Charlie's commentary, though hyperbolic, is largely correct IMHO.
Interesting news. AMD shipped more graphics cards in the second quarter than NVidia.
DX11 is pretty much irrelevant to most of the market as I'm not aware of any game that won't even run on a DX9 card, let alone requiring DX11.
The thing with integrated graphics is that it is moving from low-end into middle-ground, leaving discrete to enthusiasts ONLY.
The reason why nvidia is the only one being affected by all of this is that they are strongly tied to discrete whereas AMD and Intel together have the integrated segment of the market cornered. Yes I realize that nvidia has some IGP products, problem is that there is no incentive to buy it over the "platform" solutions.
In other words, why would you want to pay an extra $100-$500 for a discrete nvidia card when your needs are met 100% by an integrated radeon?
Plus the profit margins are probably minute on the integrated chips; so while it's nice to be shipping 100,000,000 integrated graphics chips you'll probably make the same profit by shipping 1,000,000 'enthusiast' boards. Unless you're forced to push margins down to get sales because your architecture is less efficient than the competition...
I think nv wants to create a competitor to atom. They bought LongRun technology from Transmeta and they hired former Transmeta devs. Maybe they will have got more luck next year with tegra powered smartbooks running android/google os. Of course for office pcs the market for dedicated gfx solutions is basically over. Only gamers need em and there nv had only very expensive models with dx11 support. This year ati had to win but next year is open.
Tegra (2) has a licenced ARM core. Also it supports video accelleration (even Flash 10.1 support for Android!). From the specs a very nice cpu. Of course a x86 cpu would be even more interesting.
no i do not make fun on you... its a embedded cpu nvidia buy this cpu with a company 'ALI'. Nvidia (M6117C - 386SX)
i'm waiting for a system nvidia 386sx +gtx480 LOL...
May be smaller than intel, but that doesn't mean they're small.
You've got it backwards. There's hardly any margin on high-end boards. It all goes into reclaiming the cost of development. The per-unit cost of an IGP is negligible -- the margin is huge there because there is no R&D required to keep punching out old stuff!Plus the profit margins are probably minute on the integrated chips; so while it's nice to be shipping 100,000,000 integrated graphics chips you'll probably make the same profit by shipping 1,000,000 'enthusiast' boards. Unless you're forced to push margins down to get sales because your architecture is less efficient than the competition...
Works like this;
You spend (for argument's sake) $10 million in R&D for a new chip. The actual production cost is $10 each. You sell them for $510 each, so for each chip, you dump $500 off the DEBT created in development. (yeah, yeah, there's interest on the debt too, lets just ignore that for now.) When you've sold 20000 units, you've paid off the entire debt, drop the price from $510 down to... $50. Suddenly you go from netting $0 per unit to $40 and you're selling 100x as many units!
Now obviously I changed certain parts of the picture for convenience sake. In the real world, you don't drop the R&D debt suddenly like that and drop the price overnight, it is more of a gradual reduction in the price as the R&D gets paid off. Nor is the R&D just a fixed number like this -- more like you have a constant R&D expense which you're paying off with a declining pricing structure.
The key to remember is that they ARE getting a good net on the low-end parts!