Friends don't let Friends install Used RAM | Hacker News Reader

Friends don't let Friends install Used RAM

miscdotgeek.com

22 pointsgeocrasher38 comments

Very low on detail, no analysis what caused this. I've never had issues with ram upgrades. Ram either works fine, or it reports errors in memtest (or reports errors under high load).

Chance of ram issues causing other hardware failure... ZERO

Agreed.

Odds probably the RAM was mismatched, and the BIOS did not choose conservative settings. Also could be voltage mismatch, and then your low voltage modules could fry (unlikely though).

Once upon a time I bought a new laptop and all's well. Several months later, however, it developed a problem: Within 10 minutes of connecting to the internet using the onboard modem (this was really a while ago) the OS will crash after ~10 minutes of normal operation. Various software patches and driver updates were promptly applied but to no avail, once connected no matter what actions you take it will consistently throw a BSoD without any hint of trouble after a few minutes. Considering the fact that the system was 100% stable offline it seemed fair to conclude that the modem was faulty.

Except that when I procured an external modem it reproduced the exact same issue.

A lot of time was wasted trying to figure out what happened. Several fruitless rounds of OS reinstalls later I started to swap test any removable component out of desperation (there really wasn't many on a laptop) and voila the problem went away once a particular stick of RAM had been removed. Later tests found that it will pass most tests in MemTest86+ except this a copy operation over repeated runs.

I have no idea why the symptoms were so peculiar, perhaps when using dial-up connections the OS had been repeatedly loading the network stack in the memory address close to the affected blocks and after a while it will inevitably encounter a IO error and crash. This is just one personal anecdote, but with a complex system it is certainly possible for RAM issues to appear like other kinds of failure.

Are you sure you didn't kill all of those components with some huge electrostatic charge you had on yourself? And blaming it on some other random thing? How exactly will a mystery RAM fry a GPU?

I've added some notes to the end of the article that clarify this.

On that note; friends don't let friends buy cheap thumb drives. We got a batch of them in high school and they gave us all one. Free stuff! They fried the USB controller of every single PC that they were plugged into. The replacement motherboards were not free.

Not to mention, when you manage to get a "BadUSB" -- in this case you might be doomed ... you might even don't recognize, that a strange force have taken control of your computer and internet connect.

Anecdote: Some kid brought a pen drive into our school and it blue screened the PC only when he copied a certain image file to it. Anything else was fine.

the first usb thumb drive i ever got had a capacity of 64mb. it went through multiple wash cycles after accidentally leaving it in my pockets many times over the first couple of years of its life.

recently when i was doing some cable spring cleaning i found it hiding in a box with usb and ethernet cable. I plugged it in and surprisingly its still working and the data on it are intact.

You can use used RAM just fine. Put it through a very thorough test before using the machine (and you should do that with new RAM just the same). And seeing new RAM fail within 90 days of purchase or so is not rare at all, even for very expensive brands. The risk is mostly in how you handle the modules, if you do that well and whoever removed it did that well too then there is no risk that you wouldn't have if you bought new.

The only reason you won't know about a lot of RAM failures is because ECC isn't as common as it should be.

Flagged for jumping to conclusions and not understanding the basic material in spite of being a self-declared professional in the field.

This one time, a guy died in a car accident; friends don't let friends drive.

This one time, a guy died in a bicycle accident; friends don't let friends ride.

And so on.

Why is this on HN? Seriously a blog post made up of reaction images with no concrete facts and the strangest conclusion.

His work PC is supposedly very important to him. Yet the same blogpost tells him that he unlocked cores and overclocked his CPU. Further, his PC felt "dog" slow because he installed a 32-bit OS. As if that ever made a day-night difference in performance when doing mild testing.

I hade a similar problem with one of my servers. The SSDs would stop responding at odd times. Strange lockups and other hardware wierdness. When I tried putting FreeBSD on it it would kernel panic at odd times during the instalation.

