So the diagonal longer seek times in Samsung are caused by what? Switching heads? But what does it mean in this case? Shouldn't they do not read at all during switching, and then reading at full speed?

And why in 160 GB WD on the map are chequered patterns, while in bigger 640 GB ones are none?

From another forum:

""I wonder would the <500 ms dissapear if you would done erasing delays, as mentioned here:
http://hardforum.com/showpost.php?p=103 ... tcount=381"

I tried that on the Samsung HD204UI, but MHDD locked up each time. So I tried simply zeroing the drive (that got rid of the visible defects in a WD drive that had been dropped during operation and had hundreds of bad sectors), but I still got >150ms <500 ms sectors afterwards. The funny thing is, the slow sectors were almost never the same ones, and that was also true with the 2TB Hitachi 5K3000.

I don't understand how a sector can pass factory screening when it needs 13-18 attempts before it can be read. "
http://hardforum.com/showpost.php?p=103 ... tcount=393

Why would MHDD lock up when deleting delayes?
Why the slow sectors are almost never the same even in MHDD? If it was only interference from interface (cables and other hardware) then it would seem that the results would be similar regarding slow blocks for Samsung and WD, but it was not.
Those surface tests seem to be useless... particularly in the case of HDDScan, but MHDD results are not perfect either.

I tested Samsung SP1614N also with Victoria.
The results are not always the same even for bad blocks (even when the options in the software were not changed).
Setting timeout to 'never' did not change anything.
Running something in the background will corrupt the results - when I opened 3 MB pdf file at the same time when the scan was running it caused some extra dark grey and green blocks to appear.
So my conclusion is that such testing software, which has to be run in Windows will never give reliable results regarding surface map access times.
Hence, IMHO MHDD, which has to be run in DOS should give the truest results (if run from DOS and not from command line in Windows).

A scan with block size set to 1:

[...]
These are sequential: