Quick Answer

An RTX 5070 build that freezes mid-game points to a driver timeout (TDR) or memory instability, not a faulty GPU. Check Event Viewer for Display TDR (Event ID 4101) or WHEA-Logger 18, reinstall a clean NVIDIA 57x.xx Game Ready driver, and test RAM at stock before EXPO. A clean driver pass clears most hard freezes. Because this started right after a Windows update, roll the changed driver back first and re-check the EXPO or XMP profile before touching hardware.

Read the logs before you swap anything

Open Event Viewer (Windows Logs then System) and look for Kernel-Power 41 (a dirty restart), WHEA-Logger 18 and 19 (a hardware error), and Display TDR 4101 (a driver timeout). Sort by the timestamp of the last fault so you compare like with like. In HWiNFO64, log a five-minute gaming run and check whether the freezes you see coincide with a temperature spike, a clock drop, or a voltage dip. A reboot or freeze wipes the screen evidence, so the log is the only honest record of what actually happened.

The fix sequence that actually works

Since this began after a Windows update, start by rolling the changed driver back in Device Manager (right-click the device, Properties, Driver, Roll Back), then re-apply your tested driver. Run DDU in Safe Mode, then reinstall a clean NVIDIA 57x.xx Game Ready driver. In the motherboard firmware, enable Resizable BAR and confirm the card sits in the top PCIe 5.0 x16 slot. Confirm a 650W-class power supply such as a Corsair RM750e or MSI MAG A750GL, and reseat the power connector fully. Watch temperatures: idle should sit around 36-45C, sustained load around 65-75C, and throttling kicks in near 88C.

Where SA buyers usually go wrong

Most call-outs we see locally are not faulty parts at all. Dust build-up from a dry Highveld winter, a cheap unbranded cable, a half-seated power connector, or a stale driver after a feature update cover the bulk of them. Before you budget for a replacement, work the free fixes first: a clean driver pass, a CMOS reset, a dust-out, and a known-good cable cost nothing. If a part genuinely is faulty, every component sold through Evetech carries a local warranty, so you log the RMA here rather than chasing an overseas seller. A quality power supply is the cheapest insurance: a 650W unit from around R1,500 protects a build worth far more.

FAQ

Why does my build freeze completely mid-game?

A hard freeze is usually a driver timeout (Display TDR, Event ID 4101) or memory instability. Reinstall the driver, then test memory at stock before EXPO to split the two causes apart.

Is a full freeze a sign of failing hardware?

Not always. A frozen image with sound still playing points to the GPU driver; a total lock with no sound points to memory or a power hiccup. The logs tell you which before you spend anything.

How do I test if memory is the problem?

Load optimised defaults to drop EXPO or XMP and play for a day. If the freezes stop, run MemTest86 with the profile back on; a single error means the profile, not a dead stick.

Fix the cause, not the symptom. Before you shop for parts, run DDU and reinstall a clean NVIDIA 57x.xx Game Ready driver on a clean session, then re-test on one repeatable scene. If a part is genuinely faulty, log the local warranty through Evetech rather than guessing.