[SOLVED] Will they run in dual channel?

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
So I have a
CORSAIR
VENGEANCE® RGB PRO 16GB (1 x 16GB) DDR4 DRAM 3000MHz C16, model number CMW16GX4M1D3000C16 ( https://bit.ly/3UWNWzv ) RAM
paired with a

Ryzen 5 3600 processor on a B450M motherboard.

I was just planning to upgrade my RAM to 32GB (2 x 16GB), but the identical model is not available since quite a long time now (always out of stock). So I looked for other RAMs with identical specifications and quite easily found

CORSAIR VENGEANCE® LPX 16GB (1 x 16GB) DDR4 DRAM 3000MHz C16, model number CMK16GX4M1D3000C16 ( https://bit.ly/3GmgmyZ )

The on-paper technical specs of both the RAMs are identical.

Even the model numbers are almost identical.

CMW16GX4M1D3000C16 - VENGEANCE® RGB PRO
CMK16GX4M1D3000C16 - VENGEANCE® LPX

So my questions are:

(1) Will they run in dual channel if paired together?

(2) Since I run my single channel RAM overclocked to 3200MHz without any issue, will I be able to run this mixup overclocked at 3200MHz as well? Or will the only run at stock 3000MHz together.

(3) Will dual channel at 3000MHz perform better than single channel at 3200MHz?

Thanks for any guidance.
 

Darkbreeze

Retired Mod
They might, they might not. Even if they were EXACTLY the same model, they might they might not. ONLY memory purchased together in a single kit is guaranteed to "play nice" together.

Please read section two titled "Mixed memory" at the following link. It will answer your question, in detail. If you have further questions afterwards, I'm happy to try and answer them.

 
Reactions: Dark Lord of Tech

Darkbreeze

Retired Mod
They might, they might not. Even if they were EXACTLY the same model, they might they might not. ONLY memory purchased together in a single kit is guaranteed to "play nice" together.

Please read section two titled "Mixed memory" at the following link. It will answer your question, in detail. If you have further questions afterwards, I'm happy to try and answer them.

 
Reactions: Dark Lord of Tech

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
They might, they might not. Even if they were EXACTLY the same model, they might they might not. ONLY memory purchased together in a single kit is guaranteed to "play nice" together.
Thank you so much for that quick reply. The link is very helpful as well. Going through it, I suppose I'd rather buy a single kit of dual-channel certified RAM as you mentioned above.
Thanks again with regards.
 

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
Greetings @Darkbreeze.
So I finally took my chances and bought a new CMK16GX4M1D3000C16 - VENGEANCE® LPX and paired it with my existing CMW16GX4M1D3000C16 - VENGEANCE® RGB PRO.
Good for me that they are running in dual-channel configuration as far as I can understand.
The latest version of CPU-Z (2.03) is displaying memory channel as 2x64-bit which got me little confused. So I had to check for some answers and found out that all newer versions of CPU-Z since version 2.00 are displaying CHANNEL as 2x64-bit.
I also checked with CPU-Z ver. 1.99 and few more older ones and they are displaying CHANNEL as DUAL.
The VALIDATION of every version (older and newer) I tried resulted in DUAL CHANNEL (128-bit) DDR4-SDRAM.

Posting a few images of my validation results here.

CPU-Z ver. 2.03 Results





CPU-Z ver. 1.99 Results



Just hoping if it could be of help to anyone still looking for answer.

Thank you once again for you time @Darkbreeze and everyone else caring to read this post.
 

Darkbreeze

Retired Mod
"Something" is wrong. DO you have XMP enabled?

Because based on those screenshots your memory is running at timings of 20-20-20-38-58 when they SHOULD be running at timings of 16-20-20-38-58. Likely this is because they are different sticks.

You might want to try manually setting the CL latency to 16 rather than 20 in the advanced memory timings section of your BIOS. The difference between CL16 and CL20 is pretty significant and it likely why you are not seeing any improvement, in fact, you might see a REDUCTION in performance if your original 16GB kit WAS running at CL16 before.
 

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
"Something" is wrong. DO you have XMP enabled?

