Can a weak CMOS battery prevent detection of a drive

[Norm Why]

I read that weak CMOS battery can prevent detection of a drive. I have two
SSD drives. One an old Samsung 500GB boots reliably. Drive D: is a new
Seagate Barracuda, 500GB that is not detected reliably. I've done everything
conceivable with the cables. I've read bad reviews on the Barracuda.

BIOS program says CMOS battery is 3V whereas 3.3V might have been when it
was new.

Is it worth my time to buy a new CMOS battery?

Thanks

[Norm Why]

CMOS battery? Crazy idea, I know.

It took me all day to solve this problem. My Samsung SATA 2.0 boot drive
always work. The Barracuda SATA 3.0 SSD is finicky. I thought maybe a SATA
3.0 cable would work. Wrong. SATA 2.0 and SATA 3.0 are electrically
identical. The SATA 3.0 cable I tried had an elbow that made it
troublesome. SATA 2.0 cable is not.

I thought why is Barracuda SATA 3.0 SSD is finicky? Maybe bigshot does not
want to be slave. I rearranged the cables in the six available SATA ports.
Then I renamed drive E: to drive D: in Win10 and everything worked.

I ran CrystalDiskInfo and CrystalDiskMark and everything was fantastic.

Saved me a bundle of money not buying a new PC workstation. 12GB of RAM
works good.

[Paul]

While the cables may be the same for all the standards,
the hardware driving the cable may not like the signals
or signal levels coming from the drive. Digital signals
do have their analog aspects. Eye opening and so on.

Some of the VIA Southbridge ports had their problems, in
that they didn't negotiate rate properly. And then any
prospective drives had to use "Force 150" to get them
to work. VIA eventually figured this out, so there
*is* some VIA product that works just fine. But there
will also be just a few museum pieces out there with
that flaw. Mine appears to be OK, but I haven't extensively
tested with SATA III drives to see if they all negotiate
properly to SATA I with that motherboard. The motherboard
is "retired", but is a viable option if the thing I'm
typing on ever dies.

The CMOS battery need drop to 2.3V before it's no longer
compliant with what is expected of it. 3.0V is still fine.
The Southbridge RTC is generally rated 2.0V plus you add
0.3V for the drop across the BAT54 schottky in the path.
A battery creating 2.3V, after diode drop, delivers 2.0V
to the Southbridge. And the RTC is supposed to run at
that low level.

If you've been cloning drives, you should use a good
software for it. Macrium Reflect Free will change
the identifiers on the partitions when it clones,
such that if a drive and its clone are plugged into
the same computer, there is no confusion about
which partition is which. Using "dd" isn't quite
the same, and requires manual intervention to prevent
one drive from entering the "Offline" state. Use
diskmgmt.msc (Disk Management) if a drive "disappears"
to see if its row is still present in Disk Management,
but the left-hand square is labeled "Offline". A number
of disk identifiers are allowed to be the same, but
there's at least one disk identifier that the OS
won't allow to be the same, and then the second
drive to be probed is put "Offline" for safety.

When Macrium clones, it also only transfers the occupied
clusters, which is like a "free TRIM" in a sense. "dd"
transfers from SSD to SSD would transfer all blocks,
which burns up a lot of the free pool on the destination,
and can benefit from issuing a TRIM per partition
in the Optimize panel later.

Paul

[David W. Hodgins]

On Fri, 16 Oct 2020 23:07:55 -0400, Paul <nos...@needed.invalid> wrote:
> When Macrium clones, it also only transfers the occupied
> clusters, which is like a "free TRIM" in a sense. "dd"
> transfers from SSD to SSD would transfer all blocks,
> which burns up a lot of the free pool on the destination,
> and can benefit from issuing a TRIM per partition
> in the Optimize panel later.

I manually create and format the new partitions, and then use rsync to transfer
the files. That's also how I do my primary backup of data. In run level 1, with
any remaining user processes killed I run ...
time nice -n 19 ionice -n 7 rsync -auvxSHXAP --specials --sparse --delete --exclude="lost+found" /home/dave/ /aback/home/dave/

Regards, Dave Hodgins

--
Change dwho...@nomail.afraid.org to davidw...@teksavvy.com for
email replies.

