What's the difference between these two memories ?

[Shadow]

https://www.kabum.com.br/produto/50758/memoria-kingston-hyperx-fury-4gb-1866mhz-ddr3-cl10-red-hx318c10fr-4
https://www.kabum.com.br/produto/50757/memoria-kingston-hyperx-fury-4gb-1866mhz-ddr3-cl10-black-hx318c10fb-4

Other than the price ?

I downloaded the Kingston specs

https://cdn.cnetcontent.com/07/e9/07e999d3-b9b5-4972-a145-bff10cf84f67.pdf

Which says:

Latency CL9-11
Voltage 1.35V, 1.5V

It does not help at all in knowing if I can add a stick or
two.

My current memory is KHX1866C10D3/4G
Number of banks 8
Nominal Voltage 1.50 Volts
(CPU-Z output)
TIA
[]'s
--
Don't be evil - Google 2004
We have a new policy - Google 2012

[Paul]

Shadow wrote:
> https://www.kabum.com.br/produto/50758/memoria-kingston-hyperx-fury-4gb-1866mhz-ddr3-cl10-red-hx318c10fr-4
> https://www.kabum.com.br/produto/50757/memoria-kingston-hyperx-fury-4gb-1866mhz-ddr3-cl10-black-hx318c10fb-4
>
> Other than the price ?
>
> I downloaded the Kingston specs
>
> https://cdn.cnetcontent.com/07/e9/07e999d3-b9b5-4972-a145-bff10cf84f67.pdf
>
> Which says:
>
> Latency CL9-11
> Voltage 1.35V, 1.5V
>
> It does not help at all in knowing if I can add a stick or
> two.
>
> My current memory is KHX1866C10D3/4G
> Number of banks 8
> Nominal Voltage 1.50 Volts
> (CPU-Z output)
> TIA
> []'s

Kingston has specs right on their own web site.

50758 HX318C10FR/4 CAS10 1866 1.35V RS220
50759 HX318C10FB/4 CAS10 1866 1.5V RS247

The available information only lists 1.5V products.

https://www.kingston.com/dataSheets/HX318C10FR_4.pdf

The article here suggests the heat spreader color
is for "taste" in PC construction. Four colors are
available in all. The heat spreader color is not intended
as a voltage indicator.

https://www.kingston.com/en/company/press/article/7141

If your retailer is to be believed, then I would
select the 1.5V product to match the 1.5V product I
already have.

Someone had a problem with this very thing in a post
about two days ago. System had two 1.5V DIMMs, poster
added two 1.35V DIMMs. The 1.35V DIMMs (DDR3L) are supposed to
work in 1.5V system, and that's why the keying on the
edgecard of the DIMM allows insertion. However, the
system of the poster would not work.

CH0 CH1 Note: all DIMMs are 4GB

1.5V ---- 1.5V ---- (Dual channel mode)
Failed to POST...
1.35V ---- 1.35V ----

For reasons known only to the BIOS logic, this worked.

CH0 CH1

1.5V ---- 1.35V ---- (Dual channel mode)
Works...
1.5V ---- 1.35V ----

Both of these setups run *all* slots at 1.5V.
The sticks are never "treated as individuals".

The way RAM works, is the environment of all DIMMs
is the same. They receive the Column Address Strobe
(CAS) on the same cycle. They receive the same clock
frequency. They receive the same voltage for their
power source. Since the DIMMs are on shared buses
and the memory controller for both channels comes
from the same chip, there is an incentive to make
all operating conditions identical.

Why inserting the DIMMs as in the second diagram
works, is a mystery. Since the DIMMs are all 4GB
ones of equivalent construction, it allows some
flexibility in mixing. The dual channel works in
both situations. The system is still in dual channel
mode after the change.

A question you would have to ask, is whether the
1.5V RAM you "boost" to 1.65V, whether a 1.35V
RAM can also be boosted to 1.65V. If you do
select the 1.35V RAM, I would do so only for
a system which does not have a high boost for
VDimm. As I don't know the extent of the "tolerance"
on the 1.35V DDR3L.

