By: aaron spink (aaronspink.delete@this.notearthlink.net), July 28, 2012 6:54 am
Room: Moderated Discussions
jp (jipe4153.delete@this.gmail.com) on July 28, 2012 2:00 am wrote:
> Aha, so your using personal bias against an architecture?
>
No, I just find it funny when a company cannot consistently describe the basic parameters of their own product.
> And your
> mixing different system hardware components and want to enforce the same naming
> convention.
>
They are all part of the same system, fyi.
> Btw I would be careful using the term "cores" and instead talk
> about SIMD units and FPUs.
>
No, cores is correct.
> Global node bandwidth is cute, but a majority of
> applications benefit from a high localized bandwidth. 51.2 is a small fraction
> of 250+ .
>
And the 250 doesn't matter when it is bottle-necked to <= 16. Then again, it isn't like that 250 is even used when you struggle to get 50% efficiency when running linpack.
> Aha, so your using personal bias against an architecture?
>
No, I just find it funny when a company cannot consistently describe the basic parameters of their own product.
> And your
> mixing different system hardware components and want to enforce the same naming
> convention.
>
They are all part of the same system, fyi.
> Btw I would be careful using the term "cores" and instead talk
> about SIMD units and FPUs.
>
No, cores is correct.
> Global node bandwidth is cute, but a majority of
> applications benefit from a high localized bandwidth. 51.2 is a small fraction
> of 250+ .
>
And the 250 doesn't matter when it is bottle-necked to <= 16. Then again, it isn't like that 250 is even used when you struggle to get 50% efficiency when running linpack.
