AMD EPYC 9734 vs AMD EPYC 9745 Comparison

FAQ

Q: Which processor wins more benchmark tests, the AMD EPYC 9745 or the AMD EPYC 9734?

A: The AMD EPYC 9745 wins all 17 recorded head-to-head benchmark comparisons. The AMD EPYC 9734 does not win a single test in the database.

Q: How much faster is the EPYC 9745 in multi-threaded Cinebench tests?

A: Across Cinebench R15, R20, and R23 multi-core tests, the EPYC 9745 consistently leads by 27.8%. For example, in Cinebench R23 multi-core, the EPYC 9745 scores 111,093 versus 86,943 for the EPYC 9734.

Q: What is the single-thread performance gap between these two chips?

A: In Cinebench R15 single-core, the EPYC 9745 scores 1580 versus 1237, a 27.7% advantage. In PassMark single-thread, the gap narrows to 21.5%, with scores of 2806 versus 2310.

Q: Which benchmark shows the largest performance delta between the two?

A: PassMark physics shows the biggest difference, with the EPYC 9745 scoring 17,122 versus 6,747 for the EPYC 9734, a massive 153.8% advantage. The PassMark integer math test also shows a large 48.7% delta.

Q: How do the two processors compare in average benchmark scores?

A: The EPYC 9745 has an average benchmark score of 425,973, while the EPYC 9734 averages 310,619. The EPYC 9745 sits at the 100th percentile among all CPUs, while the EPYC 9734 sits at the 99th percentile.

Q: What are the launch MSRP values for these two server processors?

A: The AMD EPYC 9745 has a launch MSRP of $12141, and the AMD EPYC 9734 has a launch MSRP of $9600.

Architecture Differences

The AMD EPYC 9745 belongs to the EPYC 9005 series and uses the Zen 5 architecture, codenamed Turin. The AMD EPYC 9734 belongs to the EPYC 9004 series and uses the Zen 4 architecture, codenamed Bergamo. Both are classified as server/workstation parts and share the AMD Socket SP5.

The manufacturing process differs significantly. The EPYC 9745 is built on a 3 nm process at TSMC, while the EPYC 9734 uses a 5 nm process, also at TSMC. The newer 3 nm node allows the EPYC 9745 to pack more transistors into a similar thermal envelope, which helps explain its performance advantage.

Core counts differ: the EPYC 9745 has 128 cores and 256 threads, while the EPYC 9734 has 112 cores and 224 threads. That is a 16-core and 32-thread advantage for the newer part. The EPYC 9734 does have a published transistor count of 71,000 million and a die size of 8x 73 mm², while the EPYC 9745 does not have published figures for those fields in the database.

Cache hierarchies show both processors share 256 MB of L3 cache, but the L1 and L2 configurations differ. The EPYC 9745 provides 80 KB of L1 per core and 1 MB of L2 per core. The EPYC 9734 provides 64 KB of L1 per core and 1 MB of L2 per core. The larger L1 per core on the EPYC 9745 likely contributes to better single-thread performance.

Clock speeds also separate the two. The EPYC 9745 has a base clock of 2.40 GHz and a boost clock of 3.70 GHz. The EPYC 9734 has a base clock of 2.20 GHz and a boost clock of 3.00 GHz. The higher boost clock on the EPYC 9745, combined with the Zen 5 architecture improvements, explains why it wins every single-thread benchmark in the database.

Both processors use DDR5 memory with a twelve-channel memory bus and support ECC memory. However, the EPYC 9745 has a memory bandwidth of 576.0 GB/s, while the EPYC 9734 has 460.8 GB/s. Both support PCIe Gen 5 with 128 lanes (CPU only). The EPYC 9745 was released on 2024-10-09, while the EPYC 9734 was released on 2023-06-12, making the 9734 the earlier design. Neither processor has an unlocked multiplier.

Head-to-Head Benchmarks

The benchmark data shows a clean sweep for the AMD EPYC 9745 across all 17 recorded comparisons. The smallest margins appear in PassMark find prime numbers, where the EPYC 9745 leads by 18.1%, and PassMark single-thread, where the lead is 21.5%. The largest margin is in PassMark physics at 153.8%.

Cinebench results are remarkably consistent. Every multi-core test, whether R15, R20, or R23, shows a 27.8% delta in favor of the EPYC 9745. The single-core Cinebench tests show a similar pattern, with R15 single-core at 27.7% and R20 and R23 single-core both at 27.8%. This consistency suggests the architectural improvements from Zen 4 to Zen 5 provide a uniform uplift across rendering workloads.

PassMark integer math shows one of the most dramatic differences: the EPYC 9745 scores 1,224,315 versus 823,150, a 48.7% advantage. This is notable because integer math often correlates with general server workloads, database operations, and compilation tasks. The floating-point math test also shows a strong 38.6% delta, with scores of 761,219 versus 549,045.

Data compression favors the EPYC 9745 by 35.5%, with scores of 3,929,890 versus 2,900,008. Extended instructions show a 36.2% delta, and random string sorting shows a 31.1% delta. Data encryption shows a 27.9% advantage for the EPYC 9745, with scores of 229,447 versus 179,390.