The 1.35V stuff is readily available, and Kingston
has resorted to "cherry picking" on the DRAM market.
While the spec sheet says 1.5V, Kingston refuses
to make a separate SKU for 1.35V in this case. Kingston
did something similar a few years back, mixing
high density and low density DIMMs under the same
SKU, in violation of their datasheet. (The datasheet
might show double-sided DIMMs, but the blister packs
of memory had single-sided DIMMs in some packages
and double-sided DIMMs in other packages. This
caused *grief* for people seeking low density RAM,
which is what the product was intended for.)

As far as I know, if 1.35V RAM is on a DIMM, the SPD
EEPROM must have an encoding for it. Consequently,
if those two products are for real on your retailer
site, they likely do not have identical SPD contents
when you examine them in CPUZ SPD table. This could
be what is causing the BIOS to fail to set up
some configurations of the mixed RAM. The BIOS was
not programmed to accept "variation" in that field
of the SPD EEPROM.

The general rule of thumb for memory is, to mix
"like with like". Don't mix server ram with
"enthusiast overclocker RAM" as the latter needs
too much voltage for stability. If you start with
server ram at stock voltage, your upgrade to
the system should also be server class RAM
(which runs at JEDEC voltage, not boosted voltage).

If you need to save a few bucks by buying the 1.35V
RAM, you can. But be prepared for the system to black
screen and not POST properly. You may have to send the
RAM back, if your BIOS does not like it. There is no
way to know in advance, whether the BIOS is ready for it.

It's also possible, if the motherboard has a BIOS upgrade,
that the BIOS could be modified to "ignore" the
voltage field. And then any configuration of DIMMs
would work with less trouble than the above example.

HTH,
Paul

[Shadow]

On Tue, 11 Dec 2018 09:43:56 -0500, Paul <nos...@needed.invalid>
wrote:

>Shadow wrote:
>> https://www.kabum.com.br/produto/50758/memoria-kingston-hyperx-fury-4gb-1866mhz-ddr3-cl10-red-hx318c10fr-4
>> https://www.kabum.com.br/produto/50757/memoria-kingston-hyperx-fury-4gb-1866mhz-ddr3-cl10-black-hx318c10fb-4
>>
>> Other than the price ?
>>
>> I downloaded the Kingston specs
>>
>> https://cdn.cnetcontent.com/07/e9/07e999d3-b9b5-4972-a145-bff10cf84f67.pdf
>>
>> Which says:
>>
>> Latency CL9-11
>> Voltage 1.35V, 1.5V
>>
>> It does not help at all in knowing if I can add a stick or
>> two.
>>
>> My current memory is KHX1866C10D3/4G
>> Number of banks 8
>> Nominal Voltage 1.50 Volts
>> (CPU-Z output)
>> TIA
>> []'s
>
>Kingston has specs right on their own web site.
>
>50758 HX318C10FR/4 CAS10 1866 1.35V RS220
>50759 HX318C10FB/4 CAS10 1866 1.5V RS247
>
>The available information only lists 1.5V products.
>
>https://www.kingston.com/dataSheets/HX318C10FR_4.pdf

Exactly the same as
https://www.kingston.com/dataSheets/HX318C10FB_4.pdf

>
>The article here suggests the heat spreader color
>is for "taste" in PC construction. Four colors are
>available in all. The heat spreader color is not intended
>as a voltage indicator.

--- slight cut

>If you need to save a few bucks by buying the 1.35V
>RAM, you can. But be prepared for the system to black
>screen and not POST properly. You may have to send the
>RAM back, if your BIOS does not like it. There is no
>way to know in advance, whether the BIOS is ready for it.

If it's not 1.5V I can send it back, since that's what they
specify on the page. That's the law out here.

Características:
- Marca: HyperX
- Modelo: HX318C10FR/4

Especificações:
- Tipo: 240-Pin DDR3 SDRAM
- Capacidade: 4GB
- Aceleração: DDR3 1866
- Cas Latência: 10
- Tensão: 1.5V <---------

>
>It's also possible, if the motherboard has a BIOS upgrade,
>that the BIOS could be modified to "ignore" the
>voltage field. And then any configuration of DIMMs
>would work with less trouble than the above example.
>
>HTH,
> Paul

It does. TY for the links and info.
I'll go for the cheaper offer. And let you know how it worked
out.

