360 Falcon: 50% cheaper to manufacture?

With the new augmentations being made power will be cut by more than half of it's original output. Yes the brick will be gone.

 

Hi all,

I've been reading this particular thread with interest. It was I who posted the original article linked to in the OP - I only found this post after checking out site stats.

I really just wanted to point out some things about the original article, it was written rather hastily at work so perhaps not worded perfectly.

The 50% figure seems to have been latched onto with a fair number of people assuming that refers to the console retail price itself - it doesn't. In the next paragraph, I said to expect a reduction on RRP of around 20%, a more modest and much more likely drop.

Thanks to the OP for the link, nice to know at least someone is reading

 

Hey Bleem - just wanted to point out that no one thinks you were talking retail price here - we're talking manufacturing costs. The expectation of 50% cost reductions is simply too great. When the Falcon story broke a couple of days ago, along with it broke the notion that Microsoft would be able to shrink the chips by 50%. BUT, that's just a hypothetical marker - to get a real-world sense of what the savings might be, you need to factor in what the actual die size is (because node shrinks don't always - in fact rarely - go to the straight halving) and what the yields are. And obviously those two chips are only a piece of the puzzle in terms of manufacturing costs; the rest of the components are affected in a cascading fashion, but it's not the same geometric halving that is even theoretically possible on the silicon, so guessing at it is much more nuanced.

EDIT: Wanted to add here a favorite slide of mine, showing what we rarely get to see - a die migration across an entire generation. Note how even within process generations, further revision refinement still gets done. And note of course how with the traditional Moore's Law node shrinks, it never goes exactly to 50%. This is for Sony and Toshiba on the CMOS process rather than MS and their work with TSMC - no fab process is identical. But it gives an idea of how varied these things can be... much moreso than simply soundbites extolling Falcon at 65nm as some cut-and-dry thing automatically halving the cost of the chips.



The EE+GS itself has of course been further refined since that slide was released.

 

Miniaturization isn't a surprise, but (if Love & Sacrifice is correct) its timeliness is. According to the article, Falcon is about more than migrating chips; it is an end product with half the power requirement of a launch unit.

Slimline PS2s were several years in the making. I think trim 360s will have a similar design chain -- one measured in years rather than months. Gradually replacing older tech with newer ones (using hybrids like Xbox 360 Elite and Halo 3 Special Edition Console) is one of the better ways of implementing change in design/manufacturing while clearing inventory and retail.

 

The more i think about it, the more odd the Elite SKU seems to be. The motherboard design seems very 'interim' and aside for a short term bump in sales I dont fully understand why it exists.

 

IMO, to get some new product in the field with the options that newer units are likely to have (i.e. HDMI) without obsoleting all the current inventory in the channel.

 

in the book The Xbox 360 Uncloaked, the design team, or cost reduction design team were planning year-2 and 3-year versions of the Xbox 360 motherboard before Xbox 360 was released.

http://blogs.mercurynews.com/aei/2005/08/a_walk_through_.html

Leslie Leland:

regarding original Xbox

regarding Xbox 360:

I suppose Falcon is year-3.




I'm assuming 65nm will be around through 2008, even with a revision of the Falcon board (year-4)

I'm guessing that, then the year-5 Xbox 360 (in 2009) will be 45nm chipset in a redesigned smaller system like PSOne and PS2 Slim.

 

Of course. That's more attractive than reality. That too, the 50% figure is a stretch based only on a sub-component of a component, which the article itself doesn't really get into because either the public is too stupid to understand, or the writer is, or both.

Decreasing the cost of the chip by 50% doesn't just mean being able to get twice the good dies for the same number of wafers (which is theoretically possible with a process shrink after maturation even if the die size doesn't quite shrink dead half). You still have to be able to bring packaging and operational costs down, neither of which seems remotely possible -- at best you might get some drop in packaging costs if you can possibly get rid of that MCM layout of the GPU.

If you want to offset those costs, you need a much more massive volume increase than a "measly" 2:1 in order to get a 50% drop in production cost. Again a shrink isn't going to get you there automatically, which the original link seems to suggest.

