Newyana2 <Newyana2@invalid.nospam> wrote:

> The bluescreen tool is interesting. I ran it. It found
> a memory dump from the crash and says it was set off
> by a realtek (audio) driver. I'm not sure that really tells
> me anything.
>
> The only option was a 256 KB dump file, but the bluescreen
> reader found whatever was there, anyway.

Just make sure it was a recent dump file created when you encountered
the BSOD while using the Win32 Disk Imager program. By default,
minidump files are not deleted, so it could be an old one you're looking
at in Nirsoft's Bluescreen View.

On 2/12/2024 4:18 PM, Newyana2 wrote:

> I built it.
>
> Intel 1-5 12400 6-Core Boxed Processor with fan $150
> MSI B760M-P Pro Intel LGA 1700 microATX Motherboard $100
> G.Skill Ripjaws V 16GB (2 x 8GB) DDR4-3200 PC4-25600 CL16 Dual Channel RAM
> $40
> Thermaltake Smart Series 700 Watt 80 Plus $60
>
> I had a case and SATA SSDs. But I also bought an M2 SSD,
> since the board accommodates it.

The processor is 6C 12T, and the new version of memtest
that has multicore option, uses the physical cores (6 of them)
rather than hyperthreading (12 of them). There are six cores,
with dual register banks, and hyperthreading merely switches
between banks (within the same core). They're not really extra
cores as such.

On page 45 here, you want XMP. This is the BIOS manual that
is supposed to be associated with your board.

https://download.msi.com/archive/mnu_exe/mb/Intel700BIOS.pdf

Since the manual doesn't have the screen pictures I might want,
I would expect something like this. I had to figure this out
on my own for my builds, since it's not in the manual as to
what trick they are pulling. I initially selected Profile 2 for
some reason. Profile 1 should be more stable.

(Tick this one)
XMP Profile 1 Profile 2

Command Rate 2 Command Rate 1

That will run your RAM at XMP DDR4-3200. It will set the RAM voltage
at 1.35V instead of 1.2V. Your memory passed test at the factory,
DDR4-3200, while at 1.35V. The thing about XMP, is the memory
voltage to use, is recorded in the SPD chip.

If you tested the two profiles, and looked at the Memtest
speed thing in the upper left hand corner,
Profile 1 gives 38GB/sec while Profile 2 gives 40GB/sec.

Profile 2, gives less Tsu to the memory chips. Surprisingly,
when I tested that on a system with four DIMMs installed,
it worked. When it should not really have worked. Normally,
high speed memory is supposed to do better with Profile 1
and a whole clock cycle of extra setup time for control/address
for example.

(Zip-English portable version)

https://www.cpuid.com/softwares/cpu-z.html

Once the OS is running, after you've set XMP profile, you
can verify what timing is being used, with CPU-Z.

*******

In this picture, is an older system. You can see the Command Rate is "2T"
in the middle one. That's my dead WinXP system.

https://i.postimg.cc/NG6s4MDn/CPUZ.gif

This is a DDR4 system, but not the main panel that lists a 2T.

https://i.postimg.cc/FR37Dh5r/My-DDR4-XMP-switched-on.gif

This is the AMD 5600G with DDR4 and a Command Rate 2T

https://i.postimg.cc/59vL8Rh5/CPUZ-before-5600-G.gif

# And with the internal GPU in use

https://i.postimg.cc/BZYR3B7x/cpuz.gif

Now, back in DDR3 era, it seems the memory had two actual profiles
where one shows 2T command rate, and the more aggressive one is 1T command rate.

