Sxotty said:
I think ATI was simply cheaper than nvidia. Nvidia was always might proud of their chipsets by the pricetag they set. Perhaps they have learned their lesson and the ULi aquisition signals they will change their ways...
Personally, I think NVIDIA just didn't have enough engineers to properly focus on both the high-end and the low-end; recently, they focused on the upper low-end (C51), and you may easily notice they haven't released anything new high-end-wise in a long time, IMO mostly because of that and the Intel nForce4. The only "refresh" they had is the 2x16 SLI chipset, which obviously was a spinoff of that R&D, since they needed to go away from a single-chip design again to make integrated possible (and it had benefits for their intel market, too).
As was noted by geo & others, NVIDIA is much more focused on high-margin products. In fact, they are so focused on it that it's for that simple reason they cancelled their development of Intel Integrated Graphics chipsets, even though it was confirmed 6 months ago or so they were progressing nicely for it. Their official reason is they felt Intel was too competitive in that segment to be able to get any substantial profit out of it.
From my POV, as I said above, they didn't have the engineering NUMBERS to diversify sufficiently and feel they'd stay competitive with high margins in all of those segments. Even if you've got the best engineers in the world, you can't expect 2 of them to work part-time on a big project and be done with it after a month.
Something that wasn't talked of much, though, is that even though ULi only cost NVIDIA about $52M, it is the
largest acquisition in their history by the sheer number of employees they've gotten through it. Indeed: ULi has 213 active employees according to
http://www.corporateinformation.com/snapshot.asp?Cusip=C7609O800
Assuming even 50% of those are marketing/sales/etc., which also matters because that's key for NVIDIA's Asia strategy, that still means they got nearly 100 new engineers out of the deal. That's more than what they got out of the 3DFX deal (because it was an IP purchase, and not an acquisition, many 3DFX employees were not given the option of joining NVIDIA, or didn't want to).
From my POV, the primary reason behind this acquisition, besides the Asia Strategy (what happened to NV's Indian Development Center they announced, anyway?), is that this will let them diversify more in the chipset market without spreading too thin. This should increase their number of chipset engineers by a good 30-40% I think, although I don't have precise numbers on how many engineers NVIDIA's chipset division currently has.
The following is purely my personal opinion of what NVIDIA should do, but I feel they should make a very clearly divided productline with distinct low/mid/high-end products. The low-end should be using single-chip designs (no south/northbridges), capable of being connected to integrated graphics chips with a couple of other features on the die too (audio? more I/O? etc.) - then you'd have 2 different southbridges, one for the mid-end and one for the high-end, and two northbridges, one for AMD and one for Intel.
This would represent 2 single-chip designs for the low-end (1 for AMD, 1 for Intel), one chip for integrated graphics & minor improvements, two chips for northbridges (1 for AMD, 1 for Intel), and two chips for southbridges (1 high-end, 1 mid-end). That's 7 chips total, which is a lot, considering NVIDIA is currently nearer to 4-5 distinct chips, but I feel that's their only way to be competitive with high margins in all segments of the industry.
It also makes it much more nicely modulable than their current solution: every time AMD/Intel introduce a new socket, they'd just make a new northbridge, with southbridges being upgraded "when needed"; and the integrated graphics could be kept compatible and upgraded "when needed", every generation too. This would be a much more interesting usage of their assets than the current rather,
IMO, relatively inefficient model that keeps them in the high-end segment of the market, and the mid-end one to a lesser extend too.
Uttar