[Paul]

But being sans-automation, your UUIDs or BLKIDs aren't fit for
purpose until you put things back together again. At least
for boot materials, this is the case.

Macrium automates the treatment of Windows and its boot materials.
It gets most of the details right. It's a prototype of
what automation should be, because ordinary users don't
have to drop to command line afterwards.

Macrium can copy EXT4 as well, which is a feature I use. But
sadly, when cloning, it doesn't have the logic to properly
handle /etc/fstab. It doesn't tidy up.

Macrium reaches into the BCD file and edits it, which is
why cloned Windows drives behave themselves. But Macrium
lacks the feature of doing the same for Linux.

I'm still waiting to find even a commercial tool that
does this for the Linux ecosystem properly. The only reason
Macrium has Windows support, is there is at least one
Windows-provided utility that is capable of rebuilding
the BCD. But it doesn't work well enough in practice
to be useful. So the Macrium designers added their
own code to do this. And it isn't always 100%
successful. An attempt to have Win2K added to the
Win7 boot menu, there were some issues with what
it did. Still not perfect, but for a lot of the
users, they might never notice the rough
edges (because who would be running Win2K except
the guy I was helping the other day...).

There are a lot of rocky shoals in the Clone Sea,
and lots of places to run aground. There's still lots
of room for improvements.

Paul

[David W. Hodgins]

On Sat, 17 Oct 2020 01:30:14 -0400, Paul <nos...@needed.invalid> wrote:
> But being sans-automation, your UUIDs or BLKIDs aren't fit for
> purpose until you put things back together again. At least
> for boot materials, this is the case.

I'm fine with manually editing fstab, in which I use labels instead of uuid ...
$ grep x7 /etc/fstab
LABEL=x7b / ext4 defaults,noatime 1 1
LABEL=x7bboot /boot ext4 defaults,noatime 1 2
manually changing /boot/grub2/install.sh and if needed, /boot/grub/device.map,
and then using chroot or systemd-nspawn to update the grub2 boot loader,
either from another install on the same system or from a live iso.

I've done it enough times, it's pretty easy to remember the steps needed, or
debug and fix if I've missed one.

[philo]

Glad you got it figured out.

To answer your question though, a 3.0 volt CMOS battery is fine.

If the battery was too low, you'd simply lose your settings.

I did *once* have a battery around 2.6 volts that would not allow the
machine to post.

[Norm Why]

Thanks for your reply.

Unfortunately, no matter what fiddling I could do, the Seagate Barracuda
continued to show unreliable symptoms. Read/write errors and failure to
detect at boot up. I ran Seagate tools that said the drive was slow, at one
time. I read a bad review of Barracuda. We decided it's a bad drive. I have
a one year warranty. I bought a 3 TB WD HDD and installed it. I am now doing
a backup. Then I will return the drive for a replacement Barracuda SSD.

SSD drives are very nice for instantaneous response, when they work.

I plan to run some large programs for medical imaging. When it comes to
medical diagnostic imaging, one needs to do it oneself.

[philo]

And you'll need a totally reliable drive for that.


Speaking of medical imaging, when I was at my dentist....when they tried
to take an x-ray, their computer crashed. After a few attempts, I told
them to try a different USB port.


All worked fine.

I should have had them deduct my technical services from my bill.

[Norm Why]

>> High

>>
>> SSD drives are very nice for instantaneous response, when they work.
>>
>> I plan to run some large programs for medical imaging. When it comes to
>> medical diagnostic imaging, one needs to do it oneself.
>>
>>
>
>
>
> And you'll need a totally reliable drive for that.
>
>
> Speaking of medical imaging, when I was at my dentist....when they tried
> to take an x-ray, their computer crashed. After a few attempts, I told
> them to try a different USB port.
>
>
> All worked fine.
>
> I should have had them deduct my technical services from my bill.

