AMD Ryzen Threadripper PRO 9965WX vs Intel Xeon 678X Comparison

The Verdict

The benchmark database paints a surprisingly one-sided picture for the Intel Xeon 678X, which takes 14 of the 16 recorded head-to-head wins against the AMD Ryzen Threadripper PRO 9965WX. The Xeon leads by narrow margins in the Cinebench suite, consistently 3.5% ahead in both single-core and multi-core tests, which suggests a well-rounded advantage rather than a workload-specific quirk. Its more decisive victories come in the PassMark suite, where it dominates floating point math by 57.6%, data encryption by 26.4%, and data compression by 25.7%. Meanwhile, the Threadripper PRO 9965WX wins only the two PassMark single-thread tests, and by a meaningful 17.4% margin. The data indicates that the Xeon 678X is the broader performer, while the Threadripper PRO 9965WX is a specialist for lightly threaded tasks.

The average benchmark score gap is substantial: the Xeon 678X records an average of 193477 across the suite, versus 147009 for the Threadripper PRO 9965WX. That is a 31.6% difference in aggregate scoring. The Xeon sits at the 99th percentile among all CPUs, while the Threadripper lands at the 98th, so both are elite parts. The nearest rival data reinforces the Xeon's standing: it trades blows with the AMD EPYC 9335 (only 0.4% behind) and leads the Intel Xeon 6741P by 0.7%, while the Threadripper PRO 9965WX sits just 0.1% ahead of the AMD EPYC 8434P. For buyers, the verdict from the numbers is clear: choose the Xeon 678X for maximum throughput across multi-threaded and mixed workloads, and choose the Threadripper PRO 9965WX only if single-thread responsiveness is the absolute priority and the 17.4% gain there outweighs losses everywhere else.

Architecture Differences

The two processors come from fundamentally different design philosophies. The Intel Xeon 678X uses the Granite Rapids architecture, built on an Intel 5 nm process with a die size of 2x 598 mm². It packs 48 cores and 96 threads, with a base clock of 2.40 GHz and a boost clock of 4.90 GHz. The AMD Ryzen Threadripper PRO 9965WX uses the Zen 5 architecture codenamed Shimada Peak, fabricated by TSMC on a 4 nm node, with a transistor count of 33,260 million spread across four chiplets of 70.6 mm² each. It offers 24 cores and 48 threads, clocked higher at 4.20 GHz base and 5.40 GHz boost.

Cache hierarchies diverge significantly. The Xeon 678X provides 112 KB of L1 per core, 2 MB of L2 per core, and a massive 192 MB of shared L3. The Threadripper PRO 9965WX has smaller per-core caches: 64 KB L1 and 1 MB L2, with 128 MB of L3. The Xeon's larger L3, combined with double the core count, likely explains its strong showing in multi-threaded and cache-sensitive tests like data compression and extended instructions. Both parts support DDR5 memory across eight channels with identical 409.6 GB/s bandwidth, and both offer 128 PCIe Gen 5 lanes from the CPU. ECC memory is supported on both, and neither has integrated graphics.

Thermal and power characteristics differ notably. The Xeon 678X has a 300 W TDP, while the Threadripper PRO 9965WX draws 350 W. The higher power envelope on the AMD part does not translate into higher multi-threaded scores, which is a striking observation given its higher clock speeds. The Xeon achieves its results with a lower TDP and more cores, indicating a different efficiency profile. Both processors have unlocked multipliers, and both are active production parts. The Xeon 678X launched in February 2026 with a launch MSRP of $3749, while the Threadripper PRO 9965WX launched in July 2025 with a launch MSRP of $2899.

FAQ

Q: Which processor has more cores and threads?

A: The Intel Xeon 678X has 48 cores and 96 threads, exactly double the 24 cores and 48 threads of the AMD Ryzen Threadripper PRO 9965WX. That core advantage is the most likely driver of its multi-threaded benchmark wins.

Q: Does the Threadripper PRO 9965WX win any benchmarks?