[Flasherly]

On Tue, 11 Dec 2018 14:45:24 -0200, Shadow <S...@dow.br> wrote:

>If it's not 1.5V I can send it back, since that's what they
>specify on the page. That's the law out here.

Was a long time ago, but last time I called Kingston I couldn't
believe the red-carpet commitment to satisfied customers. I'd a
question regarding a USB flashstick, back when something like a 2 or
4G was a big deal.

[Shadow]

On Tue, 11 Dec 2018 13:19:28 -0500, Flasherly <Flas...@live.com>
wrote:

I have something like 3 days to send it back to the shop, for
a full refund. More than that, and it's under Kingston's possibly
non-existent warranty.

[Flasherly]

On Tue, 11 Dec 2018 16:47:35 -0200, Shadow <S...@dow.br> wrote:

>I have something like 3 days to send it back to the shop, for
>a full refund. More than that, and it's under Kingston's possibly
>non-existent warranty.

Took me longer than 3 days to call for an explanation of why
physically formatted sectors might not correspond to a limit of unique
file entries I'd falsely presumed FAT capable of storing.

Initiate a live chat after preparing those contingencies for possible
recourse should any then arise.

https://www.kingston.com/en/support

[Shadow]

On Tue, 11 Dec 2018 15:05:17 -0500, Flasherly <Flas...@live.com>
wrote:

Nah, I'll just install the modules, run Memtest86 for 6 hours
or so, and if it OK's them, keep them. Otherwise, return to the shop.
They pay for the postage.

[Flasherly]

On Tue, 11 Dec 2018 20:48:45 -0200, Shadow <S...@dow.br> wrote:

>Nah, I'll just install the modules, run Memtest86 for 6 hours
>or so, and if it OK's them, keep them. Otherwise, return to the shop.
>They pay for the postage.

Memory is memory.

[Shadow]

On Tue, 11 Dec 2018 23:04:44 -0500, Flasherly <Flas...@live.com>
wrote:

I'll try to remember that.
;)

[Flasherly]

On Wed, 12 Dec 2018 07:35:58 -0200, Shadow <S...@dow.br> wrote:

>I'll try to remember that.

I had 9 chips, 8 banks or rows, of 9 chips to a row.

The power would surge: That's closest to a "crash" anybody's going to
get, other than a maybe a fly walking into a mess of vacuum tube.

Then I'd have to find the chips affected, maybe one, two, sometimes
three - each individual DRAM chips with aluminum legs, pulling them
out, a row at a time, until a reboot came up to count what was left.

Then each vacant row needed to be reseated. The bad RAM was in it if
the BIOS indicated it failed a memory count;- each chip then was
substituted with a good one, one by one, until it counted.

I kept a plastic bag of such, good spare RAM chips.

The maddening part: later on down the road, I could run into another
"crash, then could pull, from that bag of chips from prior crashes, a
substitute that works.

That was a single 1 Megabyte computer with 8 banks of sockets for RAM.

The cost of the 1 Megabyte RAM: $300US dollars.

Next computer was slotted RAM: the first chips slotted on a board to
insert into a female slot.

Never had any trouble with RAM since.

Happy testing.

[Shadow]

On Tue, 11 Dec 2018 10:23:14 -0200, Shadow <S...@dow.br> wrote:

>https://www.kabum.com.br/produto/50758/memoria-kingston-hyperx-fury-4gb-1866mhz-ddr3-cl10-red-hx318c10fr-4
>https://www.kabum.com.br/produto/50757/memoria-kingston-hyperx-fury-4gb-1866mhz-ddr3-cl10-black-hx318c10fb-4
>
> Other than the price ?
>
>I downloaded the Kingston specs
>
>https://cdn.cnetcontent.com/07/e9/07e999d3-b9b5-4972-a145-bff10cf84f67.pdf
>
>Which says:
>
>Latency CL9-11
>Voltage 1.35V, 1.5V
>
> It does not help at all in knowing if I can add a stick or
>two.
>
>My current memory is KHX1866C10D3/4G
>Number of banks 8
>Nominal Voltage 1.50 Volts
>(CPU-Z output)
> TIA
> []'s

To Paul and others that helped, TY. Installed. According to
CPU-Z hey are exactly the same model number and voltage as the "black"
model, so it appears the "heat dissipaters" or whatever are painted
different colors, but the memories are the same. Ran two rounds of
Memtest86+ and no errors.

