How it all shakes out

Intel

This is perhaps the most interesting thing of all to analyse. Let's divide Intel's issues into short-term and long-term.

Short term

Conroe: ATI CrossFire graphics working on the Intel 975X chipset has been stated as the preferred platform for Conroe. That is unlikely to change. Intel needs a dual graphics solution for Conroe to enable it to compete against Athlon 64 when it comes to gamers and high-end sales. With NVIDIA playing hardball with SLI licensing, it will most likely have no choice but to keep pushing ATI CrossFire for this generation, although it will have to come up with some interesting messaging. Alternatively, it could ask NVIDIA very nicely for an SLI driving license for its 975X chipset.

Counter-offer: Could Intel counter-offer for ATI? The AMD-ATI deal is still subject to approval from shareholders of ATI and would be off if Intel stepped in and offered a bigger amount. $6 billion is not a small amount of cash, but it would be manageable for Intel and would really cheese off the Austin guys. However, unlike AMD, Intel wouldn't really get much out of buying ATI since there isn't really anything that it brings to the table that Intel doesn't already do or have plans for.

Buying NVIDIA: Intel could snap up NVIDIA as an opposite move. This would create a clear platform dichotemy and would certainly solve NVIDIA's long term problems. However, the issue here is the same - NVIDIA wouldn't bring much to Intel's party, apart from the SLI branding. Remember that Intel sells more graphics chips than either ATI or NVIDIA in the form of integrated chips. Intel is perfectly capable of making a high-end graphics board to compete with both companies - it has just chosen to stay out of that game, so far. That may or may not continue to be the case.

Also, I don't believe that Jen-Hsun would ever sell NVIDIA to another company, unless the company was in a really bad way. He is in very much in control of NVIDIA at the moment and the company is profitable because he has made the right business moves month on month for some time now.

Ironically, ATI may have done its biggest damage to NVIDIA not through the creation of great products, but through goading Intel to get into the high-end graphics scene through its sale to AMD.

Long term

GPU on CPU: Intel has a strategy of integrating GPU functions onto the core of future CPUs. This has been the case for a while now, and shows no signs of changing. We mentioned that there will still be room for uber-high-end discrete boards, but Intel will be expanding its integrated graphics line like crazy. It will be integrated onto the CPU rather than the motherboard, and the capabilities will get better.

Intel's major sales of desktop CPUs have been to business, and businesses now need 3D graphics processing with the advent of Windows Vista, and going forward. Rather than making them buy discrete boards, thus benefitting NVIDIA and ATI, you can bet that business buyers would be happy with uprated Intel graphics directly on the CPU - it would save them money, give them the functionality they need and also keep things simple.

Intel's GPU on CPU is also modular, meaning you can effectively create 'SLI cores' - for example, a 6-core processor could have four CPU cores and two graphics cores working in 'SLI'. This technology could also be spun off into a discrete card if Intel decided it wanted to move into that space to compete with ATI-AMD.

Platforms: Intel was the first company to see the potential for platform sales, and it's not about to change that. It will continue to grow its sales and add capabilities to the platform, integrating anything that it can. As the largest seller of microprocessors in the world, and one of the most successful fabricators of silicon, it is quite happy to see extra functionality and value added to its primary product.

AMD's existence: Believe it or not, AMD's continued existence is good for Intel. Without AMD, Intel would have total domination of the chip space, and this would lead to all sorts of regulatory and competition legal issues. Having AMD around means it avoids all those sorts of problems. AMD is enough competition to avoid legal problems for Intel, without being such strong competition that it can really take Intel down. If AMD was to bowl over, another competitor - like IBM, for example - could really have a go at Intel in the personal computer space, and this could be a far bigger problem than AMD has been so far.

Unintended consequences

There are a few more consequences for the computing world in general that are worth looking at.

Physics is dead: Well, OK, not quite, but kinda. First we had PhysX; then that was going to be killed by physics on the GPU; now GPUs are being assimilated onto processors. Given that physics can already be done on the processor, that leaves dedicated physics coprocessors with a rather limited value. Games like Crysis are already using spare processor cores for physics calculation, and once GPU capability is added, there looks to be little reason not to use spare capacity there, too.

High-end gaming will become more of a niche: It's useful to keep some perspective - high-end gamers are a relatively small section of the market. Sales of super-cards like the Radeon X1900-series and GeForce 7900-series are a fraction of graphics sales, and the major money is in the low end. However, the low-mid-high end cards all share the same interface, which makes it more viable to provide high end cards.

We are not going to see bleeding-edge graphics on processors, because there will not be capability or capacity. We will see low and medium end, but not high. But this will, by definition, make high-end graphics an increasingly niche product which could push prices up and decrease configuration flexibility. When you can have a tiny system with almost everything on one chip, why add cost to the system by including a graphics expansion slot? In the long run, we could see a situation where, as gaming becomes popularised by the increased availability of 3D graphics through integration, high-end gaming becomes increasingly niche.

Less consumer choice: We currently have a situation where components from any of the four big players can be mixed and matched. We could see a situation where NVIDIA stops interacting with AMD, Intel stops interacting with ATI and the whole thing gets more awkward and incompatible.

All in all...

Things are only going to get more interesting in the coming months and years. Just as we were starting to get bored of endless product cycles and refreshes, something like this happens.

Prepare for power games.