AMD Zen 6 Benchmark Shows 16.5% IPC Performance Gain

AMD's upcoming Zen 6 architecture is showing improvements over Zen 5 in early benchmark results. A new test using an EPYC processor suggests that Zen 6 could deliver higher single-threaded performance, with several workloads showing gains of more than 23%. The results also indicate that earlier performance reports may have underestimated the improvements AMD expects from its next-generation Ryzen processors.

AMD's next-generation Ryzen processors are attracting attention following new Zen 6 benchmark results. Early reports suggested improvements over Zen 5, but the latest results indicate that the architectural gains could be higher than initially expected. OneUsmus, the developer behind the DRAM Calculator for Ryzen and Hydra software, previously shared information about upcoming AMD hardware.

AMD Zen 6 Performance Looks Better in Early Benchmarks

Many AMD owners use these tools, and the developer has access to hardware before its wider release. A new benchmark result has now emerged from an AMD EPYC 9996 processor tested in Geekbench 7. Although EPYC processors target the server market, they use the same core architecture as AMD's Ryzen processors. This makes them useful for evaluating the performance improvements expected from the upcoming Zen architecture.

EPYC 9996 is a 256-core, 512-thread processor featuring Zen 6C cores. Unlike Intel's E-cores, AMD's Zen 6C cores are compact Zen 6 cores that retain the same instruction set and offer broadly comparable IPC. EPYC 9996 isn't representative of a consumer processor in terms of core count, but its single-core performance can help show the architectural differences between Zen 5 and Zen 6.

To compare the two architectures, a user tested a Ryzen 7 9700X and fixed its clock speed at 4.1GHz, matching the benchmark's EPYC processor frequency. The processors then competed in multiple single-threaded tests at the same clock speed.

Zen 6 Delivers More Than 23% Gains in Some Workloads.

Geekbench combines multiple synthetic tests into an overall score that covers different workload types. In several individual tests, Zen 6 outperformed Zen 5, with some video encoding and photo editing workloads showing performance gains of more than 23%. Not every benchmark showed the same improvement. IPC gains vary by application, and other factors can affect certain workloads.

For example, memory latency can affect the HTML5 browser benchmark, while the EPYC processor's large L3 cache can influence some results. However, several of the tested workloads should be less affected by these factors. Overall, the EPYC processor was 16.5% faster than the Ryzen Zen 5 processor in single-threaded workloads.

This result suggests Zen 6 could offer a meaningful improvement in instructions per clock over Zen 5, though performance will vary by application and final processor specifications. The 16.5% single-threaded improvement makes Zen 6 look more promising than earlier reports suggested.

Comparison does not account for potential clock speed increases or differences in core counts.

Zen 6 is expected to improve both clock speeds and core counts. Higher frequencies could further boost single-threaded performance, while additional cores could improve workloads that benefit from parallel processing. However, these benchmark results represent an early look at the architecture rather than a direct measurement of a retail Ryzen Zen 6 processor.

Final performance will depend on the specifications, clock speeds, memory configuration, and workload. Nevertheless, the results suggest AMD is making progress with Zen 6. If the architectural improvements carry over to consumer Ryzen processors, the next generation could deliver a noticeable performance increase over Zen 5.