The only strange thing that happened is my clock sometimes
loses or gains 20-30 seconds in a day. (I use Neutron from Keir.net to
synchronize on Startup). And that never happened before I installed
the memories.The error was always a second or less.
Maybe I twisted something on the MB when I pressed the
memories in place. Weird.

[Paul]

That is 230ppm at least. That's a little high.

The computer has two time pieces. Windows time. BIOS time.

Windows runs a software clock. It depends on BCLK for traceability.

THe BIOS runs the RTC clock, which depends on the 32768Hz motherboard crystal.

The software clock can only lose time (by "missing" clock tick interrupts).

Windows time could gain or lose, based on BCLK being off.

BIOS time could gain or lose, based on 32768Hz clock.

Paul

[Shadow]

On Sun, 23 Dec 2018 19:18:59 -0500, Paul <nos...@needed.invalid>
wrote:

>> The only strange thing that happened is my clock sometimes
>> loses or gains 20-30 seconds in a day. (I use Neutron from Keir.net to
>> synchronize on Startup). And that never happened before I installed
>> the memories.The error was always a second or less.
>> Maybe I twisted something on the MB when I pressed the
>> memories in place. Weird.
>> []'s
>
>That is 230ppm at least. That's a little high.
>
>The computer has two time pieces. Windows time. BIOS time.
>
>Windows runs a software clock. It depends on BCLK for traceability.
>
>THe BIOS runs the RTC clock, which depends on the 32768Hz motherboard crystal.
>
>The software clock can only lose time (by "missing" clock tick interrupts).
>
>Windows time could gain or lose, based on BCLK being off.
>
>BIOS time could gain or lose, based on 32768Hz clock.

Last night I synched my time, turned off my network, closed
windows. When I started up today I opened Wireshark, turned the
network back on and time was spot-on (nothing showed on Wireshark
other than the usual ARP stuff and the Neutron query).
After it was on for 3 hours I checked the time and it was 4
seconds off. I disconnected the network again, etc et al, and when I
had rebooted the time was exact, no delay. I did NOT correct the time
before rebooting.
IOW the BIOS time seems to be working fine.
I always assumed Windows used the BIOS hardware clock and not
some "internal software clock".
I read somewhere that BCLK is used for overclocking. My BIOS
settings are set to default. What could be altering the BCLK time by
so much ? And could adding memory have affected it somehow ?
TIA

[Paul]

Shadow wrote:

>
> Last night I synched my time, turned off my network, closed
> windows. When I started up today I opened Wireshark, turned the
> network back on and time was spot-on (nothing showed on Wireshark
> other than the usual ARP stuff and the Neutron query).
> After it was on for 3 hours I checked the time and it was 4
> seconds off. I disconnected the network again, etc et al, and when I
> had rebooted the time was exact, no delay. I did NOT correct the time
> before rebooting.
> IOW the BIOS time seems to be working fine.
> I always assumed Windows used the BIOS hardware clock and not
> some "internal software clock".
> I read somewhere that BCLK is used for overclocking. My BIOS
> settings are set to default. What could be altering the BCLK time by
> so much ? And could adding memory have affected it somehow ?
> TIA
> []'s

Well, it's either the absolute frequency of BCLK which
is off, or, something is preventing clock tick interrupts
from being serviced.

The best references I've seen on the various clocks in a
computer, is on the VM hosting software company web sites.
They usually explain what clocks are inside a real PC,
and how the virtualized environment provides those same
clocks as "fakes". But it also teaches you about how
clocks work on the host itself.

If you have a modern multi-core Intel processor, you
can try locking the cores together temporarily as a test.
The machine can save power if the cores operate
independently on frequency, but it also causes some
complications when handing clock information from
one core to another. My other machine, the core clocks
are locked, and turbo is disabled (to prevent overheat).

