PlayStation 3 Delayed, Over $800?
AWhiteFlame writes "Cnet is reporting that a research report issued by Merrill Lynch suggests that the Sony PlayStation 3's American release may be postponed until 2007. From the article: 'The analyst firm proposed the idea that high costs and Sony's decision to use an 'ambitious new processor architecture--the Cell' is making it look like the company might not be able to meet its goal of getting the PS3 out in the U.S. this year.' Sony did not immediately respond to a request for comment." The official report (pdf) would also seem to indicate that the console will be somewhere in the neighborhood of $900 when it launches.
I'm not an insider by any means, nor a PS "fan boy," but isn't it likely that this is just very intelligent marketing by Sony? It's generally accepted that a game console launching at $900 (hell, $600), isn't going to happen in this day and age of mass market acceptance being an essential requirement of the development of any piece of electronics. This falls right in line with the Blueray machine costs . . . make it seem like astronomically expensive hardware fit for a king, and then release them at a fraction of the price, and sooner. I don't care when they release it, but I'm betting it will be this year, and at a $500 price point or lower.
Apple just did it with the Intel switch. First they've started releasing the stuff 6 months earlier than they said they would, and now their upgrading the processor clock speeds for free. Who wants to bet that wasn't in the writing already for the entire gestation of their Intel plans. If there were two companies I would compare hype-capabilities apple-to-apple (sorry), it would be Apple and Sony.
A B A C A B B
Shit, for that amount of money, I might as well just get a new PC.
Add up the numbers in the column. It only adds to $800.
This report is way way off.
Additionally:
The only thing worse here than M-L's estimate of the price of the PS3 this year is their estimate of it in 3 years.
Let's start from this year.
$230 seems high for just the CPU. I couldn't say how much, but I can say that Sony wouldn't even bother to make their console if the CPU cost half over half of the expected selling price.
The Blu-Ray drive price is WAY too high. Philips is going to ship a Blu-Ray writer drive for $500 in May. That's $500, retail. That includes retail markup, and cost of shipping to retailer. Also, Philips pays Blu-Ray license fees to produce units and Sony doesn't. And did I mention the Philips writes and the PS3 only has to read? And I can buy a quality DVD-Writer for under $40 retail right now. A Blu-Ray reader drive is a little different, but not a lot. It cannot cost much over $100, and it'll be well below that by fall, when the PS3 production ramps up (or perhaps just begins in earnest, I dunno).
6 USB ports? It will not have that many. 4 tops (2 front, 2 back). And the connector cost seems high, I'd say $3 today for USB ports, maybe $2.
For 802.11g and ethernet, Sony is using IP from Marvell that is normally used as an 802.11 access point. So it has all 3 ethernet ports and the 802.11g (and an ethernet hub) in a single chip (or less, see below). I'd say $5 for the ethernet and 802.11g together, maybe a bit more if they really leave 3 connectors on the back.
If the $100 was for a hard drive, they're the dumbest people alive. I can get a 40GB 2.5" drive for well under $100 retail. The OEM price cannot be over $50, and they could always go to under 40GB if it saves money. I'll just assume they added wrong.
I think also M-L doesn't understand that when you make a custom chip you can put a lot of stuff on it. The link (brains) for the USB, 802.11 and ethernet are probably on the main chip in the unit, bringing the cost of them down to nearly free. The 802.11 PHY/radio will probably be a separate chip, but the USB PHY is certainly on board, maybe the gigE one too.
So M-L is well over the initial price here.
Now, let's look at the future prices.
$100 for an OEM Blu-Ray reader in 3 years? Unpossible. Blu-Ray would have to be the biggest flop in the world for this to happen. My guess is you'll be able to buy a Blu-Ray writer drive for less than $60 in 3 years at retail. Look at how DVD writer prices collapsed. Readers will probably be under $40 retail. OEM prices for either will be even lower. And again, Sony doesn't have to pay license fees, so that lowers their prices even further.
