Had a 4-piece paired set of 16 GB memory modules for a total of 64 GB.
Worked fine for 6 years. Then one of them failed. Had errors during
backups during verification step, and memtest86 failed it. GSkill
replaced it, but warned that a single module even of the same model
would have the same specs, but may not match perfectly with the original
3 modules. Got the replacement, memtest86 said it passed, and it worked
okay with the other 3 modules for a couple weeks. However, during that
time, I did not power down the computer, or it wasn't powered off for
more than a day. Went on vacation, and the computer was stuck in a loop
with a "55" error message on an onboard LED status display. Looked it
up, might've been a RAM issue, and took out the newly replace module
(and another good one to keep DDR supported across the remaining 2
modules), and the computer booted okay.
My mobo (Asrock Taichi 390) has a BIOS setting for XMP which is to
overclock the memory modules, but within the manufacturer's allowed
spec. Figured I'd disable XMP to run the RAM at its rated spec, but the
only 2 settings for XMP are Auto, and Profile 1 (XMP 2.0). Well, there
is only 1 XMP profile, so Auto just picks that one. There is no setting
to disable XMP to prevent overclocking.
Any ideas on how to disable XMP to prevent overclocking the RAM? I have
not yet tried going into the mobo's BIOS overclock settings to see if I
can specify the clock and voltage on the RAM to see if my choices would override the XMP setting. Seems odd that XMP is available, but no way
to overtly disable it.
Possibly the 6-year mobo is aging on its voltage to the memory, and
upping the RAM voltage might solve the problem, but I really don't want
to overclock nor overvolt my setup.
I'd like not to lose 16 GB of RAM by omitting the new replacement
module. If that's my only choice, then I have to decide if DDR is worth
some minor performance increase, or go with the 3 originally paired
module to go with 48 GB of RAM.
.... Got the replacement, memtest86 said it passed, and it worked
okay with the other 3 modules for a couple weeks. However, during that
time, I did not power down the computer, or it wasn't powered off for
more than a day. Went on vacation, and the computer was stuck in a loop
with a "55" error message on an onboard LED status display. Looked it
up, might've been a RAM issue, and took out the newly replace module
(and another good one to keep DDR supported across the remaining 2
modules), and the computer booted okay.
Auto is JEDEC, XMP is XMP.
First of all, check your motherboard for the ability to "record a
profile". If you use custom settings, or if you have more than one
experiment ongoing, you can do things like Load Profile 1 or Load
Profile 2 or Load Profile 3 at BIOS level, and those are stored the
same way UEFI NVRAM is stored (BIOS chip). I have to remind you of
that, because I was messing around with a computer here, flashed it,
lost all my customization and then realized I'd never recorded
whatever I had. It's good to plan ahead, if you have "something to
lose" while experimenting.
The next step, is to get a readout for system level.
https://www.cpuid.com/softwares/cpu-z.html
Scroll down to the four purple entries on the left.
The ZIP English should be portable, so you can just keep
a cpuz_x64.exe in your Downloads folder. It will keep a cpuz.ini
for any settings it wants to carry from one session to the next.
You can attempt to use the Overclock controls, and select a memory
speed there. For example, say I use the overclock to select 1866
speed, the "Auto-ness" of the BIOS will work out the correct CAS,
tRCD and whatever else for you. There is no particular reason at this
point, to be computing those values yourself. In the customization
section (without looking at the manual), you might see Trcd and it
might have a choice of "Auto" for example, which means the SPD table
on the DIMM will be consulted and interpolation or extrapolation will
be used as necessary to work out a correct value.
As for your strategy, there are other ways to skin a cat.
On a couple motherboards, I've had to bump the voltage to the
IMC (Integrated Memory Controller in CPU) by one voltage division
in the settings. That could be a 25 millivolt step. Usually there
are colors associated with the voltage settings, and you can go
up a number of steps in the Yellow without a problem. The Red color
voltages are getting a bit too high for long term use. I have
had good results correcting "marginal" errors on weak controllers.
Like, say I get one error per day, bumping the memory controller
internal operating voltage a bit might be enough. Sometimes, backing
off on Trcd isn't enough, and the memory controller is the weak part
rather than the DIMM.
The memory controller could wear via ElectroMigration, but I don't
think your processor would be a candidate for that excuse. Some of the
AMD processors from their Global Foundry era, you could damage those
by overclocking, and I would read reports of "uh, it won't even run
stock any more". Other processors seem to be OK with a mild overclock
and there isn't that issue of the processor getting weak as it ages.
It's hard to say, as usual, what is going on when a memory fails.
My voltage bump strategy is usually associated with Intel, but that
does not mean that AMD could not have it happen. It hasn't happened
here "this week" :-)
Start with checking whether you can record a Profile at BIOS level, to
save your settings if they are lost. Settings do not have to be
compatible across BIOS updates, so a Profile is not "guaranteed"
to work on a reload, but it usually works. And depending on how much
cruft you've been switching off at BIOS level, it can save you some
time when changing the BIOS battery or after a BIOS flashup.
A BIOS flashup many not necessarily change your UEFI keys. The Secure Boot section has key management, and you can back up keys to a FAT formatted USB stick.
I did that a week ago on the Big Machine, while trying to get the
2023 Microsoft key into the key set. So a BIOS Profile is one thing,
key management is a second issue (when say, an OS you install messes
with the keys, you erase the OS in question, then restore the key set
from your FAT USB key).
XMP uses boosted voltage (it generally avoids going outside the datasheet limit value for the RAM pads). Auto will use a very mild boost (which is why turning down the speed is not as much of an advantage as you might think). Playing with the IMC voltage, might correct a very marginal situation,
but then, you don't have a very good idea as to what margin exists after
the correction. You can leave individual memory timings on Auto, so the
JEDEC table in the SPD on the DIMM, those values will be used. You can
use CPUZ to monitor what tRCD it happened to use, if you want to re-enter
the BIOS and take manual control of one parameter. (It's hard for
me to remember now, but the "timing" next to CAS, which I think tRCD is
next, bumping that up by one [slower] is good for four DIMM loading.)
Some BIOS offer two XMP settings. You would get Auto, XMP1, XMP2.
The first profile is Command Rate 2 (conservative, extra setup time to clock),
the second profile is Aggressive Command Rate 1 (pad must make setup time requirement, while bus is heavily loaded). They do not label the XMPs in
that case as "Conservative" or "Aggressive" and only a memtest bandwidth measurement indicates what the profile is using in that case.
If you are offered only one XMP, that should be a Command Rate 2 choice.
I don't see much profit in offering the Aggressive one, as there isn't
a lot of reason to think it would work (Intel has always liked CR2).
My AMD offered CR1 and CR2, and they both work! I measured with Memtest
and the proof of pudding is seeing an extra 2GB/sec in the mostly pointless bandwidth measurement. I turned it back to CR2 and use the Conservative setting. That should not wear anything particularly, but it does make
it harder to have error free operation at bus timing level.