Whereas on the BIOS of today, the memory seems to have just one profile
where the Command Rate is not specified, and the BIOS offers "two profiles"
where it inserts 2T (conservative) in Profile #1 and 1T (aggressive) in Profile #2.
So this picture shows how it used to work, with an actual profile for
each in the SPD chip. As a bonus, the bottom of this shows a picture from
my poor ole laptop, and the laptop memory has no XMP table in it.

https://i.postimg.cc/3JzgJJzz/XMP-Reg-Table-Memory.gif

Paul

"Paul" <nospam@needed.invalid> wrote

| On a four slot motherboard, you can test two sticks in single channel
mode.
| | channel0 channel1
| | |
| white white <=== the color of the slots, is to emphasize
how to do dual channel
| | |
| black black

Let me see if I understand. My manual says to use A2 for 1 stick,
A2/B2 for two sticks, or 4 sticks. You're saying that if I use
A1/B1 it works but in single channel mode? And a single stick
will work in any slot? Why, then, do they say A2 for a single
stick? This is the first I've dealt with dual-channel memory.
But it is dual channel, so I have it in A2/B2. To test I took out
B2. Then when I put it back it was OK.

What if I wanted to use 3 sticks? Does that work?
A2/B2 as dual channel with A1 added? I'm not thinking
of doing that. I just wondered, if I happened
to have one spare stick at some point.

I think that for now I'll see how it goes. With zero errors
on my final test using A2/B2, it seems unlikely that there's
a problem.

I did use the MS test first, which also found errors, but
it had nothing to say about where or what kind, so that
wasn't much use.

"VanguardLH" <V@nguard.LH> wrote

| > I built it.
| >
| > Intel 1-5 12400 6-Core Boxed Processor with fan $150
| > MSI B760M-P Pro Intel LGA 1700 microATX Motherboard $100
| > G.Skill Ripjaws V 16GB (2 x 8GB) DDR4-3200 PC4-25600 CL16 Dual Channel
RAM
| > $40
| > Thermaltake Smart Series 700 Watt 80 Plus $60
| | Mostly interested if you got driver updates via WU, or you got them from
| the hardware maker. You said the latter, but did you disable hardware
| updates in WU?

I disable all of Windows Update functionality once I get it set up.
With Simplewall I don't even let anything call out. But it's good to know
that it might sneak in some crap. I don't remember noticing hardware
updates options in Winaero.

I went to MSI and got the drivers. I did have a problem recently
with my laptop. I wanted to change display res and display properties
told me it was a generic MS driver. It only had one res option! It
actually took a fair amount of research to track down the Intel
driver that would work.

"VanguardLH" <V@nguard.LH> wrote

| > One unusual thing with Memtest: It said there are 12 cores, with 6
| > active. Is it possible that I actually bought a much higher powered
| > CPU and they've crippled it to sell it down-market?
| | No, you got a 6-core CPU with hypertheading enabled. Think of
| hyperthreading like the interpolated resolution you see advertised for
| scanners to bloat their specs.
| Ah. OK. As long as I'm not missing anything. :)

On 2/12/2024 7:46 PM, Newyana2 wrote:
> "Paul" <nospam@needed.invalid> wrote
>
> | On a four slot motherboard, you can test two sticks in single channel
> mode.
> |
> | channel0 channel1
> | | |
> | white white <=== the color of the slots, is to emphasize
> how to do dual channel
> | | |
> | black black
>
> Let me see if I understand. My manual says to use A2 for 1 stick,
> A2/B2 for two sticks, or 4 sticks. You're saying that if I use
> A1/B1 it works but in single channel mode? And a single stick
> will work in any slot? Why, then, do they say A2 for a single
> stick? This is the first I've dealt with dual-channel memory.
> But it is dual channel, so I have it in A2/B2. To test I took out
> B2. Then when I put it back it was OK.
>
> What if I wanted to use 3 sticks? Does that work?
> A2/B2 as dual channel with A1 added? I'm not thinking
> of doing that. I just wondered, if I happened
> to have one spare stick at some point.
>
> I think that for now I'll see how it goes. With zero errors
> on my final test using A2/B2, it seems unlikely that there's
> a problem.
>
> I did use the MS test first, which also found errors, but
> it had nothing to say about where or what kind, so that
> wasn't much use.
>
>

https://download.msi.com/archive/mnu_exe/mb/PROB760M-PDDR4.pdf

Page 25 A1---A2 B1---B2 Physical

CPU --- A1---A2 Logical
--- B1---B2
^
|
+---- Fill from bus end, towards the CPU

For single sticks alone, A2 or B2 should work.

