By: Michael S (already5chosen.delete@this.yahoo.com), August 26, 2015 2:59 am
Room: Moderated Discussions
Anon (nope.delete@this.nope.com) on August 26, 2015 1:19 am wrote:
> Michael S (already5chosen.delete@this.yahoo.com) on August 25, 2015 2:58 pm wrote:
> > Afterthoughts of Core-M vs Cherry Trail comparison.
> >
> > Until yesterday I didn't look at Intel Core-M and Cherry Trail too closely, but yesterday I did.
> > And what I saw? Top of the line Core-M (M-5Y71) is better than top of the
> > line Cherry Trail (x7-Z8700) in every single aspect except selling price:
> >
> > Single Threaded Performance - ALOT better
> > Graphics - ALOT better
> > Multithreaded FP - ALOT better
> > These two were expected
> > What I didn't expect is that Core-M is also significantly ahead in multithreaded Integer tests. So
> > far ahead that it seems that multithreaded Integer performance per Watt is likely about equal.
> >
>
> Core M is a much more expensive part,
Yes, I said that.
> with larger die area (when you take into account the
> compulsory on-package chipset),
Irrelevant for KNL replacement that I am discussing.
Also, as I said above, I don't propose Haswell or Skylake like uArch, but something with scalar part of the core more like Bridges. Scalar part 1 *Bridge-like core shouldn't be much bigger than a a scalar part of pair of *Mont-like cores on the same process.
> and over double the TDP (4.5W vs 2W).
How do you know the TDP Cherry Trail?
I was under impression that Intel only published SDP (2W) and for TDP you have to be an OEM and to sign NDA.
> Not to mention the
> Broadwell Core M is missing an ISP, which is kind of essential in a tablet these days.
>
The post was not about tablets. It was about relative merits of two families of Intel cores for HPC-oriented Xeons. Tablet chips were mentioned only as the best available public source for comparisons of the cores.
> Michael S (already5chosen.delete@this.yahoo.com) on August 25, 2015 2:58 pm wrote:
> > Afterthoughts of Core-M vs Cherry Trail comparison.
> >
> > Until yesterday I didn't look at Intel Core-M and Cherry Trail too closely, but yesterday I did.
> > And what I saw? Top of the line Core-M (M-5Y71) is better than top of the
> > line Cherry Trail (x7-Z8700) in every single aspect except selling price:
> >
> > Single Threaded Performance - ALOT better
> > Graphics - ALOT better
> > Multithreaded FP - ALOT better
> > These two were expected
> > What I didn't expect is that Core-M is also significantly ahead in multithreaded Integer tests. So
> > far ahead that it seems that multithreaded Integer performance per Watt is likely about equal.
> >
>
> Core M is a much more expensive part,
Yes, I said that.
> with larger die area (when you take into account the
> compulsory on-package chipset),
Irrelevant for KNL replacement that I am discussing.
Also, as I said above, I don't propose Haswell or Skylake like uArch, but something with scalar part of the core more like Bridges. Scalar part 1 *Bridge-like core shouldn't be much bigger than a a scalar part of pair of *Mont-like cores on the same process.
> and over double the TDP (4.5W vs 2W).
How do you know the TDP Cherry Trail?
I was under impression that Intel only published SDP (2W) and for TDP you have to be an OEM and to sign NDA.
> Not to mention the
> Broadwell Core M is missing an ISP, which is kind of essential in a tablet these days.
>
The post was not about tablets. It was about relative merits of two families of Intel cores for HPC-oriented Xeons. Tablet chips were mentioned only as the best available public source for comparisons of the cores.