How to Resolve GPU Driver Crashes on Nvidia RTX 5090?

The Nvidia RTX 5090 is one of the most powerful consumer graphics cards available today. It delivers incredible performance for gaming, content creation, and AI workloads. But many owners have reported frustrating GPU driver crashes that ruin the experience.

These crashes show up in different ways. Some users see black screens during games. Others get the dreaded nvlddmkm.sys blue screen of death. Some experience complete system freezes where monitors go dark while the PC keeps running in the background.

If you own an RTX 5090 and face these issues, you are not alone. Forum threads on Reddit, Nvidia GeForce Forums, ASUS ROG forums, and MSI community boards are filled with similar reports. The good news is that most of these crashes have identifiable causes and workable fixes.

This guide walks you through every solution, from simple driver reinstalls to advanced BIOS tweaks, so you can get your RTX 5090 running stable again.

In a Nutshell

• Driver corruption is the most common cause of RTX 5090 crashes. A clean uninstall with DDU (Display Driver Uninstaller) followed by a fresh driver install fixes the problem for many users. Always disconnect from the internet before reinstalling drivers so Windows Update does not interfere.
• PCIe Gen 5.0 compatibility problems affect a large number of RTX 5090 owners. Switching your BIOS setting from Auto or Gen 5 to PCIe Gen 4.0 costs you only about 1% performance at 4K but often eliminates crashes entirely.
• Insufficient or unstable power delivery causes many crash scenarios. The RTX 5090 can draw over 575W with power spikes reaching even higher. A quality PSU rated at 1200W or above is strongly recommended, and you should inspect your 12V power cable for signs of melting or damage.
• CPU and memory overclocking instability often gets overlooked as a cause of GPU crashes. Disabling features like PBO (Precision Boost Overdrive), DDR5 Nitro Mode, and EXPO/XMP profiles can reveal whether your RAM or CPU is the true source of the problem.
• DisplayPort cable version mismatches have been confirmed as a crash trigger. Using older DP 1.2 cables alongside DP 2.1 cables causes the nvlddmkm.sys BSOD on RTX 5090 systems. Replace all display cables with DP 2.1 versions.
• Physical seating issues with the GPU in the PCIe slot are surprisingly common because the RTX 5090 is a large, heavy card. Reseating the GPU and adding a support bracket has fixed the problem for multiple users.

Understanding Why the RTX 5090 Crashes

The RTX 5090 is the first high end consumer GPU to fully use PCIe Gen 5.0 and the Blackwell architecture. This means it pushes new boundaries in data transfer speed, power consumption, and driver requirements. Each of these factors introduces potential failure points that previous GPU generations did not have.

Driver crashes on the RTX 5090 typically fall into three categories. The first is a software crash where the Nvidia display driver (nvlddmkm.sys) stops responding. Windows tries to recover the driver through a TDR (Timeout Detection and Recovery) reset. You see a brief black screen, and sometimes the system recovers on its own.

The second type is a hard crash where all monitors go black, GPU fans spin to maximum speed, but the PC itself keeps running. You can still hear audio from Discord or games. This points to the GPU losing its display output while the card remains powered.

The third type is a full system crash or BSOD. The entire PC locks up or restarts. Event Viewer logs usually show nvlddmkm.sys as the faulting module. This type often points to deeper issues with power delivery, PCIe signaling, or hardware defects.

Understanding which type of crash you experience helps you narrow down the cause and pick the right fix from the sections below.

Perform a Clean Driver Install with DDU

A corrupted or conflicting driver installation is the single most common reason for RTX 5090 crashes. Windows Update sometimes installs an older or incompatible Nvidia driver in the background. This creates conflicts with the driver you installed manually. The solution is a complete clean install using Display Driver Uninstaller (DDU).

First, download the latest stable Nvidia driver from the official Nvidia website. Save the installer file to your desktop. Next, download DDU from the Guru3D website. Boot your PC into Safe Mode by going to Settings, then Update and Security, then Recovery, and clicking “Restart now” under Advanced Startup.

Once in Safe Mode, run DDU and select “Clean and Restart.” This removes every trace of the old Nvidia driver from your system. Before you restart, disconnect your Ethernet cable and disable your Wi Fi adapter. This step is critical because Windows Update will try to install a generic Nvidia driver the moment you connect to the internet.

