XMP 2.0 vs Memory Try It! and CL timings

Menno555

New member
PRIVATE E-2
Joined
Jul 3, 2019
Messages
13
I have a MSI Z390-A Pro (with latest bios) with a i9-9900KF CPU. And a noob when it comes to RAM tweaking. And also no overclocking.
Had to upgrade my RAM to Crucial Ballistix MAX RGB 32GB Kit (2 x 16GB) DDR4-4400 (BLM2K16G44C19U4BL).
I knew it was a bit of a gamble (and overkill) but on the Crucial site it's listed as compatible plus it was there €130 cheaper ;)
So got it today and turns out that with XMP enabled, it's not stable. Windows 10 works normal but my browser got unstable and games crashed, and so on.
But with XMP disabled and Memory Try It! @ 4400Mhz it's all stable.
And here the question: with XMP it's CL 19-19-19-46 @ 1.40V and with Memory Try It!, it's 19 - 26 - 26 - 46 @ 1.45V
Is this something that's acceptable or are there other "tricks" to consider?

Thanks in advance.
 

citay

Pro
SERGEANT
Joined
Oct 12, 2016
Messages
4,491
That kit is unusual. Normally, with the Micron ICs used exclusively on Crucial RAM, you have timings where tCL is at least one or two clock cycles lower than tRCD and tRP. Because with Micron chips, the latter two are the limiting timings, as they don't scale well with voltage (unlike tCL). So for using Micron ICs, these are quite agressive XMP timings. On high-end kits from other brands, which employ Samsung B-Die ICs, you can regularly see the first three primary timings being the same number, but for Crucial, it used to be that tCL was two lower than the others.

So this slightly unusual default profile doesn't seem to pay off for everyone,

Screenshot 2021-12-02 at 15-31-27 Crucial Ballistix MAX RGB 4400 MHz DDR4 DRAM Desktop Gaming ...png


As i write in my thread about RAM, in case of RAM problems, there are three possibilities, which can also be used in combination:
1) Raise voltage
2) Lower frequency
3) Loosen timings

Your board has chosen a combination of 1) and 3). It has recognized the Micron ICs and drastically lowered tRCD and tRP as well as increased the voltage quite a bit.

Of course you can do better manually. Maybe a slight relaxation of tRCD and tRP at the default XMP voltage are enough to make it stable, like 19-21-21. Then you have to run extensive RAM stability tests to ensure that it's stable. Crashing vs. not crashing is only a very coarse way to check it. When it crashes, you already have a big instability. If it's only slightly unstable, it won't crash, but you might still get silent file corruption (since everything goes through the RAM first, all writes to disk etc.). So you should test it according to what i listed in the thread.
 

Menno555

New member
PRIVATE E-2
Joined
Jul 3, 2019
Messages
13
Thanks for your reaction.
I just did use MemTest with batches of 2500Mb until almost all RAM was in use and they all went over 100% with 0 errors. Did read your helpful thread and will see what MemTestHelper does too.
And I'm for sure going to try you suggestion to see what happens with default voltage and 19-21-21. Although that 0.05V increase looked tiny for my inexperienced mind, judging by your reaction it isn't that small a raise.
 

citay

Pro
SERGEANT
Joined
Oct 12, 2016
Messages
4,491
DDR4 operates in quite a narrow voltage range. The official voltage is actually 1.2V, and all DDR4 that is used in OEM office PCs for example will adhere to that and not have a heatspreader either. Only the "overclocker RAM" (let's call it) has 1.35V as standard. Then you have the more extreme kits, above DDR4-4000 (which is going towards what DDR4 is capable of for daily use) with low latencies, which use up to 1.5V as standard as per their XMP profile in the most extreme kits. And around 1.5V is usually also where it may start to require some active cooling over the RAM.

So while the absolute change in voltage might be small, at 1.45V you're creeping more towards the higher end of the operating range. Don't get me wrong, it's still not dangerous per se, but you only want to have it as low as necessary, with maybe +.02V extra for stability headroom. So if it's stable at DDR4-4400 19-21-21 @ 1.4V, you could set it to 1.42V to have some stability reserves. You never want it right at the edge of stability, because then any change in the environment, like higher case temps, might push it into an instability again.
 

Menno555

New member
PRIVATE E-2
Joined
Jul 3, 2019
Messages
13
Again thanks :)
I did try some things and with 100% MemTest (that MemTestHelper is indeed handy) of 89% RAM, it gave zero errors.
This with 19-20-20-43 @ 1.410V.
But, that 43 was a mistake though because on the Crucial product page it says 43 instead of the actual 46 (also according to CPU-Z), so I had 43 in my head.
Is that last one important? Or is it just a "happy accident" on my part that works?
 

citay

Pro
SERGEANT
Joined
Oct 12, 2016
Messages
4,491
Good. The fourth timing, tRAS, is minimum tRCD + tRP, so with 19-20-20, it would have to be minimum 20+20=40. It does actually say 19-19-19-43 at Crucial for this kit, so i don't know why the XMP profile would use tRAS 46. Anyway, this timing is not so critical. On some kits, Crucial used to put it right at the minimum (for example DDR4-3200 CL16-18-18-36), on the newer kits they like to add a bit (for example DDR4-3600 CL16-18-18-38) and then on yours, a bit more. But it doesn't make a big difference, neither for performance nor for stability. You can keep it as it is if it's all working well.
 

Menno555

New member
PRIVATE E-2
Joined
Jul 3, 2019
Messages
13
Checked the label on the stick and there also it says 46. Ah well, it works :)
So thanks again for the tips and added knowledge.
 
Top