The PassMark physics result deserves special attention. The EPYC 9745 scores 17,122, while the EPYC 9734 scores 6,747. The 153.8% delta is the largest in the entire comparison and suggests that physics simulations, which often involve complex floating-point and branching workloads, benefit disproportionately from the Zen 5 architecture and higher core count.

The EPYC 9745 also leads in the aggregate PassMark multithread test by 27.8%, scoring 130,698 versus 102,286. The average benchmark score comparison reinforces this pattern, with the EPYC 9745 at 425,973 versus 310,619 for the EPYC 9734.

Specification Differences

The two processors differ in several key specification fields. Core count: the EPYC 9745 has 128 cores, the EPYC 9734 has 112. Threads: 256 versus 224. Base clock: 2.40 GHz versus 2.20 GHz. Boost clock: 3.70 GHz versus 3.00 GHz. TDP: 400 versus 340.

Architecture differs: Zen 5 versus Zen 4. Codename differs: Turin versus Bergamo. Process node differs: 3 nm versus 5 nm, both at TSMC. The EPYC 9734 has published transistor and die size figures of 71,000 million and 8x 73 mm², while the EPYC 9745 does not have those fields populated in the database.

Cache configurations differ at the L1 level: 80 KB per core versus 64 KB per core. Both share 1 MB L2 per core and 256 MB L3. Memory bandwidth differs: 576.0 GB/s versus 460.8 GB/s. Release dates differ: 2024-10-09 versus 2023-06-12. Launch MSRP differs: $12141 versus $9600. Part numbers differ: 100-000001460 versus 100-000001235.

The processors share identical memory support (DDR5), memory bus (twelve-channel), ECC support, PCIe configuration (Gen 5, 128 lanes CPU only), socket (AMD Socket SP5), manufacturer, market segment, production status, and multiplier unlock status. The EPYC 9734 has no integrated graphics, and the EPYC 9745 also lists integrated graphics as N/A.

The Verdict

The data points to one clear conclusion: the AMD EPYC 9745 outperforms the AMD EPYC 9734 in every measured metric. With 17 wins and zero losses, the EPYC 9745 demonstrates superiority across rendering, encryption, compression, math workloads, and single-thread tasks. The smallest delta is 18.1%, and the largest is 153.8%.

The EPYC 9745 achieves this through a combination of architectural advancement (Zen 5 versus Zen 4), a denser manufacturing process (3 nm versus 5 nm), more cores (128 versus 112), higher clocks (2.40/3.70 GHz versus 2.20/3.00 GHz), and greater memory bandwidth (576.0 GB/s versus 460.8 GB/s). These advantages compound into a processor that sits at the 100th percentile among all CPUs, while the EPYC 9734 sits at the 99th.

The average benchmark score difference is substantial: 425,973 versus 310,619. In relative terms, the EPYC 9745 leads its nearest rival by 3.4% over the AMD Ryzen Threadripper PRO 9995WX, while the EPYC 9734 trails its nearest rival by 0.4% behind the AMD EPYC 9575F. This places the EPYC 9745 as the higher-performing part not just against the EPYC 9734, but also within its own competitive tier.

Where Each One Wins

The AMD EPYC 9745 wins in every category recorded in the database. Its largest victory comes in PassMark physics, where the 153.8% delta indicates an exceptional advantage in simulation workloads. Its next strongest wins are in integer math (48.7%), floating-point math (38.6%), and extended instructions (36.2%). These results suggest the EPYC 9745 is particularly well-suited for compute-intensive scientific workloads, financial modeling, and any application that exercises heavy arithmetic operations.

The EPYC 9745 also excels in data compression (35.5%) and random string sorting (31.1%), pointing to strength in database compression, log processing, and data pipeline tasks. Encryption performance (27.9% ahead) covers security workloads such as TLS termination and VPN gateways. The consistent 27.8% advantage across Cinebench multi-core tests indicates solid rendering performance for visualization and simulation tasks.

The AMD EPYC 9734, while losing every benchmark, still occupies the 99th percentile of all CPUs. Its nearest rival comparisons show it trading blows in its tier: it trails the AMD EPYC 9575F by 0.4%, trails the Intel Xeon 6781P by 1.6%, trails the AMD Ryzen Threadripper PRO 9985WX by 3.2%, and leads the Intel Xeon 6774P by 3.4%. The EPYC 9734 remains a capable server processor for workloads that prioritize its 112 cores and 340 W TDP, which is lower than the EPYC 9745's 400 W TDP.

The EPYC 9734's 256 MB L3 cache and twelve-channel DDR5 support with ECC make it suitable for memory-intensive server environments. Its 5 nm process and Zen 4 architecture still deliver strong performance for its generation. However, in any direct comparison with the EPYC 9745, the newer part's advantages in architecture, process node, core count, clocks, and memory bandwidth produce decisive wins across every benchmark in the database.

For workloads that demand maximum throughput, the EPYC 9745 is the clear choice. For environments where the lower TDP and launch MSRP of the EPYC 9734 are priorities, and where the performance deltas are acceptable, the older part remains a viable option. The benchmark data gives no scenario where the EPYC 9734 outperforms the EPYC 9745.