GPU-Z v2.71.0 Adds Blackwell Monitoring and Overclocking
TechPowerUp released GPU-Z version 2.71.0 on September 23rd, 2026. The update adds hot-spot and memory temperature monitoring for NVIDIA's RTX 50-series and provides new overclocking frequency readouts for NVIDIA GPUs.
TechPowerUp has launched GPU-Z version 2.71.0, a new release of its system information and monitoring tool for graphics hardware. The update, dated September 23rd, 2026, brings specific monitoring capabilities for NVIDIA's upcoming GeForce RTX 50-series "Blackwell" architecture and expands overclocking data visibility.
The headline addition is thermal monitoring for NVIDIA's next-generation GPUs. GPU-Z can now report the hot-spot temperature, revealing the single hottest point on an RTX 50-series GPU die. It also reads temperatures from individual memory chips. By default, the utility shows the temperature of the hottest memory module, but hovering the mouse cursor over the reading displays values for every chip on the card.
Enhanced NVIDIA Overclocking Data
For users tuning NVIDIA graphics cards, the Advanced tab now includes an "overclocking" readout. This feature displays overclocked frequencies for various internal clock domains. According to TechPowerUp, this includes the crossbar (XBAR), SYS, and video clock. It also helps identify if a GPU's boost clock is locked. The update further introduces the ability to read NVIDIA SYS and L2 cache clocks directly, though these sensors are disabled by default and must be enabled in the settings to appear.
A new sensor reports PCI-Express bandwidth in gigabytes per second for NVIDIA GPUs. The developers have also added a visual cue for the logging function: if a sensor is not set to log data to a file, its dropdown arrow on the Sensors tab will appear grayed out.
Broadened GPU Support List
The release extends official support to a wide range of new and upcoming graphics processors from all major vendors. The newly supported GPUs include both consumer and professional models.
| Vendor | Newly Supported GPUs and Variants |
|---|---|
| NVIDIA | RTX 5060 (GB205), RTX 4070 10 GB, RTX PRO 4500 Blackwell Server, H100 PCIe variant, Tesla V100-SXM2-16 GB variant, P102-101 |
| AMD | Steam Machine GPU, Radeon PRO V710 (second variant) |
| Intel | Panther Lake, Lunar Lake, Arrow Lake H (285H, 265H, 255H, 235H, 225H), RPL-H Refresh (240H, 210H), Twin Lake (N355, N350, N250, N150) |
| Qualcomm | Snapdragon X2-90, X2-85, X2-45 |
Bug Fixes and Refinements
Version 2.71.0 includes several corrections to improve accuracy. A fix addresses the transistor count reporting for NVIDIA's GP102 GPU. Memory detection and reporting issues have been resolved for Intel platforms, including LPDDR5X on Meteor Lake and Arrow Lake, LPDDR5 frequency on Alder Lake, and DDR5 bandwidth on some systems. A fix was also applied to the unit count reporting for Qualcomm's Adreno 726 graphics.
The utility remains a free download from TechPowerUp's website. The full change-log details all adjustments made in this version.
