Article: AMD's Jaguar Microarchitecture
By: Linus Torvalds (torvalds.delete@this.linux-foundation.org), April 11, 2014 8:14 am
Room: Moderated Discussions
David Kanter (dkanter.delete@this.realworldtech.com) on April 10, 2014 11:24 pm wrote:
>
> I suspect a large part of this is because Jaguar doesn't have any turbo capabilities
> and the SoC memory controller and fabric is probably more primitive.
Sure, there are probably lots of explanations.
> I expect that in a competently designed SoC, Jaguar should beat out Silvermont-based
> designs. Especially in FP. It's strange to see results otherwise.
But who cares? In real life you get what you get (and Intel simply being better at turbo is part of that), and FP doesn't really matter.
It's worth pointing out that at least for some of the benchmarks (it's unclear whether just the "productivity" ones), Anandtech seems to have forced the CPU's to their rated speed. Which means that Turbo didn't get to play.
On the whole, looking at those benchmark scores, I'd probably say Jaguar perhaps did a little bit better than Silvermont. 3DPM is clearly an outlier and nobody should care about that result anyway, but the fact that Silvermont actually won the Dolphin benchmark is more interesting. Also, the lack of Turbo makes me suspect that Silvermont was somewhat held back in the name of score consistency. So ...
On the whole there's a notable lack of "slam dunk" there, and I do think the CPU side is rather comparable.
Linus
>
> I suspect a large part of this is because Jaguar doesn't have any turbo capabilities
> and the SoC memory controller and fabric is probably more primitive.
Sure, there are probably lots of explanations.
> I expect that in a competently designed SoC, Jaguar should beat out Silvermont-based
> designs. Especially in FP. It's strange to see results otherwise.
But who cares? In real life you get what you get (and Intel simply being better at turbo is part of that), and FP doesn't really matter.
It's worth pointing out that at least for some of the benchmarks (it's unclear whether just the "productivity" ones), Anandtech seems to have forced the CPU's to their rated speed. Which means that Turbo didn't get to play.
On the whole, looking at those benchmark scores, I'd probably say Jaguar perhaps did a little bit better than Silvermont. 3DPM is clearly an outlier and nobody should care about that result anyway, but the fact that Silvermont actually won the Dolphin benchmark is more interesting. Also, the lack of Turbo makes me suspect that Silvermont was somewhat held back in the name of score consistency. So ...
On the whole there's a notable lack of "slam dunk" there, and I do think the CPU side is rather comparable.
Linus