HD Tune gives me strange (wrong?) results on WD hard disks

Lemon Juice · Jun 29, 2013

I'm benchmarking my new WD hard drive WD5003AZEX and I have no idea why HD Tune gives me very strange graphs. Look at the two graphs below - done on the same drive at different times:

There's this strange almost flat line throughout most of the disk space, it looks as if something is limiting the transfer rate to about 125MB/s.

However, other tools show me very different results:

These results were done on the first 100GB (fastest) partition of the drive and the speed looks much greater than what HD Tune reports. I'd like to know what is going on here, why so different results?

WD5003AZEX is a SATA 3 drive. My motherboard is Asrock N68C-GS FX, which has no AHCI but supports SATA II. The system is Windows 7 Home Premium 32 bit.

BTW, I have an older WD drive attached and this phenomenon also occurs when benchmarking it, the speeds are just lower but the same flat line in HD Tune and the other tools reporting much higher transfer rates.

fzabkar · Jun 29, 2013

FWIW I notice that the results for the other tools are consistent with HD Tune's burst rate measurement. Perhaps HD Tune's flat line is a true measure of the sustained transfer rate whereas the other tools may be reporting the burst rate. If that is the case, then I would be looking for a bottleneck in the SATA controller or driver.

Lemon Juice · Jun 30, 2013

I was worried that it might have been a bottleneck in my sata controller but after searching for more information I found someone who experienced the same problem and it turned out to be a problem of HD Tune being capped to a certain transfer rate and unable to report the real drive speed. I have no idea in what circumstances the bug occurs but it seems to do so in my case.

I have tested real filesystem transfer rates using Disk Bench and I was able to achieve real read speed up to 146MB/s - this is much higher than the raw read speed of 125MB/s reported by HD Tune.

Finally, I tested my disk using Sandra and the results are what I would expect them to be and in line with all the tools except HD Tune:

Detailed Results
Speed at position 0% : 166.69MB/s (29.9MB/s - 166.69MB/s) (118%)
Speed at position 3% : 171.33MB/s (49.56MB/s - 171.33MB/s) (121%)
Speed at position 7% : 171.85MB/s (59.39MB/s - 171.85MB/s) (122%)
Speed at position 10% : 169.57MB/s (54.2MB/s - 169.57MB/s) (120%)
Speed at position 13% : 166.68MB/s (45.34MB/s - 166.68MB/s) (118%)
Speed at position 17% : 165MB/s (47.11MB/s - 165MB/s) (117%)
Speed at position 20% : 162MB/s (41.47MB/s - 162MB/s) (115%)
Speed at position 23% : 161.5MB/s (43.48MB/s - 161.5MB/s) (114%)
Speed at position 27% : 162.12MB/s (59MB/s - 162.12MB/s) (115%)
Speed at position 30% : 158.24MB/s (53.54MB/s - 158.24MB/s) (112%)
Speed at position 33% : 155MB/s (55MB/s - 155MB/s) (110%)
Speed at position 37% : 155MB/s (51.65MB/s - 155MB/s) (110%)
Speed at position 40% : 155MB/s (43.55MB/s - 213.84MB/s) (110%)
Speed at position 43% : 151.5MB/s (47.11MB/s - 151.5MB/s) (107%)
Speed at position 47% : 148.55MB/s (46MB/s - 148.55MB/s) (105%)
Speed at position 50% : 146MB/s (47.59MB/s - 146MB/s) (103%)
Speed at position 53% : 144.2MB/s (51.27MB/s - 144.2MB/s) (102%)
Speed at position 57% : 139MB/s (36.17MB/s - 139MB/s) (98%)
Speed at position 60% : 136.87MB/s (42.9MB/s - 136.87MB/s) (97%)
Speed at position 63% : 133.5MB/s (42.61MB/s - 133.5MB/s) (95%)
Speed at position 67% : 130.17MB/s (38.51MB/s - 130.17MB/s) (92%)
Speed at position 70% : 127MB/s (37.5MB/s - 127MB/s) (90%)
Speed at position 73% : 123.82MB/s (34.25MB/s - 123.82MB/s) (88%)
Speed at position 77% : 125.4MB/s (47.12MB/s - 125.4MB/s) (89%)
Speed at position 80% : 120.11MB/s (39.83MB/s - 120.11MB/s) (85%)
Speed at position 83% : 118.16MB/s (44.56MB/s - 118.16MB/s) (84%)
Speed at position 87% : 112.56MB/s (34.25MB/s - 112.56MB/s) (80%)
Speed at position 90% : 108.08MB/s (37.3MB/s - 108.08MB/s) (77%)
Speed at position 93% : 102.8MB/s (31.43MB/s - 102.8MB/s) (73%)
Speed at position 97% : 98.1MB/s (34.42MB/s - 98.1MB/s) (69%)
Speed at position 100% : 93.27MB/s (29MB/s - 93.27MB/s) (66%)
Random Access Time : 15ms (13.18ms - 46.49ms)
Full Stroke Access Time : 15.36ms (9.71ms - 49.16ms)