Paul

[Shadow]

On Mon, 24 Dec 2018 09:25:13 -0500, Paul <nos...@needed.invalid>
wrote:

Well, you seem to on to something. My computer was on for
nearly 3 hours running simple "one thread" programs and didn't lose a
second.
So I put it to convert a video with Wondefox Video Converter,
which uses all cores on my AMD-FX8300 (it says it's using 7 cores).
The conversion took under 5 minutes and my clock is now 12 seconds
behind.
So it has something to do with using more than one core, AND
it never happened before I added memory to the second memory bank (I
can't remember clock ever being so late but I never actually tested
it).

I'm using XP, if that matters.
I'll now boot into Linux and see if the same thing happens
there.

[Shadow]

No. In Linux even using all 8 cores at near max there is no
slowdown of the clock. I converted 3 videos using ffmpeg. The CPU
ventilator sounded like a helicopter taking off. So it must be an OS
thing.
Funny thing, I tried ffmpeg in XP and the clock went FORWARD 8
seconds.
I suppose I'll just have to carry on using Neutron at Startup,
since the time is never off by more than a minute, it won't affect me
as a user.
[]'s

I just checked with system monitor, and although it does use
core 1 most of the time, there are occasional small spikes in all the
other cores. I thought it would only use the other cores when under
stress ....

[Flasherly]

On Mon, 24 Dec 2018 16:53:54 -0200, Shadow <S...@dow.br> wrote:

>Funny thing, I tried ffmpeg in XP and the clock went FORWARD 8
>seconds.
> I suppose I'll just have to carry on using Neutron at Startup,
>since the time is never off by more than a minute, it won't affect me
>as a user.
> []'s
>
> I just checked with system monitor, and although it does use
>core 1 most of the time, there are occasional small spikes in all the
>other cores. I thought it would only use the other cores when under
>stress ....

Always on the second, that's not an option with the nature of a
computer that isn't intended for a chronometer. As with my watches,
with a built-in radio receiver, their mechanical limits are
overridden, once daily, for a percentage correction within an
individual second of accuracy. The same function is a definable
connection event, to an national atomic clock standard interface, as
provided software.

Although I'm not sure how older my Dimension 4 software version is, it
does allows definition of that event. I may have mine polling every
few minutes or less. All my primary, one wall and two wrist watches,
are equipped with transmission radio receivers. And my computer is
always to the second within sync to them for the same degree of
accuracy.

That's not to say that computer accuracy couldn't in itself be changed
time to be incorrect, a possible software occurrence that affects the
clock as an adverse situation, at least until Dimension 4 re-polls to
an atomic clock to perform a correct reset.

http://www.thinkman.com/dimension4/

From a practical standpoint, sounds to me as easy as it is possible to
approach substituting a computer at some less than a higher accuracy
than true chronographic precision.

[Shadow]

On Wed, 26 Dec 2018 18:15:30 -0500, Flasherly <Flas...@live.com>
wrote:

Sorry Flasherly, I got lost after the first phrase.
I sync to an atomic clock on startup, so it should be within
half a second of "real" time at the end of the day. But I can live
with 30 seconds off.
I was worried it might be hardware failure, but since Linux
does not have the glitch, I'm pretty sure it's just XP having fits
with the amount of CPU and memory at its disposal.
[]'s

[Flasherly]

On Thu, 27 Dec 2018 08:42:23 -0200, Shadow <S...@dow.br> wrote:

>Sorry Flasherly, I got lost after the first phrase.
> I sync to an atomic clock on startup, so it should be within
>half a second of "real" time at the end of the day. But I can live
>with 30 seconds off.
> I was worried it might be hardware failure, but since Linux
>does not have the glitch, I'm pretty sure it's just XP having fits
>with the amount of CPU and memory at its disposal.

The first written phrase, I provided, corresponds to after your
startup, and that is specific to Dimension 4, in the user settings,
for defining how often Dimension 4 performs an Atomic Clock
synchronization.

The Dimension 4 polling interval, I checked, just after the prior
post, and I've determined that my computer is about 8 seconds faster
than a standalone La Crosse radio receiver atomic clock, on the wall,
behind this monitor.

But the whole point is that if you can go beyond an OS start-up
synchronization event, a more frequent interval timed synchronization
may improve your computer's chronographic accuracy.

