hard drive too slow

[adam russell]

I have a fairly new computer with a western digital caviar SE16 250GB 7.2K
rpm 16MB hard drive. I sometimes get lag when the drive usage goes to 100%
(as shown by the nvidia monitor), and it sometimes goes on for up to 30
seconds. On advice from others I have used the perfectdisk defragger, and
the WD smart utility which passes it, though I think that just means the
smart functions function and doesnt say anything about whether they go as
fast as they should. Is there any utility that can measure the throughput
and access times to see if they are slower than spec?

 

[winston19842005]

I have a fairly new computer with a western digital caviar SE16 250GB 7.2K
rpm 16MB hard drive. I sometimes get lag when the drive usage goes to 100%
(as shown by the nvidia monitor), and it sometimes goes on for up to 30
seconds. On advice from others I have used the perfectdisk defragger, and
the WD smart utility which passes it, though I think that just means the
smart functions function and doesnt say anything about whether they go as
fast as they should. Is there any utility that can measure the throughput
and access times to see if they are slower than spec?

Windows?

One of my favorite sites (when my main system was a Windows system) was
pcpitstop.

It can tell you where you bottlenecks are, including your hard drive
transfer speed.

It once told me my new (at the time) Gateway PC had DMA disabled on the
drive, and how to enable it! I was shocked (at Gateway) and impressed with
pcpitstop and the speedup which ensued...

Check it out. http://www.pcpitstop.com/

 

[Paul]

I have a fairly new computer with a western digital caviar SE16 250GB 7.2K
rpm 16MB hard drive. I sometimes get lag when the drive usage goes to 100%
(as shown by the nvidia monitor), and it sometimes goes on for up to 30
seconds. On advice from others I have used the perfectdisk defragger, and
the WD smart utility which passes it, though I think that just means the
smart functions function and doesnt say anything about whether they go as
fast as they should. Is there any utility that can measure the throughput
and access times to see if they are slower than spec?

You can try HDTach, and tell us the beginning and ending values of the displayed
curve. Try the "long test", to smooth the curve out a bit. I just tried this
an hour ago, on a new disk I got.

http://www.simplisoftware.com/Public/index.php?request=HdTach

Paul

 

[adam russell]

You can try HDTach, and tell us the beginning and ending values of the
displayed
curve.

The curve went from 65MB/s to 35MB/sec. (Random access 13.5ms, cpu
utilization 1%, average read 55MB/sec) Im not sure if that is low. The
spec says:
Data Transfer Rate (maximum)
- Buffer to Host
300 MB/s max
- Buffer to Disk KS / AAKS
748 Mbits/s max / 972 Mbits/s max

Should I be seeing like 6x the speed Im getting or am I reading the spec
wrong? Spec for my drive is here:

http://wdc.custhelp.com/cgi-bin/wdc...3NlYXJjaF90ZXh0PTI1MDBrcw**&p_li=&p_topview=1

Hmm, thats a long one. Its a western digital model 2500KS.

 

[Paul]

The curve went from 65MB/s to 35MB/sec. (Random access 13.5ms, cpu
utilization 1%, average read 55MB/sec) Im not sure if that is low. The
spec says:
Data Transfer Rate (maximum)
- Buffer to Host
300 MB/s max
- Buffer to Disk KS / AAKS
748 Mbits/s max / 972 Mbits/s max

Should I be seeing like 6x the speed Im getting or am I reading the spec
wrong? Spec for my drive is here:

http://wdc.custhelp.com/cgi-bin/wdc.cfg/php/enduser/std_adp.php?p_faqid=1343
Hmm, thats a long one. Its a western digital model 2500KS.

(For those Western Digital FAQ links, you can snip them off just after the
p_faqid field. I had to snip your link above, so my USENET server would
accept this post.)

My old 80GB drive (the boot drive) gave 57MB/sec begin,
30MB/sec end, 15.8ms random access. (Seagate ST380011A 80GB 7200PRM)
I used the long test, and there were still a significant number
of dips in the performance curve.

