EPYC 9996 Geekbench Result Points to Up to 24% Higher Per-Clock Performance for Zen 6
An early Geekbench result for AMD's EPYC 9996 server processor puts its single-core performance close to that of a much higher-clocked Zen 5 chip. Wccftech calculates an advantage of up to 24% per clock, offering a preview of the architecture behind the next generation of Ryzen processors.
Layla Haddad
How the Geekbench comparison stacks up
The 256-core, 512-thread EPYC 9996 scored 2,847 in Geekbench's single-core test and 68,427 in the multi-core test. It ran at 4.12 GHz, its maximum boost frequency. By comparison, the 16-core Zen 5-based EPYC 4545P scores 3,064 in the single-core test with a boost clock of up to 5.5 GHz.
Wccftech's comparison of those scores and clock speeds puts the Zen 6 processor ahead by up to 24% on a per-clock basis. The result is among the first EPYC 9006 entries appearing in the Geekbench database as AMD ramps production of the already-launched Venice lineup.
A separate user comparison used a Ryzen 7 9700X locked to 4.1 GHz and paired with DDR5-6000 CL34 memory. Wccftech reports that the Zen 6 result was up to 24-25% faster, with an average advantage of 16.5%.
The EPYC 9996 has 1.0 GB of L3 cache and 256 MB of L2 cache. Its rated thermal design power is 600 watts, configurable down to 400 watts.
What the server result means for Ryzen
The tested processor uses Zen 6c cores, a denser and more efficient design than standard Zen 6 with the same instruction set architecture. Wccftech argues that optimized memory support on enthusiast desktop platforms could provide additional performance benefits for Ryzen.
Higher desktop clock speeds remain a rumor. Wccftech notes that AMD has increased clocks and power on its EPYC processors and expects Olympic Ridge desktop chips to reach higher frequencies still. The upcoming Ryzen family is also expected to combine higher core counts, more 3D V-Cache and a redesigned architecture while retaining the AM5 socket.
Another estimate comes from leaker adroc_thurston, who claims Zen 6 will improve instructions per clock by roughly 13% over Zen 5, comparable to the approximately 13% gain Zen 4 delivered over Zen 3. That remains an unconfirmed estimate; desktop Zen 6 Ryzen processors are expected in 2027.
AMD's broader Venice performance claims
AMD unveiled the EPYC 9996 on July 23, 2026, with a price of $14,904 in 1,000-unit quantities. The chip uses TSMC's 2nm process and the new SP7 socket. Its 256 cores represent a 33% increase over the 192-core EPYC 9965, the previous Turin flagship.
For the EPYC 9006 family as a whole, AMD claims up to 1.8 times the performance of the preceding Zen 5 generation, citing higher IPC and boost frequencies reaching 5.15 GHz. That frequency is a lineup maximum, not the EPYC 9996's boost clock. AMD also indicates roughly 20% more performance per core than Turin.
Venice expands memory support to 16 channels, with DDR5 at 8,000 MT/s or MRDIMM at 12,800 MT/s. Bandwidth reaches up to 1.6 TB/s per socket, compared with Turin's 614 GB/s across 12 channels. Venice is also the first server CPU to support PCIe Gen 6.
In AMD's SPECrate 2026 estimates, a two-socket Venice 9996 system scores 2,070 against 925 for NVIDIA's Vera, a 2.2-times throughput advantage. AMD puts its 96-core high-frequency Venice part about 1.2 times ahead of Vera on a per-core basis. Both comparisons used GCC 15.2.
Platform rollout
At the Venice launch, AMD accounted for 46% of server CPU revenue, according to Mercury Research, while demand for top-end EPYC processors exceeded supply.
Although the EPYC 9006 lineup has launched and is shipping in volume, the platform rollout is staged. SP7 platforms ship in the fourth quarter of 2026, followed by mainstream SP8 platforms in the first half of 2027. Venice-X and the LPDDR-based Verano host part are scheduled for the second half of 2027.
