Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
It doesn't matter how a company decides they want to make a profit on a card or not, and if they decide to factor in R&D, the fact is R&D costs money, and selling cards is their return on that investment, if one doesn't cover the other, they lose money, they are losing money, thats really as plain as it gets.
I've actually forgotten the figues but in Q1 they basically had a 50mil profit on around a 900-950billion revenue, that included a one off payment of 100mil, so really that had a 950billion revenue and 150mil profit, the next quarter they had a 100mil loss, which is a 250mil turn around which is huge in and of itself, thats after Fermi launched. They also lost some 10-15% market share, the revenue was also down from almost a billion to around 750-800mil its been a few months so the figures aren't fresh in my head.
But basically they had a quarter of a billion turn around in a single quarter on profitability, on top of a pretty significant drop in revenue.
The fact is we do know a lot of the costs, a wafer is $5k, you can get x amount of cores of size Y off the same wafer, after yields are factored in that gives you a cost per core.
Nvidia were selling a 386mm2 core for 15-20% less than AMD were selling 336mm2 5850's for, and the 5850 was signiifcantly faster, they were selling them for not far off half of what 5870's went for, though they sell significantly less than 5850's.
The new cards are 255mm2 iirc, the wafer cost is the same, that means they've quite significantly increased yields per wafer and therefore dropped cost per core down.
End of the day a pcb costs around the same for each of them, power circuitry a similar amount, 1gb memory costs what 1gb memory costs though again AMD are selling in FAR higher quantities right now which probably offers them some slight advantages, only talking a few pennies on a chip but over a couple million chips, that adds up.
IF Nvidia/AMD made their stuff in different fabs with different quality materials and unknown yield comparison, you couldn't compare them easily.
They are both made on the same process, at the same fab, smaller = cheaper, way smaller = way cheaper.
Nvidia can keep selling them at a loss to maintain some market share, though doing so has gained them only 1.5% market share back after MASSIVE losses in the past year, and lost money to do that. Its also ofset by the fact bulk orders reduce before a new generation is out, so both Nvidia and AMD will sell more in the first 3 months of any given cards lifecycle than the last 3 months. AMD 5xxx sales were wrapping up last quarter while Nvidia were ramping up and releasing new parts, 1.5% by fireselling cards cheaper than you can make them is horrible.
DOn't forget, Gibbo told us when the ridiculous 768mb version pricing started a month ago, that these prices were ONLY for 6-8 weeks.
I've actually forgotten the figues but in Q1 they basically had a 50mil profit on around a 900-950billion revenue, that included a one off payment of 100mil, so really that had a 950billion revenue and 150mil profit, the next quarter they had a 100mil loss, which is a 250mil turn around which is huge in and of itself, thats after Fermi launched. They also lost some 10-15% market share, the revenue was also down from almost a billion to around 750-800mil its been a few months so the figures aren't fresh in my head.
But basically they had a quarter of a billion turn around in a single quarter on profitability, on top of a pretty significant drop in revenue.
The fact is we do know a lot of the costs, a wafer is $5k, you can get x amount of cores of size Y off the same wafer, after yields are factored in that gives you a cost per core.
Nvidia were selling a 386mm2 core for 15-20% less than AMD were selling 336mm2 5850's for, and the 5850 was signiifcantly faster, they were selling them for not far off half of what 5870's went for, though they sell significantly less than 5850's.
The new cards are 255mm2 iirc, the wafer cost is the same, that means they've quite significantly increased yields per wafer and therefore dropped cost per core down.
End of the day a pcb costs around the same for each of them, power circuitry a similar amount, 1gb memory costs what 1gb memory costs though again AMD are selling in FAR higher quantities right now which probably offers them some slight advantages, only talking a few pennies on a chip but over a couple million chips, that adds up.
IF Nvidia/AMD made their stuff in different fabs with different quality materials and unknown yield comparison, you couldn't compare them easily.
They are both made on the same process, at the same fab, smaller = cheaper, way smaller = way cheaper.
Nvidia can keep selling them at a loss to maintain some market share, though doing so has gained them only 1.5% market share back after MASSIVE losses in the past year, and lost money to do that. Its also ofset by the fact bulk orders reduce before a new generation is out, so both Nvidia and AMD will sell more in the first 3 months of any given cards lifecycle than the last 3 months. AMD 5xxx sales were wrapping up last quarter while Nvidia were ramping up and releasing new parts, 1.5% by fireselling cards cheaper than you can make them is horrible.
DOn't forget, Gibbo told us when the ridiculous 768mb version pricing started a month ago, that these prices were ONLY for 6-8 weeks.
