Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Memory won't run at 2400?

356 views
Skip to first unread message

John Doe

unread,
Feb 13, 2017, 1:56:17 AM2/13/17
to
Motherboard, Gigabyte (GA-B250M-DS3H)
CPU, Intel Celeron G3930
RAM, Crucial Ballistix Sport LT DDR4-2400 (BLS8G4D240FSB)

All of those are supposed to facilitate memory at 2400. But
seems settings will not allow it.

https://www.flickr.com/photos/27532210@N04/?

VanguardLH

unread,
Feb 13, 2017, 2:35:07 AM2/13/17
to
One of the reasons, well probably the biggest reason, not to use the
flickr site to host photos is that they will mix yours in with others.
I was wondering what bike parts had to do with your memory timings.

Did you customize the memory settings or let the BIOS/UEFI pick them
from the SPD on the memory module(s)?

Are ALL the memory modules matched? They don't have to be purchased in
a matched set but you should get them all the same. Easiest way is to
buy them at the same time for the same brand and model. Otherwise,
there is a chance that you mixed high- and low-density modules or other
architecture differs between the modules. If you mixed slower and
faster modules together, make sure to put the slower ones in the first
slot (or first two paired for dual mode) since the BIOS is likely to
read those to get the SPD settings that apply to all the modules. If
you put the fastest in the first slot(s) then the BIOS would use the
fast SPD settings for them which would be too fast for the other but
slower memory modules.

https://en.wikipedia.org/wiki/DDR4_SDRAM
"Standard transfer rates are 1600, 1866, 2133 and 2400 MT/s.[50] (12/15,
14/15, 16/15 and 18/15 GHz clock speeds, double data rate.) 2666 and
3200 MT/s (20/15 and 24/15 GHz clock speeds) are provided for, but the
specifications are not yet complete."

I cannot see in your pic the clock rates, just the multiplier.

That you bought PC-19200 memory doesn't mean your mobo can support it.
Does the mobo's manual, what is the fastest memory specs it will
support?

http://www.gigabyte.us/Motherboard/GA-B250M-DS3H-rev-10#sp
"To support 2400 MHz or XMP memory, you must install a 7th generation
processor."

http://ark.intel.com/products/family/95544/7th-Generation-Intel-Core-i7-Processors

What CPU is in the mobo?

Paul

unread,
Feb 13, 2017, 2:49:36 AM2/13/17
to
It's even in the spec, too.

http://ark.intel.com/products/97452/Intel-Celeron-Processor-G3930-2M-Cache-2_90-GHz

The memory interface is on the processor, so this is a
CPU function, not a chipset function. The chipset is
merely a PCH (Southbridge) for the SATA ports etc.

And compared to what I paid for my E8400 back in the
day, it's a bargain.

And the spec on the Crucial site says:

Intel XMP 2.0 profiles for easy configuration
DDR4 PC4-19200 16-16-16
Unbuffered NON-ECC 1.2V 8GB

So the very first thing you'd do, is enter the BIOS
and find the XMP switch. As long as you have one or
two sticks, XMP should work for you right away. You
might have to select from one of two XMP profiles,
and the only problem I had with that, is they're
not labeled properly, so you cannot tell which is
which. Or, whether they're both "useful". The version
of XMP on mine, only works right for one stick per
channel config. So if you got 2x8GB, then switch
on the XMP, and "bam, you're done". There's no futzing
with XMP.

*******

It's in the CPU support chart, so that's not the problem.
Your G3930 uses F1 BIOS, like all the processors in the chart.

http://www.gigabyte.com/Support/Getcpulist?Type=Product&Value=6124

About all I can suggest, is flip on XMP and use the magic.

(Pg.23 in the manual)

http://download.gigabyte.us/FileList/Manual/mb_manual_ga-b250m-ds3h_e.pdf

Paul

John Doe

unread,
Feb 13, 2017, 3:39:34 AM2/13/17
to
Paul <nos...@needed.invalid> wrote:

> John Doe wrote:

>> Motherboard, Gigabyte (GA-B250M-DS3H)
>> CPU, Intel Celeron G3930
>> RAM, Crucial Ballistix Sport LT DDR4-2400 (BLS8G4D240FSB)
>>
>> All of those are supposed to facilitate memory at 2400. But
>> seems settings will not allow it.
>>
>> https://www.flickr.com/photos/27532210@N04/?
>>
>
> It's even in the spec, too.
>
> http://ark.intel.com/products/97452/Intel-Celeron-Processor-G3930-2M-Cache-2_90-GHz
>
> The memory interface is on the processor, so this is a
> CPU function, not a chipset function. The chipset is
> merely a PCH (Southbridge) for the SATA ports etc.

Right, thanks for the reminder.

> And compared to what I paid for my E8400 back in the
> day, it's a bargain.

The motherboard was $74 and the CPU was $42 (shipped FedEx Slowpoke/Smartpost).

