XMP 1.3 and 1.2 in two nearly identical sets of DDR3 RAM.

[BambOOs]

Hello!

I need some advice about XMP Profiles. Till today I had 2 x 4 GB 1600 MHz DDR3 RAM by Kingston installed in my PC. This set has XMP 1.3 Profile which gives 1600 MHz with CL9 at 1.65V (but I lowered this voltage to 1.56V and it's 100% stable for ~3 years).

Today I added set of 2 x 4GB 1600 MHz DDR3 RAM - also made by Kingston and with nearly identical parameters - 1600 MHz with XMP profile with CL9 at 1.65V.
Before inserting new set of RAM, I went to the BIOS and restored all RAM settings to Auto. Then I mounted the new set, booted up properly and went directly to the BIOS and loaded XMP Profile again with manual lowering the voltage to 1.56V. Rest of the settings are on Auto just like it was set before.

So, until now I have no issues whatsoever, BUT I discovered some differences between these two sets:

1. My older RAM has 1.3 version of XMP, but newer has 1.2 version. Are they compatible with each other?
2. In BIOS I can only see one XMP profile (1.3) from my older RAM. There is no way to select profile from newer RAM.
3. Also, I checked what CPU-Z says about memory: speed is 800, all the timings are OK (CL9 and rest), all is working in Dual Channel Mode. The only difference I see here is Command Rate: if I remember correctly it was 1T before, but now is 2T. Can it cause lower performance?
4. My newer RAM is Dual Ranked, older is Single Ranked and I wonder if it can cause some issues?
5. And another difference or maybe some king of bug? When I select each module in CPU-Z to check it's specification it says that my older RAM's Max Bandwitch is "PC3-12800 (800Mhz)" which is correct, but newer RAM's Max Bandwidtch is "PC3-10700 (667 MHz)" which is weird because RAM is rated at 1600 MHz (800MHz).

I wonder what should I do:

1. Leave it as it is set now.
2. Don't load XMP Profile and set everything manually in BIOS.

Thanks in advance.
Greetings!

 

[NdMk2o1o]

Don't load XMP Profile and set everything manually in BIOS.

So you have no issues running both sets apart from CPU-Z is saying? you might be reading XMP or current profile of one set and the jedec max bandwidth of the other when in fact they're both running 800 (1600 ddr) can you post SS of CPU-z all memory tabs for each set? also try hwinfo and see what that is saying for each channel, you might have dropped down to 2t for compatibility reasons with the sets being slightly different to each other, this can help with stability in such cases and does take a small performance penalty though nothing noticeable under everyday conditions.

Also if it ain't crashing or giving you any weird issues, I would say leave it as is, you'll probably find the faster timed sticks have loosened timings to match the slower timed ones which is what usually happens if you mismatch RAM

 

[BambOOs]

So, here are SS:

 

[NdMk2o1o]

Both have XMP of 1600 with the same main timings so no issues there, sometimes the reported max bandwidth is reported as max jedec speeds others as XMP, I had a similar mismatched pair not long ago that read like that though both ran at the higher speed they were rated for under XMP. if you SS the memory tab and not SPD it will tell you what their currently running at though from what you explained it seems both are running at 1600, if you have no issues or crashing etc, I really wouldn't waste much time trying to find fault where there probably isn't one also, iirc the reason they reverted to 2t was likely the fact you are now running 4 sticks instead of 2, which is normal for some platforms

 

[kastriot]

They are same spec 1.2 vs 1.3 doesn't mean anything and since it works don't bother with it.

 

[BambOOs]

Both have XMP of 1600 with the same main timings so no issues there, sometimes the reported max bandwidth is reported as max jedec speeds others as XMP, I had a similar mismatched pair not long ago that read like that though both ran at the higher speed they were rated for under XMP. if you SS the memory tab and not SPD it will tell you what their currently running at though from what you explained it seems both are running at 1600, if you have no issues or crashing etc, I really wouldn't waste much time trying to find fault where there probably isn't one also, iirc the reason they reverted to 2t was likely the fact you are now running 4 sticks instead of 2, which is normal for some platforms

Here is SS from memory tab:

Tommorow I will do some benchmarks:

1. With only my older set installed.
2. With only my newer set installed.
3. With current configuration.

I will check what timings (especially Command Rate) will be set with each of sets.
If 2T will be only with my newer set, then I will check the difference in performance. If it will be ~10-15% then I'll try set all things manually.

 

[NdMk2o1o]

Here is SS from memory tab:

Tommorow I will do some benchmarks:

1. With only my older set installed.
2. With only my newer set installed.
3. With current configuration.

I will check what timings (especially Command Rate) will be set with each of sets.
If 2T will be only with my newer set, then I will check the difference in performance. If it will be ~10-15% then I'll try set all things manually.

Probably +-5% if that, not worth staying at 8GB vs 16GB. Main things are clock speed and main timings, I will try a quick run of aida64 with 1t/2t and my current RAM settings see if it makes any difference, I know it had a bigger performance hit back in the LGA 775/ Phenom days, not so much now iirc.

 

[BambOOs]

Okey, so I did some benchmarks. The only difference is memory latency: with both sets it is about ~56 ns. When I install 4 modules, latency is about ~64 ns. When XMP profile is selected I can't set individual parameters like Command Rate, so the only way is to disable XMP profile and set all the things manually. But I won't do that for now, because the performance difference is marginal.