For a pair of sticks, A2+B2 = dual channel

A second pair of sticks for dual channel addition to A2+B2, is A1+B1.

Whereas B1+B2 is single channel, good for a memtest using two sticks.

You fill from the end of the bus. A2 and B2 are filled
first. And only if they are filled, do you think about
using A1 and B1. That's the basic idea.

The bus stub length, of A2 hanging off the end of A1 is short.
It depends on the signal rise time, whether the A2 stub is considered
a "transmission line" or a "bulk piece of wire". If you looked
at the eye opening on a digital storage scope running in accumulate
mode, you would see a difference in eye opening.

But we try to do the right thing anyway, because the noise
floor is hard to measure when computers don't have ECC to
measure this for us. If you had a working ECC, you could test
A1 by itself, and see if you could detect it was inferior or not.

*******

If I do this

A2 B2
X----8GB 8GB----8GB

the result is 24GB detected, A2+B2 run dual channel, B1 runs single channel.
Not all of the memory space runs at the same speed.

But this hardly makes a difference to how "snappy" the system feels.
The CPU cache, for most normal applications, has a high hit rate,
and so the memory contribution to performance is muted.

Paul

"Paul" <nospam@needed.invalid> wrote

| https://download.msi.com/archive/mnu_exe/mb/PROB760M-PDDR4.pdf
|

Thanks. I have that manual, but I did download the BIOS
manual, which I hadn't thought to look for. I set the XMP
and went from 2100 to 3200. Crazy stuff. I see that I'm
behind the times.

On 2/12/2024 7:54 PM, Newyana2 wrote:
> "VanguardLH" <V@nguard.LH> wrote
>
> | > One unusual thing with Memtest: It said there are 12 cores, with 6
> | > active. Is it possible that I actually bought a much higher powered
> | > CPU and they've crippled it to sell it down-market?
> |
> | No, you got a 6-core CPU with hypertheading enabled. Think of
> | hyperthreading like the interpolated resolution you see advertised for
> | scanners to bloat their specs.
> |
> Ah. OK. As long as I'm not missing anything. :)

Register-Bank-#1 Register-Bank=#2
| |
+---------+---------+
|
CPU core o-o-o (Out Of Order execution)
|

When the CPU makes a memory fetch, it can have one outstanding
memory operation before it blocks.

If it blocks, then without Hyperthreading, it would be sitting
on its hands, and waiting for the memory subsystem to come back
with the data. The time-to-come-back, could be 50 to 100 cycle
times. That's time that could be spent doing something.

Instead, it flips the bank switch, and loads the Program Counter
from the second bank, and tries to work on the code there. It is
executing two threads, and switching threads when memory blocks.
Sometimes the thread in the second bank, its memory fetch has come
back and is sitting in the cache (and hasn't been evicted yet).

This means the performance improvement from this "wart off the side",
you do not expect it to ever double the performance of the core logic.
Like, if I put five banks of registers on that core, it would
not be a bit different than running with two. All the extra effort has
already been harvested.

There have been multiple generations of the thing, and the above
description hardly does justice to the variants. the very first
HT had a bug in it.... and nobody noticed. That's because the
recovery time was a couple hundred microseconds.

With the wealth of resources in the processor, you may not even
notice that the scheduler is not doing the right thing, with the
resources. What is interesting, is comparing fan noise when web
surfing in Windows, versus web surfing in Linux, using Firefox
in both cases and visiting the same web site. The fan in Linux
is quieter.

Your CPU doesn't have P+E cores. It has only P cores, and
it is P that has the HT in it. This means you have fewer things
to fiddle with, and "see whether they help or not" :-) Smartphones
had the P and E core concept, and that's spread to the Intel arena.

