Hi all, i have a simple/stupid question about RAM.
Recently i have had replace my mainboard, gaining 2 more slot for RAM.
Actually, im using a 2x8gb, and im searching for another same kit for reach 32 total gb, but i dont know when i will find It.
Im using some application in jails, and service used RAM is around 6-7 GB, leaving the zfs cache with less available RAM.
Im not having problem, but i have an unused 2x4gb kit, different chip but same spec… It can be used for reach an extra 8gb of total available RAM, or breaking the “Power of 2” rule (sorry, i dont know if is correct in english ) will have a worst impact on system performance?
It’s not recommended, but if the kit is the same speed and of similar timings, you may be able to get away with it. I ran 48GiB when I was first starting out, a mismatched 2x8GiB Corsair kit and a 2x16GiB Crucial kit. I was running them well under their rated speeds and manually configured the timings to match that of the slower kit and it worked okay for about 6 months before I upgraded.
I’d still absolutely recommend getting yourself a matched kit because they have been evaluated by the manufacturer for stability. Anything else YMMV, and any nuclear fallout, damaged relationships, or otherwise world-ending calamities fall solely under your own responsibility.
Yes, im aware of this, and for sure the final “goal” is find an identical kit.
Having this kit unused, I wanted to understand if it could be a temporary solution to allocate more resources, or if the risks are greater than the benefits (or even make the situation worse).
I’d say if you care about data protection, wait until you can get a properly matched kit. Introducing variables like mixing and matching memory can cause system instability unless you’re very careful (and even if you’re very careful, things can always go wrong, best not to throw extra possibilities into the mix).
I’d take a somewhat contrary position and say that it’s highly unlikely you’d have problems with stability or data integrity. Any issues arising from mismatched (but otherwise sound, which I’m assuming) RAM are going to affect performance, not really stability. But I’d expect the performance gain from having more RAM to outweigh the loss from your system not being able to interleave that RAM as it could were it all matched.
I’m not fully knowledgeable on how mismatched memory is handled in systems at a low-level, and I can only assume that it would vary from board to board with different memory controllers & configurations etc.
If they are the same spec then yeah it is most likely fine, as I said I was running a mismatched set for a fair while without issue (after manually setting timings, as the system would refuse to boot otherwise!). I just personally wouldn’t want to run it in a production system where I’m handling a large amount of data I want to ensure remains safe in transit (at which point you’d want to be looking at ECC in any case).
In most cases it’s a better idea to focus on capacity over speed, so running the DIMMs potentially slower but with a higher capacity certainly is an option, though if it’s only a temporary solution until it’s replaced with a 32GiB kit it’s more a matter of weighing up whether or not you want to wait, i.e. is it working okay now, is there reasonable benefit to temporarily increasing capacity in an unsupported way in the amount of time it will take to obtain a 32GiB kit?
If mixing and matching I’d consider this probably the most important thing, if they’re the same model (speed & timings) they’re very likely to run without issue. I think I should have phrased ‘kit’ better as that’s typically what memory is sold under, I have no issue using multiple of the same model.
At the end of the day though there’s nothing really wrong with playing about with it and seeing if it works. While I haven’t personally had major stability issues mixing memory (even between models), personally it is just not something I’d want to run 24/7, though I totally understand how opinions can vary on that.
Really thanks to all for your valuable points of view.
I don’t have specify before, but my system dont have for sure to handle large amount of data/have mission critical data, or anyway nothing that is not well backupped.
All the ram i have, besides from different size/manufacturer, have identical JEDEC spec. The mainboard bios is also very flexible about settings, dont think will be a problem.
I think i will give a try, after have run a long memtest obv
I do agree that a kit would be better if one can find it.
They all are of the same type, speed and frequency so I decided to just go with them.
But I’d not advice it. If I lose my data then it’s on me. If someone else loses data because of me then I will never forget it until the end of my life
The one thing I didn’t see posted well was to run a RAM test. @oxyde did mention it a the end of the sentence, obviously
If you install the extra RAM, ensure you run MemTest86+ for a long time, I personally will run for over 24 hours (up to 3-5 days depending on the importance of the system) or 5 COMPLETE test cycles, whichever is longer.
I would also run a CPU Stress Test as the RAM typically directly interfaces with the CPU these days.
As for matching RAM, well what has been said is true. Use similar timings, they don’t have to be same make/model. Same Voltage as well. If it passes the RAM Test, it should be good. If you were building a new system then of course you would buy all matching RAM, and if your system was extremely important, well a good solid series of tests would be good.
) will have a worst impact on system performance?
