I'll probably stop posting for quite a while again after this, but FWIW I thought I'd comment on Ion and also Tegra.
Here's the key: On the desktop, an Atom 330@1.6GHz, unbundled, costs ~$43. A Celeron E3200@2.4GHz, unbundled, costs ~$43. On laptops, an Atom 330, unbundled, still costs ~$43. But a Celeron Dual-Core T3000@1.8GHz costs $80. The performance gap isn't as large and the latter's TDP is a full 35W.
In Q4 2009 (assuming no delay...) comes Ion2. FWIW, its specs are: 128-bit DDR3, GT216 with 16TMUs/48SPs, and Socket 775 support including VIA Nano.
On the desktop, I'm willing to bet NVIDIA will focus nearly exclusively on the Celeron E3xxx and Pentium E5xxx/E6xxx (plus VIA Nano). With Pineview and the resulting limited/inexistent supply of standalone Atoms, they don't have a choice, but this could actually be a good move either way. The costs will be higher than Pineview, and the power consumption too (although xbitlabs has measured the E3200 as taking
less than 20W; the 65W TDP is massively inflated but I wonder if it doesn't still force OEMs to use very large heatsinks for passive) - but neither massively so. And the value proposition will be very obvious, even more than with Ion1.
On laptops, it's a perfect refresh platform for the 2010 Macbook obviously. It's also very viable for relatively low-end notebooks. But in netbooks, it'd obviously need to run at very low clock speeds and it's not clear standalone Atoms will be either available or desirable by then. Therefore, NVIDIA's best shot at Back-to-School 2010 design wins would be VIA Nano - and how that turns out depends completely on how well VIA executes. Never a good thing, with all due respect to the amazing job VIA/Centaur has done given their small team size and how optimistic I am about the 40nm Nano once it does come out.
This obviously brings us to NVIDIA's true strategy for netbooks: Tegra. With 35 smartbooks design wins and a surprisingly massive number of major carriers on board, they've got a very good shot at this if carriers don't screw it up and the media isn't even more retarded than I thought (would be surprising given my current view of it). More importantly, their roadmap and execution seems strong (already sampling the handheld variant of Tegra2 today).
I know some people will disagree completely with this, but the most insightful comment I've heard from Jen-Hsun at Analyst Day 2009 - and possibly the most insightful I've heard of any *semiconductor* CEO period - is the realization that the x86 PC is no longer where most of the developer attention is going to for consumer apps. Most of it is going to the web and handhelds, and it's clear this trend is only increasing every day. I'd personally argue that besides gaming, most of the effort in PCs is going to apps that are relevant to consumers but much more so to corporate (e.g. Office); and this effort is increasingly being duplicated on other platforms.
x86 Everywhere was a good idea 10 years ago. Soon it will be a pipe dream for people who haven't grown up.
(and before people think I'm saying all this strictly in defense of NVIDIA; I'm very skeptical they can deliver Ion2 in Q4 anyway. And notice I didn't talk about Ion3 or Consumer GPGPU. Enough said. While I still believe x86 will commoditize, that certainly does NOT mean NVIDIA will ultimately benefit much unless you consider a weaker Intel a benefit in itself. I'm more optimistic about their ARM prospects for a wide variety of obvious reasons)
