Memory question + adding sticks without pairs

[luxlunae]

Last year when I built my computer my memory was originally delivered to the completely wrong address so new egg sent me a replacement. By the time the original package made its way to me I'd built the computer and didn't feel like getting back into it, so I threw the memory in a drawer.


It is:


G.SKILL Ripjaws V Series 32GB (2 x 16GB) 288-Pin DDR4 SDRAM DDR4 3200 (PC4 25600) Desktop Memory Model F4-3200C16D-32GVK - Retail


Today I finally (a year later) got around to trying to install it, and I can only fit one of the two sticks I have into the motherboard without removing the cpu cooler (not interested in doing that today!) .


So now I have 48gb of ram in my computer. A quick google search told me that although installing an unpaired memory isn't as good as installing a paired, it is better than not installing it at all. (and all three sticks of ram are identical).


Are there any negative side effects that might not have shown up in that cursory search on my phone?

[Ughmahedhurtz]

In my experience, doing this can:

* Reduce performance (usually by half on the single module as it is running in single-channel mode rather than dual-channel)
* Reduce reliability (largely because you invalidate the XMP timings, but depends on motherboard and BIOS, and is the same when you mix&match modules even if you install them in pairs); rule of thumb *should* be to set timings for the slowest module. If yours were the same speed/timings on the label, that will help a lot, and you can probably write down what the timings are that have been working for you so far and just manually set them that way after installing the third module.

Most higher-end boards shouldn't suffer any serious errors as a result but you may see some really strange performance hiccups depending on whether your app is using memory pages in the single-channel or dual-channel regions.

As always, though, trust but verify. I would pull the new third module back out where you're sitting at your original 32GB and run an overnight Prime95 stress test with settings per https://www.reddit.com/r/overclockin...ram_stability/. Then, if that passes OK (and it should lol), reinstall the extra module, change your Prime95 settings for the 48GB setup and run it overnight again. If it passes overnight, you should be good to go. If you have issues, you can try backing speeds down a bit or reducing timings, but a cursory search for recommendations for that didn't yield much beyond conjecture and "don't do that" admonitions. :P

And if you see strange stuttering or artifacting in the future, pull that third module back out for a while and see if it goes away. Memory can be really tricky when you have imbalanced timings or voltages and such.

[jak3676]

Is is possible to have dual-channel mode for 2 sticks, but then single-channel mode on the 3rd stick? Or does that make them all run in single-channel mode? Does this vary by motherboard/BIOS manufacturer? I would think your BIOS/UEFI would show the end result.

Having said that, the performance impact of single vs dual-channel mode is generally small - maybe a 1-2% difference in overall game performance for most games. However there are some games that are impacted more than others. In a quick google search, I could not find anyone who has done testing on WoW with the current BFA client. (sources: https://www.cgdirector.com/single-du...-threadripper/, https://techguided.com/single-channe...-quad-channel/, https://www.hardwaretimes.com/dual-c...ce-ryzen-3000/)

RAM is one of those items where you either have enough or you don't - i.e. adding more than you need does not improve performance. So, unless you were hitting full usage on your previous 32GB sticks (but still less than 48GB with the 3rd stick), then I'd just stick to the 32GB until you want to spend the time popping off the CPU cooler to install the 4th stick. But if you're actually using between 32-48GB of RAM, then by all means leave in the 3rd stick.

[MiRai]

Having said that, the performance impact of single vs dual-channel mode is generally small - maybe a 1-2% difference in overall game performance for most games. However there are some games that are impacted more than others. In a quick google search, I could not find anyone who has done testing on WoW with the current BFA client. (sources: https://www.cgdirector.com/single-du...-threadripper/, https://techguided.com/single-channe...-quad-channel/, https://www.hardwaretimes.com/dual-c...ce-ryzen-3000/)

I'll be absolutely honest, I only quickly scanned those articles, but more often than not the tests:

A) Don't mix channels with an odd number of DIMMs (e.g. they're just testing single vs. dual vs. quad in a closed ecosystem)
B) Don't push a lot of RAM (e.g. solo player gaming performance)

With that, the question becomes, when she pushes beyond 32GB, then what happens? While multiboxing, are some of the games utilizing dual-channel and others not? Is that even a thing can happen, or does the machine just force the lowest common denominator across all applications, which would be single-channel? I don't know the answer to those questions, but if I had to guess, I'd say there is definitely going to be a performance hit in that scenario, in one way or another. Now, whether the user can actually feel the difference is going to depend on the user and how keen they are at noticing changes in their system, but until someone actually tests* using mixed channels while utilizing a lot of RAM across several running applications, I'd recommend that people avoid it where possible.

* Actually testing it with benchmarks, not feelings (e.g. the two configurations "feel" the same).