SATA subanstially slower than PATA

[Captain Norm]

A few months ago I built a new system, AMD3500+, 1GB ram, MSI K8N Neo2
Platinum mb.

If I run GHOST 2003 (dos version), I get the following MB/Minute:

989 MB/min WD Raptor 36GB (10Krpm, udma150) to WD 2000JB (7200rpm, udma100)
1880 MB/min WD 1200JB (7200rpm, udma100) to WD 2000JB (7200rpm, udma100)

In summary, the 10Krpm udma150 drive is about half the speed of the 7200rpm
udma100 drive, with the same destination drive.

I went one step furthur & ran the same GHOST program on the same system,

2470 MB/min Maxtor 6Y200P0 (udma133) to Maxtor 6Y250P0 (also udma133).

This last test indicates the system is very capable of going faster than the
WD360GD. I would expect the WD360 to be at least as fast as the udma133
7200rpm drives.

I talked to MSI tech support, they went thru the bios settings, everything
is set optimally.

I complained to Western Digital, they sent me a re-certified WD360GD. Same
results.

I also ran SiSoft Sandra & got the following:

wd360gd 39MB/s (di), 20ms avg access (aa), 10000rpm, udma150 sata
wd2500jb 45MB/s di, 21ms aa, western digital 250GB 7200rpm, udma100 ide
maxtor 6y200p0 47MB/s di, 8ms aa, maxtor 200GB 7200rpm, udma133 ide
st3300831as 57MB/s di, 7 ms aa, seagate 300GB 7200rpm, udma150 sata (on the
same sata interface).

WD360 should be my fastest drive, but still is the slowest.

Does anyone have any idea why?

 

[Peter]

If I run GHOST 2003 (dos version),

Maybe that is the problem: SATA under DOS with Ghost DOS version?

Try to run under WinPE with the same storage drivers as for XP.
Use Ghost32.exe.

 

[Bob Willard]

A few months ago I built a new system, AMD3500+, 1GB ram, MSI K8N Neo2
Platinum mb.

If I run GHOST 2003 (dos version), I get the following MB/Minute:

989 MB/min WD Raptor 36GB (10Krpm, udma150) to WD 2000JB (7200rpm, udma100)
1880 MB/min WD 1200JB (7200rpm, udma100) to WD 2000JB (7200rpm, udma100)

In summary, the 10Krpm udma150 drive is about half the speed of the 7200rpm
udma100 drive, with the same destination drive.

I went one step furthur & ran the same GHOST program on the same system,

2470 MB/min Maxtor 6Y200P0 (udma133) to Maxtor 6Y250P0 (also udma133).

This last test indicates the system is very capable of going faster than the
WD360GD. I would expect the WD360 to be at least as fast as the udma133
7200rpm drives.

I talked to MSI tech support, they went thru the bios settings, everything
is set optimally.

I complained to Western Digital, they sent me a re-certified WD360GD. Same
results.

I also ran SiSoft Sandra & got the following:

wd360gd 39MB/s (di), 20ms avg access (aa), 10000rpm, udma150 sata
wd2500jb 45MB/s di, 21ms aa, western digital 250GB 7200rpm, udma100 ide
maxtor 6y200p0 47MB/s di, 8ms aa, maxtor 200GB 7200rpm, udma133 ide
st3300831as 57MB/s di, 7 ms aa, seagate 300GB 7200rpm, udma150 sata (on the
same sata interface).

WD360 should be my fastest drive, but still is the slowest.

Does anyone have any idea why?

If you want to benchmark a HD, get a HD benchmark. HDtach is the only one
I trust; v3.x runs under XP, and v2.x runs under W9x. Sandra is not a HD
benchmark, and neither is Ghost.

And, FWIW, the WD360 is not the fastest HD in your bunch, in terms of max.
STR: that Seagate SATA HD is newer and faster. If you want the fastest HD,
and you don't want SCSI or RAID, then the WD740 is still the king.

 

[Eric Gisin]

Ghost is a file copy, not true image copy. If the source has small files you
will get low speeds.

