Do you still test new RAM?

TV2

n00b
Joined
May 10, 2022
Messages
17
How many people here still test new RAM modules with a utility like Memtest86+ when building or upgrading a new rig?
 
I haven't tested ram since you could go to best buy and they had a hardware module tester.
 
Memtest is the first thing I run after a new build. It will confirm the modules are running where I set them (or where they should be if you let it auto select or use XMP settings) it also confirms I didnt get a (rare but happens) dead module, or a dead or bad RAM slot.

I don't run it all night, just one good pass is enough for me to feel good about the settings as well as the rest of system overall and then begin OS installation.

My .02
 
the only time I have ever tested Ram was when I was having a stability issue back in the DDR2 days. I haven't come across a bad stick since, and have built maybe 100 machines.
 
RAM warranties are pretty good so no worries there however OS's can still suffer when RAM isn't running right eventually slowing down your PC and in worst case scenarios causing us to format and reload our OS's. I'm firmly in the camp that still test RAM thoroughly at it's rated speed or XMP profile. Conversely RAM can test out bad and not be bad. How is that you say? Well the memory controller resides on the CPU with most modern processors so it could also mean you have an issue with it. In the end this is something you'd want to know sooner rather than later especially with a new build. For RAM I use HCI Design's Memtest it loads from within windows and can do one full test depending on amount of RAM you have in 4-6 hrs. If you have no errors you can rest assured you did the bare minimum to avoid headaches down the road. You should follow up the RAM test with full system stress testing for the same reasons. I use Aida64 Extreme System stability Test for that taking up another 2 or 3 hrs of your time. Not long all things considered.
 
I try to go at least one pass, unless I'm in a real hurry, but I'll run at least 30 minutes. Last time I built stuff, it wasn't uncommon for auto ram voltage to often be significantly low, which usually shows up in memtest real quick. Better there than in a borked OS install.
 
On my own rigs? Never. I'll find any issues while tweaking timings, voltage, and speed. On someone elses? Always.
 
Yes, I always test memory, first thing I do after the hardware build. Aside from enabling XMP I rarely bother with memory tweaking. Since I'm usually dealing with used or relatively low end parts, I often find that I need to lower XMP clocks just a little to get stability.
 
How many people here still test new RAM modules with a utility like Memtest86+ when building or upgrading a new rig?
I have never done this shit. I install the RAM and see if it works. If I have a problem with the system, then I'll test it.
 
Interesting.
I asked the question here because in came up on another computer forum. The OP was being advised that testing was unnecessary by another long time member and that made me wonder.
I have always tested RAM on every new rig I build (and stress test too) just because - well - that's what you do! I've dealt with enough DOA parts to not just go on faith.
I wonder if it is a generational thing?
I started building my PCs in the mid 1990's, so as the other guru suggested, I may just be a Luddite.
 
I started building my PCs in the mid 1990's, so as the other guru suggested, I may just be a Luddite.
same, ive just found that as time past, parts got more reliable and i dont bother unless something seems off. building for someone else, i do a little testing.
 
Interesting.
I asked the question here because in came up on another computer forum. The OP was being advised that testing was unnecessary by another long time member and that made me wonder.
I have always tested RAM on every new rig I build (and stress test too) just because - well - that's what you do! I've dealt with enough DOA parts to not just go on faith.
I wonder if it is a generational thing?
I started building my PCs in the mid 1990's, so as the other guru suggested, I may just be a Luddite.
Nope. I started in the 1990's as well. I don't test hardware outside the case unless its a water cooling build with hard tubing. I do not run memory tests, stress tests, or any of that shit on new hardware and I never have outside the context of hardware reviews. I've worked in high volume service centers repairing computers. I'm going to tell you, software based diagnostics are not 100% accurate to begin with. I've seen Memtest86 show errors and people incorrectly assume RAM was bad only for the problem to be something else, or even BIOS settings. Tests like OCCT Intel Burn in and others place unrealistic loads on CPU's and sometimes other hardware. Yet, I've seen hardware pass those tests only to crash 30 seconds into a Handbrake encode.

I've built myself a new computer every year since that time and sometimes more than one depending on what my needs were. I've built 100's of machines professionally and I can't tell you how many I repaired in high volume service centers. I've also been reviewing hardware for 16 years. I've probably reviewed more than 200 motherboards at this point. That's to say nothing of the machines I've deployed physically in data centers or how many workstations I've deployed in schools or in offices. I've built out infrastructure in large corporate datacenters. In my experience, DOA hardware is extremely rare. I'd estimate a DOA rate of 1-2%. Now, often customers when I worked in service centers or retail stores would claim something was DOA, but a lot of the time I'd test something it was actually fine.
 
Dan, you must either be lucky, or I'm unlucky, or your stability needs differ from mine. Out of 7 computers in my backroom cluster, 6 of them would crash in gcc or show other instability at least once every hour or so. All 6 demonstrated memtest errors; I replaced the memory in one and backed off the memory clocks by a notch or two in the others, and they now run all day long. That might be partly because I tend to have at least some used / older hardware in these boxes.
 
Dan, you must either be lucky, or I'm unlucky, or your stability needs differ from mine. Out of 7 computers in my backroom cluster, 6 of them would crash in gcc or show other instability at least once every hour or so. All 6 demonstrated memtest errors; I replaced the memory in one and backed off the memory clocks by a notch or two in the others, and they now run all day long. That might be partly because I tend to have at least some used / older hardware in these boxes.
if youre cobbling together old used parts, sure test em. new out of box, issues are rare these days...
 
Well, I usually get ECC RAM, so I will be notified of errors unless the module is so screwed up it prevents the computer from starting.

