F3-10666CL7D-4GBRH from 2011 vs 2015

[Teodan]

I've been trying to add new, "identical" memory to my PC and getting crashes to black screen/reboot. At this point I'm not sure if I've gotten bad memory twice (that MemTest can't detect - this is unlikely), the 2,4 memory slots on my motherboard are bad, or there is an incompatibility between timings for the memory which is the same part number but one pair is from 2011 and the other from 2015.

Specs:
Motherboard: MSI H67MS-E43
CPU: Intel i3-2100
GPU: PowerColor PCS+ Radeon R9 380 DirectX 12 AXR9 380 4GBD5-PPDHE (had to get a new video Bios from powercolor to get it to work because of old motherboard)
Old memory: G.Skill 2x2GB - F3-10666CL7D-4GBRH (built in 2011)
New memory: G.Skill 2x2GB - F3-10666CL7D-4GBRH (built in 2015)

I've already RMA'd one set of new memory. For that first set I ran MemTest86 and it didn't find any errors but if I took the new memory out, the computer stopped crashing so I sent it back. I wish I'd logged the steps I took so I while I think I put the new memory in the 1,3 slots and left the 2,4 slots empty to see if they are bad and I still had a crash I'm willing to try that again. However, I've found other anomalies with the 2nd new set of memory. I'm wondering whether I managed to get 2 bad sets of memory in a row, have bad 2,4 memory slots on the motherboard, or am having a problem with auto-set memory timings. This is the main thing I'm trying to figure out because if I have bad 2,4 slots that's a big problem.

I ran MemTest86 on the 2nd new set with no errors but MemTest86 said I had 8164 MB and it tested up to 9208 MB. Booting to Windows it says I only have 8 GB. My Bios says I have 8192 MB. So I don't know what is going on with those MemTest numers. If the 2,4 slots are bad, is it possible for MemTest86 to not find problems after 12 passes? Since MemTest tests in chunks I assume it will never find timing incompatibilities between physical sticks. Correct?

On memory timing - if I look at the timing settings there are things I can't explain. With auto settings, the 2011 and 2015 memory are both 9-9-9-24 vs. the 7-7-7-21 in the part spec and X.M.P support info I can read using my Bios. (I assume X.M.P info is programmed values from G.skill - like a "readme" of specs actually on the memory.) Ok, fine at least they're the same. But the 2011 memory's tRFC is 74 CLK and the 2015 memory's tRFC is 174 CLK. Could that difference be causing the crashes and the new memory modules are good?

If I look in the X.M.P Support Info for each of the 4 sticks the 2011 memory has tRFC of 60 CLK and the 2015 memory has tRFC of 174 CLK.

The other X.M.P values for 2011 and 2015 are the same except for tFAW and tRC. 2011 has tFAW of 42, tRC of 30. 2015 has tFAW of 20, tRC of 28. The Bios auto-settings sets all 4 sticks to tFAW of 20, tRC of 33.

I think it is unlikely that I got a 2nd bad set of new memory. In fact, the first one was probably good too. But I don't have any knowledge that can make me confident that the 2,4 slots are good or bad as opposed to the problem being these memory timing numbers I don't understand. I'm hoping for some help understanding what is important and what is not. The crashes don't happen right away so randomly banging away at different memory settings to see if it will resolve the crashes could take a very long time. And I could spend forever with no luck if the 2,4 slots on the motherboard are bad.

Thank you very much for any help. I'm really hoping I can just change timing settings and stop the crashes.

[GSKILL TECH]

The new chips are much different so it may be a compatibility issue. Do you have the latest BIOS for the motherboard?

[Teodan]

I do have the latest BIOS - 1.15 or 1.F depending on the numbering convention. Now I'm running the new memory only in slots 1,3 so they are probably fine like I expected. How can I solve the compatibility issue? Is it even possible to do with timing adjustments?

[emissary42]

You can try to manually set tFAW 32 and tRFC 174 for both kits and see if that does the trick, but the issue with mixing kits is that sometimes it just doesn't work out even with fully manual settings.

As a rule of thumb when mixing kits always use the higher of the two values for a certain timing, since the higher density ICs usually can't go as tight as their lower density predecessors.

(tRC is not relevant for Intel based systems.)

[Teodan]

I think I may have been reading screens in my BIOS wrong. Now I think the tFAW values are actually the same in the SPD data. Only the XMP data has different tFAW values. The tRFC is definitely different. I think it is 74 & 174 CLK in the SPD data and 60 & 174 in the XMP data.

If I go to the screen where I can actually set timing values I get a somewhat different set of parameters. There is no tRC (makes sense based on what emissary42 said). In this timing setting screen, on Auto settings, tFAW is 20 whether the 2011, 2015, or both are installed. tRFC is 74 or 174 when 2011 or 2015 is installed by itself. So I think I'm set there.

But there are a few new values on this screen. tWWDD is 3 for 2011 and 5 for 2015. Is that another higher is safer value? I can't find any explanation of what it is unlike tFAW and tRFC which I can understand why higher is better. Also, there are tRTL and tRTL2 values that are different whether 2011, 2015, or both sets are in. I can't find any explanation of those either. I'm not sure if I should leave tRTL and tRTL2 at the auto value from when both are in (36, 36), pick 2011 (34, 34), or pick 2015 (33, 34).

If I could find an engineering explanation of the parameters I could figure it out myself. Or perhaps G.Skill, knowing the differences between the Feb 2011 and Dec 2015 (from stickers on DIMMs) versions of the same part number, can tell me what to set everything to?

Thank you for your help.

[emissary42]

The tertiary timings will be calibrated automatically during POST / memory training by the MB itself. The IO-L and RTL are a rare exception in that their values can differ between modules. In general the tertiaries are not part of the SPD or XMP and unless you are tweaking for benching / maximum performance you shouldn't touch those.

[Teodan]

Ok, so I'll leave tRTL and tRTL2 alone on auto. But I'm not clear on whether tWWDD is a tertiary timing and I should leave it alone or not. I was trying to set tRFC to 174 already but either that alone didn't help or else I was doing it wrong. That's what led to me noticing tWWDD and the tRTLs.