Question Why do Ryzen DRAM OC Guides have you enable XMP?

[modex20]

I looked at a couple of videos to see what was involved in overclocking RAM on a Ryzen system. In particular I looked at these two:

In each one, they enabled their XMP profile and then started plugging in the values. This confused me for a couple of reasons. Why would the BIOS software even allow this? What is the point of enabling XMP only to override all the settings. From a UI experience perspective it seems odd that the BIOS doesn't naturally fall back into manual mode as soon as you go in and tinker with the timings.

I know that the XMP Profile loads a bunch of settings, and that you can override a couple of them. But these guides have you manually overriding everything. Wouldn't it make more sense to disable XMP here?

 

[DrMrLordX]

In each one, they enabled their XMP profile and then started plugging in the values. This confused me for a couple of reasons. Why would the BIOS software even allow this? What is the point of enabling XMP only to override all the settings.

That's something I would never do, personally. Also not all boards allow you to override XMP settings once they are loaded - if you go into manual mode, it dumps XMP mode completely. Depends on a board-by-board basis.

XMP settings can change a lot of subtimings and other oddities that may or may not be exposed to your UEFI. By enabling XMP settings, you get whatever funky settings are set by the manufacturer which could hurt performance once you dial in your own settings. Also it depends on whether it's an AMD or Intel machine. While AMD systems are more XMP-friendly now, it's still a setting regime designed primarily for Intel IMCs.

 

[Hans Gruber]

That's something I would never do, personally. Also not all boards allow you to override XMP settings once they are loaded - if you go into manual mode, it dumps XMP mode completely. Depends on a board-by-board basis.

XMP settings can change a lot of subtimings and other oddities that may or may not be exposed to your UEFI. By enabling XMP settings, you get whatever funky settings are set by the manufacturer which could hurt performance once you dial in your own settings. Also it depends on whether it's an AMD or Intel machine. While AMD systems are more XMP-friendly now, it's still a setting regime designed primarily for Intel IMCs.

The reason AMD machines were not memory friendly was because most of the early bios from different board makers sucked. People tell themselves lies to make it seem otherwise. I have an MSI board that runs everything you throw at it regardless of the memory maker of the actual dimm's. Not the resellers of each memory kit. I had a 1.0.0.4 bios beta that was absolute garbage from the pits of hell. They released an update a few days after and 1.0.04 runs everything as it did before. This is on a B350 motherboard. Most know these boards were not expected to push 3600mhz but this one can run anything I put in it.

I think they suggest running XMP settings simply to get people up and running. It was considered old school to tweak the memory timings on ram.

 

[modex20]

XMP settings can change a lot of subtimings and other oddities that may or may not be exposed to your UEFI.

That's what I had wondered about myself as well. Then I came across this article: https://www.gamersnexus.net/guides/3333-memory-timings-defined-cas-latency-trcd-trp-tras

And it had this quote

Neither SPD nor XMP contains every timing. According to a Kingston representative: “We tune the ‘Primary’ timings (CL,RCD,RP,RAS) only. Other timings are left alone at JEDEC’s recommendations, likely for the MRC.” In regards to some specific subtimings we asked about, “since they are not in the SPD (or XMP), we can’t change them even if we need to.” This lines-up with our experiences in the “Real World Impact” section of this article. Even if memory manufacturers wanted to go deeper, there is a specific and limited list of SPD entries.

So it seems like there is not enough room on the EEPROM for much more other than the primary timings and voltage.

 

[DrMrLordX]

The reason AMD machines were not memory friendly was because most of the early bios from different board makers sucked.

There's more to it than that. The settings you need for stability on Intel machine are/were not the same as the settings you needed for stability on AMD machines. This was especially true in the Ryzen and pre-Ryzen days. Different IMCs behaved differently. The trfc settings alone on my DDR4-3733 kit's XMP settings were completely inappropriate for Ryzen. Still are.

That's what I had wondered about myself as well. Then I came across this article: https://www.gamersnexus.net/guides/3333-memory-timings-defined-cas-latency-trcd-trp-tras

And it had this quote


So it seems like there is not enough room on the EEPROM for much more other than the primary timings and voltage.

Depends on the size of the EEPROMs installed on the DIMM. I remember hearing from 1usmus that XMP settings can tweak stuff that even his Ryzen Memory Calculator won't address (or stuff that might not be exposed on some older/cheaper boards). That's one of the reason I don't play with it.

 

[modex20]

Depends on the size of the EEPROMs installed on the DIMM. I remember hearing from 1usmus that XMP settings can tweak stuff that even his Ryzen Memory Calculator won't address (or stuff that might not be exposed on some older/cheaper boards). That's one of the reason I don't play with it.

That’s pretty alarming. It means that these guides are potentially spreading misinformation and putting people’s systems at risk. That’s irresponsible of them.

 

[DrMrLordX]

That’s pretty alarming. It means that these guides are potentially spreading misinformation and putting people’s systems at risk. That’s irresponsible of them.

It's not so much a risk thing as it is a "why is this so slow?" kind of thing. XMP will load up some pretty loose timings to significantly increase the chance of getting the advertised clocks. It'll also (sometimes) tweak your voltage settings in various places. You're one CMOS reset away from XMP not being a problem if anything like that happens. I'd be a little more worried about the voltage tweaks than anything else.

 

[modex20]

It's not so much a risk thing as it is a "why is this so slow?" kind of thing. XMP will load up some pretty loose timings to significantly increase the chance of getting the advertised clocks. It'll also (sometimes) tweak your voltage settings in various places. You're one CMOS reset away from XMP not being a problem if anything like that happens. I'd be a little more worried about the voltage tweaks than anything else.

Funny you should mention the voltage tweaks. In that first video the guy mentions that he can leave all the voltage settings (with the exception of DRAM voltage) at Auto. I have that exact same motherboard. For my RAM the calculator recommends SOC Voltage of 1.025 and VDDG/P voltages of 0.9. When left at Auto my board set all of those to 1.1 volts!.

 

[DrMrLordX]

Funny you should mention the voltage tweaks. In that first video the guy mentions that he can leave all the voltage settings (with the exception of DRAM voltage) at Auto. I have that exact same motherboard. For my RAM the calculator recommends SOC Voltage of 1.025 and VDDG/P voltages of 0.9. When left at Auto my board set all of those to 1.1 volts!.

Sometimes, lower VDDG/VDDP voltage actually increases stability. In some rare cases, lowering vDIMM can also increase stability. XMP can mess with things like vSoC which makes it problematic in some cases.