The Rtx 4070 Ti is experiencing crashes during every game on the Tronix Trinity.
The Rtx 4070 Ti is experiencing crashes during every game on the Tronix Trinity.
Hi everyone,
I’ve been facing some issues with my RTX 4070 Ti over the past few days. The graphics card was functioning properly until about three days ago, after which it began crashing during every game. I’ll outline everything below for clarity.
The Issue
Every game I launch crashes within seconds (sometimes minutes in lighter titles like Dota 2). When the crash occurs, only the game closes—my PC doesn’t restart or trigger a BSOD. In certain games, I also notice artifacts before the crash. Benchmarks such as Heaven or stress tests in OCCT run without problems. GPU temperatures are normal and there’s no overheating.
What I’ve Tried in Troubleshooting
Checked connectors & power supply
Switched to a 3×8-pin adapter (no native 12VHPWR cable) and re-plugged all connectors.
Updated GPU drivers, rolled back drivers, and updated the BIOS—none resolved the problem.
Performed system checks: OCCT GPU + VRAM stress test passed without errors; Windows Memory Diagnostic showed no RAM issues; MSI Afterburner confirmed stable temperatures.
Tried underclocking the GPU, but it still crashes.
Tested the GPU on another PC—it exhibited the same crashing behavior.
Crash Dump Analysis (Nsight Aftermath)
The dump indicates:
Device state: Hung, Engine reset occurred.
Application flagged: grb_vulkan.exe (Ghost Recon Breakpoint, Vulkan).
Event Viewer logs aren’t always clear about “device stopped responding,” but GPU resets match the crash times.
System Specs
CPU: Ryzen 7 7700
Motherboard: B650M-H/M.2+
RAM: Corsair Vengeance Rgb pro 32GB (16x2) DDR5 6000mhz
GPUs: RTX 4070 Ti zotacc Trinity
PSU: exgex1050 cougar (80+ Gold)
OS: Windows 11 (22H2)
Driver version: 581.15
What I Believe
Given that the card displays artifacts and crashes across multiple systems, it’s likely a hardware problem—possibly related to VRAM or GPU core failure. The adapter type (3×8-pin → 12VHPWR) might be contributing, but the artifacts usually point to the GPU itself. Stress tests can miss intermittent GPU faults, so passing in OCCT doesn’t guarantee a working unit.
Questions
Does this sound like a failing GPU to you all?
Should I consider returning the card for repair?
Are there any additional checks I should perform using a native 12VHPWR cable?
Also check for issues through the Reliability History/Monitor. This method is more user-friendly, and the timeline view might show some trends.
Run the built-in Windows troubleshooters; they can identify and resolve problems.
Execute "dism" and "sfc /scannow" to detect and repair faulty or corrupted Windows files.
Links provided: https://www.windowscentral.com/how-use-d...s-10-image, https://www.lifewire.com/how-to-use-sfc-...es-2626161
I used "Apex Legends" with Nsight Aftermath this time, and it provided a proper response to the crash. It seemed connected to missing shaders at address 0x00. I'll share the picture soon.
I just did a fresh OS install, so I think sfcscannow won't be necessary.
Before that fresh install, I tried all these sfcscannow tools and the one you mentioned.
Picture of Nsight After math Dump after launching Apex Legends and then crashing.
I reviewed the "View reliability history" and discovered critical errors.
Using your description, along with the detailed testing, I’d simply disassemble and modify it. As it remains under warranty and an RMA is available, I wouldn’t bother with that step since I wouldn’t rely on any current company selling video cards to shift blame if the warranty was still valid.
Iam bit confused, Why would the gpu need a change of paste/pads , There s no temperature issues with it.
Or is it about temps and iam missing something here?
--------
Yeah your right about the RMA thing , Only reason i haven't returned it yet is because it takes a hell lot of time with this returning process here in Pakistan ( No official vendor present in the country)
VRM lacks a temperature sensor, so pads must link directly to the heatsink. Issues with VRM pads can lead to unpredictable problems, making it difficult to identify the cause without disassembling the device.