Originally Posted by Masked
The profit margin on Tegra/Tesla etc is much different than the consumer GeForce.
They make more money off of those lines in terms of raw profit but, their biggest financial gain is because of the average consumer...
GeForce makes more due to volume.
The profit margin of Tesla/Quadro is quite high as previously stated. They probably make a few hundred to thousands for each sold.
The margin for Tergra is VERY low.... a few dollars per SoC.
Originally Posted by ketxxx
The cost of, lets say a 7950 in parts is nowhere near the retail price, its maybe £25. This is essentially what the manufacturer would eat when replacing a dead card, £25. Everything else has already been covered with the initial purchase of the product. Ever notice how you almost NEVER get a brand new item when you RMA? Theres a reason for that youknow, which I have just covered.
That's not how things works....
A GK104 and GDDR5 from Nvidia is a lot
more than $40USD.
Then the partners have to buy the rest of the Bill of Materials, pay for assembly, and shipping.
Then they have to pay for logisitics.... warehousing, shipping, etc
Then they have to pay for customer service...
Then they have to pay for marketing and advertising...
Then they have to pay for administrative support (lawyers, managers, etc).
Then they have to pay for RMA costs....
The partners are paying a few hundred dollars and take on risk for each card. Their margins are not THAT great.... NVIDIA makes most of the money. (I worked for EVGA as a support contractor for a few years.)
In your arguement, you are missing the concept of economic cost. Even if a $600 card cost you $400 to produce, you still are out of $600... not $400. Companies do not want to spend increase time and increase risk for less profit.
Originally Posted by DoktorCreepy
Its always about the TSMC yield issues with Nvidia and apparently with some people who write articles about them.
Other companies have yield issues too but, they don't complain about them every time they are making something on a new nm process.
Other companies are not producing as big/complex chips as Nvidia though. With big chips, a single fault can make an entire chip unusable. With smaller chips, the single fault might make one or a few chips unusable but the rest are sellable.
AMD almost had the same issue as Fermi but they were wise enough to have a few extra runs at the 40nm process. They discovered the limitations months before NVIDIA and were able to account for it in their design.