Robert in CA wrote:
I did a search and used the same title and found this:

https://www.abebooks.com/Windows-10-...30806756149/bd

but your correct about the versions and there's no rush.

Just for fun while were waiting for the keyboard I did a HD Tune scan
on both computers. The first is the 8500 which I did after it had sat idle
for about 30 minutes and nothing else was opened.

https://postimg.cc/bGxPRhL0

The second is the 780 which I did right after the 8500 but it had not sat idle
with nothing else opened.

https://postimg.cc/gw116Bv9

Interesting the difference in the scans with the 780 graph at the bottom
with spikes and the scattering of dots. The transfer rates are way off and so are
the burst rates yet the only difference between the two scans was the 30 minute
idle time.
So now I'm letting the 780 sit for 30 minutes with the HD Tune still
open (I would have to reopen it anyway) and here's the second scan on the
780 after 30 minutes:

https://postimg.cc/18kL8jGk

Now this scan is much more like the 8500. It's amazing the difference of letting it
sit for 30 minutes does. I looked it up an it says that the burst rate should be higher
than the max transfer rate and neither computer is. Is this of concern?


Robert

Burst rates are notoriously difficult to get right when
writing benchmarking software.

They could make a change to the way the OS works,
requiring you to make tweaks to the burst routine.

In theory, invalidating caches and warming-up caches,
should be all that is needed. For example, when memtest
measures the speed of the system, if you look at the
source code, you can see how they attempt to invalidate
caches that might throw off the results.

But in practice, some things are harder to do than
others, and getting the burst measurement right
is very hard (you don't want the OS System Read cache
answering the probe, and want only the cache RAM chip
on the disk controller doing it, which means you also
need a way of ensuring the "item" you want to read,
is already inside the cache RAM chip).

Under Windows 10, HDTune was returning slightly incorrect
results for the basic transfer curve, so even that can
be screwed up by time base problems or OS scheduler problems.
If I needed to measure a disk today, I'd try it on several
OSes, just to check the honesty of the lot.

Paul