I think most of their discrete market share comes from the days when they almost didn't have competition, ie 8800 series.
What you are discussing is overall videocard
usage by brand/model type, not current market share. I'll explain why there is a huge difference between the 2.
If in Q4 2010, 100 new discrete GPUs were sold, 50 went to NV and 50 went to AMD (both have 50/50% market share). Now let's say in Q4 2011 NV sold 54 new GPUs and AMD sold 46 GPUs for 100 new cards sold, NV just increased its market share QoQ by 4% (50 vs. 50 Q4 2010 vs. 54 / 46 Q4 2011).
Using your explanation for market share, the overall market share would be 104 NV / 200 cards sold? So it would only gain 2% (or half)?
Think about what you are saying.
Imagine if since 1985 NV sold 1 Billion cards and AMD sold 1 Billion cards. If you counted all NV and ATi/AMD videocards since inception of the 2 companies, then even if NV sold 20 million videocards last quarter and AMD sold 0, that would be a drop in the bucket against 1 billion videocards ATi/AMD sold in the last 20 years (the numbers I just used as an example, but you get the picture). The market share #s would barely move. But market share #s moved a lot. In fact, NV gained 3.7% qtr-qtr. If you included all videocards from the past sold by AMD,
how in the world could NVidia gain 3.7% market share against 1 Billion ATI/AMD cards going all the way back to 1985? That's because market share tracks
new sales in that period of time not overall usage by brand.
Market Share (%) = sales of the business, product or brand / total sales in the market at any given specified period of time (i.e, monthly, quarterly, annually).
Products that have already been sold years ago are not counted in market share report of this kind. Only sales of existing products on the market are included (which includes newer generations such as HD7000 and older such as HD5000, if they are available for sale). But GPUs from the past are not counted. Market share counts new sales. If I own an 8800GTS I bought 4 years ago, that's not counted this quarter at all. That's market share in the past. Steam counts it, but not this research report. This report tracks new sales of competitors in the last quarter vs. how they sold relative to each other the quarter 12 months ago. This is why you see such a sudden collapse for NV since they don't sell IGPs anymore.
Market share is a key indicator of market competitiveness—that is, how well a firm is doing against its competitors. This metric, supplemented by
changes in sales revenue, helps managers evaluate both primary and selective demand in their market. That is, it enables them to judge not only total market growth or decline but also trends in customers’ selections among competitors.
^^ Changes are tracked based on existing sales, not what happened in the past or how many users currently hold 5-year-old hardware.
For example, based on videocard usage statistic on Steam, NVIDIA GeForce 9800 has 8.20% market share (usage). GTX560 has 9.47%. Next quarter, when market share #s come out for GPUs, how much market share does 9800 have in JPR's report? 0% since NV doesn't sell GeForce 9800. Steam would still show ~ 8% usage for 9800.