Because based on those screenshots your memory is running at timings of 20-20-20-38-58 when they SHOULD be running at timings of 16-20-20-38-58. Likely this is because they are different sticks.

You might want to try manually setting the CL latency to 16 rather than 20 in the advanced memory timings section of your BIOS. The difference between CL16 and CL20 is pretty significant and it likely why you are not seeing any improvement, in fact, you might see a REDUCTION in performance if your original 16GB kit WAS running at CL16 before.
No, not XMP, but I have manually set the RAM Frequency to 3200mhz, as I was running my single RAM stick at 3200mhz. Actually I didn't even touched in the BIOS after installing the new RAM stick. Just checked in the task manager and it displayed 3200mhz. And then in CPU-Z. That is all.
Since both these RAM sticks are 3000mhz sticks, XMP will run them at 3000mhz and I'm running them at 3200mhz, I didn't wanted to downclock them.
Should I switch to XMP from manual 3200mhz? Do I still need to tweak the CL Latency manually if I switch to XMP @3000?
 

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
So I switched DOCP (XMP) on in BIOS and tried CPU-Z validation again and got these results now:

CPU-Z 2.03



CPU-Z 1.99



The timings are now 16-20-20-38-70 but still not 16-20-20-38-58.
Should I still tweak the latency manually?
 

Darkbreeze

Retired Mod
Now it is right. And now you should probably see some amount of gains in performance. The last timing, you can certainly TRY it at 58, but do not be extremely surprised if it doesn't like it. It's fine though, it is not a huge inconvenience to have to start over with your BIOS configuration if doesn't want to POST with that setting. You can always just remove the CMOS battery to reset the BIOS if it won't train or doesn't automatically revert, and then reconfigure it again. Also, you can play around with the settings a bit to see what works but if you make manual changes to your memory configuration and it DOES POST, it's a very good idea to test the configuration with Memtest86 to make sure it is stable.

Truthfully, considering you have two different DIMMs, you might not be ABLE to get that fifth timing down to 58 but you can certainly try.

You'll find some instructions on testing the memory configuration towards the end of my guide at the link below. I'd really only be worried about running four passes of all 7 tests in Memtest86.

 
Last edited:

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
I ran the Memtest86 as you suggested and no errors were detected. I tried all of the 13 tests for 1 complete pass with both, XMP @3000mhz and manual oc @3200mhz without any issue detected.



I also tried gaming and video rendering of the same project to test any improvement in timing.
Unfortunately, I still couldn't notice any FPS or Frametime difference in gaming even with lower CAS Latency.
However, rendering the video @3000mhz XMP with lower latency took about a minute longer than @3200mhz with higher CAS latency.

The only noticeable difference is using USERBENCHMARK software. Their RESULT page is actually showing that the RAM performance is getting DROPPED by about 3-4% when running at XMP 3000mhz.
I have actually tried both configurations (XMP @3000mhz and manual @3200mhz) 5 times each. The results every time is ranging in 84.x% on XMP and 87.x% at 3200mhz. Even touched 88.x% twice on 3200mhz.

So apparently 3200mhz at 22-22-22-53-75-1 is somehow performing better than 3000mhz at 16-20-20-38-70-1 at least on paper.
I have absolutely no clue how or what is going on here but the numbers are speaking for themselves.

Can you suggest something that could be used to actually perform a noticeable difference between the two configurations? I'm a bit curious now. But I would also like to mention that I am not experienced at all to mess with the advanced BIOS settings. So don't want to touch them any more than switching between XMP and manual frequency.

Thank you for taking interest and replying to so many queries.
 

Darkbreeze

Retired Mod
One pass is never sufficient if you've made changes of any kind, manually, to the memory configuration. If you change even ONE setting, DRAM voltage, primary, secondary or tertiary timings, clock frequency, anything to the memory configuration other than what the board automatically defaults to or the XMP/DOCP profile sets it to, you NEED to run FOUR full passes. Anything else is simply a waste of time. One pass tells you nothing. Most often errors are not detected until the third or fourth pass especially if they are not especially egregious errors.
 

