Weird discovery I can't explain at all

[johns]

I use PowerQuest Drive Image 2002 to create backup
images of my hard drive. I simply make an image of
the C-drive to a D-partition on the same drive. When
the program is running ( in DR DOS mode ), I can tell
the data transfer rate. Here's a few rates I've observed
on different machines:

HP Kayak ( fairly old, but fast ram ) .. 300 meg/min
P4 2.0 with 500 meg ram .................... 160 meg/min
AMD 64 3000 with 1 gig ddr400 ....... 1 gig/min
DELL 4600 series P4 2.8 with 500 meg ... 1.4 gig/min

HUH! You say ... a freaking cheapy like a DELL
4600 series P4 is 40 % faster than that AMD 64 ????
Yep. I've seen it with my own two eyes. I cannot
explain it. And neither one of them is exactly slow !!!
And I tested 16 of them. It is a fact. Anybody got
a clue ?

johns

 

[SteveH]

I use PowerQuest Drive Image 2002 to create backup
images of my hard drive. I simply make an image of
the C-drive to a D-partition on the same drive. When
the program is running ( in DR DOS mode ), I can tell
the data transfer rate. Here's a few rates I've observed
on different machines:

HP Kayak ( fairly old, but fast ram ) .. 300 meg/min
P4 2.0 with 500 meg ram .................... 160 meg/min
AMD 64 3000 with 1 gig ddr400 ....... 1 gig/min
DELL 4600 series P4 2.8 with 500 meg ... 1.4 gig/min

HUH! You say ... a freaking cheapy like a DELL
4600 series P4 is 40 % faster than that AMD 64 ????
Yep. I've seen it with my own two eyes. I cannot
explain it. And neither one of them is exactly slow !!!
And I tested 16 of them. It is a fact. Anybody got
a clue ?

johns

Faster HDD?
HDD Controller that works better in DOS mode?
Numbers beinng misreported?

SteveH

 

[GlowingBlueMist]

I use PowerQuest Drive Image 2002 to create backup
images of my hard drive. I simply make an image of
the C-drive to a D-partition on the same drive. When
the program is running ( in DR DOS mode ), I can tell
the data transfer rate. Here's a few rates I've observed
on different machines:

HP Kayak ( fairly old, but fast ram ) .. 300 meg/min
P4 2.0 with 500 meg ram .................... 160 meg/min
AMD 64 3000 with 1 gig ddr400 ....... 1 gig/min
DELL 4600 series P4 2.8 with 500 meg ... 1.4 gig/min

HUH! You say ... a freaking cheapy like a DELL
4600 series P4 is 40 % faster than that AMD 64 ????
Yep. I've seen it with my own two eyes. I cannot
explain it. And neither one of them is exactly slow !!!
And I tested 16 of them. It is a fact. Anybody got
a clue ?

johns

Possibly CD drive or other low speed device (when compared to the hard
drive) on same IDE ribbon cable as the hard drive. Some IDE controllers
slow down to match the slowest device on the ribbon cable regardless of
which device is currently doing a transfer.

 

[johns]

Nope. On the AMD 64, the only hard drive in the system
is a 160 gig Hitachi SATA .. and that is one of the fastest
SATAs on the market. Mobo has nVidia 3 chipset which
is also first rate. I don't think the numbers are being
misread either. I watched a wall clock to see if the time
to move about 7 gigs agreed with the software time
reported. They agreed. 7 gigs went in about 4.5 min
on the DELLs. Also note the difference between the
Kayak and the P4 2.0. Those Kayaks are AMD 600s.
They are slow, and they have 40 gig ATA drives. Yet
their image transfer rate is double the P4 2.0s. That
is because the Kayaks use super fast ram. Apparently
the DELL 4600s are doing something with their ram
that I don't know about. But what? The AMD 64s
have on board cache for ram, and are noted for their
super fast transfers. Either I have the AMD 64s
crippled in some way that I don't know about, or
DELL has a toy they ain't advertising. 1.4 gig/min
transfer rate is screaming, and it is coming from one
of their cheapest models. Big mystery ?

johns

 

[johns]

HDD Controller that works better in DOS mode?

