Intel Core i5 11600K + Core i9 11900K Linux Performance Across ~400 Benchmarks
Written by Michael Larabel in Processors on 30 March 2021. Page 22 of 22. 44 Comments

There were 317 benchmarks carried out for the CPU/system tests (not counting the iGPU or gaming results, which take the total count to around 400), of which the Core i9 11900K managed to come in first place just 13% of the time.

The geo mean for all the 300+ CPU/system benchmarks.

When taking the geometric mean of all 300+ benchmarks (see the geo mean in the respective sections of the article for the iGPU performance and then Linux gaming performance given those separate areas with the rest of the dozens of benchmarks there), the Core i9 11900K represented a 4% improvement overall compared to the Core i9 10900K. Keep in mind that is for both single and multi-threaded benchmarks combined. The Core i9 11900K suffered in many of the multi-threaded benchmarks due to the Core i9 10900K being a 10 core / 20 thread design. With the Core i5 11600K where having a 6 core / 12 thread design as with the i5-10600K, it was a 16% improvement in performance for the geo mean of those CPU/system benchmarks. The Core i5 11600K at least was able to run aligned with the Ryzen 5 5600X while the Core i9 11900K only paired with the Ryzen 7 5800X across all of those benchmark results.

When looking at the CPU package power consumption over the course of running those 300+ benchmarks, there was much greater power consumption with the Rocket Lake processors. The average CPU power consumption went from 65 Watts with the i5-10600K up to 83 Watts with the i5-11600K while the peak CPU power consumption recorded went from 137 Watts to 211 Watts. The Core i9 10900K to 11900K meanwhile went from a 95 Watt average to 101 Watts and then for the peak recorded power consumption from 240 Watts to 284 Watts. As mentioned, no Ryzen 5000 series package power consumption readings on Linux due to the mainline AMD_Energy driver lacking support for those desktop parts.

When using a Noctua heatsink, the average temperature during all of the benchmarks was 54 degrees while the i9 11900K had an average temperature of 55 degrees and a peak of 99 degrees.

Given the wild price fluctuations right now in the retail market and availability differences, performance-per-dollar metrics are not included in this article. However, go check out all of the results in full on OpenBenchmarking.org. From that page you can also punch in your own pricing for the various CPUs via the top of the page to generate your own performance-per-dollar graphs based on local/available pricing. See that linked OpenBenchmarking.org result file to enjoy the plethora of benchmarks in full.

The Gen9 vs. Gen12 Xe Graphics geo mean performance as shown earlier in the article.

The Gen12 Xe Graphics with Rocket Lake are great to see and significant compared to the Gen9 graphics that have been powering Intel desktop CPUs since Skylake. The performance improvements with the Xe Graphics are quite significant across OpenGL/Vulkan graphics workloads as well as GPU compute situations. The numbers as shown in this article are quite good and stay tuned for more benchmarks. The Xe Graphics on Linux have been working out superb with the one "force probe" workaround as noted that may be needed depending upon your kernel configuration.

The Linux gaming benchmark results geo mean, as shown earlier in the article, based on the tested games.

Beyond the Xe Graphics, it really depends on what software you use whether the Rocket Lake processors make sense or are even faster than Comet Lake if you are eyeing the Core i9 11900 series where it means losing out on two cores. Where Rocket Lake performed particularly well on Linux was deep learning software that can make use of AVX-512 VNNI / DL-BOOST. Deep learning tests such as Intel's oneDNN deep learning library, Microsoft's ONNX Runtime, the PlaidML framework, and similar software were performing very well for the i5-11600K and i9-11900K were running great -- as long as the software has seen Intel/AVX-512 optimizations. TensorFlow Lite and NCNN for example were less competitive there.

Other workloads out of those 300+ benchmarks ran where the tested Rocket Lake CPUs did perform good as well were QuantLib for quantitative finance, Zstd compression/decompression in select areas, Lc0 chess with the Eigen back-end, JPEG XL image encode/decode, JSON parsing with simdjson, and other select workloads like GraphicsMagick and Himeno. Also showing significant wins for Rocket Lake against the AMD Ryzen 5000 series and previous-generation Comet Lake processors were signal processing / software-defined radio use-cases like LuaRadio, Liquid-DSP, GNU Radio, and srsLTE. The Rocket Lake performance in those areas were good but at higher power consumption than Comet Lake.

So there are those areas where Rocket Lake is actually a viable product and performing very well (most notably with the software that can successfully leverage AVX-512). For other areas, Rocket Lake tended to have a hard time being competitive with the AMD Ryzen 5000 series. But long story short, head on over to OpenBenchmarking.org and you can look at all 300+ CPU results in full. Then for the dozens of other benchmarks on gaming the results were mixed and then for the iGPU performance that is a big win for Rocket Lake... As covered earlier in the article, for those Gen9 vs. Gen12 tests, Rocket Lake was 37~39% faster than Comet Lake Gen9. The four dozen or so Gen9 vs. Gen12 iGPU performance data can be found in this OpenBenchmarking.org result file.

Stay tuned for more Intel Rocket Lake Linux benchmarks here and via OpenBenchmarking.org over the days ahead. A variety of different focused benchmark articles are in the works as well as looking at the Rocket Lake performance on Intel's own Clear Linux, how the BSD support/performance is for these new CPUs, and more. Let me know your feedback and any future test requests via the forums.

If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal tips are also graciously accepted. Thanks for your support.


Related Articles
About The Author
Author picture

Michael Larabel is the principal author of Phoronix.com and founded the site in 2004 with a focus on enriching the Linux hardware experience. Michael has written more than 20,000 articles covering the state of Linux hardware support, Linux performance, graphics drivers, and other topics. Michael is also the lead developer of the Phoronix Test Suite, Phoromatic, and OpenBenchmarking.org automated benchmarking software. He can be followed via Twitter or contacted via MichaelLarabel.com.

Trending Linux News