Early next year AMD will release its much hyped 64-bit Hammer chips. Its country manager for Australia and New Zealand, John Robinson and technical manager Michael Apthorpe speak to Howard Dahdah about what customers can expect from its new x86-64 architecture.
Why should customers turn to AMD for 64-bit computing?JR: The 64-bit opportunity that AMD presents is a solution that enables all 32-bit applications users to seamlessly transfer themselves into 64-bit. So basically you can go into 64-bit computing and retain all your 32-bit legacy applications.
MA: It is seamless because the AMD platform is an extension of your existing 8, 16, 32-bit. The natural migration is to go from 32 to 64-bits. This architecture is built around that and is an extension of it.
JR: Our 64-bit chips are designed to cover not just servers and high-end workstations activities, but will be adopted to desktops and mobile space so we will have all form factors using 64-bit next year. This itself allows the core to be manufactured in volume and have a significant impact on pricing. There will be a natural migration because we will have the economies of scale to have volume with the associated cost reduction.Explain just what Hammer is.MA: Hammer is an internal codename for x86-64. We have two processors available in our x86-64 architecture: Athlon 64 and Opteron.
Our Opteron processor is a server-based product and our Athlon 64 is a desktop product. The internal core is the same. The two significant differences between them is that the desktop processor has a single memory controller and one 16-bit HyperTransport Interconnect. Our Opteron has a dual memory controller and three HyperTransport Interconnects. So the Opteron is more flexible and scalable for multiple processing.
HP says it is not considering a 64-bit future with AMD, instead it is aligned with Intel's IA64 architecture. What do you say?MA: At the 64-bit forum held at Comdex (in November) we had great support. IBM migrated DB2 across in two days. We also have been supported by Cray. Why would Cray, a company synonymous with supercomputing, go with ours?
JR: Cray will use 10,000 Opterons in one cluster. Cray will build a supercomputer (for Sandia National Laboratories in the US). This deal is worth $US90 million. Now that is a big call. They have obviously been using early prototypes. They have evaluated it and accepted the performance of what it can do.
What ISVs do you have onside for the Opteron chip?MA: Heaps and heaps. All the major companies.
JR: The fact is we have not yet launched the product. It is not available until the end of first quarter next year. Right now we have made a lot of noise about it.
What is in this processor that will excite IT managers?JR: This brings 64-bit computing to an environment that is from a cost point of view totally different to what anyone has seen in the past. With our 64-bit offering we won't be in the thousands of dollars (for our processors). We will be in the hundreds of dollars for exactly the same if not better performance.
When you think about it, it is the processors that make up the bulk of the differences between price in systems.
MA: The main thing about our new X86-64 bit processors is that it is a natural migration from 32-bit to 64-bit computing. The problem with large organisations is the total cost of ownership and migration to new technology. The main feature about our new Opteron processor and x86-64 is that it is a natural migration from 32-bit technology to the 64-bit world. Most of all their current software works on it, like TruAlpha, SPARC and the like. What that means is they don't have to go and buy proprietary hardware and software interfaces. All their current applications and operating systems just work on our new Opteron processor. So again, it is seamless migration from their current environment to the new world of 64-bit computing.
One of the problems large organisations are experiencing is that 4GB is the maximum amount of memory you can get in 32-bit computers and it is topping out very easily. You want to get more users on your network and run larger applications. You also want to be able to go from a 1Gb network to 10Gb network in the future, so you will need to get new hardware. The problem is that you have invested $10 million to $20 million on your current hardware and software applications; do you completely rewrite your software and throw that money away just because you don't have enough RAM? No, you will keep your 32-bit technology and suffer the consequences. With our Opteron processor, it will be a natural migration to 64-bit technology; you can just slowly introduce a few servers as you need them and keep all your existing software.
JR: The decision is straightforward. If he (IT manager) wants to go down the Itanium path he has to accept only the applications that are specifically written for IA64 and nothing else. And he has to consider all those legacy applications he has that they have worked with in the past. They will not work at all. He can't use any of that. So where does he go? Wait for new applications to be written and also pay [lots] for this new architecture, or, go to an environment that still maintains total compatibility with all existing applications and then be able to feel comfortable that 64-bit applications in x86 format are going to be scalable. There is no change of hardware or platform or whatever.
This is why we have made a lot of noise leading up to our release. A lot of people are holding back on their decision-making on which way to go [AMD or Intel].
Are there any Australian beta testers?MA: There will be shortly. It is still a [while to the] launch. The next sequence within the launch is that we are seeding it into corporations to get their feedback before the launch. Next month our local OEMs will be evaluating so that on launch day you can buy a system from your local OEM. We are preparing the ground.