For Those Who Don't Know What SPD memory speed is for

HAL_404

[H]ard|Gawd
Joined
Dec 16, 2018
Messages
1,240
this is a lengthy post; might want to first get a cup of Java or a beer ...

I read not too few comments online about how AMD MoBo's were having issues regarding MFG/Retailer posted mem speeds vs how it would post in the MoBo's BIOS via Auto settings and that when XMP was enabled ... why it still didn't show the MFG's/Retailer's posted mem specs.

Today I began to question why, when I bought DDR4 3000 memory with advertised timings of 15 - 17 - 17 - 35 why it was that it wound up posting in the Mobo's BIOS as DDR4 2166 with timings of 16 - 21 - 22 - 50

So I went to the memory MFG's website and looked up the specs and it showed as follows:

SPD Latency 15-15-15-36
SPD Speed 2133MHz
Tested Latency 15-17-17-35
Tested Speed 3000MHz

The Retailer's website only showed the tested numbers so then I thought, "ok, why did the memory MFG post two different sets of specs for the same memory module?"

Then I decided to look up what SPD meant and lo and behold I found the following description:

"When a computer is booted (started), serial presence detect (SPD) is information stored in an electrically erasable programmable read-only memory (EEPROM) chip on a synchronous dynamic random access memory (SDRAM) memory module that tells the basic input/output system (BIOS) the module's size, data width, speed, and voltage. The BIOS uses this information to configure the memory properly for maximum reliability and performance."

I had to read that definition a couple times before it would sink in and even then, I continue to disagree with the "maximum performance" part https://whatis.techtarget.com/definition/serial-presence-detect-SPD but I digress here ...

And so, I began to learn that memory speed/timings on AMD MoBo's (recent one's anyways) are automatically set to the SPD values hence the reason we will see an XMP On/Off button in the MoBo's BIOS settings (recent boards anyways) so that the TESTED values will be applied for those of us who are "enthusiast" gamers

But the XMP button doesn't work right, sets my 3000 mem to 2666 and that won't do so rather than presume the MoBo was defective I decided to manually selected DDR4 3000 from a drop-down list in BIOS under Advanced Memory Settings and then set the timings to 15 - 17 - 17 - 35 which was a chore since the MSI BIOS showed two tRCD settings (tRCDRD and tRCDWR) ... geeesh :barefoot:

Why they make doing this so difficult for the average gamer?

So my question then became, "why bother putting a XMP button in the BIOS when 'They' know it's not going to work as expected?" But yet once again I digress here ...

And if that wasn't enough to cause aggravation for the average gamer (no, not the digression part) ... then there's the tRC setting that doesn't ever seem to get mentioned anywhere in the specs ...

for those still reading and in need to know what tRC is:

" The minimum time interval between successive ACTIVE commands to the same bank is defined by tRC where tRC = tRAS + tRP "

https://www.techpowerup.com/articles/64

you'll likely need to hunt for the tRC setting under the BIOS memory Advanced settings to find it ... at least I had to

And so, if you bought or are considering buying an AMD MoBo, you are now a well informed rig builder as to why XMP likely won't work right for you either and why you'll need to manually set the Mem specs for yourself in BIOS

example:

tCL ..... tRCD ..... tRP ..... tRAS
15 ......... 17 ........ 17 ........ 35

and don't forget to set tRC ;)
 
Last edited:
Yeah it's going to be fun with my kit as XMP sets the voltage .3v too low for the rated speed. It's trying to run 3800 CL14 at 1.2 volts. Not even sure what other subtimings it set. I'm going to use the ryzen dram calculator to set everything anyhow. Thanks for the post.
 
I found the same issue on my MSI X570 board. Set the XMP on and sure it says all the timings in the profile and the speed, but it never sets the memory to that speed, in my case 3600. I manually set it and it works as expected, using the drop down, my board also set the correct timings in Auto mode. I have it bumped to 3800 with the Fabric at 1900 now though.
 