OK...checking with the second hand to the Casio. This is my watch:
https://en.wikipedia.org/wiki/File:Casio_Tough_Solar_Wave_Ceptor.jpg

I'm about 4 seconds fast on the computer, thus my computer is not as
accurate as the Casio. (I've checked the Casio to other standards,
such as a short-wave band radio receiver and England's "Big Ben" timed
broadcast. So my computer is not within 4 seconds near to a
chronograph, nor is Windows or, apparently, Dimension 4.) Checking
the BIOS clock may also be an indicated course to account;- as might
running Linux shed further light on narrowing in on a cause of the
discrepancy for accurate time keeping.

Between a computer crystal-derived signal reference and a WEB software
interface to poll a reference Atomic Clock server, there's no excuse
for this behavior. I believe it would be safe to assume, that you not
build such as a rocket-ship to blast off to the moon, not if you're
designing that trajectory based on a computer's ability to keep timed
accuracy.

[Flasherly]

On Thu, 27 Dec 2018 14:51:48 -0500, Flasherly <Flas...@live.com>
wrote:

standalone La Crosse radio receiver atomic clock,

-

La Crosse Atomic Clocks are better than a plain crystal movement: the
one-dollar modules from China I use when I build my own wall clocks.
Just not as good as a WaveCeptor, and its reviews may tend to reflect
that, somewhat on the spotty side.

[Shadow]

On Thu, 27 Dec 2018 14:51:48 -0500, Flasherly <Flas...@live.com>
wrote:

Oh, it's a freeware time-sync utility. I thought you were
talking about science fiction.

So is Neutron.

http://keir.net/neutron.html

7Kb small, and has an INI file so you can change the servers.
(the included ones are not working). If the first server doesn't
respond it tries the second, etc, for a total of 14.
I can give you my list if you want.
Keir writes good software.

[Flasherly]

On Thu, 27 Dec 2018 18:54:56 -0200, Shadow <S...@dow.br> wrote:

>
> Oh, it's a freeware time-sync utility. I thought you were
>talking about science fiction.
>
> So is Neutron.
>
>http://keir.net/neutron.html
>
> 7Kb small, and has an INI file so you can change the servers.
>(the included ones are not working). If the first server doesn't
>respond it tries the second, etc, for a total of 14.
> I can give you my list if you want.
> Keir writes good software.
> []'s

I'd have asked if you tried decreasing a N Interval to poll for a
lower integer (more often) than a default periodicity, or manually,
but there may not be that provision if there's nothing more to
configure it than the illustration.
http://keir.net/resources/scrn_neutron.png

I'd personally want that feature automated - within a definable N
Integer - as a part of a clock-aide program options. Or, you may
simply write it yourself, perhaps easily enough, with AUTOIT scripting
language for Windows.

Fiction occurs as an assumption of limitations, that a computer is a
instrument of capabilities to measure time at some discrepancy from a
scientific principle atomic clocks provide.

Yet computers may perhaps play that role in specialized instances. In
data network transfers as a part of secure protection protocols
augmented to account a high-precision timing event. Perhaps augmented
with specialty time-references or equipment outside of a common
desktop build. I wouldn't rule it out offhand.

. . .

Temperature is still only part of the story. Thermal noise is the
ultimate limitation on crystal oscillator performance (crystals are by
far the most common type of "digital clock").

The crystal itself has Brownian noise due to dissipative effects of
air resistance, anchor loss, and thermoelastic damping. Brownian noise
creates a random force that acts to disturb the crystal vibration.
This force creates random fluctuations in the exact oscillation
frequency of the crystal. The electronic oscillator circuit
responsible for compensating for the energy dissipation due to
mechanical damping also adds noise that has essentially the same
effect.

