Wednesday18 May 2022

Very few bright spots to be seen in AMD's future

Reading time is around minutes.

In the last couple of months we have talked a lot about AMD and the direction they are trying to move to. Most of what we have reported is not good news and centers on the fact that AMD’s R&D/production budget is dwindling to the point where they cannot push multiple projects at one time. They have had to consolidate their efforts to the point that they do not really have products to bring to market to make them more money. An example of this is the lack of a new GPU for the normal launch cycles. AMD does have some products in the pipeline, but these might not be enough to win them back any marketshare from Intel or NVIDIA.

One of the products that we know for certain will be hitting in 2015 and one that AMD is betting a lot on is the Fiji GPU with 2.5D HBM. This product will be featured in their higher end graphics card designs and is supposed to bring them back into line with NVIDIA performance. Right now most rumors put this launch sometime in the first part of 2H 2015, but there is a possibility that it will later in the year perhaps closer to the end of the year. This is almost a make or break product for AMD and they have to get it right out of the gate or they will lose money they simply do not have.

Another product that is expected very soon is AMD’s Godaveri Desktop APU. This is supposed to bring a better performance per watt ratio to the table and is now rumored to hit in next month (May). This rumor might be nothing more than that though as it was originally intended to be launched during Computex (June). Of course someone might have thought that pushing a rumor of an updated product could buy AMD some time and attention and this type of marketing has worked in the past. The problem is that even the top end APU, the A10 8850K is not much of an improvement. You are getting very similar performance at the same TDP 95Watts. It is not likely that VAR and OEMs will be rushing AMD’s doors to get this one and consumers are also not likely to make the move if there is little to be gained.

Last up we find “leaked” slides of what might be AMD’s next CPU core called Zen. As of this writing the slides look genuine and could represent AMD’s decision to get back to what really works. The down side of this is that they are pretty dumping all of the time and money put into Bulldozer which lived much longer than it should have anyway. Bulldozer tried to introduce new ways to process information at the core level, but all it really did was show how not to build a CPU core.

There are enough articles on the net about the new Zen cores so we will only cover the highlights hear as the rumor currently stands. Zen is expected to be more like a traditional AMD CPU core with wider floating point pipelines (2x 256-Bit FMAC units) that will be capable of fusing for a single 512-bit AVX instruction per cycle. Zen will also have a much wider integer pipeline (about 50%) which will help with processing single threaded performance per core. All of this should some out using the Samsung/Global Foundries 14nm process which means more in a smaller package.

If the timeline on this is to be believed we could see Zen as early as 1H 2016, but that is a very optimistic timeline indeed. This means that AMD’s next 18 months are very critical for them. Godaveri is little more than AMD’s need to get something new on the market in the desktop/laptop space which their Fiji GPUs will be a big deciding factor for their future in the GPU market. Meanwhile, if Zen is real AMD is looking to attempt a return to performance competition with Intel sometime in 2016. Sadly for AMD, Intel will be pushing out 10nm CPUs which could mean that Zen will get here in time to compete with Intel’s last generation.

Last modified on Tuesday, 28 April 2015 16:08

Leave a comment

Make sure you enter all the required information, indicated by an asterisk (*). HTML code is not allowed.