> David Empson wrote:
> > nospam <nos...@nospam.invalid> wrote:
> >> Erik Richard S�rensen <
tu...@tulle.dk> wrote:
> >>> Yes, and FW400 will be quite a lot faster than USB too.
> >> actually, only slightly faster.
> >
> > On an iMac G4, Firewire 400 will be in the order of twice as fast as USB
> > 2.0 (assuming the drive is fast enough).
> >
> > The iMac in question supports USB 2.0, but USB performance is limited by
> > the CPU and possibly the hardware architecture. On similar vintage Macs,
> > the best speed I ever saw from USB 2.0 was about 17 MB/s (with a
> > moderate CPU load), while Firewire was around 35 to 40 MB/s (and almost
> > no CPU load), using the same hard drive to test both interfaces, and
> > with no hubs or other USB devices connected (which would disadvantage
> > USB further).
>
> Generally I have FW400 speeds between 30-40mb/sec from internal FW400 to
> external HDs with Oxford chipsets and 30-35mb/sec to HDs with NEC
> chipsets. And using the PowerLogics Combo PCI cards on my G4 I have
> speeds as high as up to near 48mb/sec to the same HD types. Using the
> USB2.0 port on the PowerLogix card has never come over 25mb/sec, and it
> slows radically when working with other USB units. - Even using the
> keyboard and my Kensington trackball can slow down the USB....
(I see you corrected the 48 to 38 in a subsequent post. No way can USB
2.0 hit its theoretical limit.)
> I've found here that utting in an USB 2.0 hub with balanced and
> individual power feed ports between the PCI card and ext3ernal device
> will stabilize transfers to external units so they keep near 25-30mb/sec.
>
> > Late G4 models got somewhat faster, and I expect G5 models are faster
> > again, but I've only seen USB 2.0 speeds reaching 30 MB/s on Intel Macs.
>
> Hm... I have a G4 MDD (latest)
By "late G4 models" I meant PowerBooks in 2005 (contemporary with late
desktop G5 models), not PowerMacs in 2003. The overall performance of
the USB chipset or some other factors seemed to improve between early
PowerBook G4s and later ones, as I recall doing comparitive tests on a
friend's final generation PowerBook G4. I can't remember the specific
numbers we observed on that model, just that it was faster than the 17
MB/s we got with an earlier aluminium PowerBook G4.
> and it won't give higher USB speeds than any other Mac. The G5 is a bit
> faster but not much. A MacPro Xeon 4x2,66ghz is a lot faster than the G4s
> and a bit faster than the G5, but here it totally depends on the tyhpe of
> external unit used. - I've tried the USB 2.0 on the MacPro to a Mercury
> Elite AL Pro Quad RAID harddisk. This combination is faster than anything
> else I've tried. - A transfer of up to 40mb/sec - sometimes 45mb/sec
> within reach if transfering large files, somewhat slower transfering
> folders with many small files, - but still faster than the G4 and G5.
That's surprisingly fast for USB 2.0, and approaching its theoretical
limit.
I've never seen USB 2.0 achieve speeds faster than 35 MB/s on any
computer, and rarely over 31 MB/s.
> I have two NewerTech MiniStack V3 with 500gb and 1tb HDs. The 500gb disk
> is near standard when using FW400 and FW800 - very, very close to the
> max. limit. The one with the 1tb disk is definitely faster. It isn't
> anormal that speeds with FW400 is up as high as 65mb/sec (!)
That is physically impossible.
The theoretical maximum data rate over Firewire 400 is 393.216 megabits
per second, or 49.152 megabytes per second.
>- and FW800 as high as near 120mb/sec (!).
Also physically impossible. The theoretical maximum data rate over
Firewire 800 (after allowing for the 8B/10B encoding) is 98.304
megabytes per second.
Both of those limits assume zero delays and zero overhead for data
packets and issuing commands, which cannot occur in reality. (Firewire
800 is full duplex so commands and data going in opposite directions
might be able to overlap, but there will still be packet overhead and
arbitration delays.)
The bit cell timing characteristics of a particular version of Firewire
are fixed, unless it steps down to slower rates (e.g. transferring at
S100, S200 or S400 speeds over a faster Firewire bus). There is no way a
hard drive or computer could transfer data about 30% faster than the
theoretical maxium.
Some possible explanations: you were mistaken, or some of the data was
being read from RAM cache on the computer rather than being transferred
over Firewire, or a file had sparse storage and some of its nominal size
was not allocated on disk so didn't need to be copied (typically
encountered with a .sparseimage, but the same method can be used for
other files on HFS/HFS+ volumes and on some other file systems).
--
David Empson
dem...@actrix.gen.nz