> And the spec on the Crucial site says:
>
> Intel XMP 2.0 profiles for easy configuration
> DDR4 PC4-19200 16-16-16
> Unbuffered NON-ECC 1.2V 8GB
>
> So the very first thing you'd do, is enter the BIOS
> and find the XMP switch. As long as you have one or
> two sticks, XMP should work for you right away. You
> might have to select from one of two XMP profiles,
> and the only problem I had with that, is they're
> not labeled properly, so you cannot tell which is
> which. Or, whether they're both "useful". The version
> of XMP on mine, only works right for one stick per
> channel config. So if you got 2x8GB, then switch
> on the XMP, and "bam, you're done". There's no futzing
> with XMP.

There is only one choice, with no description whatsoever.

> It's in the CPU support chart, so that's not the problem.
> Your G3930 uses F1 BIOS, like all the processors in the chart.
>
> http://www.gigabyte.com/Support/Getcpulist?Type=Product&Value=6124

Even the memory stick part number is in the motherboard's RAM
support list.

When I try flipping the XMP switch, apparently the system
cannot handle it. No idea what the problem might be. The only
thing on the motherboard is the memory and a video card.
Maybe I will try removing the video card and using the on
chip video.

Everything appears to be brand-new, except the RAM stick came
unsealed. Could have been a return item.

The BIOS is up-to-date.






--

John Doe

unread,
Feb 13, 2017, 4:27:59 AM2/13/17
to
FWIW...
Correction. When using an XMP profile, it does not crash, but
nothing changes, still 2133.

Paul

unread,
Feb 13, 2017, 5:32:24 AM2/13/17
to
If you use a copy of CPUZ, what does it show ?

This is some of my DDR3-2400. The first XMP has
Command Rate 2T, and is the more "relaxed" profile.
The Command Rate 1T which is right-most in the diagram,
is wishful thinking. I don't think any RAM could
do 1T at speeds like this. The 2T is much more likely
to work.

https://s18.postimg.org/ux7r9qtt5/CPUZ_SPD.gif

I'm sure your SPD is correct, and the motherboard/BIOS
has decided it's too unstable or something.

The motherboard should *not* be offering XMP, if
it cannot adjust VDimm and related voltages. You
can see in the example, that my RAM needs 1.65V
(i.e. limited by safe operation of CPU), to work
at the speed in question. Your DDR4 is a lower
voltage RAM, but the same principles apply. Maybe
it needs 1.35V instead of 1.2V or something. To
support XMP, the design has to be adjustable to
accommodate whatever is in the parameters.

Paul

John Doe

unread,
Feb 13, 2017, 5:49:43 AM2/13/17
to

John Doe

unread,
Feb 13, 2017, 6:42:45 AM2/13/17
to
Seems strange that in the BIOS, 60-80% of the settings that are
not grayed out, that are printed in bright white and can be
selected cannot be changed.

John Doe

unread,
Feb 13, 2017, 7:23:29 AM2/13/17
to
After looking on the Internet... Gigabyte BIOS uses page up
and page down to adjust BIOS settings. They really should
specify that someplace. The only place it is mentioned in the
manual is for adjusting system date and system time.

Still no luck at going above 2133.

Paul

unread,
Feb 13, 2017, 11:55:27 AM2/13/17
to
I notice your PCI Express bus is overclocked.
Was that on purpose ? It says 135MHz. When the
bus runs PCI Express Rev.3, almost 16GB per second,
I don't really think it needs to be overclocked.
Just running at Rev.3 speeds should be more than
enough for texture transfer. The nominal input
clock should be 100MHz, AFAIK.

*******

It's almost like a memory multiplier is missing. But I
don't see an obvious reason why they would remove it.
All the specs say it should be there.

Paul

John Doe

unread,
Feb 13, 2017, 12:04:48 PM2/13/17
to
Paul <nos...@needed.invalid> wrote:

> John Doe wrote:

>> After looking on the Internet... Gigabyte BIOS uses page
>> up and page down to adjust BIOS settings. They really
>> should specify that someplace. The only place it is
>> mentioned in the manual is for adjusting system date and
>> system time.
>>
>> Still no luck at going above 2133.
>
> I notice your PCI Express bus is overclocked. Was that on
> purpose ? It says 135MHz. When the bus runs PCI Express
> Rev.3, almost 16GB per second, I don't really think it
> needs to be overclocked. Just running at Rev.3 speeds
> should be more than enough for texture transfer. The
> nominal input clock should be 100MHz, AFAIK.

I did not change that. That is a stock setting.

Not trying to overclock, just trying to get the stated speed
of the RAM.

Thanks for the attention.

> It's almost like a memory multiplier is missing. But I
> don't see an obvious reason why they would remove it. All
> the specs say it should be there.

Seems that way to me, like a setting is missing.

Maybe that is why Amazon sold the motherboards for $74
(total) before the price jumped to $86 or more.

John Doe

unread,
Feb 15, 2017, 5:09:27 AM2/15/17
to
Inaccurate specifications might be the problem. Intel lists
the CPU as supporting 2400 memory and not supporting 2400
memory, depending on which HTTP protocol page you visit...

https://ark.intel.com/products/97452/Intel-Celeron-Processor-G3930-2M-Cache-2_90-GHz

http://ark.intel.com/products/97452/Intel-Celeron-Processor-G3930-2M-Cache-2_90-GHz

The "s" version supports 2400 memory...

