My memory is acting up?
My memory is acting up?
Hello,
I've been experiencing intermittent crashes while gaming on my PC for several years now. Recently, playing Dota 2 often results in a crash to desktop without any warning signs. At times, the game freezes completely instead of crashing, and I see Nvidia driver reset entries in Event Viewer. This issue seems to occur regularly, with only occasional exceptions. One day the entire system froze, but after turning it off and on again, it started responding normally. During this event, several Nvidia driver errors were recorded in Even Viewer.
There were also crashes without any Event Viewer logs, which produced crash dumps and indicated access violations.
I encountered a similar problem a few months ago and was unable to pinpoint the cause, so I reinstalled Windows. Since then, things have stabilized, except for the past few days.
What I've tried so far:
- Updated DDU and installed the latest drivers
- Updated 566 drivers via DDU
- Ran sfc /scannow
- Turned off all app overlays except Steam
- Stopped Afterburner (aware that Riva can trigger crashes)
- Disabled certain Wallpaper Engine plugins linked to RGB
- Turned off Discord hardware acceleration
- Disabled EXPO profile last night, allowing smooth gameplay without crashes
No errors were logged during the EXPO profile changes.
Could it be that the EXPO profile becomes unstable over time? If so, how should I adjust my RAM settings to 6000MHz instead of the default 4800MHz?
PC Specifications
- Motherboard: ASRock B650 PG Lightning
- CPU: AMD Ryzen 7 7800X3D
- RAM: G.SKILL Trident Z5 Black RGB 32GB (2x16GB) DDR5 PC5-48000 6000MHz CL36 F5-6000J3636F16GX2-TZ5RK
- GPU: Gigabyte RTX 4090 Gaming OC
- Power Supply: MSI MPG Gold A850GF
- Operating System: Windows 10 Pro
- Storage: 2 x M.2 NVMe SSDs, 1 x SATA SSD, 1 x 7200RPM HDD
Software used during gaming:
Discord, Steam, Steelseries GG, FanControl, Wallpaper Engine, TranslucentTB
There seems to be an advantage in manually setting RAM to 6000MHz rather than relying on EXPO. I’m unfamiliar with these changes and don’t know what adjustments would help stability.
Edit: Before this issue, there were no driver or Windows updates.
With EXPO, you'd conserve your time and effort significantly too.
Indeed, this also applies to XMP on Intel CPUs.
Prior to proceeding to manual RAM OC, it's important to ensure your RAM is functioning properly.
If it doesn't operate without errors, it's wiser to replace the entire module instead of attempting manual adjustments, which could lead to further complications.
Regarding RAM testing, run memtest86:
Link: https://www.memtest86.com/
Instructions: https://www.memtest86.com/tech_creating-window.html
A single full pass (covering all 15 tests) is the minimum. Two passes are preferable, and four are acceptable.
Given the time required, it's best to run it overnight.
- 8GB RAM: ~1 hour per full pass → ~4 hours for 4 passes
- 16GB RAM: ~5 hours per full pass → ~20 hours for 4 passes
If no errors appear, your RAM is functioning correctly.
If errors occur, consider replacing the RAM.
For thorough testing, 32 full passes are recommended (per test #7).
Additional guidance:
https://www.memtest86.com/tech_individua...descr.html
However, most users won't undertake 32 full passes due to the time investment, especially with large RAM capacities. The consensus among support teams is that four full passes are sufficient.
Proceed with four full passes while your system runs at 4800 MT/s and verify stability.
Once confirmed stable, manual RAM OC may be worth considering if you're prepared to invest time and patience.
Manual OC is extremely time-consuming.
XMP/EXPO is another method of RAM optimization. Enabling these profiles means the manufacturer has already validated stable frequencies across various timings. They then embed these settings into an OC profile, commonly known as XMP for Intel and EXPO for AMD.
The concept behind manual OC involves:
* Defining the standard JEDEC frequency, timings, and voltage your RAM supports (e.g., G.Skill typically runs at 4800 MT/s with timings around 36-36-36-96 at 1.10V). You can check this via HWinfo64.
* Adjusting these parameters—raising frequency, tightening timings, lowering voltage, or combining changes—to see if stability improves.
* After each adjustment, thoroughly test the RAM using memtest86, which requires a full pass (15 tests) and takes considerable time. A minimum of two passes is advised, with four being acceptable. For 2x16GB configurations, expect around 5 hours per pass.
More often than not, pushing frequencies, timings, or voltages too aggressively leads to instability and failure to boot into the OS. You'll then need to recalibrate in BIOS with smaller steps.
In general, most users avoid manual OC due to the significant patience and effort required. Most opt for enabling XMP/EXPO instead, which the manufacturer has already optimized.
For those seeking maximum performance and willing to dedicate time, manual OC is an option—but it's not without challenges.
Download and install HWinfo64:
Link: https://www.hwinfo.com/download/
Capture a screenshot of the Summary page and share it here (e.g., upload to www.imgur.com).
On the Summary page, you'll see the built-in XMP/EXPO profiles along with their timings and voltages. This is essential if you plan to manually configure your RAM.
If your RAM supports multiple XMP/EXPO profiles, start by using the slower transfer profile before making any adjustments. Waiting until you reach 5600 MT/s or higher will save you considerable time and effort.