More than that Philo,

Further back in this thread I thought I said I replaced the Seagate
Barracuda with a good one. I also bought a 3TB WD HDD for backup. Prior to
swapping out the bad Barracuda, I made a backup to the 3TB WD HDD. When I
tried to restore, I ran into CPU overheating problems. Not every thermal
grease is OK. Second try was "Kryonaut Ultra High Performance Thermal
Grease" from thermal grizzly. Overheating problem was solved and restore is
working.

The CPU I have may be heat damaged. I checked, the best CPU for this
Gigabyte main board is Intel Q9650. I had an Intel Q9650 CPU but it burned
out. Temperature reached 126C. then screen went blank. I ordered another
Q9650 CPU from a reputable supplier. I'll start again.

Heat death of a CPU is slow. During slow death, Win10 is a bugger. If you
see RunTimeBroker in Task Manager, you know you're in trouble. While it
looks like a software problem it is more likely to be a cooling problem. I
have a gaming case. I could install two more fans. I ordered 4 RAM heat
spreaders. I'll do a before and after check with my IR Laser temperature
gun.

[Paul]

The screen went blank ?

The CPU has THERMTRIP. If the temperature goes high enough (implying
the heatsink fell off the CPU), the CPU sends THERMTRIP to the motherboard
and THERMTRIP causes PS_ON# to be deasserted. That shuts off the power
and removes +12V running VCore. The fans should go off.

The CPU protects itself.

I have an E8400, which is 1/2 of a 9650. It draws 11W at idle and
43W flat out. Your processor would be double that amount.

Maybe your VCore is a bit too high ? The joint between CPU and heatsink
would have to be absolutely dry, for it to get that hot. Even with
a tiny heatsink with fan running, it should do better than that on
cooling. For one thing, the CPU uses throttling to try to avoid
a temperature that high. And that's one reason why it shouldn't
be hitting 126C. It has one defense mechanism, that reduces CPU
capability as a tradeoff for temp. To hit 126C, you really need an
air gap between CPU and heatsink. Like if the push pins aren't fastened
or something.

Paul

[Norm Why]

[snippage]

"Why look a gift horse in the mouth?" Currently a Quad Q8400 is plugged into
Gigabyte MOBO. For convenience I connect using RDP. CPU use is 3%.
Temperatures are 40C., 36C., 32C. and 36C. I think core#1 shows heat damage.
Electromigration occurs when things get too hot.

[Norm Why]

>>"Why look a gift horse in the mouth?" Currently a Quad Q8400 is plugged
>>into
>>Gigabyte MOBO. For convenience I connect using RDP. CPU use is 3%.
>>Temperatures are 40C., 36C., 32C. and 36C. I think core#1 shows heat
>>damage.
>>Electromigration occurs when things get too hot.
>>
>

> You just have to purchase ssd's from a good manufacture !
>
>
> KenW

The Intel 9650 CPU I ordered was delivered and with Halnziye: Silver Thermal
Paste. Let's compare specs:

Halnziye: HY710 10g Silver Thermal Paste.

Halnziye HY710 10g Silver Thermal Paste has decent level of thermal
conductivity at 3.17W/mk. It's capable of providing long-term stability for
your machine, remaining operational through temperatures between 22 and 464
degrees Fahrenheit.

Thermal Grizzly Kryonaut. specs

Thermal Conductivity 12,5 W/mk
Viscosity 120-170 Pas
Specific Weight 3.7g/cm3
Temperature -200 °C / +350 °C
Capacity 1.5ml / 5,55 g 3ml / 11.1g

The units are Watts per meter per degree Kevin. Using a thin layer reduces
"meters". Thermal Grizzly Kryonaut comes with a spatula with which one can
apply a thin layer. Halnziye HY710 has thermal conductivity at 3.17W/mk.

Thermal Grizzly Kryonaut has Thermal Conductivity 12,5 W/mk

12.5/3.17 = 3.94 or almost 4 times the thermal conductivity, Ceteris
paribus, "holding other things constant." This is a very important
scientific concept for you who have studied multivariable calculus.

