“Great strategy is important, but execution really matters.”
AMD’s chief technology officer, Phil Hester, has spoken exclusively to HEXUS.tv to reveal the cornerstone of AMD’s product development strategy.
Expanding on the Fusion concept, which was first explored soon after the take-over of ATI, Hester told us about the Accelerated Processing Unit (APU).
“The APU idea really says that there will be families of designs that have different ratios of traditional processors, GPUs and other specialised accelerators, depending on the particular market segment you’re going at,” he said.
In practice this means having several different, specialised types of accelerator within one processing unit. The fact that such a concept incorporates many of the processing functions that would normally have been done by a number of discrete components in a PC has given rise to the label ‘PC on a chip’ for this type of technology. “A PC on a chip is what’s coming in the client space,” said Hester.
In his initial analysis of Hester’s presentation, HEXUS.net editor Tarinder Sandhu explained how such a heterogeneous APU could be tailored for specific usage models. “An APU can be architected for specific platforms: CPU-heavy for the workstation/server environment and GPU-heavy for the media-rich desktop and mobile spaces," he said.
"CPUs are fast enough to open your Word, spreadsheet or Outlook but struggle when tasked with transcoding high-definition media, something that GPUs can thrive on if the application is coded for correctly.”
As Tarinder implied, accelerator types can have a go at tasks they’re not especially good at, but they will do so inefficiently and essentially have to work much harder. By allocating tasks to those accelerators best able to do them, the work required is minimised and the energy requirements are reduced.
Energy efficiency is fast becoming as important as raw power in determining the desirability of a processor. This isn’t just to placate the greenies, but because laptops have become the dominant form factor and extending battery life is a big priority."Putting more and more cores that use up more power but don’t change the user experience is not a good thing."
“You can integrate a CPU and a GPU by having an internal PCI-E bus,” said Hester. “But we’re trying to do a much tighter integration so that we get the best possible power efficiency. Putting more and more cores that use up more power but don’t change the user experience is not a good thing.” This tighter integration apparently involves having all the accelerators on one die.
On the matter of AMD’s latest generation of processors, Hester said: “Yields and speed do not correlate [to each other]. The yields on the 65nm technology have been fine since we started ramping. There have been no issues, contrary to the FUD (definition?) that Intel might be creating.
“The matching and validation [tuning] has taken longer than we would like. That is the process that needs execution improvement.” On that note, he concluded: “Great strategy is important, but execution really matters.”
On the whole it was really encouraging to see AMD deliver such an exciting strategy and show clear vision on where it’s headed. AMD in 2007 did sometimes appear a bit rudderless as it incorporated ATI and Intel and NVIDIA raised their game. Phil Hester gives the impression of being very much at the helm.
The convergence of IT products with consumer electronics and telecommunications ones is a positive trend for AMD, with AMD/ATI built into many TVs, mobile phones, etc. Intel’s domination of the CPU space is always going to be tough to challenge, so AMD is fighting the battles it thinks it can win.Hester’s most telling comment, however, was his last. It’s all very well having cunning plans and clever ideas, but if these aren’t executed well they count for nothing. HEXUS.channel will be keeping a close eye on the progress of Hester’s APU vision. It should be fascinating.