and compiler vendors, like "this CPU has a bug that it skips the last branch when a loop is entirely on a cache line, you should avoid it by adding NOP". I/O interfaces were like that too. But it's no more allowed for Cell for the wider market and the PS3 as an entertainment computer.

The 90nm SOI yield for the Cell is expected to be very good. For the 65nm process, a subset of LongRun2 technologies is applied to the Cell. 2 weeks ago David R. Ditzel (Transmeta CTO) said the work with SCEI had almost finished.

XDR DRAM is also mature. It was a new untested technology back then but now Elpida's design has already reached the 3rd generation in 90nm. The design of RSX is a safe bet. Overall the hardware side is in a very good shape due to the delay.

This, is shallow probably is the weak point of SCE whose experience as a software platform company still.

Thanks one for your translation. I like the mature-ness of Cell. It will be great for homebrew development since support from Sony will be non-existent.
What does LongRun2 buy for Cell 65nm ? Is it used for 90nm too ?

 

[one]

Thanks one for your translation. I like the mature-ness of Cell. It will be great for homebrew development since support from Sony will be non-existent.
What does LongRun2 buy for Cell 65nm ? Is it used for 90nm too ?

The subset of LongRun2 (probably it's LongRun1 equivalent since the full LR2 is not applicable to SOI) gives it lower power consumption. Besides, if the XDR yield is good enough it's likely there are chips that can reach 3.2Ghz with lower power. Then they can lower the output of the power supply unit and cut the cost of the PSU. A PS3 gets lighter, cheaper, cooler, quieter.

The article says LR2 is for 65nm Cell.

Since the OP misses the direct link to the article
http://pc.watch.impress.co.jp/docs/2006/1027/kaigai314.htm