Upgrade issue with memory failure from 16GB to 32GB (Last updated 18th November 2020)
Upgrade issue with memory failure from 16GB to 32GB (Last updated 18th November 2020)
Hello - your support means a lot and I'm glad you're on my thesis thank you page. Contents - Background Questions PC Specifications
Background I'm a PhD researcher focused on handling big image/video files (multiple TIFFs). My 16GB memory often reached its limit. I added another 16GB of RAM, using the same kit as before (HyperX Fury 16GB, 2666MHz DDR4). After installation, I experienced several crashes and BSODs, including a kernel mode heap corruption alert. Windows worked briefly before failing completely. I was pretty sure XMP profile 1 was enabled. A search suggested possible faulty memory, so I ran the Windows Memory Diagnostic tool three times—three times it confirmed hardware issues. After replacing the RAM, diagnostics showed no problems. Further research indicated the memory might be defective and is sold by the manufacturer, who are willing to recheck if the issue is intermittent. Since then, I've been reading a lot and wondering why RAM seems so tricky. I've heard that even with identical kits and part numbers, compatibility can vary—processor limits might not support four sticks as well as two, or voltage requirements aren't met. Other possibilities include a faulty motherboard slot or a bent CPU pin that only appeared after memory was fully seated. A hardware forum post mentioned this could be a full-time task, but it seems worth trying.
Questions: Would using XMP disabled improve stability with all 32GB installed? Is capacity more important than speed for my work? Could the processor or motherboard specs be limiting performance even with identical RAM? Is there a clear sign of a hardware defect, or could it be something simpler like a bad slot or pin? How likely is a BIOS update to fix this if I'm on the latest version and no notes mention it? What would you think about running diagnostics per stick and adjusting settings individually? Are these memory modules very similar in performance?
Start by saving this file and verify each stick individually. If no issues appear, it confirms the sticks are in good condition. Only purchasing a complete matching set guarantees reliability. Ryzen tends to be particular with RAM performance. If your sticks seem intact, you can safely use them at their standard slower settings.
Yes, it's essentially reverting the XMP profile to its default setting.
I would purchase this set. You can find it here: https://www.microcenter.com/product/6188...u4b--black. Consider offering your existing unit on eBay or nearby.
You aim to turn off XMP, which can involve selecting the "Auto" option.
I checked the datasheet again—your spec sheet lists the same parameters as the factory timing: DDR4-2666 CL16-18-18 at 1.2V with the XMP profile #1. Disabling XMP shouldn’t affect performance if both versions match. It’s likely just a coincidence or a minor variation in your setup.
I believe Ryzen sets the default to 2133 MHz when auto-configured.
It relies on the specific memory's Jedec setting and the board's auto-profile selection. While many modules support multiple Jedec profiles, most 2400mhz memory uses the same profile. The variation often comes from sub-timings—tighter when XMP is active, which can lead to instability.
I addressed the issue promptly by sending back the new RAM after it had been thoroughly tested on a memory bench. The seller confirmed no problems and agreed to return it. They even rechecked it, and everything worked perfectly. For added confidence, I ran memtest86 with default settings and received a clean result. Later, I ran Windows Memory Diagnostic again, which also reported no faults. This made me even more uncertain about choosing between the two tools.
I’m torn between relying on memtest86 or opting for new RAM. I’m curious how many passes memtest86 would need to find a problem that WMD missed. Could WMD produce false positives? To investigate further, I tried booting into Windows and checked for any damage. The results were mixed, so I’m left wondering what the best course of action is.
My main concern is whether I should trust memtest86 or proceed with replacing the RAM. I’ve searched online for advice on Windows memory diagnostics and false positives, but the information seems inconsistent. I’m hoping you can help me decide based on your experience.
You're working with a 1600AF and have four sticks on it. Early versions of Zen, especially with four chips, had inconsistent results—sometimes hitting, sometimes missing, particularly if you had 4 chips. The performance jumped significantly after adding the i/o die in Zen2. I've tested nearly every Ryzen revision and likely bought around 25 of them, many of which are still selling well.