AMD and ATI are set to merge in a deal worth $5.4bn. Greg Corke takes a closer look at what this could mean for the workstation sector and the future of 3D graphics.

Could the future see AMD’s Opteron CPUs and ATI’s FireGL GPUs merge into a unified processor architecture?

Rumours surrounding AMD’s merger with ATI have been circulating for some time, but the official confirmation, made on 24th July 2006, is still causing major ripples throughout the IT sector. If approved, the deal will cost AMD a whopping $5.4bn, made up of cash and stock. AMD’s chip rival Intel and ATI’s graphics nemesis Nvidia are certainly sitting up and taking notice.

$5.4bn is a serious amount for any company, so what exactly is AMD getting for its money? AMD’s strength is in processors, but unlike Intel it doesn’t have a presence in graphics or chipsets. ATI has expertise in both of these areas and AMD has acknowledged its long term objective is to create a unified platform using a mixture of its CPUs (Central processing Units), ATI’s GPUs (Graphics Processing Units) and ATI’s motherboard chipsets, using different combinations depending on the target market. It’s been rumored that Intel is also heading towards this unified architecture for some time, and will naturally have the lead on AMD, so for this alone the merger makes perfect sense.

" AMD has acknowledged its long term objective is to create a unified platform using a mixture of its CPUs, ATI’s GPUs and ATI’s motherboard chipsets "

That approach however, is unlikely to find its way into the performance workstation sector, where cutting edge technology is demanded for both CPU and GPU. History shows that leading 3D graphics technology is not exactly small and graphics cards also output significant amounts of heat. The thought of cramming workstation class CPUs and GPUs into a single space just doesn’t seem feasible, at least in the short term.

While AMD rival Intel has invested significant resources in its own graphics technology, it does not have the performance of ATI, whose Radeon 3D gaming family and FireGL workstation range have helped push the limits in many sectors. This currently gives the AMD/ATI combination an advantage over Intel and as a result Intel may look for parity by working closer with Nvidia on high-end graphics developments, though reports of friction between the two companies could hinder this.

Indeed, the ATI/AMD merger throws up all sorts of questions over future hardware developments. Nvidia’s chipsets are standard in most motherboards used in Opteron-based workstations, ATI develops chipsets for the Intel architecture, and millions of ATI graphics cards are used in Intel motherboards worldwide. While the initial indications are that the merger will not affect relationships between Intel and ATI it is hard to believe that future developments from both Intel and AMD won’t be skewed to some extent in the long run.

Now under one roof, AMD and ATI could also help speed developments in workstation graphics technology. Both ATI and Nvidia have suffered in recent years from CPUs not actually getting much faster. Engineers can experience a boost in 3D performance in virtually all CAD applications as clock speed increases, but this hasn’t happened. Instead, Intel and AMD are following the path of adding multiple cores to their CPUs.

In the short term, one obvious solution is to get graphics cards talking to multiple CPU cores, and it’s very likely this line of development is already underway at ATI and could now be accelerated with AMD on board. This would also make a stronger case for ATI launching its dual graphics card technology, Crossfire, in to the professional arena at some point in the future. Of course, integrating a CPU and GPU into a single package could also help solve this issue, if heating and power issues can be overcome.

There are, of course, other interesting aspects to this merger in the workstation space, one being the use of GPUs in other areas. Over the past six months ATI has been very vocal about using its GPUs for processing tasks traditionally carried out by CPUs, such as video encoding and rendering. Why, after all, use a general purpose processor to do a specialist job? This makes the idea of using an array of GPUs - much like ATI did for SGI a few years back for high-end 3D graphics - to accelerate these processes a very attractive idea for future deployment in the CAD/DCC sectors and something that would not have escaped the attention of AMD.

From ATI’s perspective, the proposed merger will give it the fabrication facilities that it has lacked for so long. ATI’s FireGL division will remain a separate unit and this comes as little surprise as it has been the most profitable segment of the company for some years. For now the merger looks like a positive move for both ATI and AMD, and with the blurring of traditional roles for CPUs and GPUs, it will be very interesting to see where AMD takes processor development in the coming years.