After the restart, run the Nvidia driver installer you downloaded earlier. Choose “Custom Install” and check the box for “Clean Installation.” Once the install finishes, reconnect your internet. This process has resolved crashes for a significant number of RTX 5090 owners across Reddit and Nvidia forums.

Try an Older or Recommended Driver Version

Not every new Nvidia driver release is stable for the RTX 5090. Several users have reported that specific driver versions introduced or worsened their crashes. The 572.16 driver was identified as a major source of problems early in the RTX 5090’s lifecycle. Later versions like 576.02 and 577.00 fixed issues for some users but introduced new ones for others.

If you experience crashes on the latest driver, roll back to an earlier version that other RTX 5090 owners report as stable. Driver version 572.60 has been recommended by Gigabyte for their RTX 5090 models. Several forum users confirmed it eliminated their crashes.

To roll back, use DDU to fully remove the current driver as described above. Then install the older driver version. You can find archived Nvidia drivers on the official Nvidia driver download page by selecting your GPU model and operating system.

Keep in mind that some games require newer drivers for optimal support. You may need to balance stability with game compatibility. Check the release notes for each driver version to see what games and fixes it includes. If a specific game keeps crashing, check its support forums for driver recommendations.

Switch PCIe Mode from Gen 5.0 to Gen 4.0

PCIe Gen 5.0 signal integrity issues are a confirmed source of RTX 5090 instability. The RTX 5090 is among the first GPUs to use PCIe 5.0, and many motherboards struggle to maintain stable signaling at this speed. Switching to PCIe Gen 4.0 in your BIOS is the fastest way to test this.

Restart your PC and enter the BIOS (usually by pressing Delete or F2 during boot). Find the PCIe configuration settings. The exact location depends on your motherboard brand. On ASUS boards, look under Advanced or PCI Subsystem Settings. On MSI boards, check Settings and then Advanced. Change the PCIe slot speed from “Auto” or “Gen 5” to “Gen 4.0.”

Testing by TechPowerUp shows that running the RTX 5090 in PCIe 4.0 mode results in only a 1% performance loss at 4K resolution. Even dropping to PCIe 3.0 costs only about 4% at 4K. Most games do not use the extra bandwidth that PCIe 5.0 provides.

This fix has been confirmed by users on the ASUS ROG forum, Nvidia GeForce Forums, and Guru3D. An ASUS Z690 Hero owner reported that switching from Auto to Gen 3 or Gen 4 completely fixed their crashes. If this resolves your issue, leave the setting at Gen 4.0 until a BIOS or driver update addresses the root cause.

Check and Upgrade Your Power Supply

The RTX 5090 is an extremely power hungry GPU. It can draw up to 575W under full load, and transient power spikes can go even higher. A weak or aging PSU often cannot handle these demands, causing the GPU to lose power momentarily and crash.

Users have reported that PSUs rated at 1000W or even 1200W still caused instability. One user measured voltage spikes and dips on the 12V rail using an oscilloscope. They found that their 1200W PSU dropped to 11.44V during load transitions, which falls outside ATX specifications. Upgrading to a 1500W PSU resolved the issue.

Inspect your power cable carefully. Several RTX 5090 owners discovered that their 12VHPWR or 12V 2×6 power connector had melted on the PSU end. One user found exposed copper and melted plastic on their PCIe power cable. A damaged cable can cause intermittent crashes and poses a fire hazard.

If your PSU is rated below 1200W, consider an upgrade. If you already have a high wattage PSU, reseat all power cables and inspect them for damage. Also check that your PSU is oriented correctly in the case with its intake fan pulling cool air from outside, not hot air from inside. One user traced weeks of crashing to their PSU being installed upside down.

Reseat the GPU and Check Physical Connections

The RTX 5090 is a large, heavy graphics card. Many models weigh over 2 kilograms. This weight can cause the card to sag in the PCIe slot over time, breaking the electrical connection. A poor physical connection is one of the most overlooked causes of GPU crashes.

Power down your PC completely and unplug it from the wall. Open the case and carefully remove the RTX 5090 from its PCIe slot. Use compressed air to clean the PCIe slot and the card’s gold contacts. Reinsert the card firmly until you hear the retention clip click into place. Tighten the expansion slot screws securely.

