Maxon of Germany has released Cinebench 2026, the latest version of its industry-standard benchmarking tool, marking a significant leap forward in how modern CPU and GPU performance is measured.
The update is built on the latest version of Maxon’s acclaimed Redshift rendering engine and supports Nvidia’s new Blackwell GPU, Nvidia Hopper GPUs, as well as Apple M4 and M5 systems. The company says this is a major architectural update to the benchmark software.
Cinebench 2026
Based on the latest Redshift rendering engine, Cinebench 2026 reflects real-world 3D production workloads more accurately than ever before, applicable to 3D CAD/BIM/DCC industries. Redshift is Cinema 4D’s default rendering engine, and the benchmark now leverages the latest technological advancements in Redshift’s development.
Cinebench 2026 is now available and free.
The new benchmark supports Nvidia’s new Blackwell (5000 series) GPUs, AMD 9000 series GPUs on Windows, as well as Nvidia Hopper and Blackwell datacenter GPUs. Moreover, the latest version also supports Apple M4 and M5-based Macs.
The benchmark is also capable of evaluating the performance of SMT-enabled CPU cores, so users can directly assess the performance gains offered by SMT (simultaneous multithreading) compared to single-threaded execution. SMT refers to a processor’s ability to run multiple execution threads on a single physical core by sharing internal resources, and the term is used by AMD. Intel has its own term for this called Hyper-Threading.
More Technical Details
As most Architosh readers are aware, rendering is a highly parallel workload, and multicore processors can speed up render times dramatically. SMT and Intel’s Hyper-Threading essentially enable the operating system to see a processor’s physical cores as made up of two or more “logical” cores. For example, a 6-core CPU with SMT enabled will show 12 threads in Cinebench 2026’s multi-core test.
SMT and Hyper-Threading boost chip efficiencies. If one thread is waiting for data from memory, the core can use the idle time to process instructions for the second thread. When rendering software is well-optimized, SMT can provide performance gains of over 30% compared to running just one thread per core.
Apple Silicon Comparisons
While Apple’s M4 and M5 chips are significantly faster than Intel and AMD’s best processors at single-core performance, Apple has yet to deploy SMT. The new CPU Single Core vs Single Thread tests in Cinebench 2026 make for a useful comparison to processors from Apple, as well as to see how performance gain is enabled by SMT for AMD and Intel processors.
Because of the new architectural improvements to Cinebench 2026, the new scoring system is not compatible (comparable) with Cinebench 2024 scores.
To obtain the new free Cinebench 2026 industry-standard benchmark, go here to download.
Architosh Analysis and Commentary
Cinebench is considered a top industry-standard benchmarking software and is used and cited widely by Intel, AMD, Apple, and other chip makers. While Geekbench 6.5 is widely considered the best tool for comparing disparate chip architectures (like ARM vs x86), Cinebench is also right up there in that category due to its extensive cross-platform capability. It supports Windows on x86, Windows on ARM, and macOS (x86 and ARM). However, Geekbench covers a broad spectrum of everyday and professional computing tasks that are typically relatively short. On the flip side, Cinebench has always focused on brute-force “sustained” chip performance. Why does this distinction matter, especially for CAD and 3D users?
Geekbench 6.5 tests and subtests are rather short (under five minutes). They also include a 5-second pause between sub-tests to minimize thermal throttling (ie, prevent it from happening). This makes it the gold standard for “bursty” single-core responsiveness kinds of tests. And CAD and BIM tools, in particular, are not only dominantly single-core (single-threaded) but also the workflows are very “bursty” by nature. In CAD and 3D, there are pauses between actions as the user manipulates 2D and 3D content. This makes it a very good test for measuring how “snappy” a computer feels when opening up an app or moving a line in CAD.
Cinebench, on the other hand, measures “sustained” performance (not “bursty” performance) and will test a system’s chips to the maximum of their thermal and power limits. This is critical because in the real world, processors can throttle down performance when thermal limits are reached. And this is critical in mobile computers, where, under battery power, peak performance may be very short. If a given chip relies on “bursting” to hit peak performance and has poor cooling capability, its Cinebench score will likely plummet over time as Cinebench runs tests for 10-30 minutes. This is where we are likely to see large advantages of Apple Silicon (M4 / M5) over Intel and AMD mobile processors.
Cinebench also tests how CPUs and GPUs handle large parallelized rendering workloads. It also measures 3D ray-tracing which is extremely taxing on heavy floating-point math. And a big reason why Cinebench is so widely considered a standard is for its multicore focus, where the system’s many cores can be hammered simultaneously at full power. Because the operating system has less input compared to other benchmarks, Cinebench is a favorite of high-end workstation builders and overclockers.
We use both Geekbench and Cinebench as primary comprehensive testing tools. The first captures the “bursty” nature of general computing and CAD/BIM apps, while the latter (Cinebench) captures raw power and “sustained” performance up against a system’s thermal limits. We highly recommend both for readers to deploy when benchmarking their own systems and setups.