If you want to measure STR under DOS, use Bart's disktool or c't h2bench.

 

[Rod Speed]

If you want to benchmark a HD, get a HD benchmark. HDtach is the only one I
trust; v3.x runs under XP, and v2.x runs under W9x. Sandra is not a HD
benchmark, and neither is Ghost.

An imaging app is however a very useful measure of real thruput achieved.

 

[Captain Norm]

I ran latest shareware HD Tach, 8mb zone.
49.4MB/s SATA 36GB WD 10000rpm, udma 150
56.9MB/s SATA 300GB Seagate 7200rpm, udma 150
49.8MB/s PATA 250GB Maxtor, 7200rpm, udma 133

Higher rpm drive still the slowest.

Regarding: <<that Seagate SATA HD is newer and faster>>

Shouldn't the higher rpm make a big difference, assuming both are udma 150?

Norm Perron

 

[Rod Speed]

I ran latest shareware HD Tach, 8mb zone.
49.4MB/s SATA 36GB WD 10000rpm, udma 150
56.9MB/s SATA 300GB Seagate 7200rpm, udma 150
49.8MB/s PATA 250GB Maxtor, 7200rpm, udma 133

Higher rpm drive still the slowest.

But nothing like the variation you are getting in ghost.

Regarding: <<that Seagate SATA HD is newer and faster>>

Shouldn't the higher rpm make a big difference, assuming both are udma 150?

Nope, sectors per track are the other crucial variable.

 

[Bob Willard]

I ran latest shareware HD Tach, 8mb zone.
49.4MB/s SATA 36GB WD 10000rpm, udma 150
56.9MB/s SATA 300GB Seagate 7200rpm, udma 150
49.8MB/s PATA 250GB Maxtor, 7200rpm, udma 133

Higher rpm drive still the slowest.

Regarding: <<that Seagate SATA HD is newer and faster>>

Shouldn't the higher rpm make a big difference, assuming both are udma 150?

Norm Perron

Now that you've run HDtach, you have the real STR data; and, that data seems
to correlate well with what storagereview measured for those HDs.

Higher RPM helps access time which (in general) helps throughput with a
random-access workload, but it does not -- by itself -- help STR, since
that is measured with a sequential access pattern. STR is proportional
to the product of the number of sectors per track and the RPM. So, with
decent design, a 10K RPM HD with 720 sectors/track and a 7200 RPM HD with
1000 sectors/track should have the same STR: 7.2M sectors/minute.

Note that with currently available HDs, neither the SATA (UDMA 150 MB/s)
nor the best PATA (UDMA 133 MB/s) is a bottleneck for a HD, since no
real (non-RAID) HD has a STR anywhere near the max. bus datarate.

 

[Captain Norm]

Bingo!

I ghosted my boot drive (WD Raptor, lots of small files) to another drive,
then ghosted back to the Raptor. Now I have the same files, same sizes of
files, both defragmented. Re-running my ghost to a 200GB udma133 drive
shows:

1544 MB/min with WD Raptor 36GB,
724 MB/min with IBM 80GB udma100 7200rpm drive.

With the same data & same destination drive, Raptor is 2 X as fast as
udma100.

My original tests was not comparing apples (many small files) with apples
(large drives with pretty large mp3 files).

Now, I'm convinced the 10,000 drive was worth it. I don't need much capacity
for the boot disk, so decided on the smaller 36GB drive. Now that I read
Bob's comment <<proportional
to the product of the number of sectors per track >>, the density makes
sense. Next time, I'll go with the 72GB Raptor.

I just have 2 additional questions, as it appears the experts are listening:
Bob talked about STR, what does that mean, I never heard that term?

Does a 8MB cache, opposed to 2MB cache really make a difference, or is that
just marketing hype?

Norm Perron

 

[Rod Speed]

Bingo!