Summary: You can turn the speed down, you can run Auto the first time,
record tRCD, reenter BIOS, only modify tRCD by bumping up by
one and retest. The other direction of play, is to look for
a voltage section, take it off Green and use a bit of Yellow
in the setting for the IMC voltage. We don't have any evidence
at the moment that this is necessary. Only the chip select signal
would be slot specific, if it is always the same slot that has
a problem. If a channel has a problem, a lot more signals could
have a bad pad driver on the CPU.
When these things happen, it's really hard to say what is broken.
Getting a replacement DIMM, and it not working, is a bit unusual.
The "matched set" idea, I don't think there is a strong correlation
with that, unless a company starts shipping untested memory or using
their "bad bin" DIMMs to replace customer warranty items. I would have
expected the errors to not favor just the one DIMM, and a bad DIMM
to be replaced by a bad DIMM. But I suppose it could happen.
Based on my AMD+DDR4 experience, I suspect the RAM chips used are
excellent and don't really need a lot of speed test. Running CR1 and
CR2 with four DIMMs ? Unheard of. How can that work??? That defies
the observations of bus loading effect over the years. In the case
of DDR5, we're back to the old rules again -- four sticks runs slow.
Must back off on the clock speed in that case, and the BIOS does that
for you.
On 17/12/2025 2:34 am, VanguardLH wrote:
.... Got the replacement, memtest86 said it passed, and it worked
okay with the other 3 modules for a couple weeks. However, during that
time, I did not power down the computer, or it wasn't powered off for
more than a day. Went on vacation, and the computer was stuck in a loop
with a "55" error message on an onboard LED status display. Looked it
up, might've been a RAM issue, and took out the newly replace module
(and another good one to keep DDR supported across the remaining 2
modules), and the computer booted okay.
You talking about a Window$ lockup or MEMTest86? What OS and proggram
were you running?? Was it really hardware that is the new DDR4 RAM?
Mixing different brands and models of DDR4 could cause stability
problem, even if XMP was not turned on. But not turning on XMP
usually solved this problem, but you lost the max possible speed.
Paul <nospam@needed.invalid> wrote:
Auto is JEDEC, XMP is XMP.
So, Auto is the SPD specs reported by the RAM? Auto is currently
selected, but the memory profile listed is "Profile 1: XMP 2.0".
I had it as Auto, but after a week of being powered down the mobo would
no longer boot. It would show a 55 error, and restart without a chance
of getting into the BIOS.
If Auto means JEDEC (aka SPD) specs then I'm not overclocked now. Hmm,
so much for figuring there was a speed or spec mismatch, however slight, between the original 3 (of the 4) modules with the new replacement
module.
I had cleared the CMOS copy of BIOS settings which should load the
defaults just before I removed the 2 memory modules (the new replacement module, and an original module both of which tested okay with
memtest86). So, I can't be sure it wasn't a reset to default BIOS
settings, or removing the new replacement module (and an old module to
keep even the module count to support DDR). I'll try reinstalling the 2 modules that I removed to get booting to work to see if it now reboots
okay with all 4 modules.
I can also run memtest86 again with all 4 modules installed to see if
the computer boots okay now, and memtest86 says all modules pass. The
BIOS reports what are the voltages, but I'm wondering how accurate that
is. I have multimeters, but who know how accurate they are. I haven't dropped a multimeter at a calibration shop for several decades. Maybe
the voltage regulators no longer actually produce the expected voltage.
Why I considered looking at the OC settings to see how much I could
increment the DRAM voltage. The mobo manual doesn't mention the
increments for changing voltage.
First of all, check your motherboard for the ability to "record a
profile". If you use custom settings, or if you have more than one
experiment ongoing, you can do things like Load Profile 1 or Load
Profile 2 or Load Profile 3 at BIOS level, and those are stored the
same way UEFI NVRAM is stored (BIOS chip). I have to remind you of
that, because I was messing around with a computer here, flashed it,
lost all my customization and then realized I'd never recorded
whatever I had. It's good to plan ahead, if you have "something to
lose" while experimenting.
I recall something about multiple profiles, and maybe saving them, so I
could try that. Maybe the saved profiles would tell what specs got
saved in each one, so I could compare Auto to XMP to overclocked (well, underclocked/standard) values.
The next step, is to get a readout for system level.
https://www.cpuid.com/softwares/cpu-z.html
Scroll down to the four purple entries on the left.
The ZIP English should be portable, so you can just keep
a cpuz_x64.exe in your Downloads folder. It will keep a cpuz.ini
for any settings it wants to carry from one session to the next.
There is actually an Asrock cutomized version of CPU-Z (2.14) and the classic/stock version (2.17) still in my Downloads folder. I grabbed
the zip version (2.17) just now from their web site.
The memory I bought was at Newegg, and listed at: https://www.newegg.com/g-skill-ripjaws-v-series-64gb-ddr4-3200-cas-latency-cl16-desktop-memory-black/p/N82E16820232092?Item=N82E16820232092
CPU-Z's Memory tab doesn't list each module separately. It shows the following:
General
Type: DDR4
Size: 32GB (only 2 modules currently installed out of 4)
Channel#: Single (*)
LLC/Ring: 3996.1 MHz (bounces around a few MHz)
Timings
DRAM frequency: 1065.1 MHz
FSB:DRAM: 3:32
CAS# latency: 15
and other specs.
(*) Might that be because the 2 modules remaining in the mobo slots are
in the A1 and A2 slots instead of A1 and B1, or A2 and B2. So, I might
not be in DDR mode at the moment.
Under its SPD tab, I can pick each slot:
Slot 1: DDR4 (UDIMM) Module size: 16 GB
Bandwidth: DDR4-3200 (1600MHz) SPD: XMP 2.0 (**)
Timings table:
Shows different frequencies for:
JEDEC #5: 1033 MHz (CAS 14.0)
JEDEC #6: 1066 MHz (CAS 15.0)
JEDEC #7: 1066 MHz (CAS 16.0)
XMP-3200: 3200 MHz (CAS 16.0)
From the bandwidth and SPD specs above, looks like Auto selected the XMP
2.0 profile.
(**) In the BIOS, Auto was selected yet it looks like XMP got used.
You can attempt to use the Overclock controls, and select a memory
speed there. For example, say I use the overclock to select 1866
speed, the "Auto-ness" of the BIOS will work out the correct CAS,
tRCD and whatever else for you. There is no particular reason at this
point, to be computing those values yourself. In the customization
section (without looking at the manual), you might see Trcd and it
might have a choice of "Auto" for example, which means the SPD table
on the DIMM will be consulted and interpolation or extrapolation will
be used as necessary to work out a correct value.
As for your strategy, there are other ways to skin a cat.
On a couple motherboards, I've had to bump the voltage to the
IMC (Integrated Memory Controller in CPU) by one voltage division
in the settings. That could be a 25 millivolt step. Usually there
are colors associated with the voltage settings, and you can go
up a number of steps in the Yellow without a problem. The Red color
voltages are getting a bit too high for long term use. I have
had good results correcting "marginal" errors on weak controllers.