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
One pass is never sufficient if you've made changes of any kind, manually, to the memory configuration. If you change even ONE setting, DRAM voltage, primary, secondary or tertiary timings, clock frequency, anything to the memory configuration other than what the board automatically defaults to or the XMP/DOCP profile sets it to, you NEED to run FOUR full passes. Anything else is simply a waste of time. One pass tells you nothing. Most often errors are not detected until the third or fourth pass especially if they are not especially egregious errors.
Greetings @Darkbreeze. Actually it took 1 hour 20 minutes to complete one pass so I didn't tried to go further as I have too many random power outages in my area. But I will try to run the tests overnight counting on my luck to see what comes out.
By the way, I didn't touched anything in the voltage section. It is all set by motherboard defaults. Actually I didn't touched anything at all in the bios other than RAM Frequency set from Auto to 3200.
The voltage is automatically being set to 1.350v at 3200mhz by the motherboard itself. And both these RAM sticks have tested voltage 1.35v in their respective technical specs mentioned on official website.

And yes, most importantly, DRAM Latency is set to Auto in both cases (XMP and manual) with no option to change the values.
I tried turning on XMP (low latency) and the then set the frequency manually to 3200 but then there is the infamous BSOD every time. Switching back to 3000 on XMP or 3200 manually are working fine however.
 

Darkbreeze

Retired Mod
Yeah, if you manually change the frequency you just about MUST change the timings. Timings that work for one frequency are not likely going to work for another frequency so if you enable XMP which puts a lock on the static configuration of the timings, but then change the speed/frequency, that is just about never going to work.

My recommendation would be that you simply reset the BIOS completely, save, exit, back into BIOS, enable the XMP/DOCP profile and leave the timings and frequency at whatever the profile dictates they are supposed to be, save settings, exit and leave it that way. Especially since you have two different DIMMs.

The MUCH better option, would be to either find another identical DIMM to one of the DIMMs you have now, to put with it, or, even much better, sell those DIMMs and buy ONE KIT, that has two identical DIMMs in it that are factory matched. That is how things are going to work the way they are supposed to.
 

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
I did as you suggested and reset the complete BIOS to its default and now Its running at default DOCP profile and nothing else is touched in the OC section. Everything is set to its default. But the DRAM timings are still locked.

The result is same 3000mhz @ 16-20-20-38-70



and I still can't figure the performance difference between 3000mhz at 16-20-20-38-70 and 3200mhz at 22-22-22-53-75.
Is the difference even noticeable in day to day work? Everything is as smooth as it was earlier. No lags, no crashes, no BSOD, no noticeable boost, neither any noticeable drop.

Can you suggest
something, any particular game or any other benchmarking tool that can clearly show me the difference brought up by the latency or frequency? I want to test both, XMP (low frequency with lower latency) vs manual OC (high frequency with higher latency).
Is it even possible to notice the difference in normal usage?
 

Darkbreeze

Retired Mod

Realbench is usually a good indicator of memory performance but so are most synthetic benchmarks like Cinebench, 3dmark, etc.
 

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
Thank you @Darkbreeze. I will compare Realbench, Cinebench R23 and 3dmark scores on both configurations 5 times each and will finally make my decision on which configuration to use based on the average results of all 3 tools.
Thanks again for giving you time and advice. Regards.
 

ajay12131

Distinguished
Jan 28, 2014
33
1
18,535
0
Hello @Darkbreeze. I have finally found the best working configuration for my RAMs with clearly noticeable performance gain in everything including gaming. There is a clearly visible smoothness in overall system performance. Even though the latency is even higher than 3200mhz, still I can actually feel the performance gain.

This is what finally worked for me:



The frequency is set to 3266mhz and the FCLK to 1600mhz. Everything else including voltage is set to AUTO.

The latencies are now even higher at 24-23-23-54-77. But there is actually a very noticeable performance gain.




Though I cannot tell the FPS difference in games, but there is clearly noticeable smoothness in games like Red Dead Redemption 2, Spider-man Miles Morales, Elden Ring and AC Valhalla even at lower FPS as 30 which is absolutely visible in terms of graphics. These are the only four games I have tested yet and all four are running smoother than earlier.
 

ASK THE COMMUNITY