Multiple users on the Reddit PcBuildHelp forum confirmed that reseating the GPU fixed their crashes. One user found that the motherboard’s heatsinks and heat shields were physically blocking the card from seating fully. Another user fabricated a support bracket for the back end of the card to prevent future sagging.

If your case uses a PCIe riser cable, try connecting the GPU directly to the motherboard instead. Riser cables can degrade PCIe 5.0 signals and cause crashes. Several ASUS ROG forum users reported that removing the riser cable and connecting the card directly to the motherboard resolved their instability.

Disable CPU and Memory Overclocking

CPU and memory overclocks that were stable with your previous GPU may not be stable with the RTX 5090. The RTX 5090 places higher demands on the entire system, including the CPU, memory controller, and PCIe bus. A marginally stable overclock can tip into instability under the extra load.

Enter your BIOS and load “Optimized Defaults” to reset all settings to stock. This disables PBO (Precision Boost Overdrive) on AMD systems and removes any Intel boost overrides. It also disables EXPO (AMD) or XMP (Intel) memory profiles. Test your system for stability at these default settings.

An ASUS ROG forum user with a 9800X3D reported that their RTX 5090 crashed constantly until they disabled PBO and their CPU curve optimizer. Another user found that DDR5 Nitro Mode (called “DDR5 RAM Boost” on Intel boards) was the specific cause. This setting is enabled by default even when you do not enable overclocking manually, so check for it specifically.

If your system is stable at default settings, re enable your overclocks one at a time. Start with the CPU, test for a day, then add back the memory profile. This method isolates which overclock is causing the problem. Many users find they can keep their CPU overclock but need to disable or lower their memory overclock.

Update Your Motherboard BIOS

Motherboard manufacturers regularly release BIOS updates to fix compatibility issues with new hardware. The RTX 5090’s PCIe 5.0 interface and high power demands have exposed bugs in older BIOS versions. An outdated BIOS can cause initialization failures, reduced PCIe lane counts, and driver crashes.

Visit your motherboard manufacturer’s support page and check for BIOS updates released after January 2025. Gigabyte released a BIOS fix specifically for PCIe lane initialization problems with RTX 50 series GPUs. ASUS and MSI have released similar updates for their X870 and Z790 boards.

Follow the manufacturer’s instructions carefully to flash your BIOS. Most modern motherboards support USB BIOS Flashback, which makes the process safe and simple. Do not interrupt the flashing process or your motherboard may become unbootable.

After updating the BIOS, load Optimized Defaults before configuring any custom settings. Some users discovered that BIOS settings carried over from the old version were incompatible with the new one. Starting fresh eliminates this risk. Also check for VBIOS updates from your GPU manufacturer. MSI released a new VBIOS for their RTX 5090 Gaming Trio that fixed black screen crashes.

Replace or Match Your DisplayPort Cables

A surprising but well documented cause of RTX 5090 crashes involves DisplayPort cable version mismatches. The RTX 5090 has stricter requirements for display cable quality than previous GPU generations. Using old DP 1.2 cables alongside newer DP 2.1 cables can trigger nvlddmkm.sys BSOD crashes.

An MSI forum user spent three days troubleshooting before discovering that their old DP 1.2 cable on a secondary monitor was the sole cause of their crashes. Once they replaced all cables with DP 2.1 certified versions, the crashes stopped completely. The user noted that the GPU handled games at 580W with no issue but crashed loading Netflix because of the cable mismatch.

Check every display cable connected to your RTX 5090. If you use multiple monitors, make sure all cables are the same version. Replace any cable that is DP 1.2 or older with a DP 2.1 cable. If you use HDMI, use HDMI 2.1 cables. Long cable runs (over 3 meters) are especially prone to signal degradation and may need certified high speed cables.

Some users also found that specific HDMI and DisplayPort combinations caused crashes. If you have a multi monitor setup, try disconnecting all but your primary display to see if the crashes stop. Then add monitors back one at a time to identify the offending connection.

Reduce GPU Power Limit and Clock Speed

If other fixes have not resolved your crashes, reducing the GPU’s power limit can provide stability while you wait for a driver fix. Factory overclocked RTX 5090 models from brands like Gigabyte, ASUS, and MSI push the card beyond its reference specifications. Pulling back on these boost clocks can eliminate crash triggers.

