Thursday, 25 April 2013 16:30

AMD to Dive Deeper Into The ARM Pond Soon...

Written by

Reading time is around minutes.

A couple of days ago the internet lit up with AMD’s announcement of their new Jaguar SoCs. These new G-series embedded processors were boasting some rather healthy improvements in performance including a claim of a 113% increase in performance over older x86 SoCs and a rather bold claim of a 125% increase in performance over Intel’s Atom. But while the new embedded APU with its 8000 series GPU should have been big news all on its own, what really caught people’s attention was a little x in the corner. As it turns out this X is intended to denote an x86 version of the G-Series SoC…

Clearly if you are going to highlight the architecture of something you intend to have an alternative to that. So if there is going to be an “x86” version then there is going to be something else. That was the question that a couple of sites asked and AMD was ready with an Answer, ARM. AMD is already licensing ARM designs for use in their Opteron server CPUs. The ARM cores will be responsible for handing advanced encryption directly in the CPU. This is a fairly low cost way of implementing this feature instead of spending money on R&D to add this into future CPUs.

Unfortunately for AMD almost everything they do has to be about money. Even before the ATi purchase AMD was not exactly rolling in R&D money. They very often worked to create R&D groups and consortiums to allow them to substantially reduce their costs to bring new technologies to the market. Unfortunately for them after the ATI buy they did not have the money to even maintain these ties and dropped out of all of the groups that we knew about. This is when the long silence fell over AMD’s new technologies and we saw very little improvement from them. The changes that were made we cost effective to produce and intended to offer AMD the best bang for the buck and did not always reciprocate that to the consumer.

Now, after Intel bailed them out with a very large settlement AMD started getting back to their old ways. They looked at what technologies and partners would benefit them most. They put together another consortium just like the old days with the HSA (Heterogeneous System Architecture) Foundation. The major players are also big players in the mobile market and have experience with ARM IP. Interestingly enough more than a few of them also have their own foundry operations. This means that AMD can offset not only the cost of some of the R&D but of fabrication costs of test parts etc. It is a very smart move for a company that is basically cash poor.

Now moving forward AMD can use ARM ingenuity in mobile devices, desktops, servers and any other place they need it while working with others to build a shared IP portfolio. It is a shadow of the AMD we all used to know and love when the Opteron and Athlon64 came out, but at least they are playing to their strength. Now some will tell you that AMD is behind the power curve when it comes to integrating ARM Core IP and their GPUs (or other components for that matter) and they are partially right. However, take a look at their partner list in the HSA. Many of them are companies that AMD has had a strong R&D relationship with in the past. There is a lot of shared information there and again AMD can reduce the money and time they would need to do this heavy lifting simply by using this brain trust. If the members of the HSA see a value in helping AMD to integrate the 8000 series into their ARM SoCs (and we have a feeling they will) then AMD will get the help they need from that integration powwow.

Still there is a little blue fly in the ointment called Intel. We are fairly confident that no one considered the possibility of Intel making a fast AND energy efficient SoC in the time span that they have. AMD has now underestimated Intel twice and in the process they have drug ARM along with them. Now AMD is going to be behind the curve by about 18-24 months when they can finally put a low power SoC with 8000 series graphics on the market. They have to rely on ARM and others more heavily than they probably intended. AMD is already very late to the thin and light, tablet and phone party while Intel is enjoying fairly good sales. Still with no money to spend to invent their own new products from the ground up AMD has to enter partnerships and buy technology IP when they can to build a better offering. AMD has made the right move here, but I am concerned that it might be too late to make the difference they want and need. We hope that AMD can ramp up and start turning things around soon. The alternative is to slide into obscurity and no one at AMD wants that.

Tell us what you think in our Forum

Read 2351 times Last modified on Thursday, 02 May 2013 14:29

Leave a comment

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