The remedy for an overheating CPU is to use Thermal Grizzly Kryonaut thermal
grease and to use as thin a layer as possible.

The Intel 9650 CPU came from a reputable company who sold me 4x 2GB DDR2
RAM. I only needed 2 modules. When I first plugged in the two RAM modules
(now for a total of 12GB), it created a heat crisis for the old Q9650 that
crashed and eventually burnt up. If I only knew what I know now at first.

Now I will have a Q9650 in a GIGABYTE main board, 6 SSD ports, 2x 500GB SSD
drives and one 3TB WD HDD. Also a bunch of other stuff that is not part of
this discussion.

Why such a proliferation of inadequate thermal grease products? Gresham's
law is a monetary principle stating that "bad money drives out good".
Likewise, so many people settle for bad grease. I've spent a lot of money on
bad grease.

[Norm Why]

The good grease worked for me. On the new Q9650 all the cores are below 40C.
It can take two weeks for grease to 'set'. That means the idle state gets
cooler and cooler.

[Norm Why]

[snippage]

>>
>> Why such a proliferation of inadequate thermal grease products? Gresham's
>> law is a monetary principle stating that "bad money drives out good".
>> Likewise, so many people settle for bad grease. I've spent a lot of money
>> on bad grease.
>>
>
> The good grease worked for me. On the new Q9650 all the cores are below
> 40C. It can take two weeks for grease to 'set'. That means the idle state
> gets cooler and cooler.

Typical low CPU load core temperatures of the Q9650 are:
31C., 31C., 28C. and 32C.
Thermal Grizzly Kryonaut thermal grease works great! However, the MOBO
manual suggests that optional fans be installed to avoid overheating. The
manual can be found by Google'n 'Gigabyte GA-EP45-DS3L .PDF'

I have a Cougar MX330-X gaming case whose manual can be found here:
https://cougargaming.com/fileadmin/downloads/USERS_MANUAL/MX330_X_usermanual.pdf

Note that 5 optional fans can be installed. CPU and Fan N are installed and
their 4-pin fan headers are occupied. According to the MOBO manual there are
2 3-pin fan headers up for grabs.

I am confused. How do I get support for maximum number of fans. Where to
buy?

Thanks

According to the BIOS, installed fans can be set to sound an alarm if fans
fail or CPU exceeds 80C. However, the beep speaker has been silent for a
long time. There is an little known beep code. The absence of a beep
confirmation at boot time indicates something is wrong. Maybe what is wrong
is the beep speaker?

[Norm Why]

[spippage]

> Typical low CPU load core temperatures of the Q9650 are:
> 31C., 31C., 28C. and 32C.
> Thermal Grizzly Kryonaut thermal grease works great! However, the MOBO
> manual suggests that optional fans be installed to avoid overheating. The
> manual can be found by Google'n 'Gigabyte GA-EP45-DS3L .PDF'
>
> I have a Cougar MX330-X gaming case whose manual can be found here:
> https://cougargaming.com/fileadmin/downloads/USERS_MANUAL/MX330_X_usermanual.pdf
>
> Note that 5 optional fans can be installed. CPU and Fan N are installed
> and their 4-pin fan headers are occupied. According to the MOBO manual
> there are 2 3-pin fan headers up for grabs.

[snip]

> According to the BIOS, installed fans can be set to sound an alarm if fans
> fail or CPU exceeds 80C. However, the beep speaker has been silent for a
> long time. There is an little known beep code. The absence of a beep
> confirmation at boot time indicates something is wrong. Maybe what is
> wrong is the beep speaker?
>

I should also mention that on one side of the MX330-X case are cutouts for
switches labeled 'FAN1' and "FAN2'. The 2 3-pin fan headers on the MOBO seem
to report to the BIOS. Switched fan power would not so report. The question
is which fans should be connected to the 2 3-pin fan headers and which
should be power switched? And where to get the power?