I ghosted my boot drive (WD Raptor, lots of small files) to another drive,
then ghosted back to the Raptor. Now I have the same files, same sizes of
files, both defragmented. Re-running my ghost to a 200GB udma133 drive
shows:

1544 MB/min with WD Raptor 36GB,
724 MB/min with IBM 80GB udma100 7200rpm drive.

With the same data & same destination drive, Raptor is 2 X as fast as
udma100.

My original tests was not comparing apples (many small files) with apples
(large drives with pretty large mp3 files).

Now, I'm convinced the 10,000 drive was worth it. I don't need much capacity
for the boot disk, so decided on the smaller 36GB drive. Now that I read
Bob's comment <<proportional
to the product of the number of sectors per track >>, the density makes
sense. Next time, I'll go with the 72GB Raptor.

I just have 2 additional questions, as it appears the experts are listening:
Bob talked about STR, what does that mean, I never heard that term?

Does a 8MB cache, opposed to 2MB cache really
make a difference, or is that just marketing hype?

Its mostly hype, most wouldnt be able to pick it
without using a benchmark, but it can be close to
free, so it makes sense to have it in that situation.

 

[Bob Willard]

Bingo!

I ghosted my boot drive (WD Raptor, lots of small files) to another drive,
then ghosted back to the Raptor. Now I have the same files, same sizes of
files, both defragmented. Re-running my ghost to a 200GB udma133 drive
shows:

1544 MB/min with WD Raptor 36GB,
724 MB/min with IBM 80GB udma100 7200rpm drive.

With the same data & same destination drive, Raptor is 2 X as fast as
udma100.

My original tests was not comparing apples (many small files) with apples
(large drives with pretty large mp3 files).

Now, I'm convinced the 10,000 drive was worth it. I don't need much capacity
for the boot disk, so decided on the smaller 36GB drive. Now that I read
Bob's comment <<proportional
to the product of the number of sectors per track >>, the density makes
sense. Next time, I'll go with the 72GB Raptor.

I just have 2 additional questions, as it appears the experts are listening:
Bob talked about STR, what does that mean, I never heard that term?

Does a 8MB cache, opposed to 2MB cache really make a difference, or is that
just marketing hype?

Norm Perron

STR = Sustained Transfer Rate. For HDs, it means the transfer rate achieved
for long transfers with a sequential access pattern and, if a single number
is reported, it is for transfers on outer cylinders where the STR is highest.

STR is a more meaningful number than the peak transfer rate in evaluating
HDs, since STR involves transfers between system RAM and the actual magnetic
media, while peak transfer rate normally means transfers between system RAM
and the HD's cache. And, peak transfer rate is sometimes slower than the
quoted (max.) transfer rate of the bus (PATA/SATA/SCSI/etc.).

That said, the best benchmark of a system is the performance of the workload
you actually use. But many SOHO users either don't know the characteristics
of their actual workload or cannot easily create a benchmark which matches
those characteristics.

 

[Folkert Rienstra]

Wrecked by that p.o.s. Mozilla news client's quoting restored.

STR = Sustained Transfer Rate. For HDs, it means the transfer rate achieved
for long transfers with a sequential access pattern and, if a single number
is reported, it is for transfers on outer cylinders where the STR is highest.

Obviously not if it says "average" STR.

STR is a more meaningful number than the peak transfer rate in evaluating HDs,

since STR involves transfers between system RAM and the actual magnetic media,

while peak transfer rate normally means transfers between system RAM and the
HD's cache.

Which is usually 'known' as "burst rate" (which is equally wrong).

 

[Ron Reaugh]

Can you say gibber?

Wrecked by that p.o.s. Mozilla news client's quoting restored.

highest.

Obviously not if it says "average" STR.
evaluating HDs,
media,

Which is usually 'known' as "burst rate" (which is equally wrong).
(PATA/SATA/SCSI/etc.).

 

[Ron Reaugh]

There were early versions of Raptors including all 36GB and some 74 GB
versions that were slower than the second version that only includes 74GB
Raptors.

 

[chrisv]

Can you say gibber?

In your case, "clueless gibber" is more accurate, Rod^Hn.

How about you settle on one name, Rod^Hn?