My brand new 80GB drive (thinner than the old one) gave 76MB/sec begin,
37MB/sec end, 14.8ms random access. (Seagate ST380215A 80GB 7200RPM)
The curve for this drive was much smoother, and showed "zoned" behavior -
a curve with stairsteps in it. There was only one downward "spike"
near the end of the disk. A much smoother curve, implying no sector
replacements or multiple read attempts as may have happened on my
boot disk.

Many of the specs in print, are useless without further information.
The 300MB/sec is the SATA cable rate, and assumes there are no other
bottlenecks at the physical level. The memory cache on the controller
PCB, may be able to handle the 300MB/sec, or a somewhat lower
figure. Or, the storage interface chip on the motherboard, may not be
able to sustain that rate. There are many possibilities. Some people
seem to think the cache on the hard drive controller board is a big deal,
but I cannot really tell.

The drive cannot sustain a read or write at 300MB/sec, and is limited
by the media rate. The diameter of the platter changes with track
being accessed, and the data rate shown in HDTach reflects that.
Also, the head rate quoted in the WD spec, would be for some kind of
encoded symbols. There might be 11 bits or 12 bits, to represent
a byte of user data, so the spec they quote is deceptive. Also, in
the page you provided, the KS model and the AAKS model, have an
exact 1.3 relationship between rates, but I'm not at all sure what
that means. I don't know what the differences would be, between the
two of them.

So your test results, have better random access than mine, and
your transfer rate is in between my two generations of drives.

You may have to look elsewhere, for performance issues. The way
the file system works, the driver timing out when trying to talk
to the disk, and having to retry, stuff like that. Not all issues
that can arise on the computer, are easy to benchmark. While HDTach
is nice (because it is free), there are some other tests, such as
what the statistical spread on access times is. If there is the
odd "long access", that implies an operation did not complete on
the first attempt. Having to wait for the platter to rotate one
more time, to retry a sector read, is what could put those dips
in the transfer rate curve.

I'm a little bit surprised, how "jumpy" the plot is for my boot
drive, but since the boot drive has not complained to me about
it, I'm not concerned

In the "good old days", someone might advise doing a "low level
format", as a way of starting all over again, then restore from
your backups. But a low level format now, doesn't do the same
thing it used to. In the old days, one surface was reserved for
servo info, and a low level format meant rewriting everything
on the other surfaces. At the current time, servo info is
"embedded" with the sector formatting on each surface, and that
means it cannot be erased and rewritten. That stuff is written
at the factory, and never gets changed after that. I'm not even
sure, whether a low level format wipes all record of "grown defects",
so that the spared out sectors are "unspared" or not. It should, at
least, take all the sectors with a "pending" status, and
do something to them, so at least those questionable sectors,
that might take many tries to read, would have their
status changed.

In your situation, I don't know where I'd turn next. When I
have a computer problem, I usually try to study it for a
while, to understand what might be the underlying cause.
But certainly, if time is money, just swapping in another
drive, and transferring the contents over, may be a more
pragmatic solution, than the way I might attack it

I have run into some motherboards, where from day one,
each attempt by the user, to access the disk, is met by
5 seconds of "silence", where nothing happens. Then the
disk access completes. Needless to say, this is
extremely annoying, and for most people, would mean
a death sentence for the motherboard in question. I
have never seen one of those kinds of problems
resolved. It could be a driver problem, but I've
never seen any follow up posts with any good news.

Paul

 

[kony]

The curve went from 65MB/s to 35MB/sec. (Random access 13.5ms, cpu
utilization 1%, average read 55MB/sec) Im not sure if that is low. The
spec says:
Data Transfer Rate (maximum)
- Buffer to Host
300 MB/s max
- Buffer to Disk KS / AAKS
748 Mbits/s max / 972 Mbits/s max

Should I be seeing like 6x the speed Im getting or am I reading the spec
wrong? Spec for my drive is here:

http://wdc.custhelp.com/cgi-bin/wdc...3NlYXJjaF90ZXh0PTI1MDBrcw**&p_li=&p_topview=1

Hmm, thats a long one. Its a western digital model 2500KS.

The performance numbers are within the expected range, at
this point you might consider problems with OS, background
apps running, drive fitness problems (run the HDD
manufacturer's diagnostics), bus/cable/drive response
problems (see Event Viewer in Windows). If you have a spare
data cable try it, or if not then reseat the present cable
at both ends.

