AMD's survival tactics?

So AMD has finally bought ATI. This looks more like a move for survival rather than a tactical move. With the new Intel processor lineup completely thrashing the top of the line AMD processors at less than a third of their costs (Not to mention the earlier top end processors from Intel which the new processors have just made a mockery of), AMD seemed to be doomed. AMD has no new processor launches lined up for the near future. Even lowering the prices seemed infeasible as the new Intel processors achieve the performance of the AMD processors at a fraction of their cost.

Meanwhile the AMD-ATI merger has made quite a few eyes look up. What could be the implications of this merger? There are many short term as well as long term implications of this merger.

The short term effect is mainly the potential damage to partners. Of these, Nvidia is the pain for AMD followed by ATI for Intel. Taking a look at ATI and Intel - Intel buys quite a few ATI chips and graphics cards and also has a license for crossfire, the SLI equivalent from ATI. In all probability this will end very soon. But that is not very closely related to the deal. ATI's chipsets were just a temporary solution to Intel's demand forecasting weakness. Since that has been more or less made up for, the love between ATI and Intel was going to end soon anyway. So not too many losses here.

However the Crossfire + Conroe is likely to be the most sought after gaming platform in the short term. This is because it will take a while before Nvidia brings out its SLI chipsets that support the Conroe processors. The affair between Nvidia and Intel has not been very impressive in the past to say the least. So you can definitely expect Intel to accelerate its internal GPU projects as far as possible. The giant will not want to be bogged down by a cribbing bride.

Hey how about Intel buying out Nvidia? Sound good right? But that is very unlikely. There is really no value addition that Nvidia can do to Intel. There is nothing much that Nvidia can do which Intel cannot do internally. Short term - it would be expensive and not a very smooth affair, long term - why spend money on something that you can make yourself?

Coming to ATI - It will have some short term loss of business from the Intel - crossfire platform if Intel decides to give-up the license. Intel would then have to do without any dual GPU offering for a while. That is unlikely. So don't be surprised if you see an Intel box with AMD crossfire written on it!!!

Now that brings us to AMD-Nvidia. Nvidia has been a close partner with AMD and the AMD+SLI has been the platform of choice for gamers for a long time. So this deal would be a terrible loss for both sides. So AMD is unlikely to give up its successful relationship so easily. They will definitely have support for Nvidia chips too, at least for a while. So the deal is unlikely to cause much of a bother for Nvidia.

Now coming to the long term effects. It is most likely that the x86 architecture is going to take a radical turn. Today we are having dual core processors and with both Intel and AMD looking to launch quad core chips within a year, we look to be headed towards a multi-core future. With software becoming more and more thread and core aware, this sure seems like the future of computing. We are soon going to see mini-cores.

Few years back, processor makers (read Intel and AMD) were churning out chips that were hot, fast, out of order execution beasts that ran a thread a fast as it could go. Programmers were stupid creatures that had to depend on the hardware to do the job for them. Then came hyperthreading where programmers finally got something better from the hardware. Soon dual core CPU's were introduced. Then came Sun with the introduction of Niagara - This is a server segment processor though. It has a massive 8 cores that are capable of 4 threads each. With today's server technology usually being the Desktop technology of tomorrow, we can soon expect similar things in the Desktop arena. Intel is not to be left behind in such things. They are looking at developing a 32 core system with massive multi-threading - Yes you read it right, 32 cores!! However that is quite a while off. Also, it will also take some time before Desktop software can actually make use of such massive threading. They need to also find a workaround to the memory speed bottleneck for that.

The world of multicores is something radically different from that of single cores. You develop each core at a fraction of the cost that it will take to develop a single core processor (with a possibly lower performance) and then you integrate all these into a single chip that makes a mockery of the single core chip in multithreaded applications. This will also help chip makers to speed up the design upgrade cycle.

Besides, you add in GPU functionality to the chip and you have the perfect recipe for a great computer! Well I am not referring to adding a GPU on the same chip, I am referring to introducing the GPU functionality into the x86 pipeline. Say something like a GPU execution unit in the pipeline that makes mince meat of graphics instructions. Sounds wonderful?

Basically GPUs’ don't have much of a future in the industry. AMD and ATI seem to have realized that. Probably Intel and Nvidia are not too far behind. This industry is really interesting. Watch out for newer, better, radically different products being offered at a never before seen pace.

Disclaimer: This blog is a collection of my readings from around the web with a few of my own additions.