Benchmark Status
Result ID : WDC WD5003AZEX-00K1GA0 (500.1GB, SATA600, 3.5", 7200rpm)
Computer : ASRock N68C-GS FX
Platform Compliance : x86
System Timer : 3.58MHz
Use Overlapped I/O : Yes
I/O Queue Depth : 2 request(s)
Block Size : 1MB
Bytes Per Sector : 4kB

Volume Information
Capacity : 465.76GB

Physical Disk
Model : WDC WD5003AZEX-00K1GA0
Version : 80.0
Serial Number : WD-WC**********
Interface : SATA600
Rotational Speed : 7200rpm
Removable Drive : No
Queueing On : Yes
Page Size : 4kB

So now I'm 99% convinced that everything is fine with my drive and sata controller and it's HD Tune's fault. Maybe it has some incompatibility with my hardware.

choucove · Jun 30, 2013

Actually, your HDTune results seem to be stating about the same thing as all the rest. The HDTune graph shows your throughput as a curve over time, while the other utilities are basically just showing the maximum output from the total time of the test. So instead of viewing the average curve, you're just getting the number from the maximum result. That's why I like HDTune best, as you are getting a very good metric of the performance of the drive as a whole, not just a min/max number. We've been able to physically see drives failing with this utility.

I've never come across an issue with HDTune being limited in bandwidth support. I've used it in testing SSDs and even RAMDisks reporting upwards of 8,000 MB/s in that scenario. Now, HDTune was really designed for physical drives in mind, but I haven't seen a situation where it was causing the limitation.

So, in your HDTune results your maximum MB/s is between 136 MB/s and 158 MB/s based on your two tests. This is pretty close in line with the results from CrystalDiskMark and Sandra. Again, notice that Sandra is reporting the maximum throughput value first, but also includes the range at each point, such as: 166.69MB/s (29.9MB/s - 166.69MB/s) (118%). Think of HDTune as a representation perhaps of "averaged" performance, or the most realistic actual performance, where as the other utilities are displaying the "maximum" performance over the timeframe.

Lemon Juice · Jun 30, 2013

choucove :

Hm, this is interesting and I'm trying to wrap my head around it since maybe I don't know how these tools work under the hood. So you are saying that CrystalDiskMark is displaying the maximum throughput and not the average? That would be very strange because logically thinking when I set it to read 1000MB then I'm expecting it to test how fast it will read 1000MB (average), I'm not interested in a single peak within that range. But I may be wrong, the help file doesn't mention anything about it. The same applies to ATTO Disk Benchmark.

As to Sandra - this was the first time I used this tool so I don't really know how to interpret its results yet, indeed it reports ranges so the main value might actually be the maximum.

But what I think disproves your assumtions is the results from Disk Bench. This is the most real-life benchmarking tool as can be because it simply measures the time a file is read in the same way any application would - so it certainly doesn't report any peak values but actual average read speed of a chosen file. Here are the results of reading four 2GB files one after another, the test was done just after system start so no caches distorted the measurements:

Read File Bench started...

