• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Program for basic memory testing

Joined
Feb 22, 2016
Messages
2,341 (0.69/day)
Processor Intel i5 8400
Motherboard Asus Prime H370M-Plus/CSM
Cooling Scythe Big Shuriken & Noctua NF-A15 HS-PWM chromax.black.swap
Memory 8GB Crucial Ballistix Sport LT DDR4-2400
Video Card(s) ROG-STRIX-GTX1060-O6G-GAMING
Storage 1TB 980 Pro
Display(s) Samsung UN55KU6300F
Case Cooler Master MasterCase Pro 3
Power Supply Super Flower Leadex III 750w
Software W11 Pro
New, used, DDR4 RAM I'd like to run some basic stability and error testing on. It's pretty warm here, but if the realistic option involves 18 hour of leaving a computer on and working hard. Would rather know if I bought bad RAM today instead of after it produces system issues.

Unsure which program is currently believed to have the most reliable results with current Windows architecture. So far I've found TPU MemTest and Mem Test86.
 
Last edited:
There is MemTest86, which is owned by PassMark. I don't know if anything important has been left out of the free version vs the $49 paid version.
And there is MemTest86+ which is still a free open-source product. I used to use MemTest86+ in the DDR1/DDR2 days and I would still use it if needed. There was a long period when 86+ wasn't being updated, but the latest version is 2023-05-07.
 
OCCT has a memory test. otherwise Intel Burn In Test can stress a whole system as well as Aida64/extreme has its own stress testing options.
 
OCCT has a memory test. otherwise Intel Burn In Test can stress a whole system as well as Aida64/extreme has its own stress testing options.


Thank you, bolded text is essentially what I was attempting to convey. Stress testing during a heat wave is of immensely small appeal.

And there is MemTest86+ which is still a free open-source product. I used to use MemTest86+ in the DDR1/DDR2 days and I would still use it if needed. There was a long period when 86+ wasn't being updated, but the latest version is 2023-05-07.

This looks like the winning option. Simplicity was desirable though I wasn't sure it was attainable.

My benchmark was opening +300 youtube videos simultaneously until all four sticks were being used. After rebooting I experienced a small hiccup in a game: OS and game loaded without anything seeming off. First 20 seconds or so of in-game were very choppy and then suddenly cleared without a single other occurrence of any kind. Bringing up concerns the only system change, which I'd just heavily loaded, was showing signs of malfunctioning the OS/BIOS had cleared a path beyond.
 
New, used, DDR4 RAM I'd like to run some basic stability and error testing on. It's pretty warm here, but if the realistic option involves 18 hour of leaving a computer on and working hard. Would rather know if I bought bad RAM today instead of after it produces system issues.

Unsure which program is currently believed to have the most reliable results with current Windows architecture. So far I've found TPU MemTest and Mem Test86.
I run Passmark Memtest86 for initial stability testing. When OC'ing I run partial tests first to quickly check for settings that aren't working to retune the ram faster. Then when getting what a think are stable settings, run 4 passses before using OCCT for 1hr stress testing. This also heats up the ram so I can get an idea of it's peak temps and literally test for warm boot failure because I've experienced ram OC's on the bleeding edge that were perfectly fine on cold boot but utter failure after warm boot.
 
Use the built in 'Windows Memory Diagnostic' App? which has you reboot.

The great thing about testing without the OS running is that more of the memory can be checked.
 
Last edited:
I guess the real question is what are you doing with the rig? In the basic sense, sure, stable is stable.

The reason I ask, is that memtest is a great option, but I have passed memtest then fried a socket on attempt into windows. (to be fair I was overclocking)
I also know from a lot of testing that things that can pass superpi, 3dmark, whatever you want, will fail in things like Handbrake or in compressing files.

In essence, my advise is to test the piss out of it on a spare drive with what you do regularly. If its happy, let it ride.
 
How does MemTest test the memory it and the OS occupy?
 
How does MemTest test the memory it and the OS occupy?

Memtest86 and Memtest86+ ARE operating systems.

Windows never boots. The test programs control the whole computer and therefore test a of RAM.

Other programs, like SuperPi have the problem you're suggesting. But good IT technicians will use a specialized tool like Memtest86
 
Counterintuitive that RAM is so much faster than any hard drive yet takes days to accurately test... using the Windows diagnostic. TV lost signal when I turned it off so I had to soft reset. Will probably throw a battery of options at it over the next few days while I'm sleeping.

8 hours is about my patience level for things to do indoors that don't involve this machine, a box fan close enough it's practically grazing skin, and A/C on high.
 
Counterintuitive that RAM is so much faster than any hard drive yet takes days to accurately test... using the Windows diagnostic. TV lost signal when I turned it off so I had to soft reset. Will probably throw a battery of options at it over the next few days while I'm sleeping.