My memory is no where close to running at XMP. If I enable XMP, it will not even post and requires a CMOS reset. It needs 1.43v to do 3733. XMP is 1.35 at 4000. Not that I would want to try to run 4000 with asynchronus fclk, but it doesn't work anyway. Ryzen DRAM calculator settings don't work either unless I make a bunch of changes. The RAM is B Die, but not on the QVL, so I guess I can't complain.
 
Neat, I set XMP and it runs fine, and applies the settings correctly.

Ryzen 3600X with G.Skill Trident Z Neo... 3600mhz on an X570 board.
 
Well, for starters, I would check to see whether or not gear down mode is enabled since you're trying to run an odd numbered CAS latency. I'd run it at either 14 or 16 instead of 15.
 
I also noticed recently the same issue with GSkill Trident Z NEO 3000Mhz. I have a MSI X570 Gaming Plus Mobo.
I visited the site of GSkill and notice the difference in the dashboard between speed and tested speed. Thought innocently that I have bought 3000Mhz RAM.
I also noticed with HWINFO64 my ram was set to 2133Mhz. So I went back into the BIOS and yes XMP profile was enable, and the frequency shown in the configuration page was 3000Mhz. But on the home page of the BIOS it was 2133Mhz.
I have set the option from a list, as the OP mentioned in his first post, DDR4-3000Mhz.
Now HWFINFO64 is displaying the correct values.

Thanks for this post. It confirms to me that something is wrong with XMP profiles.
 
It all depends upon the compatibility with your motherboard. Some RAM and Motherboard combos work together perfectly. Some do not. If you want to be sure, most motherboards have a QVL which lists the RAM verified to work at its advertised speeds, with that motherboard. That does not mean that no other RAM will work correctly.

Additionally, bios updates for motherboards often improve memory compatibility. If your RAM doesn't work at XMP and you haven't updated your bios, you should. Its very possible it will improve the experience you have with your memory.
 
I also noticed recently the same issue with GSkill Trident Z NEO 3000Mhz. I have a MSI X570 Gaming Plus Mobo.
I visited the site of GSkill and notice the difference in the dashboard between speed and tested speed. Thought innocently that I have bought 3000Mhz RAM.
I also noticed with HWINFO64 my ram was set to 2133Mhz. So I went back into the BIOS and yes XMP profile was enable, and the frequency shown in the configuration page was 3000Mhz. But on the home page of the BIOS it was 2133Mhz.
I have set the option from a list, as the OP mentioned in his first post, DDR4-3000Mhz.
Now HWFINFO64 is displaying the correct values.

Thanks for this post. It confirms to me that something is wrong with XMP profiles.

Did you recently update bios? My memory did that on my msi x570 gaming pro carbon.
I had to reset my bios and when I turned xmp on again it set the correct memory setting.
 
this is a lengthy post; might want to first get a cup of Java or a beer ...

I read not too few comments online about how AMD MoBo's were having issues regarding MFG/Retailer posted mem speeds vs how it would post in the MoBo's BIOS via Auto settings and that when XMP was enabled ... why it still didn't show the MFG's/Retailer's posted mem specs.

Today I began to question why, when I bought DDR4 3000 memory with advertised timings of 15 - 17 - 17 - 35 why it was that it wound up posting in the Mobo's BIOS as DDR4 2166 with timings of 16 - 21 - 22 - 50

So I went to the memory MFG's website and looked up the specs and it showed as follows:

SPD Latency 15-15-15-36
SPD Speed 2133MHz
Tested Latency 15-17-17-35
Tested Speed 3000MHz

The Retailer's website only showed the tested numbers so then I thought, "ok, why did the memory MFG post two different sets of specs for the same memory module?"

Then I decided to look up what SPD meant and lo and behold I found the following description:

"When a computer is booted (started), serial presence detect (SPD) is information stored in an electrically erasable programmable read-only memory (EEPROM) chip on a synchronous dynamic random access memory (SDRAM) memory module that tells the basic input/output system (BIOS) the module's size, data width, speed, and voltage. The BIOS uses this information to configure the memory properly for maximum reliability and performance."

I had to read that definition a couple times before it would sink in and even then, I continue to disagree with the "maximum performance" part https://whatis.techtarget.com/definition/serial-presence-detect-SPD but I digress here ...