Read file: C:\temp\ssd.tc
Size: 2147483648 bytes
Time: 13766 ms
Transfer Rate: 148,772 MB/s

Read File Bench ended

Read File Bench started...

Read file: C:\temp\ssd-b.tc
Size: 2147483648 bytes
Time: 13828 ms
Transfer Rate: 148,105 MB/s

Read File Bench ended

Read File Bench started...

Read file: C:\temp\ssd-c.tc
Size: 2147483648 bytes
Time: 13875 ms
Transfer Rate: 147,604 MB/s

Read File Bench ended

Read File Bench started...

Read file: C:\temp\ssd-d.tc
Size: 2147483648 bytes
Time: 13891 ms
Transfer Rate: 147,434 MB/s

Read File Bench ended

As you can see I can consistently achieve 147-148 MB/s real transfer speeds so the readings of 125 MB/s from HD Tune can't be correct. I don't think it's possible that Disk Bench managed to get advantage of the few tiny spikes every time. And I'm sure that when using Disk Bench there is some small filesystem overhead so the raw throughput must be a little higher - something around 170 MB/s seems very likely to me.

Another thing is I have another drive connected to the same motherboard so if there was a driver or controller limit then it should be the same for both drives but it's not:

(the test was short-stroked to 200GB because this disk has a few bad sectors so the full benchmark won't run)

This all looks strange to me but what makes me think it's really HD Tune that is screwing the results is what Disk Bench reports. I think it's not possible that the OS could read a file faster than the actual throughput allows.

choucove · Jul 1, 2013

You are right that there is a pretty good discrepancy between the two reports given a single mechanical hard drive. The HDTune curve for the original hard drive looks about on par with what is expected (and what I personally have tested.) Those WD Black 500 GB hard drives can offer quite a bit of throughput for a mechanical drive. I've got six of those here at my office right now that I used for file transfers and test systems, and running HDTune on them even from the same computer can give me a variation of up to 20 MB/s of throughput. So perhaps this is an issue with HDTune, but all in all it is in line with the expected average of around 125 MB/s to 140 MB/s that these drives are known and rated for.

It is possible that the HDTune of course might be giving you slightly skewed results. They are close to the expected values, but if it is not getting the exact accurate picture you believe, then perhaps contacting their support would be the only way for sure to know. Still, you've run a whole host of different benchmarking tests here and they all fall within the upper expected values for the drive performance at around 140 MB/s so I would say that is going to be closer to your projected average drive throughput.

Lemon Juice · Jul 1, 2013

choucove :

But as far as I know a flat line like this throughout most of the disk space is not what is expected from a mechanical HD, normally there is a progressively declining curve. Here are other benchmarks of this drive I found on the internet and this is roughly what I would expect the curve to look like:

My motherboard is pretty basic so I don't expect such good results but a capped flat line at 125MB/s looks very suspicious to me, this is the only reason I tried so many other benchmarking tools and none of them reports as poor results as HD Tune.

choucove :

Yeah, I after a few days of owning the drive I can say I'm impressed with its performance. After installing Win7, installing many programs and defragmenting the drive the system boots in about 25 seconds to being fully responsive, the best result I timed was 22 seconds. No need to buy as SSD for a general purpose PC unless someone has money to spend. And I was pleasantly surprised that this drive is definitely faster than my 4-year old WD 250 GB RE3, which is supposed to also be a high performance drive. As we can see the technology gets improved in spite of the same rotational speed.

So the results from the tools other than HD Tune plus my subjective feeling of how this drive performs when I use my system tell me that (fortunately) HD Tune underestimates the drive.

Search

HD Tune gives me strange (wrong?) results on WD hard disks

Lemon Juice

Honorable

Lemon Juice

fzabkar

Judicious

Lemon Juice

Honorable

choucove

Distinguished

Lemon Juice

Honorable

choucove

Distinguished

Lemon Juice

Honorable

TRENDING THREADS

Latest posts

Moderators online

Share this page