Klimax (danklima.delete@this.gmail.com) on August 16, 2014 7:52 am wrote:
> juanrga (nospam.delete@this.juanrga.com) on August 16, 2014 4:39 am wrote:
> > Klimax (danklima.delete@this.gmail.com) on August 16, 2014 3:56 am wrote:
> > > juanrga (nospam.delete@this.juanrga.com) on August 16, 2014 3:15 am wrote:
> > > > Maynard Handley (name99.delete@this.name99.org) on August 15, 2014 4:41 pm wrote:
> > > > > David Kanter (dkanter.delete@this.realworldtech.com) on August 15, 2014 3:01 pm wrote:
> > > > > > juanrga (nospam.delete@this.juanrga.com) on August 15, 2014 11:39 am wrote:
> > > > > > > David Kanter (dkanter.delete@this.realworldtech.com) on August 15, 2014 9:41 am wrote:
> > > > > > >
> > > > > > > > I also am skeptical that anyone can match Intel in performance while trailing behind by a node. To do that,
> > > > > > > > it is necessary to offer a very different product with a
> > > > > > > > different system architecture. E.g., target workloads
> > > > > > > > where cache does not help and simply slap down more memory controllers and cores (hint: that's a GPU!).
> > > > > >
> > > > > > > The ISA advantage will be greatly reduced in the top-end
> > > > > > > side of the performance spectrum, but will not vanish.
> > > > > > > Keller mentioned during Core Day conference that his K12
> > > > > > > core will have a "bigger engine" than its x86 sister
> > > > > > > thanks to the advantages of ARMv8 over x86-64, which allows to spend more transistors on compute.
> > > > > >
> > > > > > I happen to know the differences between those two designs. I'm not really sure it's
> > > > > > going to translate into a significant performance delta. My guess is maybe 10%.
> > > > >
> > > > > 10% is two years of Intel's current 5% improvement a year.
> > > >
> > > > And what if final number is more close to 20--30%?
> > > Terminally unproven and terminally without any evidence. Just
> > > supposition on side of ARM proponents. Nothing more.
> > >
> >
> > It was a mere question not a bold statement in need of silicon benchmarks. But
> > I agree you have not proven anything still except your dislike by ARM servers.
>
> Problem is, there is not even evidence there could be found 20%.(As a theoretical possibility)
> Much less that even 10% are there. (One should definitely keep on mind, that ancient depreciated
> x86 instructions got moved into microcode and that most sane compilers will avoid most problematic
> instructions. Which will definitely skew instruction mix in real workload and may miss most
> of x86 problems, because of used versions of instructions)
>
> As for me, I definitely don't like slideware/vaporware. Also people who go
> on and on how great slideware will be when it will finally get released.


The 10% figure was given by David. I already asked him to provide details on how he got that number.

Another poster mentioned 20--30% advantage for ARM servers. In my reply to him, I noticed that his guesses agree with the claims of 90W ARM SoCs providing 80--90% of the performance of 140W Xeons.

That reply was submitted days ago. Recently I got a Microprocessor Report article about ARM servers, they estimate the ~80W Cavium SoC will offer a score of 350 on SPECint_2006, whereas they give a 320 score for the 95W Xeon E5-2470 v2. The Intel chip has a node advantage of about 30%.

AMD data on Opteron Seattle against Opteron Jaguar. AMD claims the ARM core is faster and efficient that its own jaguar core. Data disclosed by AMD suggest the ARM-based Opteron is ~2.5x more efficient. Assuming the x86 tax remains constant when scaling up the cores then the tax would be reduced to about 26% for bigger, server-class, cores.

And so on.