Like, say I get one error per day, bumping the memory controller
internal operating voltage a bit might be enough. Sometimes, backing
off on Trcd isn't enough, and the memory controller is the weak part
rather than the DIMM.
The memory controller could wear via ElectroMigration, but I don't
think your processor would be a candidate for that excuse. Some of the
AMD processors from their Global Foundry era, you could damage those
by overclocking, and I would read reports of "uh, it won't even run
stock any more". Other processors seem to be OK with a mild overclock
and there isn't that issue of the processor getting weak as it ages.
It's hard to say, as usual, what is going on when a memory fails.
My voltage bump strategy is usually associated with Intel, but that
does not mean that AMD could not have it happen. It hasn't happened
here "this week" :-)
Start with checking whether you can record a Profile at BIOS level, to
save your settings if they are lost. Settings do not have to be
compatible across BIOS updates, so a Profile is not "guaranteed"
to work on a reload, but it usually works. And depending on how much
cruft you've been switching off at BIOS level, it can save you some
time when changing the BIOS battery or after a BIOS flashup.
A BIOS flashup many not necessarily change your UEFI keys. The Secure Boot >> section has key management, and you can back up keys to a FAT formatted USB stick.
I did that a week ago on the Big Machine, while trying to get the
2023 Microsoft key into the key set. So a BIOS Profile is one thing,
key management is a second issue (when say, an OS you install messes
with the keys, you erase the OS in question, then restore the key set
from your FAT USB key).
XMP uses boosted voltage (it generally avoids going outside the datasheet
limit value for the RAM pads). Auto will use a very mild boost (which is why >> turning down the speed is not as much of an advantage as you might think). >> Playing with the IMC voltage, might correct a very marginal situation,
but then, you don't have a very good idea as to what margin exists after
the correction. You can leave individual memory timings on Auto, so the
JEDEC table in the SPD on the DIMM, those values will be used. You can
use CPUZ to monitor what tRCD it happened to use, if you want to re-enter
the BIOS and take manual control of one parameter. (It's hard for
me to remember now, but the "timing" next to CAS, which I think tRCD is
next, bumping that up by one [slower] is good for four DIMM loading.)
Some BIOS offer two XMP settings. You would get Auto, XMP1, XMP2.
The first profile is Command Rate 2 (conservative, extra setup time to clock),
the second profile is Aggressive Command Rate 1 (pad must make setup time
requirement, while bus is heavily loaded). They do not label the XMPs in
that case as "Conservative" or "Aggressive" and only a memtest bandwidth
measurement indicates what the profile is using in that case.
If you are offered only one XMP, that should be a Command Rate 2 choice.
I don't see much profit in offering the Aggressive one, as there isn't
a lot of reason to think it would work (Intel has always liked CR2).
My AMD offered CR1 and CR2, and they both work! I measured with Memtest
and the proof of pudding is seeing an extra 2GB/sec in the mostly pointless >> bandwidth measurement. I turned it back to CR2 and use the Conservative
setting. That should not wear anything particularly, but it does make
it harder to have error free operation at bus timing level.
Summary: You can turn the speed down, you can run Auto the first time,
record tRCD, reenter BIOS, only modify tRCD by bumping up by
one and retest. The other direction of play, is to look for
a voltage section, take it off Green and use a bit of Yellow
in the setting for the IMC voltage. We don't have any evidence
at the moment that this is necessary. Only the chip select signal >> would be slot specific, if it is always the same slot that has
a problem. If a channel has a problem, a lot more signals could
have a bad pad driver on the CPU.
When these things happen, it's really hard to say what is broken. >> Getting a replacement DIMM, and it not working, is a bit unusual. >> The "matched set" idea, I don't think there is a strong correlation >> with that, unless a company starts shipping untested memory or using
their "bad bin" DIMMs to replace customer warranty items. I would have
expected the errors to not favor just the one DIMM, and a bad DIMM >> to be replaced by a bad DIMM. But I suppose it could happen.
Based on my AMD+DDR4 experience, I suspect the RAM chips used are >> excellent and don't really need a lot of speed test. Running CR1 and
CR2 with four DIMMs ? Unheard of. How can that work??? That defies >> the observations of bus loading effect over the years. In the case >> of DDR5, we're back to the old rules again -- four sticks runs slow.
Must back off on the clock speed in that case, and the BIOS does that
for you.
CPU-Z listed a tRC value for each module occupying a slot, but not a
tRCD value. Back under its Memory tab, "RAS# to CAS# Delay (tRCD)" is listed, and at 15.0 clocks which matches on the above per-module timings
of "JEDEC #6: 1066 MHz (CAS 15.0)". So, now I'm confused if XMP2.0 is
being used as shown per module under the SPD tab, or if JEDEC #6 is
being used as the real SPD values the BIOS uses. The mobo manual under
OC settings mentions a tRCDtRP described as "Row Precharge: The number
of clock cycles required between the issuing of the precharge command
and opening the next row."
At this point after powering off (yanking the power cord, too), I'll
first move the working 2 modules into slots A1 and B1, and check with
CPU-Z if it shows Channel # as dual (which I presume means DDR mode).
I'd like to get back DDR mode. Per the mobo manual:
Dual Channel Memory Configuration
Priority DDR4_A1 DDR4_A2 DDR4_B1 DDR4_B2
1 Populated Populated
2 Populated Populated Populated Populated
Looks like slots A2 and B2 for now with only 2 modules to get DDR mode.
So, I'll move the one in A1 over to B2 giving me A2 and B2 populated.
Then check if CPU-Z says dual or DDR for Channel # under its Memory tab.
It's confusing if the BIOS' Auto mode is using SPD with JEDEC or XMP.
Under CPU-Z's SPD tab, it shows SPD ext = XMP 2.0, and that's with the
BIOS configured to Auto. Could be this Asrock mobo with its easy
settings doesn't let me get away from XMP. That's why I figure I may
have to use the overclock tweaks to override. But first I'll experiment
with bumping the DRAM voltage. With all 4 slots populated, maybe the
voltage regulators just can't pump out as much as current as they used
to.
"Mr. Man-wai Chang" <toylet.toylet@gmail.com> wrote:
On 17/12/2025 2:34 am, VanguardLH wrote:
.... Got the replacement, memtest86 said it passed, and it worked
okay with the other 3 modules for a couple weeks. However, during that
time, I did not power down the computer, or it wasn't powered off for
more than a day. Went on vacation, and the computer was stuck in a loop >>> with a "55" error message on an onboard LED status display. Looked it
up, might've been a RAM issue, and took out the newly replace module
(and another good one to keep DDR supported across the remaining 2
modules), and the computer booted okay.
You talking about a Window$ lockup or MEMTest86? What OS and proggram
were you running?? Was it really hardware that is the new DDR4 RAM?
OS and BIOS not even involved. Cannot get to BIOS and even less to OS.