8 hours is about my patience level for things to do indoors that don't involve this machine, a box fan close enough it's practically grazing skin, and A/C on high.

Its because RAM errors are exceptionally rare. Even when your RAM is broken, its rare to "catch" the RAM acting up.

Your computer can perform "one test round" of RAM within 2 seconds (ex: 32GB of RAM at 3200 MHz is read/written from at 25GB per second per channel, and typically computers are dual-channel so 50GB per second is the practice at 3200 MHz).

But when I ran into my last issue, in a 24-hour test I would only have ~10 errors.

So for literally millions of test-cycles, only a dozen times or so would there actually be an error. We're talking 99.9999%+ reliable. But guess what? This only led to ~2 hours or less of stability in practice, because RAM only needs to mess up once for your whole computer to blue-screen of death or worse (ex: a critical write error during Windows Update could permanently destroy your Windows installation or device drivers).

--------------

The only way to ensure you really get that 99.9999999%+ reliability needed (to ensure your computer goes error free for more than a few hours at a time) is to do over 24-hours of constant testing. Its very important to get this correct, so its worth 24+ hours of testing (or more if you're paranoid).
 
The only way to ensure you really get that 99.9999999%+ reliability needed (to ensure your computer goes error free for more than a few hours at a time) is to do over 24-hours of constant testing. Its very important to get this correct, so its worth 24+ hours of testing (or more if you're paranoid).

I was aware and looking for an easier option if it existed. Had errors showed up during 1/3 day test that was aborted you can bet that a refund would be incoming though.

My question revisiting whether any program had quietly become indispensable in the last decade was answered. What I did in the time period so close to installation to quickly fill the RAM would've been much better handled in a dedicated program. Memory training might explain the event that worried me enough to seriously look for errors. Or dynamic optimization within the game adapting to a fairly large hardware change.

All of the answers were deeply appreciated.
 
If you are not booting into windows Memtest86+ is the only choice because it runs from a USB drive. If you are in windows, go with memtest5. Links in the text.

My benchmark was opening +300 youtube videos simultaneously until all four sticks were being used. After rebooting I experienced a small hiccup in a game: OS and game loaded without anything seeming off. First 20 seconds or so of in-game were very choppy and then suddenly cleared without a single other occurrence of any kind.
That's not really a good test.
 
If you are not booting into windows Memtest86+ is the only choice because it runs from a USB drive. If you are in windows, go with memtest5. Links in the text.

The inbuilt W11 diagnostic actually reboots into a GUI that suspends the OS from loading anything into memory. My concern was distinguishing basic memory test from program derived to act as a torture test. With as much cooling as it has and as slow as this RAM is my needs are firmly on the other side of OC. Beyond that I was unconcerned with making a bootable drive or not so long as the program was efficient at finding memory errors or other issues.

That's not really a good test.

Obviously with only 8GB of memory and W11 using half of that I'd always been afraid of accidentally locking up by selecting the open all links option in my browser's bookmark manager.

Was it stupid, assuredly. Certainly easier than setting up RAMDISK.
 
The inbuilt W11 diagnostic actually reboots into a GUI that suspends the OS from loading anything into memory.

If really clever it can move itself around (or force itself to run from the hard drive) so all the memory can be tested; but I don't know how challenging this would be to program.

As I recall it is just text reporting and not a GUI, but that is just me being finicky.
 
Last edited:
I was wondering if a Kernel was even running, or if the Kernel can be driven to swap space, or even moved around in RAM (the best solution).
 
Last edited:
HCI memtest or Karhu ram test. both are used in the pro overclocking scene for a reason. $5 for HCI pro license or $10 for karhu personal license. I use both to test my ram tuning. The other options mentioned here are good too but might be a bit dated.


Best of luck to you with tuning!
 
As I recall it is just text reporting and not a GUI, but that is just me being finicky.

Every software person I know refers to this user definable, step above command line, arrangement as a GUI. Anyone familiar with a keyboard should be able to follow the on screen instructions and explanations to TAB + Arrow Key around inside it.

I was wondering if a Kernel was even running, or if the Kernel can be driven to swap space, or even moved around in RAM (the best solution).

You hit upon main question I had about the various paid software. I suspect Microsoft went the economical direction of loading a bootable in their consumer diagnostic tool.

In essence, my advise is to test the piss out of it on a spare drive with what you do regularly. If its happy, let it ride.

There is a chance I'm missing something here. Granted I typically structure my threads into a highly simplified rejuvenation of age old topics that will produce a fresh search result. :)

Reasoning you are exploring for using a spare drive escapes me at the moment. I have a few ideas, but would rather let you expand upon what this meant if you would be so kind.
 
Back
Top