Open MSI Afterburner or your GPU manufacturer’s tuning software. Lower the power limit to 90% or 95%. This prevents the extreme power spikes that can overwhelm your PSU or trigger overcurrent protection. One user on Reddit found that dropping the power limit to 98% stopped all crashes in Call of Duty.

You can also reduce the GPU core clock by 100 to 150 MHz using the frequency offset slider. A Gigabyte Aorus 5090 owner reduced their clock by 110 MHz and lowered power consumption to 90%. Combined with disabling RGB syncing software, this eliminated all crashes for over two weeks of heavy use.

Remember that these third party RTX 5090 cards come factory overclocked above Nvidia’s reference speeds. Reducing the clock by 100 MHz simply brings the card closer to reference performance. The actual impact on frame rates is minimal, typically under 3%. This is a reasonable trade off for a stable system until Nvidia releases a driver that properly manages boost behavior.

Disable RGB Syncing and Background Software

Background software can interfere with GPU driver stability in ways that are hard to predict. RGB syncing software like ASUS Armoury Crate, Gigabyte Control Center, and similar tools communicate with the GPU at a low level. These background processes can conflict with the Nvidia driver and cause crashes.

One Reddit user confirmed that disabling RGB syncing features in both Armoury Crate and Gigabyte Control Center was a critical part of fixing their RTX 5090 crashes. They had tried everything else, including driver rollbacks, power limit reductions, and BIOS changes. Only after also disabling RGB syncing did the crashes stop.

Uninstall or disable the following types of software for testing: GPU monitoring overlays like Rivatuner Statistics Server, manufacturer RGB software, hardware monitoring tools that poll the GPU frequently, and any game capture or streaming overlays. An RTX 5090 owner reported that Rivatuner overlay alone caused crashes in every game, even though the games ran fine without it.

If your system becomes stable after removing background software, add each program back one at a time to find the specific culprit. You may need to wait for software updates that add proper RTX 5090 compatibility. Check the developer’s website for updated versions before reinstalling.

Run Hardware Diagnostics to Rule Out Defects

If you have tried every software fix and your RTX 5090 still crashes, the card itself may be defective. Reports of hardware level failures include blown capacitors, VRM overheating, and damaged ICs. These defects cannot be fixed with driver or BIOS changes.

Run a GPU stress test using tools like 3DMark, FurMark, or OCCT. Monitor the GPU temperature, clock speed, and power draw using HWiNFO64 during the test. Watch for temperatures above 85C on the GPU core or unusual voltage readings on the 12V rail. Dips below 11.8V suggest a power delivery problem.

Also run a memory test using MemTest86+ to rule out faulty system RAM. Bad memory causes seemingly random crashes that mimic GPU driver issues. Let the test run for at least 4 to 8 hours or until it completes several passes.

If the GPU crashes during stress tests even after all the fixes above, contact your GPU manufacturer to begin an RMA (Return Merchandise Authorization) process. Several users on the ASUS ROG forum and MSI community boards have successfully RMA’d their cards after confirming hardware defects. Document your troubleshooting steps before contacting support, as this speeds up the approval process.

Adjust Windows Power Settings

Windows power management settings can conflict with the RTX 5090’s power states and cause driver crashes. Setting Windows and Nvidia to prefer maximum performance eliminates power state transitions that can destabilize the GPU.

Open the Windows Control Panel and go to Power Options. Select the “High Performance” plan. If it is not visible, click “Show additional plans.” This prevents Windows from aggressively stepping down CPU and GPU power states.

Next, open the Nvidia Control Panel (not the Nvidia App). Go to “Manage 3D Settings” and find “Power Management Mode.” Change it from “Optimal Power” or “Normal” to “Prefer Maximum Performance.” This keeps the GPU running at higher clock speeds rather than rapidly switching between power states.

One user on the Nvidia GeForce Forums reported that this single change fixed their frame generation crashes on the RTX 5090. The rapid transitions between low and high power states were triggering driver instability. By keeping the GPU in a consistently high power state, the transitions that caused crashes were eliminated.

Also check your PCIe power saving settings in the Windows Power Options advanced settings. Expand “PCI Express” and then “Link State Power Management.” Set this to “Off.” This prevents Windows from reducing PCIe link speeds to save power, which can cause communication failures between the GPU and motherboard.