When booting, a LED status readout shows 55 which indicates something
wrong with RAM, then the computer restarts. It is stuck in a tight loop about a RAM error, so I couldn't even get in BIOS settings.
Mixing different brands and models of DDR4 could cause stability
problem, even if XMP was not turned on. But not turning on XMP
usually solved this problem, but you lost the max possible speed.
As mentioned, not mixing different brands or even different models
within a brand. Original a 4-pc set of memory modules were purchased.
One module went bad, and got warranty replaced with the same brand and
model. Memtest86 passed all 4 modules, including with the new
replacement module.
In the discussion with Paul in another subthread, getting out of XMP (overclocking) mode may not be possible in the BIOS using the easy
settings. Whether I picked Auto or the XMP profile (there was only 1),
looks like the mobo forces XMP mode; i.e., Auto picks XMP 2.0, and my
other choice is an XMP 2.0 profile. May have to go into the overclock settings in BIOS to force the mobo to stop using XMP mode, and possibly
bump the DRAM voltage since all 4 slots are populated (more load on the
same voltage bus).
Worked for 6 years, a module went bad, got an exact replacement, all 4 populating all 4 slots passed memtest86, worked okay for a couple weeks, powered down for a week, came back, and was stuck in a very tight boot
fail loop that never approached letting me get into the BIOS (so
obviously no OS never yet got involved).
From my reading, DDR (dual-channel) mode is more important for
performance than is XMP. I'll have to configure the memory modules in
the correct mobo slots to get DDR mode. That means using only 2 of the
4 16GB modules dropping me from 64GB to 32GB. If I try using a 3rd one
from the original 4-pc set, I'll lose DDR mode. In losing 1 module, I actually lose 2 while trying to keep DDR mode.
Compared to DDR mode, XMP's performance boost is minimal. There isn't
much I run that would have a perceptible performance gain with XMP. I
play few video games, and they don't tax the CPU, GPU, or memory. First
is to reconfigure to get DDR mode active. That'll use 2 modules in the proper slots on the mobo. Then I'll play with all 4 modules, but I'd
like the BIOS to use the JEDEC specs from SPD instead XMP specs from SPD
on the memory modules. I'd have DDR, and no overlocking of the DRAM. However, I may have to tweak the DRAM voltage a step up with all 4 slots populated. It ran okay with the SPD XMP settings for 6 years, but electronics will fatigue over time, like voltage regulators not pumping
out enough current resulting in lowered voltages. My multimeters are
not sufficiently accurate nor certified to ensure any voltage I measure
are accurate compared to what the BIOS reports, but then voltage
monitoring by the mobo might not be super accurate, either.
As mentioned, not mixing different brands or even different models
within a brand. Original a 4-pc set of memory modules were purchased.
One module went bad, and got warranty replaced with the same brand and
model. Memtest86 passed all 4 modules, including with the new
replacement module.
In the discussion with Paul in another subthread, getting out of XMP (overclocking) mode may not be possible in the BIOS using the easy
settings. Whether I picked Auto or the XMP profile (there was only 1),
looks like the mobo forces XMP mode; i.e., Auto picks XMP 2.0, and my
other choice is an XMP 2.0 profile. May have to go into the overclock settings in BIOS to force the mobo to stop using XMP mode, and possibly
bump the DRAM voltage since all 4 slots are populated (more load on the
same voltage bus).
Replacement RAM might have (slightly?) different timing from the one you purchased eariler. Production lines and components change from time to time.
Did you forget the good, old method called Clear CMOS? That's the best
method to reinstall all RAM modules, and maybe everything.
Van is getting an error of "55", which is apparently
"RAM not found", and this implies something is seriously
wrong now. It's possible no user-level tricks will
fix this. We'll see I guess.
Paul wrote:
Van is getting an error of "55", which is apparently "RAM not found",
and this implies something is seriously wrong now. It's possible no
user-level tricks will fix this. We'll see I guess.
Clearing CMOS usually turn off all overclocking settings and set all
DDR4 slots back to JDEC (no XMP). The debug LED might be wrong.
I admit that I have never used expensive boards that have debug LEDs. :)
Ah, I forgot something when I performed the annual dusting out of the computer using compressed air cans. My house is dusty, and had pets
until recently. I forgot to replace the CR2032 CMOS battery. It was in
my notes reminding me of the yearly clean, but I forgot to replace it.
If a CMOS clear was what actually ended up fixing my setup, maybe the
CMOS coin cell battery is weak, and settings got corrupted in the the
CMOS copy of the BIOS EEPROM settings after a week without any power.
It's the CMOS copy used on boot. I just remembered that I forgot to
replace the CR2032 coin cell battery. I replace at about 3-year
intervals although the battery should be good for 5 years.
The 55 error ("The Memory could not be detected.") which goes away after yanking out the modules in the last 2 slots makes me a suspect an aging problem with the mobo. Currently I degraded to 2 sticks in the first 2
slots (A1 and A2), but that is not a valid config to support DDR. I'll
move the stick from A1 to B2 since A2+B2 is what they say to use with 2 sticks to get DDR mode. That will include the 3rd slot where was the replacement memory module. Then I'll run memtest86 on the 2-stick 32 GB A2+B2 setup. Then I'll try with all 4 slots populated (CMOS has already
been cleared), and run memtest86 again. But memtest86 passed all 4
sticks a couple weeks ago after I got the replacement memory module.
On 17/12/2025 8:16 pm, VanguardLH wrote:
The 55 error ("The Memory could not be detected.") which goes away after
yanking out the modules in the last 2 slots makes me a suspect an aging
problem with the mobo.ÿ Currently I degraded to 2 sticks in the first 2
slots (A1 and A2), but that is not a valid config to support DDR.ÿ I'll
move the stick from A1 to B2 since A2+B2 is what they say to use with 2
sticks to get DDR mode.ÿ That will include the 3rd slot where was the
replacement memory module.ÿ Then I'll run memtest86 on the 2-stick 32 GB
A2+B2 setup.ÿ Then I'll try with all 4 slots populated (CMOS has already
been cleared), and run memtest86 again.ÿ But memtest86 passed all 4
sticks a couple weeks ago after I got the replacement memory module.
Just go back to basics, start with 1 stick of DDR, then 2, then 3 and 4. Check the motherboard manual (maybe via manufacturer's website) if you forgot how to use dual-channel.
On Wed, 12/17/2025 8:24 AM, Mr. Man-wai Chang wrote:
On 17/12/2025 8:16 pm, VanguardLH wrote:
The 55 error ("The Memory could not be detected.") which goes away after >>> yanking out the modules in the last 2 slots makes me a suspect an aging
problem with the mobo.ÿ Currently I degraded to 2 sticks in the first 2
slots (A1 and A2), but that is not a valid config to support DDR.ÿ I'll
move the stick from A1 to B2 since A2+B2 is what they say to use with 2
sticks to get DDR mode.ÿ That will include the 3rd slot where was the
replacement memory module.ÿ Then I'll run memtest86 on the 2-stick 32 GB >>> A2+B2 setup.ÿ Then I'll try with all 4 slots populated (CMOS has already >>> been cleared), and run memtest86 again.ÿ But memtest86 passed all 4
sticks a couple weeks ago after I got the replacement memory module.