Mine is the same. All my processors have only one core type inside.
I've never had a P+E flavor of processor.

Paul

"Paul" <nospam@needed.invalid> wrote

| Register-Bank-#1 Register-Bank=#2
| | |
| +---------+---------+
| |
| CPU core o-o-o (Out Of Order execution)
| |
| | When the CPU makes a memory fetch, it can have one outstanding
| memory operation before it blocks.
| | If it blocks, then without Hyperthreading, it would be sitting
| on its hands, and waiting for the memory subsystem to come back
| with the data. The time-to-come-back, could be 50 to 100 cycle
| times. That's time that could be spent doing something.
| | Instead, it flips the bank switch, and loads the Program Counter
| from the second bank, and tries to work on the code there. It is
| executing two threads, and switching threads when memory blocks.
| Sometimes the thread in the second bank, its memory fetch has come
| back and is sitting in the cache (and hasn't been evicted yet).
|

I think I understand that. The dishwasher gets one slice
of time, in the system's looping workload slices allotted to various
processes, to receive dishes to wash, but he has two sinks to work
out of. I was confused because while CPUID says there are 6 cores
and 12 threads, Memtest says there are 12 cores, 6 active.

If it is taking over an hour to copy 1 gig, you probably have a h/w problem

"philo" <philo@www.novabbs.com> wrote

| If it is taking over an hour to copy 1 gig, you probably have a h/w
problem

I thought that was odd, too, but it seems to be something
about how the specific RPi image program copies and writes
to the SD card. It told me to expect a long time, and I don't
have any problems otherwise. Everything is extremely fast.

I ended up reading the image and writing to two new SD
cards. I then tested one of those in the RPi4 and it booted
without issues.

On 2/13/2024 9:39 AM, Newyana2 wrote:
> "Paul" <nospam@needed.invalid> wrote
>
> | https://download.msi.com/archive/mnu_exe/mb/PROB760M-PDDR4.pdf
> |
>
> Thanks. I have that manual, but I did download the BIOS
> manual, which I hadn't thought to look for. I set the XMP
> and went from 2100 to 3200. Crazy stuff. I see that I'm
> behind the times.
>
>

I couldn't use the XMP on the Test Machine (4930K),
but on the new build, that's a great feature. It
exceeded expectations (it can do things you would not
predict would work -- like running CR 1T with four
DIMMs, that should not be possible).

Yes, hitting that switch, it takes care of the settings
for you, and in this generation, "it just works". Which
is a relief compared to the hell on the ten year old machine.
I ended up doing manual tuning and it took all week.

It's a real labor saver.

Paul

Newyana2 <Newyana2@invalid.nospam> wrote:

> "Paul" <nospam@needed.invalid> wrote
>
>| https://download.msi.com/archive/mnu_exe/mb/PROB760M-PDDR4.pdf
>
> Thanks. I have that manual, but I did download the BIOS
> manual, which I hadn't thought to look for. I set the XMP
> and went from 2100 to 3200. Crazy stuff. I see that I'm
> behind the times.

When I asked if you were overclocking, you said "not doing any fancy
business" and "no use for overclocking". But now you're playing with
XMP (Intel Extreme Memory Profile).

https://www.intel.com/content/www/us/en/gaming/extreme-memory-profile-xmp.html
https://www.pcgamer.com/what-are-xmp-profiles-and-how-do-i-use-them/

In my mobo's manual regarding XMP, it says:

Load XMP Setting
Load XMP settings to overclock the memory and perform beyond standard
specifications.

Hopefully I disabled that setting. I want a reliable and stable
computer. The only time I'd see something of overclocking would be in
video games where the framerate would be higher, but it's already high
enough that an even higher rate won't be noticed. However, I don't play
super-fast super-high graphics super-complex texture video games.

Once you decide to overclock the memory, you better get a utility to
measure temperatures. And your memory modules should have heatsinks
(real heatsinks, not pretty label placards). Also make sure no cabling
is blocking or diverting airflow over the memory modules. You might
need to add more fans than what came with the case. Heat and longevity
for electronics are the anti-thesis of each other. Overclock = more
heat = shorter life. I let my mobo read the SPD from the memory
modules, and use those settings published as the ratings for them.