Thanks.







--

Paul

unread,
Feb 15, 2017, 11:53:48 AM2/15/17
to
John Doe wrote:
> Inaccurate specifications might be the problem. Intel lists
> the CPU as supporting 2400 memory and not supporting 2400
> memory, depending on which HTTP protocol page you visit...
>
> https://ark.intel.com/products/97452/Intel-Celeron-Processor-G3930-2M-Cache-2_90-GHz
>
> http://ark.intel.com/products/97452/Intel-Celeron-Processor-G3930-2M-Cache-2_90-GHz
>
> The "s" version supports 2400 memory...
>
> Thanks.

Download the Intel Processor Identification Utility,
to verify the processor details. There were two versions,
one for Pentium III, and the later one is the one that
would be constantly updated for modern stuff.

A Skylake in that family, was only rated for 2133, whereas
Kaby Lake is supposed to be 2400.

Paul

John Doe

unread,
Feb 16, 2017, 2:35:12 AM2/16/17
to
Paul <nos...@needed.invalid> wrote:

> John Doe wrote:

>> Inaccurate specifications might be the problem. Intel lists
>> the CPU as supporting 2400 memory and not supporting 2400
>> memory, depending on which HTTP protocol page you visit...
>>
>> https://ark.intel.com/products/97452/Intel-Celeron-Processor-
G3930-2M-Cache-2_90-GHz
>>
>> http://ark.intel.com/products/97452/Intel-Celeron-Processor-
G3930-2M-Cache-2_90-GHz
>>
>> The "s" version supports 2400 memory...
>>
>> Thanks.
>
> Download the Intel Processor Identification Utility,
> to verify the processor details.

I just did, but apparently it has not been updated for the
most recent CPUs.

> There were two versions, one for Pentium III, and the later
> one is the one that would be constantly updated for modern
> stuff.

If you looked at the (HTTP) link you first posted, it does
not show support for 2400 memory.

Coincidentally, I did not look at the link you posted either.
Because I had already seen the same link under (HTTPS).

> A Skylake in that family, was only rated for 2133, whereas
> Kaby Lake is supposed to be 2400.

That is not what the HTTP link says for the exact same
processor.

Anyways... No big deal here. I just needed a CPU that can be
upgraded and will allow swapping for troubleshooting.

Thanks.

John Doe

unread,
Feb 22, 2017, 10:59:35 AM2/22/17
to
They just changed it so that both say 2133 memory only.



--

I wrote:

> Inaccurate specifications might be the problem. Intel lists
> the CPU as supporting 2400 memory and not supporting 2400
> memory, depending on which HTTP protocol page you visit...
>

John Doe

unread,
Feb 24, 2017, 8:13:31 AM2/24/17
to
Paul <nos...@needed.invalid> wrote:

> John Doe wrote:

>> FWIW...
>> Correction. When using an XMP profile, it does not crash,
>> but nothing changes, still 2133.
>
> If you use a copy of CPUZ, what does it show ?
>
> This is some of my DDR3-2400. The first XMP has
> Command Rate 2T, and is the more "relaxed" profile.
> The Command Rate 1T which is right-most in the diagram,
> is wishful thinking. I don't think any RAM could
> do 1T at speeds like this. The 2T is much more likely
> to work.

Still running at 2133...
The timing at 2400 is supposed to be
16 16 16 39 but I have had it at
12 13 13 35 and using 1T, with no apparent problem during
ordinary use. Those values were just stuck in there without
further experimentation. So maybe it could be tighter.

https://www.flickr.com/photos/27532210@N04/?

However, seems like 1T does not provide anywhere near the
performance boost I recall it provided many years ago (the
last time I tried changing that).




--

Paul

unread,
Feb 24, 2017, 3:23:38 PM2/24/17
to
John Doe wrote:
> Paul <nos...@needed.invalid> wrote:
>
>> John Doe wrote:
>
>>> FWIW...
>>> Correction. When using an XMP profile, it does not crash,
>>> but nothing changes, still 2133.
>> If you use a copy of CPUZ, what does it show ?
>>
>> This is some of my DDR3-2400. The first XMP has
>> Command Rate 2T, and is the more "relaxed" profile.
>> The Command Rate 1T which is right-most in the diagram,
>> is wishful thinking. I don't think any RAM could
>> do 1T at speeds like this. The 2T is much more likely
>> to work.
>
> Still running at 2133...
> The timing at 2400 is supposed to be
> 16 16 16 39 but I have had it at
> 12 13 13 35 and using 1T, with no apparent problem during
> ordinary use. Those values were just stuck in there without
> further experimentation. So maybe it could be tighter.
>
> https://www.flickr.com/photos/27532210@N04/?
>
> However, seems like 1T does not provide anywhere near the
> performance boost I recall it provided many years ago (the
> last time I tried changing that).

Well, look at the other cycle counts. How does a command
rate of 2T, compare to 16 ? LOL. Command Rate might have
been more fun in the CAS2 era.

Another possibility, is the Command Rate setting doesn't
actually work. And the hardware is fixed. The policy on
Command Rate on Intel was always a bit fuzzy.

Paul

0 new messages