Just go back to basics, start with 1 stick of DDR, then 2, then 3 and 4. Check the motherboard manual (maybe via manufacturer's website) if you forgot how to use dual-channel.
Regarding your hypothesis that clearing the CMOS "loads Setup Defaults",
the Asrock board appears to be failing at this task (as seen in a
Reddit thread).
It is doing something else. Something else naughty. Some people are replacing hardware after a "55", which implies the firmware is not of high quality.
It'll be interesting to see if a CMOS battery makes any difference
at all. I would not bet on this helping, but it is still a good
idea to fit a new cell.
There are systems that successfully run, day after day, with
a completely flat CMOS battery. Having a good battery, is not
always required, but there is no question that some SuperIO
have a VBAT measurement pin, where the chip does something
with a battery voltage measurement. I just have not seen
a SuperIO chip spec where any sort of "feature" with respect
to VBAT is mentioned (there is no text that says "system
will not start with 0 volt battery").
Regarding your hypothesis that clearing the CMOS "loads Setup Defaults",In the old days, BIOS battery usually affected only the clock. But motherboards these days are way more sophiscated and have far more
the Asrock board appears to be failing at this task (as seen in a
Reddit thread).
It is doing something else. Something else naughty. Some people are replacing hardware after a "55", which implies the firmware is not of high quality.
It'll be interesting to see if a CMOS battery makes any difference
at all. I would not bet on this helping, but it is still a good
idea to fit a new cell.
There are systems that successfully run, day after day, with
a completely flat CMOS battery. Having a good battery ....
On 17/12/2025 10:50 pm, Paul wrote:
In the old days, BIOS battery usually affected only the clock.
Regarding your hypothesis that clearing the CMOS "loads Setup Defaults",
the Asrock board appears to be failing at this task (as seen in a
Reddit thread).
It is doing something else. Something else naughty. Some people are replacing
hardware after a "55", which implies the firmware is not of high quality.
It'll be interesting to see if a CMOS battery makes any difference
at all. I would not bet on this helping, but it is still a good
idea to fit a new cell.
There are systems that successfully run, day after day, with
a completely flat CMOS battery. Having a good battery ....
But motherboards these days are way more sophiscated and have
far more settings than just date and time to remember.
After some testing, I found NONE of the memory modules (original 3, or
the 1 warranty replacement, all the same brand and model) would run
using XMP. Even with just 1 module, none of the modules that claim XMP support would work. I'd set the BIOS to XMP mode, cold boot, and there
was a really long wait until the POST screen appeared; however, it noted there had been many boot attempts whereupon safe settings got used to
finally get to the POST screen. If Auto was selected, 1, 2, or 3 of the original modules would let the computer boot. The warranty unit was
farked: couldn't boot even with it alone.
XMP wanted to use 3200 MHz (1600 MHz x 2) for the target DRAM clock.
Auto (JEDEC) uses 2133 Mhz (well, 1066 MHz x 2). The product specs, and CPU-z, say the max bandwidth for the G.Skill modules is 3200 (1600 MHz),
but none of the modules would run at that clock rate. XMP was out, and
Auto (JEDEC) had to get used. Although that means running at 66% of the claimed max clock rate, it does run with the slower clock. Claiming XMP support at 3200 but refusing to boot even to the POST screen is an
unusable setup.
"Mr. Man-wai Chang" <toylet.toylet@gmail.com> wrote:
Paul wrote:
Van is getting an error of "55", which is apparently "RAM not found",
and this implies something is seriously wrong now. It's possible no
user-level tricks will fix this. We'll see I guess.
Clearing CMOS usually turn off all overclocking settings and set all
DDR4 slots back to JDEC (no XMP). The debug LED might be wrong.
I admit that I have never used expensive boards that have debug LEDs. :)
The status LED mostly shows 55 when the computer was quick cycling (not
even getting to where I could enter the BIOS screens). However,
sometimes it showed 7F. When starting up, the LED flashes between
several codes, but I'm not sure which are valid since the LED elements
are just flickering alot which might look like codes. It did once or
twice show 7F (before restarting) which means "Chipset initialization
error. Please press reset or clear CMOS" (that's a push button in the
rear pannel). There is a 2-pin CMOS header to short, but I use the
button to clear CMOS.
The 55 error ("The Memory could not be detected.") which goes away after yanking out the modules in the last 2 slots makes me a suspect an aging problem with the mobo. Currently I degraded to 2 sticks in the first 2
slots (A1 and A2), but that is not a valid config to support DDR. I'll
move the stick from A1 to B2 since A2+B2 is what they say to use with 2 sticks to get DDR mode. That will include the 3rd slot where was the replacement memory module. Then I'll run memtest86 on the 2-stick 32 GB A2+B2 setup. Then I'll try with all 4 slots populated (CMOS has already
been cleared), and run memtest86 again. But memtest86 passed all 4
sticks a couple weeks ago after I got the replacement memory module.
CMOS has already been cleared. The only change I then made in the BIOS
was to get rid of the Asrock banner page that obliterates the POST
screen which I'd rather see during boot. It was quite evident the CMOS
clear happened, because I used to have BIOS configure for a system
password; i.e., I needed to enter a password to get further than the
boot, and start loading the OS. Now the system password isn't required anymore, and I might leave it that way. It was assurance that
regardless how the OS was configured to retry a failed boot that the
computer wouldn't end up keep cycling on a failed boot (by hardware, or
with the OS or whatever it loaded).
Ah, I forgot something when I performed the annual dusting out of the computer using compressed air cans. My house is dusty, and had pets
until recently. I forgot to replace the CR2032 CMOS battery. It was in
my notes reminding me of the yearly clean, but I forgot to replace it.
If a CMOS clear was what actually ended up fixing my setup, maybe the
CMOS coin cell battery is weak, and settings got corrupted in the the
CMOS copy of the BIOS EEPROM settings after a week without any power.
It's the CMOS copy used on boot. I just remembered that I forgot to
replace the CR2032 coin cell battery. I replace at about 3-year
intervals although the battery should be good for 5 years.
After some testing, I found NONE of the memory modules (original 3, or
the 1 warranty replacement, all the same brand and model) would run
using XMP. Even with just 1 module, none of the modules that claim XMP support would work. I'd set the BIOS to XMP mode, cold boot, and there
was a really long wait until the POST screen appeared; however, it noted there had been many boot attempts whereupon safe settings got used to
finally get to the POST screen. If Auto was selected, 1, 2, or 3 of the original modules would let the computer boot. The warranty unit was
farked: couldn't boot even with it alone.
XMP wanted to use 3200 MHz (1600 MHz x 2) for the target DRAM clock.