Turns out the CPU aux power connector was not hooked up. I must have knocked it loose while working on the innards a while back. The machine has been rock solid ever since I reconnected it.

I wouldn't read this expecting a condemnation of (or evidence against) installing using RAM. It's a comedy of errors brought about by the hero's hubris (and tiredness). The author is not really suggesting that you should never install used RAM.

Consider it a longform version of https://xkcd.com/349/ .

You get me.

One question on behalf of this: Has anybody experiences with mixing RAM types today? My problem is, that new computers most of the time come with 4GB pre-installed (1 module), but I want to have 8GB. Get the same RAM (DDR3) is rather unlikely -- so, would you add a different RAM module to the preexisting memory module or is it better to replace it with a pair of new RAM modules?

Any experiences, opinions?

Yeah, I've gone through 4GB -> 4GB + 8GB -> 8GB + 8GB upgrade route a few times. It works if you're short on money and can't upgrade at once.

But yeah, buy decent, trusted RAM, and run it through memtest overnight.

If you want dual channel(twice the memory bandwidth), the modules should be equal.

In theory, yes. In practice, you get ~20% more bandwidth and almost no improvement in application performance.

If it doesn't work, switch the order. The timings of the modules might be different.

My experience comes from an Ivy Bridge machine with a B75 chipset and 4 RAM slots, YMMV.

I had a single 2GB 1333MHz stick and bought another one second-hand. The sticks had different densities (one had 4 and the other 8 memory chips on) but timings were the same. The system would not boot with the sticks on the same bank (same slot color) but when I moved one stick to a different bank, the system booted and BIOS indicated the memory was operating in dual channel mode at correct timings. I didn't experience any issues with the system for 1 year.

If your machine only has two RAM slots with a single bank, then it's a crapshoot on whether the system will boot. With two different banks you can usually mix non-matching memory.

Huh, wierd, to my knowledge, in order for dual channel to work, they have to be on the same bank.

Yes, I would have expected the same.

Thanks to you and all others!

Unfortunately, the computer I am about to buy has only one bank (smaller size computer). I also was thinking about using the dual channel mode -- but as the speedup would not be that great, I could do without. But I guess too, usage of different modules on the same bank might be a crapshot.

The guy who sells me ram suggested I don't buy two of the same 4gb modules but the 8GB(2x4) kit of the same brand for £1 more as they're tested together. Read into that what you will.

Just get the additional RAM from the same vendor (or Crucial or Kingston). Your mainboard will automatically change the settings to the lowest common detominator so it will work just fine.

And what exactly was the problem? How exactly the RAM caused this?

I don't know either. The article is a graphomaniac mess with little structured substance to it, and even less proper troubleshooting discipline present in the author's steps.

That was kind of the point of the article really. I screwed up. Made mistakes. The frenzied approach to the article reflects the experience I put myself through. I added a summary that hopefully drives home the point better and offers some hindsight to it. Thanks for your comments, it made me realize what I left out.

The only big risk with new RAM is if it has a lower speed than the other RAM in your system, as (correct me if I'm wrong) that can slow down your other RAM modules.

I don't get it. What caused the problems? The bad ram or the 32 bit windows? It makes no sense that a 32 bit windows only acts up after trying bad RAM.

A bit of a misleading title, in the end he had a 32bit installation of Windows... he upgraded to the 64bit version and everything was fine. No RAM to blame.

I think that was just with his fresh install which caused additional problems, he still had other issues with his original install, which I assume was 64 bit. It's not exactly clear from his writing though.

Also this makes me wonder why 32 bit versions of Windows 10 even exist. Are 32 bit x86 CPUs even being made? I thought they were phased out a long time ago, and any hardware from that era would struggle to run Windows 10.

TLDR: ESD followed by cargo cult thinking

Correlation does not imply causation.

Used RAM is fine. Article not worth your time.

Bullshit article

Reply on news.ycombinator.com