A Geekbench 7.1 listing attributed to AMD’s 256-core EPYC 9996 offers an early look at Zen 6c performance: 2,847 points in single-core and 68,427 in multi-core. Its clearest comparison is a Ryzen 7 9700X fixed at 4.1 GHz, which scored 2,444 single-core points. That puts the EPYC about 16% ahead in this benchmark, but differences between the systems—and the fact that the EPYC uses compact Zen 6c cores—make this a limited result, not a verdict on future desktop Ryzen chips.
1
2
EPYC 9996 benchmark details
The reported result was for Geekbench 7.1.0 on a Dell PowerEdge R9815 running Linux. Leaker HXL, also known as @9550pro, spotted the listing. The processor is reported to have 256 Zen 6c cores and 512 threads.
1
2
Geekbench recorded the chip at about 4.12 GHz; the listed maximum boost is 4.1 GHz. Reports also describe 1 GiB of L3 cache and 256 MB of L2 cache. AMD lists a 600 W default CPU power rating, with a configurable range of 400–600 W.
15
16
19
21
The reporting identifies Linux but does not consistently establish a precise distribution or kernel. One secondary account specifies Ubuntu 26.04 LTS, while the original listing summaries say only Linux. One report labels the benchmark’s instruction set as AVX2; that detail is not independently confirmed in the available listing summaries.
1
28
30
How the per-clock estimates compare
| Processor |
Geekbench 7.1 single-core score |
Reported clock |
What the comparison shows |
| EPYC 9996, Zen 6c |
2,847 |
About 4.12 GHz; 4.1 GHz maximum boost |
Leaked reference result. 1 19 |
| Ryzen 7 9700X, Zen 5 |
2,444 |
Fixed at 4.1 GHz |
Closest-to-matched-clock comparison. 1 |
| EPYC 4545P, Zen 5 |
3,064 |
About 5.45 GHz; up to 5.5 GHz boost |
Faster raw score at a much higher clock. 2 |
Against the Ryzen 7 9700X: Comparing the scores directly gives (2,847 ÷ 2,444) − 1, or about 16.5%. The clocks are close but not identical; adjusting for 4.12 GHz versus 4.1 GHz gives roughly 15.9%. In other words, “about 16%” is a reasonable description of this Geekbench result, not a precise measurement of architectural IPC.
1
19
Against the EPYC 4545P: The 9996 scores about 7% lower outright—2,847 versus 3,064—but ran at a substantially lower reported clock. Dividing each score by its reported clock yields an estimated advantage of roughly 23% per GHz for the 9996. Using rounded clock figures can bring the calculation close to 24%, which explains the “up to 24%” claim. It remains an estimate made from results at different clocks, not a same-frequency test.
2
5
These calculations describe Geekbench score relative to reported clock speed. They are not direct measurements of instructions retired per clock, and they do not show how the processors compare across a wider range of workloads.
What the subtests add—and what they don’t
Reporting on the 9700X comparison says Zen 6c led in all 15 single-core subtests. The reported spread runs from an 8.2% lead in File Compression to 25.7% in HTML5 Browser; Photo Editing is reported at 24.8%.
1
4
Those figures suggest the lead was not confined to one test, but the available reporting does not provide a complete, verifiable score table for all 15 subtests. The results also come from a comparison between different processor platforms, so they cannot isolate core architecture from every other factor.
1
2
Why this doesn’t settle desktop Ryzen performance
The EPYC 9996 is a server processor using compact Zen 6c cores. The result does not test a standard Zen 6 desktop chip, and it cannot establish how a future Ryzen processor will perform.
2
3
The systems also differ: the EPYC result came from a Dell server, while the Ryzen comparison used a desktop processor fixed at 4.1 GHz. Matching the nominal clock helps make the single-core comparison more informative, but it does not make the test a controlled, same-platform experiment. Memory, firmware, power behavior, operating-system configuration, and benchmark conditions can also affect results.
1
2
A single Geekbench result cannot confirm gaming performance, broad application gains, Windows performance, or desktop multi-core scaling. It is best read as an early benchmark-specific signal for Zen 6c, not a promise that future Ryzen processors will be 16%–24% faster across the board.
3
4
What remains uncertain about the next desktop Ryzen family
The EPYC 9996 is associated with AMD’s EPYC 9006 “Venice” server generation, and reporting describes the series as launching in July 2026. That server launch does not confirm the timing, final branding, specifications, or availability of a future Zen 6 desktop family.
1
35
The name Olympic Ridge and possible Ryzen 10000 branding appear in secondary coverage, but the benchmark does not verify those details. Nor does it establish desktop clock speeds, memory support or tuning, a maximum of 24 cores, or a retail release date. Those remain separate claims requiring confirmation.
3
4
For now, the strongest conclusion is narrow: the EPYC 9996’s leaked Geekbench score suggests about a 16% matched-clock single-core lead over the tested Ryzen 7 9700X, while a separate calculation puts it around 23%–24% per GHz versus the faster-clocked EPYC 4545P. Neither comparison confirms a universal Zen 6 IPC uplift or predicts desktop performance.