One of the HDDs unexpectedly slower than others

[ibauer]

Hi,

I've just bought a new PC. There are 3 HDDs used separately (no RAID):

HDD#1: WD7501AALS-00E8B (750 GB)
HDD#2: WD1001FALS-55J7B (1 TB)
HDD#3: WD1001FALS-55J7B (1 TB)

My problem is that HDD#1 seems not to perform as fast as expected. Here are benchmark results from HD Tune Pro 4.00 trial (average transfer rate, access time):

HDD#1: 67.0 MB/s, 12.5 ms
HDD#2: 88.1 MB/s, 12.5 ms
HDD#3: 86.4 MB/s, 12.2 ms

According to the specs both of these HDDs should have the same transfer rate. Could anyone please advise why they haven't?

My PC spec:
CPU: Intel Core i5-750
RAM: 8 GB DDR3-1333
MOBO: Gigabyte GA-P55M-UD2 REV 1.0
Latest BIOS
Both Intel and Onboard SATA mode set to AHCI
Using Native SATA mode
OS: Windows 7 Pro x64
Clean Windows installation, no antivirus/antimalware or other software installed (only drivers + HD Tune Pro 4.00 trial)
HDD#1 has 2 partitions: system partition (100 MB) and the rest of disk space is taken by Windows
Latest drivers from Gigabyte site (Intel chipset driver, Intel SATA driver, GA SATA driver and others)
Writing to disk cache is enabled for all 3 HDDs in Device manager
Full scan using CHKDSK and HD Tune Pro revealed no disk errors for all 3 HDDs
HD Tune says the stabdard is ATA8-ACS - SATA II, UDMA Mode 6 (both supported and active) for all 3 HDDs
AAM is enabled and the value is 254 (0xFE) for all 3 HDDs

I have spent several days finding out the culprit, but no luck whartsoever. Please kindly advise. Thank you in advance.

 

[cjl]

That's not surprising at all. The WD 750 is slower than the 1000 because it has 250GB platters, while the 1000 has 333 GB platters. They absolutely do not have the same rated transfer rates. Everything you posted seems normal.

 

[geofelt]

The drives are surprisingly similar. Here is the spec sheet for the black series:
http://www.wdc.com/wdproducts/library/SpecSheet/ENG/2879-701276.pdf

Note that the maximum sustained data transfer rate is the same, 106 MB/s.

I think the difference can be explained by the partitioning of the drive. It is not clear which partition was tested. The fastest parts of a drive are on the outer cylinders where the first partition is placed.

 

[cjl]

Geofelt: they may have the same spec, but given the difference in platter size, I'd be surprised if they actually tested the same. I'd say the numbers in the OP are spot on.

 

[geofelt]

Geofelt: they may have the same spec, but given the difference in platter size, I'd be surprised if they actually tested the same. I'd say the numbers in the OP are spot on.

Where does one find the platter size and density numbers? I did not find them on the WD web site.

It seems to me that if the maximum sustained data transfer rate is the same, and the rotation rate is the same, then at the maximum point, the linear densities must be the same.

I agree that the OP has a normal situation.

 

[cjl]

I don't know where they're listed, but I know that the WD 500 and 750 models have 250GB per platter (except for some brand new 500GB blues),and the 640 and 1TB models have 333GB per platter (more accurately, 320GB per platter for the 640 model and 333 for the 1TB).

 

[JackNaylorPE]

It seems to me that if the maximum sustained data transfer rate is the same, and the rotation rate is the same, then at the maximum point, the linear densities must be the same.

CJ is correct......the issue is areal density .... with platter of identical size, at the same rpm.....on a 750 GB drive w/ 250 GB platters, data is passing the heads at a slower rate than a 320 GB or 500 GB platter.

 

[geofelt]

CJ is correct......the issue is areal density .... with platter of identical size, at the same rpm.....on a 750 GB drive w/ 250 GB platters, data is passing the heads at a slower rate than a 320 GB or 500 GB platter.

I think it is linear density which is a bit different. That is the number of bits on the track which pass by the read/write heads in one revolution. If the transfer rates are the same, then the amount of data on a track must be the same. If the tracks are spaced closer together, then the areal density can be greater without increasing the data transfer rate.

 

[cjl]

True, but typically, the linear density increases with higher areal density. This is reflected in both the benchmarks above and every other benchmark I've seen of these drives online as well.

 

[Pointertovoid]

And it tells as well that within one family, some disk sizes are slower, for the manufacturer uses a lower density. Though this doesn't always reflect in the specs!

Denser platters are also better for access time, as the same data (the OS, the application, the paging file...) fits on a smaller area where the arm moves more quickly.

 

[geofelt]

True, but typically, the linear density increases with higher areal density. This is reflected in both the benchmarks above and every other benchmark I've seen of these drives online as well.

I might have thought so also.

The 750 has 5 heads vs. 6 on the 1tb drive, indicating that some of the diference in capacity is from using 5 surfaces instead of 6. That does not fully explain the difference, only part of it.

I am always skeptical of synthetic benchmarks because they do not document what part of the drive they are testing, and it CAN make a difference.
SSD benchmarks are worse because they do not test the real world performance after the drive has been filled. If you can, it is much better to test your application with YOUR data on YOUR system. In the case of the OP, I suspect that he will not be able to tell the real world difference among the drives.

 

[ibauer]

Thanks a lot guys for all your replies, much appreciated. It looks like the figures I got from the HDD benchmark are normal and expected (due to a lower areal density of the platters). Now I'm wondering what drive to use as my system drive (OS, paging file and programs). The access times are about the same amongst the drives, but the transfer rate is different. Would it be better to use one of the 1 TB HDDs as the system drive? The other 1 TB drive would be used for my data and virtual machines (software development and testing) and the 750 GB drive as a backup disk. Does it make sense?

Currently the HDDs are connected to SATA ports as following:

Channel#0: HDD#1
Channel#1: HDD#2
Channel#2: HDD#3

Is it better to reconnect one of the 1 TB HDDs to channel 0 if it's going to be the system disk? Or does not it matter at all?

Thanks again.

 

[geofelt]

It does not matter which SATA port the drives are connected to.

Since each drive has plenty of capacity, it really does not matter which one contains the OS and applications. From a symmetry point of view, I would pick the 750 for the OS, even though the 1tb drives might be a touch better for performance for the OS.

For best performance, be on the lookout for a SSD with TRIM support for the OS. SSD prices and technology are rapidly changing

 

[ibauer]

Thanks, geofelt. I will reinstall Win 7 once more, using the 750 GB disk for OS, but this time I'm going to wipe out the 100 MB hidden system partition on the system drive and see whether or not it has any impact on the transfer rate (the other two 1TB drives had only one partition).

With regard to your suggestion about SSD disk - I know their transfer rates and especially access times are much better compared to common platter drives, but unfortunately they are beyond my budget.

 

[geofelt]

Thanks, geofelt. I will reinstall Win 7 once more, using the 750 GB disk for OS, but this time I'm going to wipe out the 100 MB hidden system partition on the system drive and see whether or not it has any impact on the transfer rate (the other two 1TB drives had only one partition).

With regard to your suggestion about SSD disk - I know their transfer rates and especially access times are much better compared to common platter drives, but unfortunately they are beyond my budget.

I had noticed the 100mb reserved space. After researching, It is used by windows-7 for recovery from boot problems and others, and I think you should not mess with it. 100mb is trivial to a 750gb hard drive, and I am not going to remove it from my puny 160gb SSD.

 

[ibauer]

Yes, you're right. I'll keep the reserved system partition where it is and better not to tinker with it at all. Thanks again.