AMD's graphic move

You can get a lot for $5bn, and AMD has certainly got a lot with ATI. As well as finding itself in charge of enough technology to produce its own complete Centrino-like platforms, with new interests in mobile and consumer electronics, AMD has disrupted the dynamics of every major commercial relationship in PC chips. Suddenly, Apples have AMD inside, Nvidia has had its garden comprehensively re-landscaped, and Intel has even more motivation to produce those high-performance graphics processors it's so badly needed for so long.

But the best thing AMD has gained through the deal is a future. The days when you built a computer out of discrete blocks — processor, chip set, graphics, sound, all in their own boxes — are running out. If you can't do the lot in one go, you won't be able to compete.

The future of processors is multicore — there is simply nowhere else to go — and that's setting the scene for very tight integration. The big argument isn't whether this will happen but what shape it will take, whether all the cores on a chip will be identical to each other. As usual in such debates, the most likely outcome will be a bit of both. Expect processors to have seas of similar cores, with specialist computation units dealing with specialist tasks. Expect one of those to be very high-performance graphics, and one the processor maker has to build itself. You can't get a systems integrator to plug one in. Intel knows this, which is why it's been winding down its relationship with ATI and ramping up its internal development efforts.

Until now, it hasn't been attractive to build high-performance graphics into a processor. These things are hot, big and memory intensive, three attributes that sit badly alongside a hot, big, memory-intensive central processor. But all three problems are being sorted out on the path to massively multicore units; extremely fast and wide memory buses coupled with large on-chip caches make it feasible to handle display data as well as everything else. And there's always the intriguing idea that many of the functions of a graphics processor have other uses, something that gets easier to exploit if they're broken out and made accessible alongside other CPU resources.

With ATI, AMD's getting ready to play this game. It knows about memory buses, ATI knows about power management and, while Intel is ahead in shrinking silicon circuits, AMD knows where to go and how to get there. Along the way, it'll get some intermediate benefits — a graphics processor that interoperates directly with Hypertransport should outperform one that has to use PCI-E, and all the pieces are in place for such a beast. It could even be packaged as a high-end gamer platform that could knock spots off anything Intel can manage with Viiv, and AMD is very partial to such coups.

Another key technology where AMD has suddenly grown extra tentacles is system management. Intel has its Active Management Technology, iAMT, which is labouring under a huge disadvantage. Unusually for Intel, it's not open — it contains trade secrets and other IP that you have to sign up to get. AMD has said that with ATI, it will produce completely open management interfaces which will be entirely compatible with open source software. Many major challenges lie ahead for management software — how to integrate it with virtualisation, how to instrument and control distributed processor and storage resources, how to homogenise heterogenous systems — and only open interfaces make such problems solvable. You don't want hidden secrets with system management.

It's not all rosy. AMD still has to assimilate ATI, a company with a very different management style. But unusually for an acquisition, there are few questions about economy of scale or increased efficiency, things that are usually given as reasons for a takeover but so often don't pan out. AMD and ATI are so complementary that with few exceptions, the benefits come not from paring out overlap but from building on synergy. For once, that word is accurate. AMD will be much more effective with ATI on board — and Intel will have an even keener fight on its hands. For those of us who look forward to ever more efficient, powerful and lower-cost computing, it's good news all the way.