Auto (JEDEC) uses 2133 Mhz (well, 1066 MHz x 2). The product specs, and CPU-z, say the max bandwidth for the G.Skill modules is 3200 (1600 MHz),
but none of the modules would run at that clock rate. XMP was out, and
Auto (JEDEC) had to get used. Although that means running at 66% of the claimed max clock rate, it does run with the slower clock. Claiming XMP support at 3200 but refusing to boot even to the POST screen is an
unusable setup.
The new module (warranty replacement) was the same brand and model, but seemed a bad unit sent to replace the old one that went bad. With Auto (JEDEC), and the warranty module by itself or when it was used with any
of the original modules, the computer wouldn't boot. I'd see the
diagnostic LED keep cycling through multiple codes when the warranty
module was involved. So, I have the 3 modules from the original 4-pc
matched set, and the warranty replacement that screws up the boot.
Back when I had just 2 original modules in the A1 & A2 slots, CPU-z's
Memory tab said Channel # = Single. I expected that since DDR works
across the banks, not within a bank. The manual says to use the A2 and
B2 slots for 2-module DDR, but the computer wouldn't boot at all with
that config. So, I tested using the A1 and B1 slots, the computer
booted, and Dual was shown in BIOS, and CPU-z said Channel # = Dual.
So, I was down to 2 of the original modules in A1 and B1 using DDR mode running at 2133 (1066x2) for a total of 32GB of RAM. I'm down to half
the original 64 GB RAM.
With it working again, now I had a warranty module that wouldn't play
nice with the 3 modules from the original 4-pc matched set, and was
flaky when it was the only one in a slot (sometimes the computer would
boot to the POST screen, but sometimes it was stuck in the tight reboot
cycle with various codes flashing on the diagnostic LEDs). Guess I got
a defective warranty replacement. But then the originals don't work
with XMP despite their specs saying so, or the mobo is farked (i.e., the modules might support XMP, but the mobo won't). Possibly upping the
DRAM voltage might get XMP working for the 3 original modules, but
that's a later project. That the warranty module won't even work alone
and in Auto (JEDEC) means I'm not wasting time on it, anymore.
So, I'm left with the 3 original modules from the 4-pc matched set that
will work. However, I always thought an even number of modules were
needed for dual-channel support, but I only have 3 good sticks now. I
could use 2 of them for 32 GB, one in each bank, to get dual mode. Yet,
as a test, I put in all 3 of the working 16 GB sticks for a total of
48GB of RAM. I expected to lose dual channel mode, but BIOS and CPU-z
say different. Could it be they detect dual-channel for 2 of the sticks
(A1 and B1), and the 3rd (A2) is operating in single-channel mode, or
however they indicate dual channel is wrong or misleading.
"It is unable to activate Dual Channel Memory Technology with only one
or three
memory module installed."
That's what the manual says, but it was wrong about the slot config for
dual channel: A2+B2 wouldn't let the computer boot, but A1+B1 would.
I'm pretty sure 3 modules (A1+A2+B1) should disable dual-channel mode.
The manual also shows the wrong slot: shows a release lever on both ends
of the slot, but only one end has the release lever.
Any suggestions on a RAM benchmark tool that will test performance of
each memory module to see if the ones in the A1+B1 slots are faster than
one in the A2 slot? Even a tool, like PassMark's memory benchmark tool (https://www.passmark.com/products/performancetest/pt_advmem.php, could
not find a comparison page between trial and payware versions) looks to
mash together its values, so you couldn't tell with multiple sticks
installed the speed of each stick, if one module was faster, or the 2
modules in matching dual slots were faster than the 1 module in an
unmatched slot.
I suppose I could go with 2 sticks in the matching dual slots (A1 + B1),
run a memory benchmark, and then reinstall the 3rd stick in an unmatched
slot (A2) to rerun the memory benchmark.
I'd like to not lose usable memory modules, but it seems 3 sticks would result in loss of dual-channel model, so I'd have to go down to 2 slots,
and down to 32GB of RAM. The remaining original module would sit inside
an anti-static bag in a drawer. It's possible the warranty module was
bad to start with, but I'm trying to do another warranty replacement.
I'm not spending $400 to get another 4-pc matched 16GB sticks to get
back again to 64GB.
I don't know how I'll survive dropping from 64GB down to 32GB although
RAM consumption is typically only between 5 to 20GB (16GB of which is allocated to a VM when it is running). (smirk)
VanguardLH wrote:
After some testing, I found NONE of the memory modules (original 3, or
the 1 warranty replacement, all the same brand and model) would run
using XMP. Even with just 1 module, none of the modules that claim XMP
support would work.
You sure that those modules really have XMP? Maybe they were incorrectly labelled to fool buyers? Did you buy them from reputable shops?
VanguardLH wrote:
XMP wanted to use 3200 MHz (1600 MHz x 2) for the target DRAM clock.
Auto (JEDEC) uses 2133 Mhz (well, 1066 MHz x 2). The product specs, and
CPU-z, say the max bandwidth for the G.Skill modules is 3200 (1600 MHz),
but none of the modules would run at that clock rate. XMP was out, and
Auto (JEDEC) had to get used. Although that means running at 66% of the
claimed max clock rate, it does run with the slower clock. Claiming XMP
support at 3200 but refusing to boot even to the POST screen is an
unusable setup.
in my Z370 motherboard, I need to *MANUALLY* select a XMP profile. XMP
is NOT activated automatically, even after DDR4 detection is done when
the motherboard POST and boot into Windows 10.
So make sure you activate XMP in the BIOS.
Asrock Z390 Taichi Physical Virtual
------A1--A2 ------A1--A2 Bus
(CPU) Ends
-------------B1--B2 ------B1--B2 Here
With Flex Memory, the sticks in the End Slots (A2+B2) run in dual
channel.
With a third stick in either of A1 or B1, the board still works. The
stick runs in Single Channel. All the sticks use the same timings and
clock choice. The BIOS must adjust the mess, according to the loading
rules.
Note that, when you run "A1 and A2", that's two loads on the bus. And depending on the memory technology, the automatic (JEDEC) clock
selection can be set lower in that case, than if only one DIMM was
per channel. Three DIMMs gives you more RAM, but all the sticks get
scaled by a small loading factor. Now, DDR5 really extracts a large
penalty for this, but DDR4 seems not quite as bad.
It seems a bit like your IMC is weak, and maybe one of the supply
voltages on there is not correct or high enough. On my X48 (a
Northbridge, not an IMC), I needed to bump the voltage feeding the MC
(Memory Controller) block, to bring it back to error free. This can
be called VCCSA, which powers two things.
Since almost no motherboards mount and label "voltage monitoring points",
we cannot do a lap of those and check. The BIOS controls may show the
control that sets the VID value on a chip. But the measured value to
verify it is working, that capability might not be there.
I still think we have to work with what you've got, and adjust things
so one or two sticks are "happy".
