Asus XMP I vs II

[mackerel]

tldr: seeking experiences with Asus XMP modes I vs II?

Trust Asus to complicate things... in short, 11700k + B560M-A, Corsair Vengeance Pro 2x16GB 3200. The same CPU and ram worked fine in a Z490 board previously at 3200 gear 1, but that had different problems elsewhere.

I was lazy and hadn't set XMP since building. Last night, I got around to it, and was facing a choice of Asus XMP I or II modes. For those not familiar, mode I is Asus using XMP primary timings, and Asus choose the rest. Mode II is using the ram values. The Asus documentation seems to push mode I using words like that is what they "validate" with. Anyway, I had some odd behaviour in that mode ending with a crash in game, and I reverted to XMP off for now. It was also in gear 1 at the time. If I manually set Gear 2, it sets the ram speed to auto. I didn't try booting with that.

I haven't got around to trying other settings yet. Maybe XMP II gear 1 is better? Maybe I really do need gear 2?

 

[EarthDog]

No issues here with gear1/xmp1 and asus boards (z590i, apex) at 3600. It's stable enough to run 30 min stability test and ~2 hours worth of scripted testing. I assume you're on the latest bios for the board?

 

[Woomack]

Some cheaper motherboards can have problems with that. For me, everything works fine on MSI Z590I Unify and was fine on ASRock Z590 motherboards up to 3600-3733 Gear 1. However, Biostar Z590I Valkyrie is not even booting at 3600 and Gear 1. Max that can boot is 3466.
ASUS XMP modes are to adjust additional sub-timings, in case something doesn't work. Gigabyte had something like that too as their motherboards had problems with the XMP of some RAM brands like Corsair. I'm not really using these modes on ASUS. It almost always works fine at the main XMP profile without any adjustments.

 

[mackerel]

Mobo is on latest bios. Forgot to say, the modules are dual rank.

I just started doing testing using aida64 selecting ram only for stress test.

XMP off, ran for 1 minute ok. I don't expect any problems here, just a quick look.
XMP I, gear 1, error after 30 seconds. This is good, as it seems easy to reproduce.
XMP II, gear 1, error after 9 seconds. Ok, the problem isn't the subtimings between the two mobo XMP modes at least. The problem is common to both.
XMP II, gear 2. So far clear after 12 minutes running.

When I switched from XMP I to II, it showed the timing differences:
RAS# to RAS# Delay L [Auto]->[9]
RAS# to RAS# Delay S [Auto]->[6]
REF Cycle Time [Auto]->[559]
REF Cycle Time 2 [Auto]->[415]
REF Cycle Time 4 [Auto]->[255]
FOUR ACT WIN Time [Auto]->[36]

I have no idea what the Auto values were without going back and looking for it. Doesn't seem worthwhile. Above may serve as reference in case anyone else wonders what settings might differ.

When switching to gear 2 (staying in XMP II), like I saw last time the speed went to Auto. I manually selected 3200 to be safe. On save and exit, it showed some silly high value as the before speed. It wasn't double though, something like 5800???

So I guess I can blame the mobo ram/timing/something implementation here? Again, it did run 3200 gear 1 on a Z490 mobo and was compute stable. If gear 2 works I'll just use that and move on.

 

[mackerel]

It seems stable in gear 2. I've run aida64 mem test for over an hour, and a custom Prime95 mem-centric test for another hour. No errors. Finally did 20 mins of P95 blend, ok. Will move onto more interesting things than try to tinker any more perf improvements.

 

[EarthDog]

Maybe try raising VccIO Memory input voltage (second IO voltage) and see if that helps. I had issues with memory stability at DDR4 4000 @ 1:2 on a couple of boards (namely ASRock), adding 0.1V of that worked. Not sure if it will help you or not.

 

[magellan]

Maybe try raising VccIO Memory input voltage (second IO voltage) and see if that helps. I had issues with memory stability at DDR4 4000 @ 1:2 on a couple of boards (namely ASRock), adding 0.1V of that worked. Not sure if it will help you or not.

I've experienced that problem on my z390 Asrock Taichi, 0.1V made the difference between stability and instability at 4005 MT/s, but I'm not sure if it was VCCIO or VCCSA that was the key voltage.

Does Asrock have the worst rep. in terms of memory overclocking capability?

 

[mackerel]

Maybe try raising VccIO Memory input voltage (second IO voltage) and see if that helps. I had issues with memory stability at DDR4 4000 @ 1:2 on a couple of boards (namely ASRock), adding 0.1V of that worked. Not sure if it will help you or not.

As said it seems working ok in gear 2. I'd rather spend my time gaming than tinkering in the hopes of getting gear 1 working.

I've experienced that problem on my z390 Asrock Taichi, 0.1V made the difference between stability and instability at 4005 MT/s, but I'm not sure if it was VCCIO or VCCSA that was the key voltage.

Does Asrock have the worst rep. in terms of memory overclocking capability?

Dunno about memory overclocking, but they don't seem that great on CPU stability in general. I've had two of their boards where I've had problems, and eventually determined it was the mobo. Z370 something, 8086k without OC only stable if I enforce TDP power limit. Prime95 shows it more quickly but crashes can happen in gaming too. B450 ITX something, just unstable with a 3700X. Increasing Vcore was the only thing that helped but didn't eliminate problems. Lowering power limit actually made it worse. Two boards might not be a big sample but I have no confidence in their power delivery. I sold that mobo on to someone who wanted to try it with an APU, reports it works fine. Guess the 3700X was too much CPU for it to handle.

 

[magellan]

Dunno about memory overclocking, but they don't seem that great on CPU stability in general. I've had two of their boards where I've had problems, and eventually determined it was the mobo. Z370 something, 8086k without OC only stable if I enforce TDP power limit. Prime95 shows it more quickly but crashes can happen in gaming too. B450 ITX something, just unstable with a 3700X. Increasing Vcore was the only thing that helped but didn't eliminate problems. Lowering power limit actually made it worse. Two boards might not be a big sample but I have no confidence in their power delivery. I sold that mobo on to someone who wanted to try it with an APU, reports it works fine. Guess the 3700X was too much CPU for it to handle.

I read most z370 motherboards didn't have the VRM's to handle overclocks of top-end Coffeelake CPU's.

 

[mackerel]

I read most z370 motherboards didn't have the VRM's to handle overclocks of top-end Coffeelake CPU's.

The 8086k was running stock, and shouldn't be any more stress than the 8700k that initially launched.