By: lurker (lurker9000.delete@this.realemail.mail), November 1, 2015 3:46 am
Room: Moderated Discussions
Poindexter (cherullo.delete@this.gmail.com) on October 31, 2015 5:37 pm wrote:
> Yeah, but to assert that the architecture is unbalanced just by looking at the number of ports,
> doing some back-of-the-envelope math about it, without access to a single simulation, is just a
> *bit* too far, don't you think? Makes it look like every AMD engineer is plain incompetent.
No argument there, just writing off the entire core with what little detail we have is too early. My friend who said he worked on Zen/K12 did say their focus was to be competitive with intel, so I'm sure the current port numbers are like they are, because they intended it to be that way.
> I just did some static instruction count on LAPACK (default package on Ubuntu) - it doesn't use
> packed AVX instructions and it's full of LEAs (think it's due Fortran's calling convention). Looks
> just the kind of code where Zen's FPU can issue 4 instructions per cycle while the integer side
> is also on full tilt. Zen may be much better than Haswell for scientific computing.
>
> It also may be great for office work, for laptops, games, all kinds of web servers, so on and so forth.
I have no doubt Zen will be better than Haswell or even Skylake at some workloads. I think in the end per clock performance between Zen and Haswell/Skylake should be close for general workloads, but I think final performance will be decided by how high Zen will clock.
> Yeah, but to assert that the architecture is unbalanced just by looking at the number of ports,
> doing some back-of-the-envelope math about it, without access to a single simulation, is just a
> *bit* too far, don't you think? Makes it look like every AMD engineer is plain incompetent.
No argument there, just writing off the entire core with what little detail we have is too early. My friend who said he worked on Zen/K12 did say their focus was to be competitive with intel, so I'm sure the current port numbers are like they are, because they intended it to be that way.
> I just did some static instruction count on LAPACK (default package on Ubuntu) - it doesn't use
> packed AVX instructions and it's full of LEAs (think it's due Fortran's calling convention). Looks
> just the kind of code where Zen's FPU can issue 4 instructions per cycle while the integer side
> is also on full tilt. Zen may be much better than Haswell for scientific computing.
>
> It also may be great for office work, for laptops, games, all kinds of web servers, so on and so forth.
I have no doubt Zen will be better than Haswell or even Skylake at some workloads. I think in the end per clock performance between Zen and Haswell/Skylake should be close for general workloads, but I think final performance will be decided by how high Zen will clock.