As one of the articles mention, run CPU-z to look at its SPD tab (spec
table recorded in memory module), and check XMP 2 or 3 is listed for the
SPD extensions of your memory modules. Click the down-arrow to select
each populated mobo slot to verify each module is XMP rated. I always
buy memory in sets, so they are matched. I populate each memory slot on
the mobo to get the max system memory available that the mobo supports.
Currently I have 64 GB. I would hope the BIOS firmware would check the
SPD specs of all memory modules to ensure all of them support XMP.

On 2/14/2024 4:30 AM, VanguardLH wrote:
> Newyana2 <Newyana2@invalid.nospam> wrote:
>
>> "Paul" <nospam@needed.invalid> wrote
>>
>> | https://download.msi.com/archive/mnu_exe/mb/PROB760M-PDDR4.pdf
>>
>> Thanks. I have that manual, but I did download the BIOS
>> manual, which I hadn't thought to look for. I set the XMP
>> and went from 2100 to 3200. Crazy stuff. I see that I'm
>> behind the times.
>
> When I asked if you were overclocking, you said "not doing any fancy
> business" and "no use for overclocking". But now you're playing with
> XMP (Intel Extreme Memory Profile).
>
> https://www.intel.com/content/www/us/en/gaming/extreme-memory-profile-xmp.html
> https://www.pcgamer.com/what-are-xmp-profiles-and-how-do-i-use-them/
>
> In my mobo's manual regarding XMP, it says:
>
> Load XMP Setting
> Load XMP settings to overclock the memory and perform beyond standard
> specifications.
>
> Hopefully I disabled that setting. I want a reliable and stable
> computer.

You don't know what you're missing.

1) The RAM is tested for that speed, at the factory.

2) The official information for the product (from Intel or AMD),
states what is supported. For DDR4, it is 3200 on one platform,
and 3600 on the other.

If you go too much higher than the CPU maker stated-spec, then
the gear changer engages, and the internal clock is turned down
by a factor of 2 or a factor of 4. You don't really want to go that
far. This is related to the IMC clock, which is 1/2 the DDR rate
or 1600 MHz. Any value of that clock, up to that value is allowed,
without the gear changer moving off 1:1 and spoiling things.

JEDEC has a spec they had in mind for the family.

However, other elements in the industry are willing to certify
operation at the (relatively wimpy) 3200 and 3600 values.

My platform could use 3600, but I selected 3200 when I bought it,
because that could be run either on an Intel system or on an
AMD system (with XMP engaged).

There seems to be plenty of margin, based on accidentally selecting
Command Rate 1T with four sticks... and it worked like it was nothing.
Previous generations, you might have to ease off on tRAS by one unit,
plus use Command Rate 2T, to get it to work. No such problem with DDR4.

And since Maya is using just two sticks (1 DIMM per channel), this
should be dead easy for XMP. Even more margin should be present.
You could drive a truck through the timing. The SI should be excellent.

I really wish I had an ECC on this thing, so I could show you it
really is zero. The memory is clean.

This isn't DDR400 and this isn't DDR2-800 either.
It's a *lot* better than that.

Paul

VanguardLH wrote:

> When I asked if you were overclocking, you said "not doing any fancy
> business" and "no use for overclocking". But now you're playing with
> XMP (Intel Extreme Memory Profile).

But it's officially sanctioned overclocking, I don't remember eer having
an issue using XMP ... manually setting the timings OTOH ...

"VanguardLH" <V@nguard.LH> wrote

|
https://www.intel.com/content/www/us/en/gaming/extreme-memory-profile-xmp.html
| https://www.pcgamer.com/what-are-xmp-profiles-and-how-do-i-use-them/
| | In my mobo's manual regarding XMP, it says:
| | Load XMP Setting
| Load XMP settings to overclock the memory and perform beyond standard
| specifications.
|

As Paul explained, it's not actually overclocking. I looked
it up myself and found the same kinds of statements. It's
a default setting of reduced speed (in my case 2/3 speed)
in order to make sure it works on a wide variety of hardware.