Strictly speaking, XMP only works with two sticks. They don't usually
make a profile for two sticks and a profile for four sticks. They
make a profile for two sticks at CR2 (the option that works) and two
sticks at CR1 (which in a lot of cases is too aggressive). On some
boards, the board may insist on going back to JEDEC with four sticks.
I think you are going to have to go into the overclocker section,
and define a clock speed you like for the memory. Then scroll until
you find the voltage section, and make some "tiny" adjustments.
Keep notes as to what you've changed. Run a memtest with the USB stick memtest.
The C: disk drive doesn't have to be connected while you do this. See if
it responds positively to a small bump.
https://download.asrock.com/Manual/Z390%20Taichi.pdf
(Page 70 Section 4.5 OC Tweaker Screen)
Page 71 Performance Mode [disabled]
Page 73 Load XMP Settings [not if you can avoid it at the moment] - no options shown
Page 73 DRAM Clock [1066 for 2133 JEDEC] "applies if ASrock Timing Optimization disabled"
Leave all other memory timings at "Auto", so the BIOS works out the values for you.
Page 80 If you adjust DRAM voltage, stop before it turns red at >1.35V or so.
(Check and see if DRAM Activating Power supply tracks DRAM Voltage at all)
Normally, the enthusiast DDR4 doesn't need boost, even at high clocks.
And your test condition is 2133, boost should not be needed here. It's not the old days.
Page 81 VCCSA (system agent and IMC logic gates)
needs 25mV more voltage [Yellow], try 12.5mv more or 25mv more, then test it.
Take tiny steps. We're compensating for a "worn" IMC rather than "duff RAM".
At least until it comes time to put your "return-ME" RAM into the board :-)
We don't know what is up with that sucker at the moment.
https://www.reddit.com/r/overclocking/comments/v9vb1i/whats_the_difference_between_vdd_imc_and_vccsa/
Your processor is 9th generation, likely a 9900 or so.
We need to find 9th generation-specific values for best advice.
On 19/12/2025 5:10 pm, VanguardLH wrote:
After some testing, I found NONE of the memory modules (original 3, or
the 1 warranty replacement, all the same brand and model) would run
using XMP.ÿ Even with just 1 module, none of the modules that claim XMP
support would work.ÿ I'd set the BIOS to XMP mode, cold boot, and there
was a really long wait until the POST screen appeared; however, it noted
there had been many boot attempts whereupon safe settings got used to
finally get to the POST screen.ÿ If Auto was selected, 1, 2, or 3 of the
original modules would let the computer boot.ÿ The warranty unit was
farked: couldn't boot even with it alone.
You sure that those modules really have XMP? Maybe they were incorrectly labelled to fool buyers? Did you buy them from reputable shops?
Without using the warranty replacement module that generates too many
boot failures, and just involving 2 or 3 of the original memory modules:
2 sticks: A1+A2 = computer boots
A1+B1 = computer boots
A2+B2 = boot fails (stuck in tight boot cycling)
3 sticks: A1+B1 and A2 = computer boots
A1+B1 and B2 (didn't test this config)
None of the above configs would boot with XMP enabled. With XMP
enabled, the boot time was severely elongated, and if the POST screen
showed up it said something about too many boot retries, and resetting
to defaults (so XMP got turned back to Auto) after which another reboot brought up the computer.
failure. Possibly the CMOS table was corrupt due to a weak battery, so that's been replaced. When CMOS was cleared, the computer booted okay,
but the clear reset the BIOS back to Auto (JEDEC), the default, instead
of using XMP. XMP is not the default, but it worked for a long time
before when I changed to XMP.
Now I have to ensure XMP is *not* enabled since none of the memory
modules (3 originals, 1 warranty replacement) work with XMP, anymore. A slower memory clock with a bootable computer is far better than a
computer that won't boot.
On 20/12/2025 12:07 am, VanguardLH wrote:
failure. Possibly the CMOS table was corrupt due to a weak battery, so
that's been replaced. When CMOS was cleared, the computer booted okay,
but the clear reset the BIOS back to Auto (JEDEC), the default, instead
of using XMP. XMP is not the default, but it worked for a long time
before when I changed to XMP.
Now I have to ensure XMP is *not* enabled since none of the memory
modules (3 originals, 1 warranty replacement) work with XMP, anymore. A
slower memory clock with a bootable computer is far better than a
computer that won't boot.
That's really strange then for a motherboard suddenly lost XMP
capability. Could it be overheating?
Plus, when testing, the computer hasn't ran long enough to even get
warm. Remember when booting fails, like with the failed original module
or with the warranty replacement that is also bad, that it gets stuck in
a tight cycling that doesn't even let me go to BIOS settings. With XMP,
boot is severely prolonged, and then the boot screen says defaults were
used after many failed boot attempts.
VanguardLH wrote:
Plus, when testing, the computer hasn't ran long enough to even get
warm. Remember when booting fails, like with the failed original module
or with the warranty replacement that is also bad, that it gets stuck in
a tight cycling that doesn't even let me go to BIOS settings. With XMP,
boot is severely prolonged, and then the boot screen says defaults were
used after many failed boot attempts.
Once getting all memory modules working without XMP, you can overclock
DDR4 manually by adjusting frequency and voltage in small steps. XMP is
not the only way to overclock DDR4. But this method is more
time-consuming than selecting XMP profiles, though you got finer control.
Asrock Taichi 390 xmp fail - Google Serach <https://www.google.com/search?q=Asrock+Taichi+390+xmp+fail>
Asrock Taichi 390 overclock ram - Google Search <https://www.google.com/search?q=Asrock+Taichi+390+overclock+ram>
I'll be testing again later while ensuring XMP is disabled. Other than picking XMP (which selects an XMP 2.0 profile, the only one), I'll make
sure it is set to Auto which apparently has the BIOS use the SPD
reported by the memory modules (i.e., their rated specs without XMP).
On 22/12/2025 11:35 am, VanguardLH wrote:
I'll be testing again later while ensuring XMP is disabled.ÿ Other than
picking XMP (which selects an XMP 2.0 profile, the only one), I'll make
sure it is set to Auto which apparently has the BIOS use the SPD
reported by the memory modules (i.e., their rated specs without XMP).
Just wanna remind you of the good, old way of overclocking. You do NOT have to rely on XMP.
On Mon, 12/22/2025 9:24 AM, Mr. Man-wai Chang wrote:
On 22/12/2025 11:35 am, VanguardLH wrote:
I'll be testing again later while ensuring XMP is disabled.ÿ Other than
picking XMP (which selects an XMP 2.0 profile, the only one), I'll make
sure it is set to Auto which apparently has the BIOS use the SPD
reported by the memory modules (i.e., their rated specs without XMP).
Just wanna remind you of the good, old way of overclocking. You do NOT have to rely on XMP.
I do not know if "good" is the right word for it.
The last time I did this on the 4930K, it took a *week*
to tune it. Each Memtest takes two hours, depending on how
close it is to being the correct settings.
I would not class the exercise as "exciting" or "satisfying".
It's dog work.