Non-ECC I would test in memtest before running a valuable OS install with it. Bad RAM can scramble the disk contents real quick.

Having said that, I had an increasing number of instances where memtest variants would just hang.
 
if youre cobbling together old used parts, sure test em. new out of box, issues are rare these days...

With the ECC unreg you can buy on Newegg these days (forgot the brand name) I had 2 out of 4 kits with errors recently.

Never had problems with used RAM.
 
Dan, you must either be lucky, or I'm unlucky, or your stability needs differ from mine. Out of 7 computers in my backroom cluster, 6 of them would crash in gcc or show other instability at least once every hour or so. All 6 demonstrated memtest errors; I replaced the memory in one and backed off the memory clocks by a notch or two in the others, and they now run all day long. That might be partly because I tend to have at least some used / older hardware in these boxes.
I think you are unlucky, and dealing with used or older hardware is probably a contributing factor. As for stability, I demand stuff be as stable as possible. I fix broken crap all day. I don't tolerate my own systems crashing.
 
Sometimes. I have experienced where modules failed in real applications and passed memtest. For something critical it's good to pass several different tests. For my own stuff, not so much.
 
This has been fascinating. I've learned something new.
I can certainly understand that if your trying to make money or do it for a job you would not want to waste time on an additional task that is of limited value. Carry on!

As a person who builds my rigs more as a hobby, and repair PCs as favors, it doesn't bother me in the least to run these tests. I build them on the workbench in the basement and take my time about it. I can set it all up, fire it up, and run the tests while I go out and do other things during the day. Check out the results when I get back. All part of the fun.

Testing proved helpful a while back. It was a build that populated all 4 RAM slots and it was throwing up errors. Testing 1 stick at a time or 2 at a time showed none. Obviously a motherboard problem. I increase the RAM voltage a couple of clicks and they ran clean for 10 passes. Stress tested OK for a day. Decided to keep the board. That rig ran without problems for 10 years before it was retired.
 
It depends. If the RAM is just going to be run at stock or XMP, then no. If it's some 3200-rated stuff that I intend to run much faster, then yeah, I'll dial it up and run some tests to make sure I know where I should draw the line.
 
I can certainly understand that if your trying to make money or do it for a job you would not want to waste time on an additional task that is of limited value. Carry on!
no, thats when you do do thorough testing.

It depends. If the RAM is just going to be run at stock or XMP, then no. If it's some 3200-rated stuff that I intend to run much faster, then yeah, I'll dial it up and run some tests to make sure I know where I should draw the line.
^^ that. i gave my 3400 ram about 10min on memtest when i bumped it to 3600. when i ran it stock i didnt bother.
 
I'm going to tell you, software based diagnostics are not 100% accurate to begin with. I've seen Memtest86 show errors and people incorrectly assume RAM was bad only for the problem to be something else, or even BIOS settings.

That's been my experience as well. If a memory test fails, it doesn't necessarily mean there's anything wrong with the sticks. If I suspect a problem with RAM, it makes way more sense to me to just swap in known-good sticks and see if the problem goes away.
 
I've never tested RAM. I have 3200 sticks that can be overclocked to 3466 safely through software that came with my prebuilt PC. I've never tried to push it above that because I honestly don't know how to do it and I'm not sure how much above the 3466 I could go before I start having adverse effects.
 
I've never tested RAM. I have 3200 sticks that can be overclocked to 3466 safely through software that came with my prebuilt PC. I've never tried to push it above that because I honestly don't know how to do it and I'm not sure how much above the 3466 I could go before I start having adverse effects.

It's like overclocking anything else. Find the limit, and then dial it back one notch.
 
You might be able to raise some timings instead. Start a new thread if you want some help.
Just checked in the BIOS and I can't modify anything for the RAM there. I appreciate your suggestion though.
 
Just checked in the BIOS and I can't modify anything for the RAM there. I appreciate your suggestion though.
HP Omen 30L
this system? yeah thats why, oems lock everything down 99% of the time. ryzen master MIGHT let you control it, if they havent locked that out too...
 
I do for mine but I do manual OC so more likely to run into errors than someone using stock or XMP.
 
HP Omen 30L
this system? yeah thats why, oems lock everything down 99% of the time. ryzen master MIGHT let you control it, if they havent locked that out too...
It looks like I can with Ryzen Master, but I don't know what I'm doing. I'll have to do some research and then ask questions in a new thread later. Thanks!
 
It looks like I can with Ryzen Master, but I don't know what I'm doing. I'll have to do some research and then ask questions in a new thread later. Thanks!
no prob. neither do i, i did mine in bios. but if you start a thread im sure others will be able help out.
 
only if a problem appears... my wife's pc (my old pc) which is ancient now (i7-4770K.. with 16GB (8x2 corsair DDR3) and a gigabyte mb Z87X-UD3H ... was being super flakey and rebooting and having other problems.. ran memtest... no errors.. problem was apparently (unless my wife hasn't told me it has been crashing still?? she will sometimes not tell me this stuff for awhile..) a dying /dead cmos battery... never honestly had a system long enough for this to happen before...

So yeah if there is a problem something I will probably try at some point is memory testing.
 
I had some weird issues on the current build over a year ago. Just random reboots. I couldn't figure out why. I ran memtest86+ and eventually started seeing failures. Do you know how long 64GB takes to test? So, then I tested one stick at a time. Of course, the errors showed up on the last stick. It was G.Skill and they replaced it under warranty. Since then, any time I make a RAM settings change, I run Memtest to make sure it's stable.
 
If you value your time you should be using ECC memory. Tests rarely find problems. Don't even bother...just swap out the memory if you're having weird problems.
 
Back
Top