Re: performance comparison of ARM 266 MHz vs i386 800 MHz
On Tue, Oct 17, 2006 at 11:15:06PM +0200, Yakov Lerner wrote:
> We need to port existing linux unserlevel soft from i386 to ARM.
> Our tasks are lot of string searchign like lot of strstr().
> I need to understand performance ratio factor for such tasks
> between ARM 266 MHz and Intel 800 MHz(via), very roughly.
> I don't have access to the target ARM box so I cannot make direct
> experiments and measurements.
> How can I roughly and theoretically come up with some factor ?
> Please help me.
Well my experience is with a 400MHz arm (PXA255 xscale) and a 266MHz
Geode GX1, doing the same job. For floating point the Geode is of
course way faster since it actually has an FPU, while the arm has to
emulate floating point. For integer work the arm is much faster. Our
system has 4 100Mbit ethernet ports on it (pcnet32's), and with the
Geode, a 35Mbit full duplex connection is about the limit that the cpu
can push through, while I managed to do a 95Mbit full duplex connection
on the arm without reaching the cpu limit. And of course the arm did
that while using about 1/4 of the power (and the geode doesn't use
A Via Eden (which I assume it is if 800MHz) is also not a very impresive
CPU (just as the geode gx1 isn't very impresive being based on the cyrix
mediagx). And actual intel would have a much better performance, as
would something like a geode lx or geode nx.
So the answer really becomes, that if your software does no floating
point calculations, then the arm will be quite a fast chip, and
otherwise it will be very slow. I would not be surprised if a 266mhz
arm is at least close to the performance of an 800mhz via eden when
doing integer calculations.