My DDR3-2400, when I was finished it was at DDR3-1866.
4 sticks XMP DDR3-2400 (settings as on the box)
8 sticks no-XMP DDR3-1866 (pretty close to the JEDEC values)
(No excessive VDimm boost)
That's a four channel machine (the extra channels largely being
a waste). All that the extra channels do, is allow a higher
total RAM capacity on the machine. The four channels do not read
at the same time, for example. It responds as if it was a dual channel machine.
But that's not why I bought it. I bought it for (2) x16 real slots for PCIe. A slot for a video card. A slot for "toys" to be added. For example, you could fit an NVMe RAID in the second slot (no, I'm not doing that!).
Paul
At this point, I'll either ensure the BIOS is set to Auto, and *not* to
XMP, to see if I get all 4 modules working by passing memtest96, and
watching for errors in programs, like when Macrium Reflect was failing
the verification step for backups....
On 23/12/2025 3:55 am, VanguardLH wrote:
At this point, I'll either ensure the BIOS is set to Auto, and *not* to
XMP, to see if I get all 4 modules working by passing memtest96, and
watching for errors in programs, like when Macrium Reflect was failing
the verification step for backups....
What exactly are your G.Skill DDR4 rated?
The DDR4-3200 (XMP) I bought have a baseline frequence of 2400MHz. Yours (2133MHz) are definitely a bit slow.
Mr. Man-wai Chang wrote:
VanguardLH wrote:
At this point, I'll either ensure the BIOS is set to Auto, and *not*
to XMP, to see if I get all 4 modules working by passing memtest96,
and watching for errors in programs, like when Macrium Reflect was
failing the verification step for backups....
What exactly are your G.Skill DDR4 rated?
The DDR4-3200 (XMP) I bought have a baseline frequence of 2400MHz.
Yours (2133MHz) are definitely a bit slow.
Sorry, missed the link you posted!! Shouldn't they have a baseline speed
of 2400MHz?
G.SKILL Ripjaws V Series 64GB (4 x 16GB) 288-Pin PC RAM DDR4 3200 (PC4 25600) Desktop Memory Model F4-3200C16Q-64GVK <https://www.newegg.com/g-skill-ripjaws-v-series-64gb-ddr4-3200-cas-latency-cl16-desktop-memory-black/p/N82E16820232092?Item=N82E16820232092>
On 23/12/2025 5:47 pm, Mr. Man-wai Chang wrote:
On 23/12/2025 3:55 am, VanguardLH wrote:
At this point, I'll either ensure the BIOS is set to Auto, and *not* to
XMP, to see if I get all 4 modules working by passing memtest96, and
watching for errors in programs, like when Macrium Reflect was failing
the verification step for backups....
What exactly are your G.Skill DDR4 rated?
The DDR4-3200 (XMP) I bought have a baseline frequence of 2400MHz. Yours
(2133MHz) are definitely a bit slow.
Sorry, missed the link you posted!! Shouldn't they have a baseline speed of 2400MHz?
G.SKILL Ripjaws V Series 64GB (4 x 16GB) 288-Pin PC RAM DDR4 3200 (PC4 25600) Desktop Memory Model F4-3200C16Q-64GVK
<https://www.newegg.com/g-skill-ripjaws-v-series-64gb-ddr4-3200-cas-latency-cl16-desktop-memory-black/p/N82E16820232092?Item=N82E16820232092>
They're the same as the ones I've got.
2133 as slowest JEDEC.
3200 as XMP rate DDR4-3200 (1600MHz)
On 24/12/2025 1:24 am, Paul wrote:
They're the same as the ones I've got.
2133 as slowest JEDEC.
3200 as XMP rateÿÿÿÿÿÿ DDR4-3200 (1600MHz)
Late reply:
I checked my DDR4-3200 moduels. Their JEDEC (#8) is rated 2400, as reported by CPU-Z, if I didn't misunderstand what you called "slowest JEDEC".
Following is the relevant section from CPU-Z:
DIMM #ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ 1
ÿÿÿÿSMBus addressÿÿÿÿÿÿÿ 0x50
ÿÿÿÿMemory typeÿÿÿÿÿÿÿ DDR4
ÿÿÿÿModule formatÿÿÿÿÿÿÿ UDIMM
ÿÿÿÿ.....
ÿÿÿÿSizeÿÿÿÿÿÿÿÿÿÿÿ 8192 MBytes
ÿÿÿÿMax bandwidthÿÿÿÿÿÿÿ DDR4-3200 (1600 MHz)
ÿÿÿÿMax JEDECÿÿÿÿÿÿÿ DDR4-2400 (1200 MHz)
ÿÿÿÿ....
ÿÿÿÿNominal Voltageÿÿÿÿÿÿÿ 1.20 Volts
ÿÿÿÿ....
JEDEC timings tableÿÿÿÿÿÿÿ CL-tRCD-tRP-tRAS-tRC @ frequency
ÿÿÿÿJEDEC #1ÿÿÿÿÿÿÿ 9.0-9-9-22-31 @ 666 MHz
ÿÿÿÿJEDEC #2ÿÿÿÿÿÿÿ 10.0-11-11-25-35 @ 766 MHz
ÿÿÿÿJEDEC #3ÿÿÿÿÿÿÿ 11.0-12-12-27-38 @ 833 MHz
ÿÿÿÿJEDEC #4ÿÿÿÿÿÿÿ 12.0-12-12-29-41 @ 900 MHz
ÿÿÿÿJEDEC #5ÿÿÿÿÿÿÿ 13.0-13-13-31-44 @ 966 MHz
ÿÿÿÿJEDEC #6ÿÿÿÿÿÿÿ 14.0-15-15-35-49 @ 1066 MHz
ÿÿÿÿJEDEC #7ÿÿÿÿÿÿÿ 15.0-16-16-37-52 @ 1133 MHz
ÿÿÿÿJEDEC #8ÿÿÿÿÿÿÿ 16.0-16-16-39-55 @ 1200 MHz
ÿÿÿÿJEDEC #9ÿÿÿÿÿÿÿ 18.0-16-16-39-55 @ 1200 MHz
ÿÿÿÿJEDEC #10ÿÿÿÿÿÿÿ 20.0-16-16-39-55 @ 1200 MHz
XMP profileÿÿÿÿÿÿÿÿÿÿÿ XMP-3200
ÿÿÿÿSpecificationÿÿÿÿÿÿÿ DDR4-3200
ÿÿÿÿVDD Voltageÿÿÿÿÿÿÿ 1.350 Volts
ÿÿÿÿ.....
| Sysop: | Jacob Catayoc |
|---|---|
| Location: | Pasay City, Metro Manila, Philippines |
| Users: | 5 |
| Nodes: | 4 (0 / 4) |
| Uptime: | 22:23:38 |
| Calls: | 117 |
| Calls today: | 117 |
| Files: | 367 |
| D/L today: |
560 files (257M bytes) |
| Messages: | 70,893 |
| Posted today: | 26 |
