I've run a bunch of preclears in the year or two I've been doing unRaid (love this product!).  I've watched pre-clear run and wondered about the results.  Take this latest result:

 

http://my.jetscreenshot.com/12412/20160806-j2z3-77kb.jpg[/img]

 

Pre-read @ 93MB/s

Zero-time @120MB/s

Post-read @ 62MB/s

 

Performance on the drive is excellent

http://my.jetscreenshot.com/12412/20160806-lrbf-22kb.jpg[/img]

 

I'm using the SUPERMICRO AOC-SAS2LP-MV8 PCI-Express x8 Low Profile SAS RAID Controller (which also has excellent performance (300MB/s)) in the PCI-E X16 slot on my ASUS P8Z77-V LK.  So that should be as good as it gets.

 

Looking at my system I have the following:

http://my.jetscreenshot.com/12412/20160806-lfz0-112kb.jpg[/img]

 

So the actual Pre-Clear numbers I'm getting seem counter-intuitive.  I would think that reads, generally would be faster than writes.  Ergo the Pre-read and the Post-read should be faster than the Zero-time?  What am I not understanding?  I get that the bottleneck for the entire system is probably the drives anyway, but thought I'd ask...

 

A side question is should one pay attention to which drives are on which controllers?  i.e. separate faster drives on the fastest ports?

 

 

 

The lower capacity disks are slower because they have lower density platters.

 

Why do you think writes ought to be slower than reads? How are you measuring them?

 

I see you're using Joe's original pre-clear script. It's post-read is slow because it doesn't read the disk sequentially - it deliberately seeks to a "distant" track between reads.

 

Your parity disk doesn't have a file system.

I'm guessing the 3rd screenshot is a spreadsheet you entered manually. As John_M noted, parity has no filesystem. Also, the sdX is not fixed and is assigned at boot time. It is not necessarily consistent so probably not useful to put in your spreadsheet.

 

So the actual Pre-Clear numbers I'm getting seem counter-intuitive.  I would think that reads, generally would be faster than writes.  Ergo the Pre-read and the Post-read should be faster than the Zero-time?  What am I not understanding?  I get that the bottleneck for the entire system is probably the drives anyway, but thought I'd ask...

 

 

The preclear script not only reads your disk drive, it also stresses it. The reads are done in chunks that are read twice and, in case of post-read, it also sums that chunk and verify if its zeroed. So it's normal it be slower than the write, which is done in a single operation.