FW: PowerMac vs. Intel for PowerPC Development Host

3 messages, 3 authors, 2003-02-20 · open the first message on its own page

FW: PowerMac vs. Intel for PowerPC Development Host

From: Wells, Charles <hidden>
Date: 2003-02-20 15:41:08

-----Original Message-----
From: Wells, Charles
Sent: Thursday, February 20, 2003 9:48 AM
To: 'Chris Wedgwood'
Subject: RE: PowerMac vs. Intel for PowerPC Development Host

While I don't want to get into the Mac vs. Intel debate, I have an
observations about Paul's numbers.

Building kernels requires a lot of disk accesses.  The Toshiba/Hitachi/IBM
style hard disks that are used in most modern laptops (Mac or Intel) are
notoriously slow compared to full-sized desktop hard drives (particularly
the high performance SCSI drives).  It wouldn't be meaningful to compare two
systems without describing the access time performance of the hard disks.

While I don't have the numbers in front of me, I believe the the fastest
machine I have access to for building kernels is my trusty old 450 MHz.
desktop G3 at home.  It has 3 very fast IBM SCSI drives.  In my book, how
fast a kernel compiles is as much a matter of how much you're willing to
spend on hard drives as it is CPU architectural considerations.  Leave the
"CPU clock speed comparisons" to the marketing guys.

-----Original Message-----
From: Chris Wedgwood [mailto:cw@f00f.org]
Sent: Thursday, February 20, 2003 12:46 AM
To: Paul Mackerras; Eugene Surovegin
Cc: brian.auld@adic.com; linuxppc-embedded@lists.linuxppc.org
Subject: Re: PowerMac vs. Intel for PowerPC Development Host

On Thu, Feb 20, 2003 at 12:22:52PM +1100, Paul Mackerras wrote:
As an actual data point: my measurements show that a dual 1GHz G4
powermac is more than twice as fast at compiling PPC kernels as a
1.7GHz P4 (single cpu).
Wow... that's *much* better that I would have guessed.  Is the
compiler the same for each?
You can't fairly compare compiling a PPC kernel on a PPC box with
compiling an x86 kernel on an x86 box.  GCC does more work compiling
for PPC than for x86.
I wasn't trying to compare fairly, mostly get the lowest possible
compile time.


On Wed, Feb 19, 2003 at 09:10:48PM -0800, Eugene Surovegin wrote:
I tried to compile the _same_ PPC kernel for one of boards (440GP
based).
Cross compiling on PIII 755Mhz running SuSE        - 16 min
Cross compiling on G4 PowerBook 1Ghz running YDL   - 6 min
I don't have a PPC host to reasonably compare with, but a 440GP kernel
(linuxppc_2_4_devel) for me builds in 3 minutes 29s (with gcc-2.95).

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: FW: PowerMac vs. Intel for PowerPC Development Host

From: Paul Mackerras <hidden>
Date: 2003-02-20 23:05:10

Wells, Charles writes:
While I don't want to get into the Mac vs. Intel debate, I have an
observations about Paul's numbers.

Building kernels requires a lot of disk accesses.  The Toshiba/Hitachi/IBM
style hard disks that are used in most modern laptops (Mac or Intel) are
notoriously slow compared to full-sized desktop hard drives (particularly
the high performance SCSI drives).  It wouldn't be meaningful to compare two
systems without describing the access time performance of the hard disks.
OK, the hard disk on the G4 is an 80GB Seagate ST380021A, while the
hard disk on the P4 (which is not a laptop) is a 40GB Seagate
ST340016A, so they should be pretty similar.  On a `hdparm -tT' test,
the G4 gets 39.75 MB/s from disk whereas the P4 gets 31.98 MB/s.  In
any case, after the first compilation, it hardly hits the disk at all
on either machine.

Paul.

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: FW: PowerMac vs. Intel for PowerPC Development Host

From: Chris Wedgwood <hidden>
Date: 2003-02-20 23:15:02

On Fri, Feb 21, 2003 at 10:05:10AM +1100, Paul Mackerras wrote:
OK, the hard disk on the G4 is an 80GB Seagate ST380021A, while the
hard disk on the P4 (which is not a laptop) is a 40GB Seagate
ST340016A, so they should be pretty similar.  On a `hdparm -tT'
test, the G4 gets 39.75 MB/s from disk whereas the P4 gets 31.98
MB/s.  In any case, after the first compilation, it hardly hits the
disk at all on either machine.
I can't see any reasonable machine hitting the disk hard, pretty much
any files required during compilation should buffer/page cache.  When
I compile here, the disk is almost completely inactive except for the
odd shirt burst of dirty pages being flushed every 5s or so.


  --cw

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help