A Yet-To-Be-Merged Kernel Patch May Boost Kabylake Graphics In Some Cases

There's a patch pending for the Intel DRM driver that in extreme select cases can boost the graphics performance by up to 60% but for most OpenGL workloads the gains will be much smaller.

With the rather low Kaby Lake HD Graphics 630 Linux performance even with the Linux 4.10 kernel and Mesa 17.0-devel code still being a mystery as to why it's currently slower than Skylake, a patch was mentioned this weekend as a possible culprit.

The patch was drm/i915: Remove WaDisableLSQCROPERFforOCL KBL workaround. Francisco Jerez commented about the small patch, "Improves performance of the GFXBench4 gl_manhattan31 benchmark by 60%, and gl_4 (AKA car chase) by 14% on a KBL GT2 running Mesa master -- This is followed by a regression of 35% and 10% respectively for the same benchmarks and platform caused by my recent patch series switching userspace to use the dataport constant cache instead of the sampler to implement uniform pull constant loads, which caused us to hit more heavily the L3 cache (and on platforms other than KBL had the opposite effect of improving performance of the same two benchmarks). The overall effect on KBL of this change combined with the recent userspace change is respectively 4.6% and 2.6%. SynMark2 OglShMapPcf was affected by the constant cache changes (though it improved as it did on other platforms rather than regressing), but is not significantly affected by this patch (with statistical significance of 5% and sample size 20)."

Intel's rolling-release Clear Linux distribution had pulled in this patch with their 12730 update. So I ran some benchmarks with my Core i5 7600K system...

But the performance difference with the patched kernel ended up being very minimal. So, unfortunately, no solid explanation yet why the Kaby Lake Linux graphics performance leaves a lot to be desired on Linux 4.10 + Mesa 17.0-devel... I'll likely be firing up a Windows 10 install on that box later today to see clearly the Windows vs. Linux KBL performance on the same hardware. Next week from Intel I am still said to be receiving a Core i7 7700K and then I also did order a Core i3 7100 for some interesting low-end benchmarks, which should be arriving Monday~Tuesday.