AMD EPYC 9565 vs AMD Ryzen Threadripper 9970X Comparison

The AMD EPYC 9565 and AMD Ryzen Threadripper 9970X are both Zen 5 parts, but the data positions them at opposite ends of the compute spectrum. The EPYC 9565 wins 9 of the 11 head-to-head benchmarks, while the Threadripper 9970X takes only 2, both single-thread tests. The EPYC's average benchmark score of 285471 places it 2% above the Threadripper's 279778, a slim margin that belies the massive differences in core count and memory architecture. The Threadripper counters with a 5.40 GHz boost clock and a 4530 single-thread score, making it the clear choice for latency-sensitive workloads, while the EPYC's 72 cores and 384 MB of L3 cache dominate every throughput-oriented test.

The Verdict

The data dictates a clear split based on workload type. The AMD EPYC 9565 is the definitive choice for server and workstation environments where parallel throughput is paramount. Its 72 cores and 144 threads deliver a 293.8% advantage in the find prime numbers test and a 163.9% lead in physics calculations over the Threadripper 9970X. Any workload that scales across many cores — data compression, encryption, floating-point math — will see substantially higher absolute scores on the EPYC. The EPYC also offers 128 PCIe Gen 5 lanes and twelve-channel memory, making it the superior platform for multi-GPU servers or memory-bandwidth-hungry databases.

The AMD Ryzen Threadripper 9970X is the pick for single-threaded performance and high-frequency responsiveness. Its 4.00 GHz base clock and 5.40 GHz boost clock are both higher than the EPYC's 3.15 GHz and 4.30 GHz, and its passmark single-thread score of 4530 is 18.4% higher. For desktop workflows that rely on fewer, faster cores — legacy applications, certain simulation tools, or interactive 3D modeling — the Threadripper will feel snappier. Its 350 W TDP and unlocked multiplier also make it a more flexible platform for a desktop workstation, whereas the EPYC's 400 W TDP and locked multiplier target fixed server deployments.

The average benchmark scores are nearly identical — 285471 vs 279778 — but they hide the divergent strengths. The EPYC's 99th percentile ranking among all CPUs matches the Threadripper's, but the EPYC achieves this via raw core count while the Threadripper does so via clock speed. If the workload is parallel, buy the EPYC 9565. If the workload is single-threaded or lightly threaded, buy the Threadripper 9970X. There is no universal winner; the data supports a strict division of labor.

FAQ

Q: Which processor has more cores and threads?

A: The AMD EPYC 9565 has 72 cores and 144 threads, while the AMD Ryzen Threadripper 9970X has 32 cores and 64 threads. The EPYC has more than double the core count.

Q: What is the single-thread performance difference?

A: The Threadripper 9970X scores 4530 in the passmark single-thread test, which is 18.4% higher than the EPYC 9565's score of 3696. The Threadripper also has a higher boost clock of 5.40 GHz compared to 4.30 GHz.

Q: How does the cache configuration differ?

A: The EPYC 9565 features 384 MB of shared L3 cache and 80 KB of L1 cache per core. The Threadripper 9970X has 128 MB of L3 cache and 64 KB of L1 cache per core. Both have 1 MB of L2 cache per core.

Q: Which processor offers more memory bandwidth?

A: The EPYC 9565 provides 576.0 GB/s of memory bandwidth via a twelve-channel memory bus. The Threadripper 9970X offers 204.8 GB/s via a quad-channel bus. The EPYC's bandwidth is nearly three times higher.

Q: Are both processors on the same manufacturing process?

A: Yes, both the EPYC 9565 and Threadripper 9970X are built on TSMC's 4 nm process node. They both use the Zen 5 architecture but have different codenames: Turin for the EPYC and Shimada Peak for the Threadripper.

Q: What is the difference in PCIe lane count?

A: The EPYC 9565 supports 128 PCIe Gen 5 lanes (CPU only), while the Threadripper 9970X supports 80 PCIe Gen 5 lanes (CPU only). The EPYC provides 60% more PCIe lanes.

Architecture Differences

Both processors share the Zen 5 architecture and the 4 nm TSMC process node, but they diverge significantly in chip layout and memory hierarchy. The EPYC 9565, codenamed Turin, uses 12 compute dies, each measuring 70.6 mm², totaling 99,780 million transistors. The Threadripper 9970X, codenamed Shimada Peak, uses only 4 of these same 70.6 mm² dies, with 33,260 million transistors. This die-count difference directly explains the core count gap: the EPYC's 72 cores versus the Threadripper's 32 cores.

The cache hierarchy is also distinct. The EPYC 9565 allocates 80 KB of L1 cache per core, while the Threadripper uses 64 KB per core. Both have 1 MB of L2 per core, but the L3 cache differs drastically: 384 MB shared on the EPYC versus 128 MB on the Threadripper. This 3x difference in L3 capacity is critical for large working sets in server workloads. The EPYC's twelve-channel memory bus versus the Threadripper's quad-channel bus further separates them, with the EPYC achieving 576.0 GB/s bandwidth compared to 204.8 GB/s.

