Question Standard JEDEC vs XMP

[Endre]

Hello! I have 2 sticks of memory: Kingston HyperX Fury (16GB, DDR4-2666, timings: 16-18-18-39 @1.2V). If I activate the XMP Profile 1, the frequency, timings and voltage remain the same, being PnP memories. But still! By activating the XMP profile, I’ve seen a tiny improvement in the benchmarks.

Can anyone tell me why?

Is there something else “beneath the surface” besides frequency and timings, that’s being optimized?

 

[Darkbreeze]

Yes, secondary and tertiary timings. Some of these are set by profile. Some are auto configured by the motherboard UNLESS you manually change them, but typically if they are part of the profile settings they will probably be at least somewhat different than what the motherboard configures them for it if is required to do so.

Overall, you probably wont' see a lot of difference depending on the default settings based on what motherboard and CPU you have. If you had much higher speed memory, then you would see a significant difference as the default 2133-2666mhz + timings would be significantly different than for the XMP 3000mhz + profile.

What is your CPU and motherboard model?

 

[ShawnWager]

So glad I never abandoned my x58...

 

[Endre]

Yes, secondary and tertiary timings. Some of these are set by profile. Some are auto configured by the motherboard UNLESS you manually change them, but typically if they are part of the profile settings they will probably be at least somewhat different than what the motherboard configures them for it if is required to do so.

Overall, you probably wont' see a lot of difference depending on the default settings based on what motherboard and CPU you have. If you had much higher speed memory, then you would see a significant difference as the default 2133-2666mhz + timings would be significantly different than for the XMP 3000mhz + profile.

What is your CPU and motherboard model?

Motherboard: Gigabyte Z390 Aorus Master rev. 1.0
CPU: Intel i7-9700K (3.6GHz, 8-Cores, 8-Threads)

 

[Darkbreeze]

So, 2666mhz is the default JEDEC speed for that platform with those components. Enabling XMP on 2666mhz sticks, with that configuration, is probably not going to change much of anything because the profile is already supported by default. Some few settings specific to those particular modules might change based on XMP vs JEDEC but it will not likely be enough to see or notice on anything unless you are being incredibly nit picky with the synthetic benchmarks.

If they were 3000mhz sticks, then you'd probably notice a minor improvement depending on what you were doing. In most cases, even if the XMP profile is the same as what the memory defaults to for that platform, it's usually a good idea to enable the XMP profile anyhow, simply for reasons of stability based on preferred secondary and tertiary timings which can definitely have an affect on stability.

 

[Endre]

Yes, secondary and tertiary timings. Some of these are set by profile. Some are auto configured by the motherboard UNLESS you manually change them, but typically if they are part of the profile settings they will probably be at least somewhat different than what the motherboard configures them for it if is required to do so.

Overall, you probably wont' see a lot of difference depending on the default settings based on what motherboard and CPU you have. If you had much higher speed memory, then you would see a significant difference as the default 2133-2666mhz + timings would be significantly different than for the XMP 3000mhz + profile.

What is your CPU and motherboard model?

 

[Endre]

I’ve made my researches today. I loaded the default BIOS settings, then rebooted, and I wrote down all the numbers listed as subtimings in the BIOS.
Then I loaded the XMP Profile 1, rebooted, and checked to see if the subtimings are identical as the default ones. They are!
There’s absolutely no difference in the BIOS settings between the default settings and XMP profile 1 for this memory: Kingston HyperX Fury, HX426C16FB/16.

There’s still an unanswered question to why is there still a tiny improvement in the benchmarks when I load the XMP Profile 1...

 

[Darkbreeze]

Because you are ONLY looking at the primary timings, not the secondary or tertiary timings AND not all boards shows those options or settings in the BIOS. Some boards show ONLY the primary timings, or the primary timings and a partial list of secondary timings but no tertiary timings.

Secondary and tertiary timings don't make a HUGE impact on things, in most cases, but they can make AN impact on performance and depending on the setting it could be significant enough to notice on synthetic benchmarks or very memory dependent applications.

https://www.overclock.net/forum/18051-memory/1630388-comprehensive-memory-overclocking-guide.html

 

[Endre]

Because you are ONLY looking at the primary timings, not the secondary or tertiary timings AND not all boards shows those options or settings in the BIOS. Some boards show ONLY the primary timings, or the primary timings and a partial list of secondary timings but no tertiary timings.

Secondary and tertiary timings don't make a HUGE impact on things, in most cases, but they can make AN impact on performance and depending on the setting it could be significant enough to notice on synthetic benchmarks or very memory dependent applications.

https://www.overclock.net/forum/18051-memory/1630388-comprehensive-memory-overclocking-guide.html

Gigabyte Z390 Aorus Master displays 3 levels of memory timings: 1-primary, 2-tertiary, 3-misc.
I looked at all of them and wrote them down.
They are the same in both cases: Default (JEDEC) and XMP Profile 1.

 

[Darkbreeze]

Then it is because the JEDEC standard allows for the same configuration based on Z390 default support as what your sticks are anyhow. So either way obviously should be fine, and the same. If the speed, timings and voltage are the same, and everything is running in dual channel in both configurations, then there can be no difference in performance other than allowed variance based on any two compared runs of the benchmark.