When to RMA Your RTX 5090

After working through all the fixes above, some RTX 5090 cards still crash because of genuine hardware defects. Knowing when to stop troubleshooting and start the RMA process saves you time and frustration.

Signs that point to a hardware defect include: artifacting (visual glitches like colored pixels or distorted textures), crashes that happen even at reduced power limits and stock BIOS settings, the GPU not being detected in BIOS after a crash, and burn marks or unusual smells coming from the card or its power connector.

Contact your GPU manufacturer’s support team and provide a detailed list of every troubleshooting step you have taken. Include your system specifications, the driver versions you tested, and any error codes from Event Viewer or blue screen dump files. The more documentation you provide, the faster your RMA will be approved.

Several users have reported successful RMA outcomes. An MSI user confirmed that the manufacturer found an actual hardware issue and replaced the card. Be aware that RMA processing times vary by manufacturer. ASUS in particular has received mixed reviews for their RMA process, so document everything and keep copies of all correspondence. If your card is within the return window of the retailer, a direct return may be faster than an RMA.

Frequently Asked Questions

What is the nvlddmkm.sys error on the RTX 5090?

The nvlddmkm.sys file is the Nvidia Windows kernel mode display driver. When this driver crashes or stops responding, Windows triggers a Timeout Detection and Recovery (TDR) event. This can appear as a brief black screen followed by recovery, or a full blue screen of death (BSOD) with the error code VIDEO_TDR_FAILURE. On the RTX 5090, this error is commonly caused by driver corruption, PCIe signaling issues, power delivery problems, or incompatible DisplayPort cables. A clean driver install using DDU is the recommended first step to fix this error.

Does switching to PCIe Gen 4.0 hurt RTX 5090 performance?

No, the performance impact is minimal. Independent testing shows that the RTX 5090 loses only about 1% of its gaming performance at 4K when running in PCIe 4.0 mode instead of 5.0. At lower resolutions, the difference may be slightly larger but still under 3% in most games. PCIe 5.0 offers 64 GB/s of bandwidth compared to 32 GB/s for PCIe 4.0, but most current games do not saturate even the PCIe 4.0 bandwidth. This makes Gen 4.0 mode an excellent stability fix with almost no real world trade off.

How much PSU wattage does the RTX 5090 need?

Nvidia recommends a minimum of 1000W PSU for RTX 5090 systems. However, real world experience from users suggests that 1200W is a safer minimum, and 1500W or higher provides the best stability. The RTX 5090 can draw over 575W under load, and transient power spikes can exceed this significantly. When combined with a high end CPU and other components, total system power draw can easily reach 800 to 1000W. A PSU with adequate headroom prevents voltage sags that trigger GPU crashes.

Can a faulty DisplayPort cable really crash my RTX 5090?

Yes. This has been confirmed by multiple users. The RTX 5090 is more sensitive to cable quality than previous generation GPUs. Using an older DP 1.2 cable alongside a DP 2.1 cable has been proven to cause nvlddmkm.sys BSOD crashes. The issue appears related to how the GPU handles mixed cable standards across multiple display outputs. Replacing all cables with DP 2.1 or HDMI 2.1 certified versions resolves the problem. This is one of the easiest and cheapest fixes to try.

Should I disable XMP or EXPO memory profiles to fix RTX 5090 crashes?

It is worth testing. XMP (Intel) and EXPO (AMD) memory profiles overclock your RAM beyond its base specification. While these profiles are usually stable, the added system stress from an RTX 5090 can expose marginal instability. Several users confirmed that disabling EXPO and DDR5 Nitro Mode fixed their GPU crashes. Load your BIOS Optimized Defaults and test for stability before re enabling memory overclocks. If the crashes stop, your memory profile was likely the cause.

How do I check if my RTX 5090 has a hardware defect?

Run a GPU stress test using 3DMark or OCCT while monitoring temperatures and voltages with HWiNFO64. Watch for GPU core temperatures above 85C, 12V rail voltage drops below 11.8V, or visual artifacting. If the card crashes during stress tests even after a clean driver install at stock BIOS settings with PCIe Gen 4.0 and a known good PSU, the card likely has a hardware defect and should be returned to the manufacturer for RMA or exchanged at the retailer.