Also, refer to your RAM specifications:
https://www.gskill.com/specificatio...-T...cification
And watch a short video explaining RAM timings and functionality:
Video: https://www.youtube.com/watch?v=Yed-a9vqTYc
Thank you for your feedback. I haven't overclocked anything before, so beyond stability checks, I'd prefer to avoid that process which would require more time. For memtest, I plan to run it tonight to observe the results. Should I enable EXPO during the test or not—I'm not sure if it makes a difference? Also, based on what I know, memory is refreshed roughly every 10 power cycles. Since I've turned off sleep features and shut down my PC each night, it's fully powered off, not just partially asleep. My current uptime is about three hours. Could a faulty memory retraining process cause this issue? In theory, would turning EXPO on and off help refresh the memory and fix the problem until the next occurrence?
It is important to maintain EXPO disabled to establish a proper baseline. Operate at JEDEC speeds of 4800 MT/s. There are three approaches you can consider for resolving your problem, each with its own time and effort requirements:
1. Execute another, slower EXPO profile (provided your RAM supports multiple profiles).
2. Consult HWinfo64 to retrieve the RAM profile timings and input the necessary details manually—such as transfer rate, timings, and voltage. Manually adjusting these values can sometimes resolve the issue when the EXPO profile fails.
3. Perform a manual OC of the RAM, entering all parameters in small increments to precisely adjust the RAM settings (timings).
Memory training is limited to a single instance, provided the CPU/MoBo and RAM are compatible and don’t interfere during initial power-on. Only certain CPU/MoBo-RAM combinations are impacted by this process.
For further guidance:
https://www.howtogeek.com/what-is-ddr5-memory-training/
When incompatibility occurs, memory training runs automatically on first power-on to determine the optimal transfer speed, timings, and voltage for your RAM. Once completed, it won’t be necessary to retrain the memory.
Note: My two builds (details available in my profile) did not undergo memory training at startup because I selected RAM from MoBo memory QVL. This training becomes more relevant if you purchase RAM not listed in MoBo memory QVL.
I verified that your RAM should function at 6000 MT/s without issues, unless a fault has been detected by memtest86. The same applies to my builds—no sleep or hibernation is used; instead, the system shuts down completely at the end of each day.
Alright, then I guess I'll run memtest without EXPO tonight and if there are no errors, I'll run it with EXPO tomorrow.
Also, not sure if related, but I noticed my BIOS is a few versions behind when looking at the Asrock link. It's currently at 3.30. Later available versions are these:
3.40:
1. Update AGESA to ComboAM5 1.2.0.3f.
2. Improve memory compatibility and system stability.
3. Enhance CPU operating stability.
3.50:
Update AGESA to ComboAM5 1.2.0.3g.
4.03:
Update AGESA to ComboAM5 1.2.7.1 for upcoming CPU compatibility.
Is it worth updating it? Could those updates potentially have any positive impact on my current issue?
👍 Updating BIOS isn't something you can do on a whim. BIOS update, as such, is only viable when you know for a fact that newer BIOS fixes the specific issue you have with your PC. (E.g to keep my R7 9800X3D safe, i had to update my AsRock MoBo BIOS from issue prone 3.20, to (then) latest 3.40.) If there are no issues, there is no reason, what-so-ever, to update BIOS. Just because you "can" update BIOS doesn't mean that you "have to". Or in other words: "If it ain't broke - don't fix it." Most MoBos never get their BIOS updated and work fine until they are obsolete. Also, do note that when BIOS update would be interrupted for whatever reason (e.g power loss), your MoBo will be bricked since PC won't run when MoBo has corrupt BIOS. And only fix is MoBo replacement. If lucky, you can roll back the BIOS update, if your MoBo has that feature. As for your case, 3.40 seems to have benefits in terms of memory (RAM). But given that the RAM worked fine for a while, i don't think your issue comes down to the BIOS version. Usually, it is suggested to update directly to the latest version, while skipping the versions in-between. However, with AsRock 4.03, i'm uneasy suggesting to update to that version. Since some time ago, there was a person here in TH forums, who bricked their AsRock MoBo when installing 4.03. Luckily, they were able to revert back to 3.50 and save their MoBo. Topic: https://forums. So, your call. Personally, i'd leave BIOS update (to e.g 3.40) as a last resort due to the risks it carries.
Thank you, I'll postpone the bios update until later. I'll share my findings once I've checked the RAM, but expect it to take a few days since I'm using the PC during the day.
I reviewed the reliability monitor and noticed MS Defender was updated on the same day the crashes began. The error appears to be a general one without a specific cause.
Description:
A hardware issue is preventing Windows from functioning properly.
Problem Event Name: LiveKernelEvent
Code: 141
Parameter 1: ffffe20b98a6e460
Parameter 2: fffff804919df9b0
Parameter 3: 0
Parameter 4: 0
OS version: 10_0_19045
Service Pack: 0_0
Product: 256_1
OS Version: 10.0.19045.2.0.0.256.48
Locale ID: 1033
It seems Defender is being updated daily, which makes me wonder if it could be linked to the problem.
It's possible to try it by starting a live GNU/Linux system (without installing) and checking if the same problems appear there. This would help eliminate any issues related to the operating system itself. The second option is to format the C:/ drive and perform a fresh Windows installation, which should remove software-related problems like bloatware and malware (except firmware rootkits). If you'd like instructions on booting into a GNU/Linux distribution, I can provide them in my next message.
What could have shifted since everything seemed normal?
Especially in the past few days after you reinstalled windows.
I wouldn’t expect something like RAM to fail suddenly.
The benefit of memtest86 lies in its independence from Windows.
If a complete test runs without errors, I’d proceed with other troubleshooting steps.
A bios update should handle any unclear RAM or stability problems first.
Please try running your games without Discord or any other essential app.
I no longer enjoy using memtest. Running it for over ten hours works well, but with Windows Prime95's big FFTs it keeps failing and does so much faster.