Originally posted by: toadeater
Nvidia, AMD are going to beat Intel. Nehalem is going to have trouble becoming popular and influencing the desktop market.
1. Intel needs game physics to be done on the CPU for anyone to care. Nvidia and AMD want it on the GPU, they hold all the cards so far.
2. Hybrid SLI/Crossfire means you can use a cheap secondary GPU for physics and co-processing.
3. GPU co-processing is going to speed up many tasks tremendously. Apple is adding native support for this to OS X, Windows will get similar eventually.
4. Nehalem won't drastically speed up everyday tasks and games. Consumers are going to ask "WTF?"
5. Nehalem is expensive and proprietary. Consumers are going to say "It doesn't fit in my 775 board."
6. 775 socket FSB-based CPUs still have room left for improvement, and are more than fast enough for most users!
Nehalem is the right thing for Intel to do, they should have gotten rid of FSB clocking long ago as AMD did, but Intel may have waited a bit too long to do it. There isn't really anything out there for consumers, gamers, and business users that requires significantly more powerful CPUs. The bottleneck is slow storage and lack of RAM. For games, the GPU will continue to be far more important than the CPU, in part because games will be even more console-centric in the future.
1. I hardly see how AMD and NV hold all the cards in terms of physics. You do know that Intel bought Havok, right? NV and Intel hold the physics cards by virtue of controlling the two big middleware providers. MS may hold some cards if they make physics part of a DirectX spec, but that won't happen for a while.
2. That's what they've been talking about for 2 years now (rumours of the Forceware 90 drivers supporting physics on a second card were around in
May 2006), and NV just announced they were in fact putting physics on the GPU itself (GTX 260/280) instead of using a separate card.
3. Agreed, assuming it gets implemented in a way which supports both AMD and NV, which would probably require MS intervention for Windows, although Apple are trying to do it (with their own stuff) for OSX.
4. You're not (IMO) going to speed up everyday tasks with a GPU which you can't also speed up with Nehalem etc. I don't see how you can argue that GPU's are going to be the way of the future because they can speed up specific tasks, and then claim Nehalem isn't good because it can do those same things upon release without needing extra coding.
5. Nehalem won't be expensive forever. Also, newer Core 2 processors don't fit in old 775 boards due to chipset issues. We've made transitions before, many times, from one socket to another. This is nothing new or troubling, and for sure won't have any impact on the general population who buy pre-built systems (the 775 issue, not the price one).
6. I'm not sure what you're tying to get at here. As I said in (5), socket isn't everything. If Intel make improvements that aren't supported by older chipsets (which some surely will), then you're out of luck on your old 775 board anyway!
Also don't forget that we are moving to DDR3, which older boards won't support either (I know, I know, we're at least 12 months away from parity between price or shipments in all probability, but we're also probably 9+ months from mainstream Nehalem).
Things will always be fast enough for most users, but that doesn't mean they aren't allowed to get faster.
You seem to be mixing markets. On one hand you complain about price and users won't want to upgrade, or they are fine with what they have, then you talk about another market, the enthusiasts or specialised groups who might make use of GPGPU features. The group who want GPGPU will also want Nehalem, and the group who don't need Nehalem also don't need GPGPU.
There are always things which need more power, always, and Intel are trying to address other problems such as slow storage with their own SSD drives, and desktop enthusiast Nehalem supports more RAM (triple channel, 6 slots) than most current desktop systems as well.