John B. Smith wrote:
On Mon, 03 Sep 2018 16:13:21 +0100, Peter Johnson
wrote:

On Sun, 02 Sep 2018 10:51:53 -0500, Charlie Hoffpauir
wrote:

My home built is getting a bit old, but within the last weeks has
noticable slowed in many operations. For example, in Photoshop, using
the erasure tool; before, erasures were instanteous. Erased materal
instantly disappeared from the scene. Now, it sometimes disappears,
and sometimes the dissappearance lags the erasure in what seems an
erernity, but is probably a second or so. The same lag appears in
simple operations in other programs as well, Thumbs up, my program for
cataloging images, and Rootsmagic, my genealogy program.

Here's my system. I'm guessing it has to do with the Intel microcode
"fixes". If so, is there any way to improve performance? Perhaps a
more modern processor? If it's the microcode, can the fix be removed?

Windows 10 Pro ver 1803 Latest update KB4100347 included the Intel
microcode fixes (8/26/2018)


EVGA GeForce GTX 960 04G-P4-3962-KR 4GB SC GAMING (8/7/2015)

Intel Core i5-4690K Devil's Canyon Quad-Core 3.5 GHz LGA 1150 88W
BX80646I54690K Desktop Processor Intel HD Graphics 4600 (7/10/2015)

GIGABYTE GA-Z97X-UD3H-BK (rev. 1.0) LGA 1150 Intel Z97 HDMI SATA 6Gb/s
USB 3.0 ATX Intel Motherboard (7/10/2015)

G.SKILL Ripjaws X Series 16GB (2 x 8GB) 240-Pin DDR3 SDRAM DDR3 1600
(PC3 12800) Desktop Memory Model F3-1600C9D-16GXM (7/10/2015)

Intel 730 Series 2.5" 480GB SATA 6Gb/s MLC Internal Solid State Drive
(SSD) SSDSC2BP480G4R5 (12/3/2014)(The Intel SSD is a carryover from
the drive in my previous system... do SSD's slow down? Computer
Management says it's 446.668 GB with 288.85 GB free)

I also have 2 3 TB spinning drives for data and backups. I don't think
either one is a factor in this, as they've been in the system long
before this started.
You have made sure that the drives aren't full?

I guess this can't happen in Win10 (I'm in XP) but here's Paul's take
on mine:

"You're in PIO mode.

This shouldn't happen.

When there is a measurable error rate, the driver scheme
changes transfer rates, in an attempt to reduce the error
rate. But it doesn't take that many "gear down" attempts
by the driver, until it's in polled transfer mode,
a word is transferred at a time by the CPU. That
destroys transfer rate performance.

When you look in the appropriate dialog, you'll see
that DMA is no longer listed, and it's changed to PIO.
But I can tell just by your transfer rate, what just
happened. "4" is a popular number - that's what I'm using
as evidence."

You look for it in Device Manager IDE controllers. My HD Tune transfer
rate was down to 4.

The original behavior came from another OS. There is
some sort of "SCSI domain validation" process, where the
bus speed (SCSI cable) rate is reduced, until the CRC
errors stop.

Microsoft apparently thought this was wonderful, and added
the behavior to the IDE ribbon cable interface. The
rate would "gear down", until it dropped all the way
to PIO (polled input/output) rates of 4MB/sec to 5MB/sec
or so. In WinXP, there was a procedure for restoring the
speed. "Rediscovering" the hardware, can bring the
speed up to normal.

That sort of thing is likely still supported on SATA
(which is Serial ATA), but the thing is, the error
condition that causes gear down won't be happening,
and also, the "ATA transfer rate thing" has no meaning
in SATA. SATA simply transfers at the cable rate.
If the drive and mobo negotiate SATA II rates, then
data is transferred at that rate. If a utility reports
the ATA information for a SATA interface, in might say
"Ultra100" or "Ultra133" but these are meaningless to
the physical process of transferring SATA data. And thus,
if it dropped to PIO, you might never notice. It may not
be possible to actually pump a word at a time, across a
SATA cable. I don't know if the packet format supports
single word transfers like that or not. I don't think
I've ever seen SATA do such a thing (drop to 4MB/sec
as a domain validation response).

Even if SATA dropped to 4MB/sec, the packets themselves
still transfer at 3Gbit/sec or 6Gbit/sec, so "gearing down"
would be useless. Whereas on a ribbon cable, the actual
clock rate on the cable is dropping when these things
happen, which improves the odds of error free cable
transfers. Running at 4MB/sec, the clock for the ribbon
cable would be molasses-slow, leaving plenty of settling
time before each data transfer.

*******

If there are enough sector reallocations on a hard drive,
on a section of disk in current usage, you might notice
the read rate drop to speeds like that. SMART will warn
you about this condition. However, SMART works best if
the reallocations are spread evenly across the disk - if
that happens, the SMART warning appears around the same
time you notice the read performance is bad.

If, on the other hand, a massive number of sectors
are reallocated in a small strip of the disk drive,
then you suffer a performance problem, and SMART
reads "0" for reallocations. I had that happen here
on a Seagate 500GB drive. I ended up retiring the
drive, even though the SMART was "clean". An HDTune
benchmark run can show you the drive has problems
of that nature.

Paul