to maintain a 90% AMD 10% nvidia market share, AMD needs to sell 90/10 = 9 cards per every 1 card nvidia sells.
to maintain an 80% AMD 20% nvidia it needs to sell 80/20 = 4 cards per every 1 nvidia is selling.
if AMD is selling 2 cards per every 1 card nvidia is selling, then nvidia's market share will continue to grow until it is 66.67% AMD and 33.33% nvidia. (a simple limit going to infinity, 66.67/33.33 = 2).
if AMD is selling 1 card per every 1 card nvidia is selling, then it will move towards a 50% market share each.
What does this math mean? it means that we know AMD does NOT outsell nvidia by MORE than 4 to 1. it can still be outselling nvidia by any lower figure, or nvidia could be outselling AMD, we cannot tell from the figures given.
Looks like again minimal adjustment to the DX11 marketshare numbers. It's interesting how much getting to market first has cemented AMD's holding the lionshare of the DX11 market even with all the releases NV has gone forward with, they still cannot make any significant change under their DX11 umbrella.
I disagree with your assessment that 2% change in a single month is "minimal". The DX11 market is limited to high end upgrades which are a small portion of the market, the majority are still invested in DX10. By the time DX11 will become mainstream these numbers could be anywhere.
Looking forward to 28nm cards next year and to see who comes to table first with those releases. My money is on AMD again, particularly if they are going to have Global Foundries as well as TSMC to pull from.
Interesting that you say "particularly". This indicates additional reasons, what are those? personally I have absolutely no clue which one of them will be first to market.
But I have to point out that using two foundries includes additional risks too, not just benefits.
The benefits are that if one has it's next process tech delayed or under perform, then they have a backup to fall unto. Furthermore, if both succeed well and their design is superior to their competitors, then they can more easily meet market demand (prevent loss of sales to competitors due to simple lack of product).
The drawbacks are that development for two different processes can cause delays in itself due to the extra work of fine tuning the design to work with either of said processes, and using two foundries increases the chance of information leaking, which could give a dangerous edge to their competitors.
Said foundries might also require assurances that THEIR technology isn't leaked to their competitors, and thus require segregation of development teams from each other. Which would further delay development and up costs.