Here are the influential voices leading the conversations where nonprofits and technology overlap.
BizTech editor-in-chief Lee Copeland spoke recently with Kevin Knox, vice president of commercial business at Advanced Micro Devices, about the chip-maker’s Barcelona quad-core server line, built off the 10h architecture, and about how AMD tweaked the processor to better support the horsepower needed for virtualization. In Part 2 of this interview, Copeland talks to Knox about AMD’s green initiatives.
BizTech: We have recently surveyed a lot of our BizTech readers about what they’re doing around green IT. And it seems like it would be a low-hanging fruit, but what we’re hearing from a lot of folks is though, while they’d like to do something green-oriented, unless there’s significant cost savings that they can document, it’s just not something that they’re going to do. Do you have any suggestions on how IT managers might be able to get behind this better?
Kevin Knox: Yes. That’s an excellent question. A couple of years ago, AMD started an organization — kind of an industry consortium — called The Green Grid consortium. And this is precisely what The Green Grid is doing. It’s really looking for a way to allow people to more effectively measure power inside of their IT environment. It’s also establishing forums for best-practices sharing. The key is that we all share this information because I think, collectively, as an industry we need to (a) raise awareness, and (b) start to address the issue of power. It is not going away any time soon. In fact, I would suggest to you that it’s going to get worse over time.
BizTech: In terms of coming up with a way to better measure this so that people can compare apples to apples, rather than an apples-to-oranges environment, do you see a power number or a standard that evolves that would help enable that?
Knox: Yes. You’re alluding to frankly an area that’s been pretty confusing in the industry for a variety of reasons. The bottom line is, you need to make sure you’re doing an apples-to-apples comparison at the system level, right? And, there’s really two pieces to that.
At the systems level, it’s important not to just look at a processor, you know, just as an example. If you look at the AMD architecture and you compare that to our competitors’ architecture, there’s things like the memory controller, which we’ve embedded in the processor, but that’s an additional power requirement for the other architecture.
The second point, which I think is extremely important, is to make sure you’re doing an apples-to-apples comparison about something that matters, like the power at the wall. It’s the amount of power that that specific server’s going to require from an outlet or from a wall perspective.
There’s a lot of things around average power, maximum power, and a lot of times things get confused, but we think at the wall is probably the most important and most meaningful for end-user customers because, frankly, that’s what they’re going to be doing — or paying — from a power perspective, from a power-bill perspective.
BizTech: That’s most definitely a bit of actionable insight that I think IT managers would appreciate. Let’s shift gears. In terms of a company that is currently using dual-core systems, when does it make sense for them to start thinking about shifting to a quad-core environment?
Knox: The benefits of shifting to quad-core are going to be pretty immediate. And, regardless of the applications, there’s a lot of other benefits around virtualization and power and better performance; it’s a new architectural design, [more] than just the switch from dual-core to quad-core.
Specifically, on the quad-core question, there are certain types of applications that take advantage of quad-core and that are written to take advantage of multiple threads inside of the software.
We think there are immediate benefits of moving to quad-core technology there, but, you know, it’s pretty much across the board. You’ll see benefits from quad-core even if you’re not taking advantage of the specific quad-core functionality just because of the new architecture that we’re introducing.
BizTech: Another question is around availability. I’ve heard that there have been some challenges getting these chips, you know, into the hands of IT managers.
Knox: It is an issue that we’ve had, in that we’re introducing Barcelona and we actually uncovered a very specific issue called a PLB problem.
BizTech: Is it the errata?
Knox: Yes, this is the errata that we were able to identify. Now, you know, I do want to make sure that everyone understands this is actually an errata that we uncovered inside our labs in a very, very extreme test environment. So, this was not something that we saw in general-production environments inside end-user accounts. Based on the fact that we were able to identify that errata, we made the decision that we were going to basically put out another stepping of the processor, called the B3 processor, to address this errata in the silicon itself.
Actually with the B2, there were actually some workarounds in BIOS and at the operating-system level. We thought that was introducing a lot of complexity into the environment. So, we decided we were going to address it in silicon. We made that decision, and we think it was the right decision to basically err on the conservative side. We will be shipping B3 silicon to our partners at the end of March, and we will see systems in production in end-user environments in the April timeframe.
We’re actually pretty excited about the time that we’re going to have between us shipping products to the time that the end users will see those products. And a lot of that, frankly, has to do with the learnings that we’ve had with some of the issues in the errata in the B2 silicon.
BizTech: So, this will be really available for people to take a look at the end of this month?
Knox: It’s really the OEM introduction schedules that will dictate when they’re available, but there will be production systems in [later this month] available for end-user consumption.
BizTech: Are there some changes that will effect performance in the B3 chips that managers will notice?
Knox: The answer is no. They should see zero performance difference between the B2 and B3. Essentially, what B3 did was it addressed the errata problems that existed in B2.
BizTech: Last, one of the challenges that a lot of IT managers will face is how to really make comparisons with new types of chips and using different types of benchmarks, because it’s been so easy to focus on clock speed — and that was really important, particularly when running a single application in a single-hardware environment situation. But as you move to virtualization and other areas, it’s harder to make clear what those performance benefits are and put, you know, numbers against them.
Knox: We certainly agree with that, and it can’t just be AMD. It’s got to be the industry as a whole.
BizTech: Do you see a point where there would be a specification that deals with some of these new enhancements that are happening at the processor level that we don’t have the ability to adequately measure right now?
Knox: The direction that we’re looking at is to have it be a flexible decision criteria, because one company may say power’s important and another may say virtualization is important. The ability to put different weightings based on the specific requirements of the organization [is] really the direction that we need to move. Similarly virtualization is not going to be the only decision criteria but it will be one of the decision criteria, and for one organization it may be really important, another it may not be so important. So, I think we need to evolve to allow more customization of the criteria, and it’s really up to us as an industry to be able to create that framework to allow more effective decisions to be made.