$60 for the main chip in 3 years seem high too. It'll be on 65nm or lower then, yields will be way up, chip size down, and they might even combine chips (like the GS and EE were combined into a single chip on PS2 in under 3 years). I couldn't say how high though. Maybe it'll be $50, but include the functions of some of the other chips in it.
$30 for 512MB of RAM 3 years from now. Seriously? That's way off. GDDR3 will not be special anymore, and Sony won't be paying much premium for XDR, since they'll have enough volume to make a market in it. Right now you can get 32M of mobile SDRAM for $4 in big quantities, 64M of mobile SDRAM for $5. And I'm to think 512MB of commodity RAM will be $30 in 3 years? Nope.
Again, they don't know the PS3 uses a single set of IP for Ethernet and WiFi, $7 between the two 3 years from now is way too high. I'd say $2 for the PHYs, links will certainly be on with another chip.
$5 for Bluetooth in 3 years? It won't drop at all? Smooth move.
These companies stink at estimating parts costs. Just remember, these are stock brokers, not engineers, not parts buyers. They just don't have any clue at all.
http://lkml.org/lkml/2005/8/20/95
This is counter-intuitive, so you don't lose points. But they lose more if he does buy it. Why? Because in the initial frenzy for the machine, there won't be enough. If someone buys one like he suggested, that's a machine no one else can buy. He's keeping it off the market, in effect. Depending on how demand goes for games, Sony may not be able to justify making any more of them.
Where as if he lets a hardcore gamer buy it, sales might be so brisk of games, that Sony decides they will eventually make a profit, keeps going.
So during the debut of the thing, it's entirely possible that him buying the machine could hurt them worse than not buying it (since he has no control to keep everyone from buying his unit). What you say only becomes true if he can convince others not to buy it either, an unlikely proposition.
Keep in mind that the online platform is a huge part of both PS3 and Xbox360. We know that this won't be a free service, this is something players have to pay for on a monthly basis, just like your existing MMOs. Most likely you pay $60 a game, then another $15 or $20 a month to give you access to all the online game services or whatever. Not everyone will be willing to pay that at first, but as the game library builds and as more people see how cool it is at their friends' house or whatever, the subscriptions will increase. Plus, manufacturing costs will not always remain $900 a unit. They'll go down, just like they always do with PC hardware. I think Sony can easily take a $400-500 hit per unit long enough to get the costs down and get the online business booming.
That is incorrect. Compilers can reduce branching by emitting predicate instructions. OoOE is not useful for branch-prediction, anyway. It's useful for increasing instruction-level parallelism by removing stalls caused by certain types of data-dependency. The PPE in the Cell, more like the T1 but also like the POWER5 relies on CMT to deal with potential stalls by scheduling a second thread to be run when the current thread would stall. What the SPE doesn't have beyond OoOE is branch prediction (the PPE does have a branch predictor), which is where compiler hinting and the use of predicate instructions are important. This is also where the large register file of the SPE is useful for loop unrolling. If a branch should be taken and isn't, then this just stalls the pipeline (the SPE doesn't have a cache, it reads from the local store which isn't flushed on a stall, considering it's 256KB of data/instruction storage). The VMX unit on the PPE is useful for performing double-precision floating-point math (each SPE does double-precision float an order of magnitude slower than packe single-precision), and will also probably be a workhorse for programs that aren't specially designed to run within the SPEs. It also provides binary-compatability with the current PowerPC, where the SPEs do not.
And your other comment about the DMAing to main memory being slow is wrong. Each SPE can DMA from main memory at incredibly high rates, it's simply limited by latency.
Plus, everyone knows that the first people to buy something like that Samsung player is paying a large premium. I would be amazed if that player cost them over $500 to manufacture.
And of course, Sony will benefit from economies of scale on the PS3 faster than that Samsung player will.
The PS3 will be sold at a loss. But I bet it will be less than $200 per unit (I'm guessing at a $400 price point myself).
Comment forecast: Bits of genius surrounded by a sea of mediocrity.