Thanks.

[Norm Why]

>>"Why look a gift horse in the mouth?" Currently a Quad Q8400 is plugged
>>into
>>Gigabyte MOBO. For convenience I connect using RDP. CPU use is 3%.
>>Temperatures are 40C., 36C., 32C. and 36C. I think core#1 shows heat
>>damage.
>>Electromigration occurs when things get too hot.
>>
>

> You just have to purchase ssd's from a good manufacture !
>
>
> KenW

I discovered that it is advised to turn off Indexing Services on an SSD.
Windows does not do this automatically. Indexing imposes to much I/O on an
SSD.

[Paul]

A motherboard can have fan headers that only provide power (+12V and GND).
Check your manual for details like that.

A motherboard will have at least three channels with RPM counters.
(+12V, GND, RPM). The RPM signal is 2ppr, two pulses per revolution.

THe most exotic headers, are ones with speed control. In the old days,
the three pin connector could have voltage control of the +12V pin.
There would be a 100uF cap next to the fan header plus a transistor.

On modern motherboards, the header is a four pin, and has +12V GND RPM PWM.
The PWM is a pulse width modulation control that sets the fan speed.
The RPM has the 2ppr thing. The +12V in this case stays at +12V.

You save the fan headers with working RPM pin, for the fans that have
an RPM signal :-)

The cheapest case cooling fans, have no RPM signal. They just have
+12V and GND. You connect them to unmonitored headers, or get
a Molex to fan adapter and power the fan that way,

The next level of fan has an RPM signal. You can plug those to
a three or four pin header equipped with RPM.

The CPU fan header pretty well always has RPM. The CPU cooler
has the third wire with the RPM signal on it. The more modern
CPU fans have all four wires +12V GND RPM PWM, even if you're
not using the PWM and have decided to run the fan at full speed.

Based on the above, you have a pool of electromechanical devices,
you have your various fan headers, you have molex connectors, now
you head off and do your planning.

Nothing says you need seven fans, to have a "lucky" computer.

My newest machine, has one fan on the back (to establish a
low delta_T between inside case air temp and external ambient).
It has a second fan positioned as a blower over VCore, as
VCore doesn't have proper cooling and runs at 65C surface temp.

The CPU cooler is huge, and it has a fan with both
RPM and PWM.

And that's it.

If the case air is getting too hot (warm cloud around
CPU cooler), then I use the 37.5mm thick fan and put it
in the back of the computer for exhaust.

Generally you aim for a "direction" of airflow. Like input
on the front, exhaust on the back. And arrange the fans
so they're not working against one another. For example,
it's pretty hard to install a "side" fan in the door, and
arrange it so it has a deterministic airflow. I did do a
seven fan setup once, and one of the fans, if you held
a tissue next to the fan, there was zero airflow, which
tells you the fans were working against one another.

Paul

[wasbit]

"Paul" <nos...@needed.invalid> wrote in message
news:rol56f$jb$1...@dont-email.me...

After we have meticulously set up the air flow, you can get smoke pellets &
matches for testing flues, we then go & stick the PC back in its cubby hole
under the desk & undo all the good work.
Cynic.! Moi? Never. :)

--
Regards
wasbit

[Norm Why]

[snip]

Thanks Paul,
See also:
Helpful Guide to the Best PWM Fan Hub Controller
https://nerdtechy.com/best-pwm-fan-controller

[Norm Why]

>> You just have to purchase ssd's from a good manufacture !
>>
>>
>> KenW
>
> I discovered that it is advised to turn off Indexing Services on an SSD.
> Windows does not do this automatically. Indexing imposes to much I/O on an
> SSD.

Also, I found if I put a HDD delay in the BIOS, the SDD is more likely to be
detected. Now it is one second, next I will try 2 seconds. Barracuda is 100%
busy after boot. Maybe not fully detected.

[Norm Why]

> You just have to purchase ssd's from a good manufacture !
>
>
> KenW

Yes, after great annoyance I think you are right.