I don't really understand all these details, but on my XP box
I have DDR3 rated at 1866. CPU-Z says it's 938. But in the
same box on Win10 it also lists half speed. So I'm guessing that
the XP RAM is running at full speed. I then booted into the XP
BIOS and switched the RAM clock speed from Auto to the top
option of 1867, rebooted, and CPU-Z now shows the same
numbers. So it seems that putting a governor on RAM is a
new thing. Like giving you a half-size washing machine tub
so that the official ratings on the motor life will look better.

On my new Win10 box it was running the RAM rated for
3200 at 2100. (Win10 also lets you confirm speed in the task
manager performance window. XP doesn't show that. So I'm
just going by CPU-Z and the BIOS setting say.)

On 2/14/2024 8:26 AM, Newyana2 wrote:
> "VanguardLH" <V@nguard.LH> wrote
>
> |
> https://www.intel.com/content/www/us/en/gaming/extreme-memory-profile-xmp.html
> | https://www.pcgamer.com/what-are-xmp-profiles-and-how-do-i-use-them/
> |
> | In my mobo's manual regarding XMP, it says:
> |
> | Load XMP Setting
> | Load XMP settings to overclock the memory and perform beyond standard
> | specifications.
> |
>
> As Paul explained, it's not actually overclocking. I looked
> it up myself and found the same kinds of statements. It's
> a default setting of reduced speed (in my case 2/3 speed)
> in order to make sure it works on a wide variety of hardware.
>
> I don't really understand all these details, but on my XP box
> I have DDR3 rated at 1866. CPU-Z says it's 938. But in the
> same box on Win10 it also lists half speed. So I'm guessing that
> the XP RAM is running at full speed. I then booted into the XP
> BIOS and switched the RAM clock speed from Auto to the top
> option of 1867, rebooted, and CPU-Z now shows the same
> numbers. So it seems that putting a governor on RAM is a
> new thing. Like giving you a half-size washing machine tub
> so that the official ratings on the motor life will look better.
>
> On my new Win10 box it was running the RAM rated for
> 3200 at 2100. (Win10 also lets you confirm speed in the task
> manager performance window. XP doesn't show that. So I'm
> just going by CPU-Z and the BIOS setting say.)
>
XMP RAM has a relatively large JEDEC table.

The bottom end of the JEDEC table is to ensure the machine
runs OK, as "OutOfBoxExperience". This is to ensure the
user can get into the machine.

If the user gets into trouble, and messed up the settings,
there are various ways to "reset CMOS", including cycling
the power on the machine three times. Some BIOS are designed
to recover that way. Once the CMOS is reset, the DRAM speed
drops to some entry in the JEDEC table, and the user can get
back into the BIOS. Since the machine was running in XMP mode
about 30 seconds after the lights came on, I don't even know
what the JEDEC default is :-) Pretty funny. Paul was in a rush
for Glory.