That is a thought, but how? Could it be the 8-bit ?
I remember that Apple always swore that 8-bit
transfers would smoke 16 or 32-bit, and they claimed
the Z80As were never going to make it for that reason.
That may very well be it. I doubt seriously if the DELLs
could top the AMD 64 in some kind of within OS
benchmark, but they are screaming in DOS mode.
My AMD 64s show a bench under 3DMark2002 of
19,800. That is fast !!!! I haven't run 3DMark on
the DELLs .. yet I'll be some kind of ill if they go
better than about 4500 with the integrated videos in
them.

johns

 

[David Maynard]

Nope. On the AMD 64, the only hard drive in the system
is a 160 gig Hitachi SATA .. and that is one of the fastest
SATAs on the market. Mobo has nVidia 3 chipset which
is also first rate. I don't think the numbers are being
misread either. I watched a wall clock to see if the time
to move about 7 gigs agreed with the software time
reported. They agreed. 7 gigs went in about 4.5 min
on the DELLs. Also note the difference between the
Kayak and the P4 2.0. Those Kayaks are AMD 600s.
They are slow, and they have 40 gig ATA drives. Yet
their image transfer rate is double the P4 2.0s. That
is because the Kayaks use super fast ram. Apparently
the DELL 4600s are doing something with their ram
that I don't know about. But what? The AMD 64s
have on board cache for ram, and are noted for their
super fast transfers. Either I have the AMD 64s
crippled in some way that I don't know about, or
DELL has a toy they ain't advertising. 1.4 gig/min
transfer rate is screaming, and it is coming from one
of their cheapest models. Big mystery ?

I don't know why you're fixated on the RAM, much less it's 'speed', in a
hard drive speed comparison.

The data transfer rates are going to be primarily determined by the
IDE/SATA channel and the hard drive, assuming everything else can keep up
(which is a rather trivial matter with today's machines), and on the hard
drive by the partitioning arrangement as well as how fragmented it is (head
seek times).

Now, if the RAM were so minuscule that it had to read/seek/write/seek/read
in excruciatingly small packets then the amount of RAM (but not speed)
might have an impact but the fact there's no correlation whatsoever between
RAM size and your disk data rates shows that isn't the determining factor here.

As for RAM speed, look at your numbers. Even an ancient BX chipset with
PC100 RAM does over 400MB/s (system rate) and your fastest disk clone was
23.3MB/s. Double that for the read and subsequent write. I.E. that old
'slow' RAM is easily 8.6 times faster than your fastest disk rate and that
should be plenty to keep up with a 11.7% hard drive DMA load leaving the
processor with 88.3% to do little more than make a read/write request and
wait for the hard drive to complete it. The RAM is effectively transparent,
not to mention that even your slowest system should have a much higher RAM
bandwidth than an old BX. About the only thing 'fast' RAM in those systems
would do is alter how many wait cycles are wasted and as Scotty once said
when an alien set the Enterprise course into a tight circle, "and at Warp 8
we're going no where mighty fast. (I.E., when you're going no where Warp 1
will do just as nicely) That's what a super fast processor with super fast
RAM would get you: a whole lot of waitin' goin' on (at Warp 8).

 

[johns]

Well, you'er explaining the obvious ... so explain the
question I asked. Why the difference in the image
creation / restore rates? The only hint I have at all
is differences in ram data rates, but even I know that
ram speed is orders of magnitude greater that the
figures I'm quoting for image creation rates. My interest
is that this business of re-imaging a hard drive is
really productive in my CAD labs in the case of vandalism, viruses, etc.
These 16 DELLS are of the
$238 ( after rebate .. monitor not included ) variety,
and I can re-image an entire lab before the students
have unpacked their books. I was looking at the
AMD 64s for this very reason, but they are $800
ea. ( my build ). I can save some big bucks considering
I have 7 of these labs, but which machine? The DELLS
just took the lead .. if I can get some hot video cards
in them for that price. I have ATI 9800s in the '64s.
Also consider the fact that the average home builder /
user is easily whacked by the bastards out there.
Reimaging in their case is a real blessing .. especially if
it can be done quickly before the other spouse catches
them. I speak from experience.

johns

 

[David Maynard]

Well, you'er explaining the obvious ... so explain the
question I asked. Why the difference in the image
creation / restore rates? The only hint I have at all
is differences in ram data rates, but even I know that
ram speed is orders of magnitude greater that the
figures I'm quoting for image creation rates.

Well, if it is, as you say, "obvious" then I'd suggest the cause likely
lies elsewhere.

What the actual cause is I couldn't say because I don't have the machines
here to investigate. It could be something as simple as the hard drives the
systems have or, as I mentioned, fragmentation.

Or it could be that the disk imaging program you're using has better DOS
drivers for some IDE/SATA controllers/chipsets than it does for others. Or
it could be that some IDE/SATA controllers/chipsets have inherently better
DOS performance while others rely more heavily on 'windows drivers'. Or it
could be something else entirely.

Some of those things, like which hard drive and fragmentation, are
potentially 'fixable'. Others might not be.