AMD has published HIPRT 3.2, the latest version of its ray tracing library for HIP, and the release notes name the integrated graphics of its upcoming Zen 6 "Medusa" processors outright. Under new features, AMD lists Medusa support for three GPU architecture IDs: gfx1170, gfx1171, and gfx115E. HIPRT is mostly used by rendering apps like Blender, so this says nothing about gaming performance. Still, the link between these IDs and Medusa has so far rested on compiler patches and leaks, and now the Medusa name sits right in AMD's own changelog.

The gfx117x IDs belong to RDNA 4m, which first showed up in LLVM back in February. AMD added GFX1171 and GFX1172 targets in March, merged GFX1171 into Mesa 26.3 in August, and pushed RDNA 4m firmware later that month. Despite the name, RDNA 4m sits in the GFX11 family with RDNA 3 and 3.5, plus a few RDNA 4 style additions like FP8/BF8 conversion and new matrix instructions. The gfx115E entry is the more curious one. GFX115x covers the RDNA 3.5 graphics in Strix Point and Strix Halo, so it looks like at least one Medusa chip could stick with RDNA 3.5. The same release fixes a bug that truncated architecture numbers containing a letter, gfx115E included.

With the advent of the Steam Deck and the bump in Linux adoption that we've seen since then, a number of Linux distributions have appeared that make tweaks to the Linux kernel in the hopes of gaining performance. CachyOS is one such Linux distribution, and it appears that those kernel performance tweaks are paying off, according to some recent testing by Phoronix.

In a comparison between CachyOS, AlmaLinux 10.2, Ubuntu 26.04, Fedora Workstation 44, and Windows 11, CachyOS consistently performed at or near the top of the list, with CachyOS in performance mode claiming first place in the geometric mean of the results of a workstation benchmark. Windows 11 Pro was right at the bottom of the chart, coming in 14.9% slower than CachyOS in performance mode.

It seems as though the launch of Valve's Steam Deck has resulted in an almost non-stop slew of comparisons between Linux and Windows as a gaming operating system. In one recent test, Meta PCs sought out to test the two operating systems head-to-head to get some hard numbers and potentially settle the debate, however, the results brought up an interesting debate about what makes one operating system better than another.

Benchmarks were conducted on an all-AMD system running an AMD Ryzen 7 9850X3D, an XFX Swift AMD Radeon RX 9070 XT 16 GB, and 16 GB of RAM, with the AMD Linux drivers effectively making this a best-case scenario for Linux. The benchmarks were run on the latest version of Windows 11 and Fedora Linux 44, games were tested at the same resolution, and results were an average across three runs. All games tested were also the Windows versions running via Valve's Proton compatibility layer.

AMD's latest Adrenalin Edition 26.5.1 WHQL Drivers are causing problems for Blender users, according to the latest Blender issue tracker. Blender 5.1.1 is crashing with these drivers, particularly when using the Cycles path-tracing render engine. The issue appears to be a ROCm runtime mismatch. Blender 5.1.1 uses ROCm 6 runtimes, but the latest AMD driver 26.5.1 only includes ROCm 7, leading to compatibility issues that prevent the application from running smoothly. Multiple users have reported crashes, and we now understand the background.

Developer involved in AMD evosystem Sahar A. Kashi confirmed that when a factory reset is performed and the system is installed from scratch, the amdhip64_6.dll file is removed, causing Blender Cycles to attempt loading ROCm 7 kernels. However, since Blender version 5.1.1 is compiled only for ROCm 6, this runtime mismatch results in application crashes. To resolve this issue on AMD GPU systems, users should install the older 26.3.1 drivers and use them until Blender version 5.2 is released, and the Cycles path-tracing render engine achieves full ROCm 7 compatibility.

Since NVIDIA introduced its ray tracing and other AI-enhanced tech with its RTX graphics cards, Blender has been using the tech to enhance 3D rendering and modelling workloads, with NVIDIA OptiX bringing massive improvements to previews in Blender's Cycles Engine. Apparently, DLSS will be the next big NVIDIA technology that Blender will integrate into its Cycles engine to make render previews quicker, more accurate, and less resource-intensive. It effectively uses the same methods as DLSS in gaming applications, reducing the load on the system by rendering a lower resolution version of the scene and then approximating and upscaling the image after the fact. It likely won't have the necessary fidelity to do production-quality renders, but it will almost certainly help preview renders more quickly and with higher precision than the current Eevee and OptiX solutions.

At Siggraph, NVIDIA gave Blender artists and creators hands-on time with a testing build of Blender 5.0 featuring DLSS upscaling and denoising in the viewport, effectively combining the speed and efficiency of Blender's Eevee rendering system with the improved realism and accuracy of the Cycles engine. Blender's upcoming DLSS implementation in Cycles was shown off in comparison clips by Andrew Price on Instagram and Jonathan Lampel on X, and, while it still looks like there is a lot of work to be done, DLSS certainly appears to help generate more realistic render previews much more quickly than Cycles alone. It also doesn't seem to suffer from the same hard limitations on certain materials and transparency effects as Eevee does—although DLSS does still create some artifacts on glass, other transmissive surfaces, and there was some ghosting present, but it is not relying on screen-space techniques to approximate transparency, like Eevee does, so there will conceivably improvements to this ahead of the official Blender 5.0 launch.