Part number F4-3200C16-32GVK
Nominal Voltage 1.20 Volts
EPP no
XMP yes, rev. 2.0
AMP no
JEDEC timings table CL-tRCD-tRP-tRAS-tRC @ frequency
JEDEC #1 10.0-11-11-24-34 @ 733 MHz <=== DDR4-1466
JEDEC #2 11.0-11-11-26-37 @ 800 MHz
JEDEC #3 12.0-12-12-28-40 @ 866 MHz
JEDEC #4 13.0-13-13-30-43 @ 933 MHz
JEDEC #5 14.0-15-15-34-48 @ 1033 MHz
JEDEC #6 15.0-16-16-36-51 @ 1100 MHz
JEDEC #7 16.0-17-17-38-54 @ 1166 MHz
JEDEC #8 17.0-17-17-40-57 @ 1233 MHz
JEDEC #9 18.0-18-18-42-60 @ 1300 MHz
JEDEC #10 19.0-19-19-43-61 @ 1333 MHz
JEDEC #11 20.0-19-19-43-61 @ 1333 MHz <=== DDR4-2666
XMP profile XMP-3200
Specification DDR4-3200 <=== DDR4-3200
VDD Voltage 1.350 Volts <=== at a specified voltage
Min Cycle time 0.625 ns (1600 MHz)
Max CL 16.0
Min tRP 11.25 ns
Min tRCD 11.25 ns
Min tRAS 23.75 ns
Min tRC 35.00 ns
Min tRRD 2.50 ns
XMP timings table CL-tRCD-tRP-tRAS-tRC-CR @ frequency (voltage)
XMP #1 16.0-18-18-38-56-n.a @ 1600 MHz (1.350 Volts)
^^^
Notice Command Rate is neither 1T nor 2T,
and the BIOS has a Profile defined for each.
Profile 1 uses 2T. Profile 2 (aggressive) uses 1T.

Some RAM have had two profiles, and in some cases,
they were the same set of values. This is the RAM in the
Test Machine, and it spells out two XMP profiles in gory detail,
with Command Rate 2T on the first profile, Command Rate 1T (aggressive)
on the second profile. This uses a different revision of the XMP spec.
This is also only intended for dual channel A2-B2 as a configuration,
rather than being a four stick configuration. The chipset on the
test machine, has a large number of IO on it and to control SSO, the
drive on the memory bus can't be quite as aggressive as the machine above uses.
That's why CR 2T should pretty well be used at all times on this generation
of equipment. To give more Setup time (Tsu) to clock edge.

Part number F3-2400C10-8GTX
Number of banks 8
Nominal Voltage 1.50 Volts
EPP no
XMP yes, rev. 1.3
AMP no
JEDEC timings table CL-tRCD-tRP-tRAS-tRC @ frequency
JEDEC #1 5.0-5-5-14-19 @ 380 MHz
JEDEC #2 6.0-6-6-17-23 @ 457 MHz
JEDEC #3 7.0-7-7-20-27 @ 533 MHz
JEDEC #4 8.0-8-8-22-30 @ 609 MHz
JEDEC #5 9.0-9-9-24-33 @ 666 MHz
JEDEC #6 10.0-9-9-24-33 @ 666 MHz <=== DDR3-1333 is the default startup speed, first run
XMP profile XMP-2400 Any time the BIOS resets itself, I end you at DDR3-1333
Specification PC3-19200
VDD Voltage 1.650 Volts
Min Cycle time 0.833 ns (1200 MHz)
Max CL 10.0
Min tRP 9.38 ns
Min tRCD 9.38 ns
Min tWR 13.25 ns
Min tRAS 25.38 ns
Min tRC 35.31 ns
Min tRFC 260.00 ns
Min tRTP 7.50 ns
Min tRRD 5.00 ns
Command Rate 2T
XMP timings table CL-tRCD-tRP-tRAS-tRC-CR @ frequency (voltage)
XMP #1 10.0-12-12-31-43-2T @ 1200 MHz (1.650 Volts) <=== CAS 10, CR=2T
XMP profile XMP-2400
Specification PC3-19200
VDD Voltage 1.650 Volts
Min Cycle time 0.833 ns (1200 MHz)
Max CL 10.0
Min tRP 9.69 ns
Min tRCD 9.69 ns
Min tWR 13.33 ns
Min tRAS 25.52 ns
Min tRC 35.52 ns
Min tRFC 260.42 ns
Min tRTP 7.50 ns
Min tRRD 5.00 ns
Command Rate 1T
XMP timings table CL-tRCD-tRP-tRAS-tRC-CR @ frequency (voltage)
XMP #1 10.0-12-12-31-43-1T @ 1200 MHz (1.650 Volts) <=== CAS 10, CR=1T (this will crash)

Paul