This still isn't quite the whole picture. Random changes in frequency
lead to a random walk in the period between two zero crossings. You
can think of this process as what happens when you flip a quarter and
keep track of the total heads and tails count. The odds of any flip
giving heads or tails is 50%. If you add 1 to your count for every
head and subtract one for every tail, the standard deviation of the
count is unbounded as time increases toward infinity. Similarly,
random fluctuations in the oscillation frequency "accumulate" over
time to lead to timing drift of the reference.


https://docs.microsoft.com/en-us/windows-server/networking/windows-time-service/accurate-time

Impact of increased polling and clock update frequency

In order to provide more accurate time, the defaults for polling
frequencies and clock updates are increased which allow us to make
small adjustments more frequently. This will cause more UDP/NTP
traffic, however, these packets are small so there should be very
little or no impact over broadband links. The benefit, however, is
that time should be better on a wider variety of hardware and
environments.

For battery backed devices, increasing the polling frequency can cause
issues. Battery devices don’t store the time while turned off. When
they resume, it may require frequent corrections to the clock.
Increasing the polling frequency will cause the clock to become
unstable and could also use more power. Microsoft recommends you do
not change the client default settings.

Domain Controllers should be minimally impacted even with the
multiplied effect of the increased updates from NTP Clients in an AD
Domain. NTP has a much smaller resource consumption as compared to
other protocols and a marginal impact. You are more likely to reach
limits for other domain functionality before being impacted by the
increased settings for Windows Server 2016. Active Directory does use
secure NTP, which tends to sync time less accurately than simple NTP,
but we’ve verified it will scale up to clients two stratum away from
the PDC.

As a conservative plan, you should reserve 100 NTP requests per second
per core. For instance, a domain made up of 4 DCs with 4 cores each,
you should be able to serve 1600 NTP requests per second. If you have
10k clients configured to sync time once every 64 seconds, and the
requests are received uniformly over time, you would see 10,000/64 or
around 160 requests/second, spread across all DCs. This falls easily
within our 1600 NTP requests/sec based on this example. These are
conservative planning recommendations and of course have a large
dependency on your network, processor speeds and loads, so as always
baseline and test in your environments.

It is also important to note that if your DCs are running with a
considerable CPU load, greater than 40%, this will almost certainly
add noise to NTP responses and affect your time accuracy in your
domain. Again, you need to test in your environment to understand the
actual results.

[Shadow]

On Thu, 27 Dec 2018 20:07:37 -0500, Flasherly <Flas...@live.com>
wrote:

>On Thu, 27 Dec 2018 18:54:56 -0200, Shadow <S...@dow.br> wrote:
>
>>
>> Oh, it's a freeware time-sync utility. I thought you were
>>talking about science fiction.
>>
>> So is Neutron.
>>
>>http://keir.net/neutron.html
>>
>> 7Kb small, and has an INI file so you can change the servers.
>>(the included ones are not working). If the first server doesn't
>>respond it tries the second, etc, for a total of 14.
>> I can give you my list if you want.
>> Keir writes good software.
>> []'s
>
>I'd have asked if you tried decreasing a N Interval to poll for a
>lower integer (more often) than a default periodicity, or manually,
>but there may not be that provision if there's nothing more to
>configure it than the illustration.
>http://keir.net/resources/scrn_neutron.png
>
>I'd personally want that feature automated

You could just use Task Scheduler, though I don't like it
running in the background, so I disabled the service.

http://www.blackviper.com/windows-services/task-scheduler/

[Flasherly]

On Fri, 28 Dec 2018 08:29:01 -0200, Shadow <S...@dow.br> wrote:

>You could just use Task Scheduler, though I don't like it
>running in the background, so I disabled the service.
>
>http://www.blackviper.com/windows-services/task-scheduler/

Learning an OS well enough to get around can seem at times less
daunting than defending or hardening it. And getting around, in the
case of Microsoft, has been as much better left to 3rd-party
developers, generously approximated by free or near program utilities.
Hardly an ingenuous claim or premise when, subsequently abandoned or
rendered obsolete from manipulation by tech industry, intent on
wholesaling individuals into an advertising market of cornered cattle
yards stocked with handheld-sets of social media.

Auto-It is somewhat elegant for running on top of Microsoft in a
script form. I've even run into programs, subsequently distributed
for freeware, written initially within Auto-It language conventions,
as is adequately explained within the distribution source for an
included documentary file. A recursive call to an timed event and a
configuration allowance for a clean exit should be relatively simple;-
worst that could happen is it could get "gummy" during an OS shutdown.