 

[adam russell]

The performance numbers are within the expected range, .. <advice snipped>

Could you help me understand how you know the 55MB/sec average is within
expected range? From the specs I cant see where the conclusion comes from.
Or is it just a personal experience judgement?

 

[kony]

Could you help me understand how you know the 55MB/sec average is within
expected range? From the specs I cant see where the conclusion comes from.
Or is it just a personal experience judgement?

That's what a modern desktop drive is expected to achieve.
Whether some other drive is slightly faster is irrelevant as
you would not perceive slowness from this minor difference.
Performance degradation enough to feel it's problematic will
come from another source.

 

[Paul]

That's what a modern desktop drive is expected to achieve.
Whether some other drive is slightly faster is irrelevant as
you would not perceive slowness from this minor difference.
Performance degradation enough to feel it's problematic will
come from another source.

Let's try my measurements as an example.

My old disk, is this one:

http://www.seagate.com/support/disc/specs/ata/st380011a.html

INTERNAL TRANSFER RATE (Mbytes/sec) ______up to 85.4
SUSTAINED TRANSFER RATE (MB/sec)__________up to 58

My measured sustained peak (at the beginning of the disk) was 57MB/sec,
which agrees very closely with the "sustained" number. The
"sustained" value is what we can see as users.

The "internal" transfer rate, is bits coming off the disk in encoded
form, with sector headers (overhead) and the like. I doubt I could
do justice to explaining the ratio. Instead, I'll just divide 85.4
by 58 and I get 1.47.

Now, we'll try the WD2500KS results. 748 Mbits/s is the "internal" rate
for that drive. 748/8 = 93.5MB/sec internal. The measured result was
65MB/sec at the beginning of the disk. 93.5 divided by 65 = 1.44, and
that suggests to me that the measured 65MB/sec is consistent with
the manufacturers 748 internal number. So the raw transfer rate of the
disk appears "normal" from that perspective.

It could be that operations leading up to the driver carrying out an
operation on the disk, are slower than they should be. One thing
the disk benchmark tells me, is you are media limited (which is normal).
If you get a declining curve, that is media limited. If you get a flat
line, and a reduced transfer rate, that tells you that either a lower
UDMA mode is being used, or you've slipped into PIO mode. PIO mode is
a flat line, at about 4MB/sec transfer rate.

The other info to be obtained from HDTach, is the "smoothness" of the plot.
Now, I have to be careful, not to abuse this observation, because again,
the person carrying out the benchmark, could have background tasks, or
other system problems (spurious interrupts), that could pollute the
results. But I was pretty impressed with the fact, that my old disk
is "bumpy", while the new disk, with the exception of one tiny spot,
was smooth. That could be evidence of either sector substitution, or
even repeated attempts to read a weak sector.

Is there anything in the Event Viewer to suggest a problem ? What operations
are slow ?

Paul

 

[Terrence Quinn]

I have a fairly new computer with a western digital caviar SE16 250GB
7.2K rpm 16MB hard drive. I sometimes get lag when the drive usage
goes to 100% (as shown by the nvidia monitor), and it sometimes goes
on for up to 30 seconds. On advice from others I have used the
perfectdisk defragger, and the WD smart utility which passes it,
though I think that just means the smart functions function and
doesnt say anything about whether they go as fast as they should. Is
there any utility that can measure the throughput and access times to
see if they are slower than spec?

Have you upgraded the Nvidia chipset drivers at any point? If so, reinstall
the same version then reboot. If this fails try slightly older drivers. I've
seen several instances where benchmarks and SMART data looked fine but AV
playback was horrible.

 

[adam russell]

Let's try my measurements as an example.

My old disk, is this one:

http://www.seagate.com/support/disc/specs/ata/st380011a.html

INTERNAL TRANSFER RATE (Mbytes/sec) ______up to 85.4
SUSTAINED TRANSFER RATE (MB/sec)__________up to 58

My measured sustained peak (at the beginning of the disk) was 57MB/sec,
which agrees very closely with the "sustained" number. The
"sustained" value is what we can see as users.