The socket and platform targets diverge as well. The EPYC 9565 uses AMD Socket SP5, a server socket, while the Threadripper uses AMD Socket sTR5, a high-end desktop socket. The EPYC's market segment is listed as "Server/Workstation," while the Threadripper is classified as "Desktop." The EPYC has a locked multiplier, whereas the Threadripper has an unlocked multiplier, reflecting their respective deployment contexts. Both support DDR5 memory and ECC, but the EPYC's twelve-channel configuration is designed for maximum bandwidth in multi-socket servers, while the Threadripper's quad-channel is tailored for desktop workstations.

Specification Differences

The core and thread counts are the most obvious divergence: the EPYC 9565 has 72 cores and 144 threads, versus the Threadripper 9970X's 32 cores and 64 threads. Clock speeds also differ, with the EPYC running at 3.15 GHz base and 4.30 GHz boost, while the Threadripper runs at 4.00 GHz base and 5.40 GHz boost. The TDP reflects this: the EPYC is rated at 400 W, the Threadripper at 350 W.

Memory support differs in channel count and bandwidth. The EPYC 9565 uses a twelve-channel memory bus with 576.0 GB/s bandwidth, while the Threadripper 9970X uses a quad-channel bus with 204.8 GB/s. Both support DDR5 and ECC memory. PCIe lanes also differ: the EPYC offers 128 Gen 5 lanes (CPU only), while the Threadripper offers 80 Gen 5 lanes (CPU only).

The cache specifications are different in L1 and L3. The EPYC has 80 KB of L1 per core and 384 MB of shared L3, while the Threadripper has 64 KB of L1 per core and 128 MB of L3. L2 cache is identical at 1 MB per core. The transistor count and die count differ substantially: 99,780 million transistors across 12 dies for the EPYC, versus 33,260 million across 4 dies for the Threadripper. The release dates are also different, with the EPYC launching on 2024-10-09 and the Threadripper on 2025-07-29. The launch MSRP for the EPYC 9565 is $10486, and for the Threadripper 9970X it is $2499. The EPYC is locked, while the Threadripper has an unlocked multiplier.

Head-to-Head Benchmarks

The EPYC 9565 dominates the multi-threaded tests with large margins. In the passmark find prime numbers test, the EPYC scores 2422 versus the Threadripper's 615, a 293.8% advantage. The physics test shows a 163.9% lead, with scores of 18036 and 6835 respectively. Floating-point math also favors the EPYC heavily: 549422 versus 309719, a 77.4% difference. These three tests highlight the EPYC's core-count advantage in compute-heavy parallel workloads.

The data compression and encryption tests continue the trend. The EPYC scores 2579631 in data compression, 46.7% higher than the Threadripper's 1757998. In data encryption, the EPYC's 141936 beats the Threadripper's 86765 by 63.6%. Extended instructions show a 47.2% lead for the EPYC (209595 vs 142342), and integer math is 54.3% higher (717948 vs 465378). Random string sorting favors the EPYC by 52.5% (291941 vs 191445), and the multithread test shows a 25.9% lead (135221 vs 107399).

The Threadripper 9970X wins only the single-thread tests. In passmark single-thread, it scores 4530 versus the EPYC's 3696, an 18.4% advantage. The same result appears in the passmark singlethread test, with identical scores. These wins are modest compared to the EPYC's margins but are consistent across both single-thread variants. The winsA count of 9 and winsB count of 2 accurately reflect this benchmark distribution.

Where Each One Wins

The EPYC 9565 is the winner in every multi-threaded and parallel workload. Its 293.8% lead in prime number finding and 163.9% lead in physics make it ideal for scientific computing, financial modeling, and any CPU-bound simulation that scales across cores. The 77.4% advantage in floating-point math suits engineering analysis and rendering workloads. The 63.6% lead in encryption and 46.7% in data compression make it the better choice for database servers, file servers, and security applications. The 128 PCIe Gen 5 lanes and 576.0 GB/s memory bandwidth further cement its position for multi-GPU compute nodes or high-throughput storage arrays.

The Threadripper 9970X wins in single-threaded performance and latency-sensitive tasks. Its 18.4% single-thread advantage makes it superior for legacy software that uses one or two cores, interactive modeling tools, and code compilation with high serial dependencies. The higher 5.40 GHz boost clock provides faster response in applications that cannot leverage the EPYC's 72 cores. The unlocked multiplier allows for overclocking, which could extend its single-thread lead further. For a desktop workstation where the user values per-core speed over absolute throughput, the Threadripper 9970X is the data-backed choice. The EPYC 9565 wins where core count and memory bandwidth matter most; the Threadripper wins where clock speed and single-thread agility are the priority.