|
Prev: NTP: (Was: Re: performance of hardware dynamic scheduling)
Next: A race-condition in a SUN paper by Detlefs on reference counting...
From: Nick Maclaren on 4 Apr 2008 12:52 In article <ft5jkh$hof$1(a)gwaiyur.mb-net.net>, Elcaro Nosille <Elcaro.Nosille(a)googlemail.com> writes: |> Wilco Dijkstra schrieb: |> |> > See also: http://laptoping.com/intel-atom-benchmark.html |> > which shows Atom/Silverthorne is less than half the speed |> > of a Celeron-M... |> |> Super-PI is single-threaded and Atom has two threads which give |> a stong performance-increase over single-threaded software if |> the SW could be parallelized. This is because Atom is a simple |> in-order design. Well, maybe, but I suspect that SuperPI is also limited by the multiplier unit(s). Depending on how Silverthorne has been designed, it is possible that a single thread can use all of them, and running two threads of SuperPI would go no faster. Without knowing more about the details than I do, I can't guess. Regards, Nick Maclaren.
From: Peter Matthias on 4 Apr 2008 13:15 Jon Harrop wrote: > > In the light of recent discussions, I thought this article would be of > interest: > > http://www.linuxdevices.com/news/NS5492118276.html This is an all marketing article. This time from ARM side. My guess is in not that distant future all routers, NAS and so on will be based on Atom. Also all TVs will include Atom. Simply because ARM is fragmented. Chinese manufacturers will make cheap standardized hardware and the open source folks will make the software for it. Peter
From: Nick Maclaren on 4 Apr 2008 13:27 In article <ft5mii$on7$1(a)gwaiyur.mb-net.net>, Elcaro Nosille <Elcaro.Nosille(a)googlemail.com> writes: |> |> > Well, maybe, but I suspect that SuperPI is also |> > limited by the multiplier unit(s). |> |> Maybe, but I think these are pipelined and could nearly |> double their throughput with the second thread because |> each thread is in-order. Provided that there are enough, yes. |> > Depending on how Silverthorne has been designed, it is |> > possible that a single thread can use all of them, and |> > running two threads of SuperPI would go no faster. |> |> Atom is in-order and I think there aren't any execution |> -units that could be used in parallel by two threads. That may well be the case. I am merely expressing reservations, because there are reasonable ways it could have been designed that could deliver either result. I simply can't guess. If you actually know that even SSE instructions can't use all of the multiplication units, then that is different. But it makes the claims of "SMT" a bit odd, if that is the case. Not impossible, of course. Regards, Nick Maclaren.
From: YANSWBVCG on 4 Apr 2008 14:17 Peter Matthias <Peter.Matthias(a)ivalid.invalid> wrote: > Jon Harrop wrote: > >> >> In the light of recent discussions, I thought this article would be of >> interest: >> >> http://www.linuxdevices.com/news/NS5492118276.html > > This is an all marketing article. This time from ARM side. My guess is in > not that distant future all routers, NAS and so on will be based on Atom. > Also all TVs will include Atom. Simply because ARM is fragmented. Chinese > manufacturers will make cheap standardized hardware and the open source > folks will make the software for it. > > Peter The big plus of Atom is x86-64 compatibility. Having experimented with ARM and Windows CE, I no longer have any real interest in anything not x86-64 bit compatible. I'd really like to have an HP 50g calculator using an Atom chip.
From: Paul Gotch on 4 Apr 2008 16:02
Peter Matthias <Peter.Matthias(a)ivalid.invalid> wrote: > This is an all marketing article. This time from ARM side. My guess is in > not that distant future all routers, NAS and so on will be based on Atom. No because routers/network equipment want big endian data for efficently handling IP packets which is why a hell of a lot of them are MIPS or PowerPC since ARM only recently got a byte invariant big endian mode. Now it has one you are starting to see ARMs tip up in some neworking equipment. > Also all TVs will include Atom. No because the BOMs for TVs are incredibly cost sensitive they will not use a chip that costs $6, it would have to be $1 and TVs have no use for x86 compatability as no consumer electronics manufacturer in their right mind is going to ship a TV which you can run user programs on, -p -- "Unix is user friendly, it's just picky about who its friends are." - Anonymous -------------------------------------------------------------------- |