two of my four WD2000 drives are slow

[dwater]

Hi,

I'm using 'Disk Utility' version 2.30.1 on Ubuntu 10.4.

I have 8 WD2000 disks; 4 are SATA and 4 are PATA.

When I use the utility to benchmark the drive performance, they all give a read performance of almost 60MB/s - apart from 2 of them, which are half that speed.

I have switched the drives with the other two PATA drives and the performance drop follows the drives, which seems to imply it's not the interfaces they're plugged into.

I noticed that the firmware was different for the two slow drives compared to the fast ones, but so is the model number.

SLOW (both are the same):
WDC WD2000JB-00EVA0 - firmware 15.05R51

FAST:
WDC WD2000JB-00GVA0 - firmware 08.02D80
WDC WD2000JB-00KFA0 - firmware 08.05J80

My naive idea was to attempt to flash the slow drives with the firmware from the fast drives, but I've no idea about the wisdom of such, nor how to actually do that.

I looked around on the forum, and learned some stuff, but nothing to guide me to do what I want.

Both the slow drives are 'spares' on a RAID5 array at the moment, so I can do what I like to them. The fast ones are active on the same array. I am planning to move all the data on the array onto a new array, after which I can do what I like with all of them and so I thought I might 'risk' playing with the firmware a bit.

Does anyone have any advice?

Max.

[fzabkar]

Could we see the performance graphs, both fast and slow?

[dwater]

Could we see the performance graphs, both fast and slow?

Sure :
slow
fast

[fzabkar]

The "slow" graph looks the right shape, ie it has a monotonically decreasing transfer rate as you traverse the surface from the outermost zone to the innermost one. The ratio of min/max transfer rates (0.6) looks about the same for all drives, suggesting that all drives have the same stroke. The graph confirms that the slow drives are not being throttled by the interface. If they were, then you would see a flat line.

I also don't believe the lower speed is due to a lower data density, ie more platters. AIUI, 7200 RPM models with 500GB platters have a maximum sustained read transfer rate of 130 MB/s.

By comparison, the data density for the fast WD2000JB models would be ...

(60 / 130)^2 x 500 = 107 GB per platter

This suggests that the fast drives have 2 platters and 4 heads.

If we apply the same logic to the slow drives, we get a data density of ...

(32 / 130)^2 x 500 = 30 GB per platter

This would suggest that the slow 200GB drives have 6 or 7 platters. This seems absurd.

[dwater]

OK, I understand what you wrote, but I wonder where this leaves me.

You mention 500GB - these are 200GB drives - does that make any difference to your analysis? I guess not.

So, yeah, wondering if your advice would be to flash the slower drive(s) to the firmware of one of the faster ones.

..or just not bother..

Thanks.

Max.

[drc]

flash the slower drive(s) to the firmware of one of the faster ones

No such thing.

[fzabkar]

OK, I understand what you wrote, but I wonder where this leaves me.

You mention 500GB - these are 200GB drives - does that make any difference to your analysis? I guess not.

I only mentioned a 500GB drive for comparison purposes.

You can estimate the data density of an unknown drive by comparing its maximum sustained data transfer rate against a known drive. I use a "standard" 500GB 7200RPM SATA drive as my "yardstick".

All other things being equal, the relationship I apply is ...

(data density A) / (data density B) = [(transfer rate A) / (transfer rate B)] ^2

As for what you can do, I really don't know. It would help if I could understand why you are getting the results that you are, but I haven't a clue. Are you testing your drives one at a time? Have you eliminated any possible RAID issue ???

The last time I saw anything like this was 20 years ago when you could low level format a drive and select different interleave factors to suit slower controllers. In your case the results could be explained by a 2:1 interleave, but of course this is absurd.

[Russwinters]

How are you testing these drives?

None of them are being used for the OS while your testing them? Correct?

The sudden down spikes on the slow drive are what is worrisome. Either there is some other operation polling the drive for data, causing the benchmark to read it as a downspike in speed, or there are areas of the disk that have sectors that are taking much longer to read then they should.


Also, in addition to the first question. How do you have all these disks connected to your system while testing? It's interesting the slow disk caps out at 31MB/s, even in the "good" areas.

Could be dying heads.

[Buzgar Andrei-Ioan]

where can i download WDC WD2000JB-00GVA0 - firmware 08.02D80 ?

[data-medics]

It really doesn't work like that. While there may be firmware updates for certain drives to fix bugs, you can't just decide to make a sideways firmware change. The reason some drives use a different firmware than others of the same model often is due to the fact that they used different parts internally. Even if you could force an "update" you'd almost certainly just brick the drive.

See if the manufacturer has an update utility for your drive and just use that. If you're still having issues then it's most likely either hardware issues or something else going on.