And so, I began to learn that memory speed/timings on AMD MoBo's (recent one's anyways) are automatically set to the SPD values hence the reason we will see an XMP On/Off button in the MoBo's BIOS settings (recent boards anyways) so that the TESTED values will be applied for those of us who are "enthusiast" gamers

But the XMP button doesn't work right, sets my 3000 mem to 2666 and that won't do so rather than presume the MoBo was defective I decided to manually selected DDR4 3000 from a drop-down list in BIOS under Advanced Memory Settings and then set the timings to 15 - 17 - 17 - 35 which was a chore since the MSI BIOS showed two tRCD settings (tRCDRD and tRCDWR) ... geeesh :barefoot:

Why they make doing this so difficult for the average gamer?

So my question then became, "why bother putting a XMP button in the BIOS when 'They' know it's not going to work as expected?" But yet once again I digress here ...

And if that wasn't enough to cause aggravation for the average gamer (no, not the digression part) ... then there's the tRC setting that doesn't ever seem to get mentioned anywhere in the specs ...

for those still reading and in need to know what tRC is:

" The minimum time interval between successive ACTIVE commands to the same bank is defined by tRC where tRC = tRAS + tRP "

https://www.techpowerup.com/articles/64

you'll likely need to hunt for the tRC setting under the BIOS memory Advanced settings to find it ... at least I had to

And so, if you bought or are considering buying an AMD MoBo, you are now a well informed rig builder as to why XMP likely won't work right for you either and why you'll need to manually set the Mem specs for yourself in BIOS

example:

tCL ..... tRCD ..... tRP ..... tRAS
15 ......... 17 ........ 17 ........ 35

and don't forget to set tRC ;)

Hi to all,

HAL...I think we have the same ddr4 modules.
LPX 3000 cL15. XMP profile disabled.
I´m running them stable at 3600 Mhz cL18. These modules are awesome for their price. Have the Hynix Afr running at this speed on 1T - 1.4V - quad channel.
All over the web they said that these modules run at max 3400 mhz stable but with the right settings and right mobo, the overclock capability is brutal.
Of course, they can´t handle cL14 at 3600 but who cares, at cL18 without even fine-tuning them is Epic for an A-die.
On my meg x399 creation, 4x 8Gb ddr4, 1920X Td, their performance is superb.

cachemem.-3600-HynixAFR3000cl15png.png
ZenTimings_3600-AFR.png
 
I feel like XMP has always been a cranky bitch. When I built my system (z170 board, i5 6600k) I put in some 3200Mhz DDR4. With XMP on and running at the memory stock speed, I had no end of crashes. It was BIOS problems on the main board. It took about a full year of bios updates for gigabyte to get it right. Works great now but man it was a long wait.

Not sure why it's still a problem.
 
I feel like XMP has always been a cranky bitch. When I built my system (z170 board, i5 6600k) I put in some 3200Mhz DDR4. With XMP on and running at the memory stock speed, I had no end of crashes. It was BIOS problems on the main board. It took about a full year of bios updates for gigabyte to get it right. Works great now but man it was a long wait.

Not sure why it's still a problem.
From my experience XMP issues on Intel platforms aren't that troublesome. But it sounds to me that in your instance, Gigabyte is to blame, not the XMP standard.
 
From my experience XMP issues on Intel platforms aren't that troublesome. But it sounds to me that in your instance, Gigabyte is to blame, not the XMP standard.

Yes I agree with that statement. The XMP standard works, the problem is that the main board manufacturers have to use it right. That seems to be asking a lot for some of them.

The strange part was that I have a friend who had an indenical system. Only difference was he bought the 3000Mhz ram where I got the 3200Mhz. His was same brand, capacity and timings just slightly slower. He had no issues at all. Doesn't seem to make sense does it?

But I can confirm that gigabyte eventually fixed it whatever it was. I'm now running 4 sticks of 3200Mhz ram in the same system. Just had to wait a year for those idiots to put down the crack pipe.
 
Back
Top