The "internal" transfer rate, is bits coming off the disk in encoded
form, with sector headers (overhead) and the like. I doubt I could
do justice to explaining the ratio. Instead, I'll just divide 85.4
by 58 and I get 1.47.

Now, we'll try the WD2500KS results. 748 Mbits/s is the "internal" rate
for that drive. 748/8 = 93.5MB/sec internal. The measured result was
65MB/sec at the beginning of the disk. 93.5 divided by 65 = 1.44, and
that suggests to me that the measured 65MB/sec is consistent with
the manufacturers 748 internal number. So the raw transfer rate of the
disk appears "normal" from that perspective.

It could be that operations leading up to the driver carrying out an
operation on the disk, are slower than they should be. One thing
the disk benchmark tells me, is you are media limited (which is normal).
If you get a declining curve, that is media limited. If you get a flat
line, and a reduced transfer rate, that tells you that either a lower
UDMA mode is being used, or you've slipped into PIO mode. PIO mode is
a flat line, at about 4MB/sec transfer rate.

The other info to be obtained from HDTach, is the "smoothness" of the
plot.
Now, I have to be careful, not to abuse this observation, because again,
the person carrying out the benchmark, could have background tasks, or
other system problems (spurious interrupts), that could pollute the
results. But I was pretty impressed with the fact, that my old disk
is "bumpy", while the new disk, with the exception of one tiny spot,
was smooth. That could be evidence of either sector substitution, or
even repeated attempts to read a weak sector.

Is there anything in the Event Viewer to suggest a problem ? What
operations
are slow ?

Thanks for that explanation.
Its mostly associated with a game (vanguard), so I guess that may be the
culprit. Reading too much from hard drive? But Ive also seen it go long
after vg is shut down and no longer listed in task manager, and sometimes
when I am not gaming at all but just using IE. I dont think there is
anything in event viewer that looks likely. I see a few errors listed here
and there but nothing that looks associated with the problem. Ill take
another look when it occurs again.

 

[Paul]

Thanks for that explanation.
Its mostly associated with a game (vanguard), so I guess that may be the
culprit. Reading too much from hard drive? But Ive also seen it go long
after vg is shut down and no longer listed in task manager, and sometimes
when I am not gaming at all but just using IE. I dont think there is
anything in event viewer that looks likely. I see a few errors listed here
and there but nothing that looks associated with the problem. Ill take
another look when it occurs again.

Does Vanguard use more memory, than physically exists in the computer ?
Maybe you are in "swap country" ? On a game I use here, I can "alt-tab"
out in mid-game, then use Task Manager, to review memory consumption.
If a game had a memory leak, performance could get worse as time goes
by.

Paul

 

[adam russell]

Does Vanguard use more memory, than physically exists in the computer ?
Maybe you are in "swap country" ? On a game I use here, I can "alt-tab"
out in mid-game, then use Task Manager, to review memory consumption.
If a game had a memory leak, performance could get worse as time goes
by.

There was a memory leak up to the last patch, but I am confident thats not
the culprit. Nvidia monitor shows me % used for memory, HD, and cpu. At
the times when HD was at 100%, memory was still at a reasonable level. Task
manager also agreed that mem use was normal. The game does use an enormous
amount (20GB) of hard drive space so maybe it does read too much, but that
doesnt explain to me why the problem also occurs outside the game. Unless I
guess if windows might not clear virtual memory till later and that process
might be causing slowdowns?

 

[kony]

There was a memory leak up to the last patch, but I am confident thats not
the culprit. Nvidia monitor shows me % used for memory, HD, and cpu. At
the times when HD was at 100%, memory was still at a reasonable level. Task
manager also agreed that mem use was normal. The game does use an enormous
amount (20GB) of hard drive space so maybe it does read too much, but that
doesnt explain to me why the problem also occurs outside the game. Unless I
guess if windows might not clear virtual memory till later and that process
might be causing slowdowns?

The problem would occur outside the game because windows is
paging out OS/other files to make the memory space for the
game, then these files have to be re-read.

The best way to improve perceived HDD speed is add more
memory to the system, so it's caching more files
perpetually... though on a game with 20GB worth of files you
may still have some lag, it'll depend on how it's written,
how many loads & reloads. 20GB for a whole game might still
only be 1GB per level, for example.