For that matter, MS would probably like to be turning a profit afterwards even if the cost of production does go way down. The core may go down in price a lot more than the premium and elite packages. Even if people start to veer to the core system, they can still sell the peripherals at absurd markups. So I'm not so sure I'd expect any major retail price reduction.

 

Exactly.

Isn't this the same thing that Sony is doing with the 60gb / 80gb+game scenario?

 

true strange article,
the wii version 'the wiimote nob now costs $2 to make instead of $6 (wii now 3x cheaper to make)
the ps3 version blu-diodes price drop from$200 to $20 "huge font -> ps3 is now 10x cheaper to manufacture",
(on a related note, ive read sony are now selling ps3s at a profit, yet ive not seen this mentioned on this forum)

 

I dont get the slide. EE #3 on 150nm decreased in size more than half from EE#3 at 180nm? Isn't 150nm only a half node? Yet some of the full node change decreases were far less?

Also, I know you hate me for this, but Deadmeat on OA likes to say that Sony processes are very bad, and that their 65 nm Cell is barely smaller than 90nm Cell. I think that was what he said. He attributed it more to Sony than any process. Of course, I know he likes to spread FUD about Sony, but I'll be interested if Cell and RSX dont seem to shrink as much as other chips at 65nm.

 

Yeah, what I think is that a shrink they didn't fully get "right" on 180nm (since traditionally you'd expect more), they did get right when they went further to 150nm. In that case although 180nm was the full node step, you could think of the 150nm version as the full node shrink and the 180nm as an intermediate step in terms of what they were able to achieve.

But you're correct in your reasoning.

LOL, well... yes, I do hate you bringing up Deadmeat, but... hey - anything's open for discussion.

Sony processes aren't *bad* I'd say, but they have history of issues at certain nodes... as do all companies. Anyway it's true that Cell on 65nm presently isn't that much smaller than Cell at 90nm - as you can imagine when IBM released the die size of Cell at 65nm, there was a thread on it here. And further to that, notice I said IBM. A company that has its own bad history with process shrinks, and the one whose 65nm HPC Cell we have figures on, so not sure why Deadmeat would try to finger Sony in that... well ok I know exactly why he would, but it's not Sony's "fault" is what I'm saying. I expect that Cell will receive a further revision at some point on 65nm to get closer to what it theoretically should be able to in terms of densities. I also expect that the 65nm Cell Sony is fabbing at Nagasaki is a bit smaller than the 65nm Cell IBM is fabbing for the HPC market; for the record, that Cell is 212mm^2 vs 235mm^2 for the original... very small size gains.

...which, leads us back to the point of the thread. At the very minimum, knowing that MS is going to 65nm is not enough. At the very minimum to get a real sense of their upper bound on savings, we need to know the die size. If and when someone busts open a Falcon 360, we'll get that information.

 

It may well be that they are, but not in all regions. The price from region to region varies by as much as $400 US. In some cases, though, it's not really that the price of the unit from Sony's perspective has gone up, but that 3rd-party importers are garnishing on exorbitant fees.

 

Yup, seems only Intel gets it right most every time, but then they specialize.

 

actually... We pay the equivalent of USD $955 for the 60gb PS3.. So 450+ difference.
But thats getting off topic.
I want me a cool and quite 360! Now!

 

Well their transition to 90nm was actually pretty messy, all things considered. But that's as much the fault of the NetBurst architecture itself as anything else.

 

I dunno if the recent 360 sales are accurate...but they need this....

 

IMO they have needed this since around April. MS has been just content out selling the PS3 even though the 360 weekly number have fallen big time. I fully expect the PS3 to wipe the floor with the 360 come July NPD. The only questions are can Sony sustain its momentum and when the hell will MS give in to a price drop. It really sucks because I think the 360 market would explode in NA with a price cut with this incredible fall line up.

 

So, when do we think the price drop is hitting now?

After or before Halo?

My Spider senses were tingeling because MS wanted PGR4 to be out in September too...

 

I also want to know if the Halo 360 is a Falcon or not