A: Yes, it wins the two PassMark single-thread tests, scoring 4551 versus the Xeon's 3758, a 17.4% advantage. It also has a higher boost clock at 5.40 GHz versus 4.90 GHz, which aligns with that single-thread result.

Q: How large is the Xeon 678X's lead in floating point performance?

A: The Xeon 678X scores 362070 in PassMark floating point math, compared to 229685 for the Threadripper PRO 9965WX, a 57.6% lead. This is the largest margin in the entire head-to-head comparison.

Q: Are both processors in the same performance percentile?

A: The Xeon 678X is at the 99th percentile among all CPUs, while the Threadripper PRO 9965WX is at the 98th percentile. Both are elite parts, but the Xeon sits one percentile higher.

Q: What memory bandwidth do the two support?

A: Both support DDR5 memory with an eight-channel bus, delivering identical 409.6 GB/s bandwidth. Neither has an advantage in raw memory throughput.

Q: Which chip has more L3 cache?

A: The Xeon 678X has 192 MB of shared L3 cache, while the Threadripper PRO 9965WX has 128 MB. The Xeon's larger L3 is 50% bigger, which may contribute to its wins in compression and encryption workloads.

Specification Differences

The key specification differences between the two processors are stark. The Xeon 678X has 48 cores versus 24, and 96 threads versus 48. Its base clock is lower at 2.40 GHz compared to 4.20 GHz, but its boost clock is also lower at 4.90 GHz versus 5.40 GHz. The TDP is lower on the Xeon at 300 W versus 350 W. The process node differs: Intel 5 nm for the Xeon, TSMC 4 nm for the Threadripper. Die size is dramatically different, with the Xeon using 2x 598 mm² and the Threadripper using 4x 70.6 mm². Transistor counts are only listed for the Threadripper at 33,260 million. Cache layouts differ: 112 KB L1 per core versus 64 KB, 2 MB L2 per core versus 1 MB, and 192 MB shared L3 versus 128 MB. The Xeon uses Intel Socket 4710, while the Threadripper uses AMD Socket sTR5. The Xeon is from the Granite Rapids architecture, the Threadripper from Zen 5 codenamed Shimada Peak. The Xeon's part number is SA2CX, the Threadripper's is 100-000000724. Release dates differ by roughly half a year, and the launch MSRP is $3749 for the Xeon versus $2899 for the Threadripper.

Head-to-Head Benchmarks

The Cinebench results show a consistent pattern. In Cinebench R15 multi-core, the Xeon 678X scores 8444 against 8162, a 3.5% lead. Single-core R15 is similarly tight: 1192 versus 1152, again 3.5%. Cinebench R20 multi-core sees 35185 versus 34009, and R20 single-core sees 4967 versus 4801, both at 3.5%. Cinebench R23 multi-core continues the trend with 83775 versus 80976. Across all five Cinebench tests, the Xeon wins by the same 3.5% delta, which suggests a stable architectural advantage rather than a workload-specific one.

The PassMark suite reveals where the Xeon really pulls away. Data compression is a decisive 25.7% win (1690896 versus 1345230), and data encryption is nearly identical at 26.4% (83598 versus 66155). Extended instructions show a 30% gap (141431 versus 108753). Find prime numbers is a 34.3% win (1010 versus 752). The largest margin comes in floating point math: 362070 versus 229685, a 57.6% blowout. Integer math is a 16% win (405075 versus 349195). Multithreaded PassMark shows a 6.4% lead (98559 versus 92604). Physics is close at 3.7% (7809 versus 7529), and random string sorting is a 9.7% win (164102 versus 149617).

The Threadripper PRO 9965WX's only wins come in the two identical PassMark single-thread tests, scoring 4551 versus 3758, a 17.4% advantage. This is a meaningful margin, but it is confined to a narrow slice of the benchmark suite. The data shows a clear split: the Xeon wins every multi-threaded and most single-threaded tests, while the Threadripper only wins the dedicated single-thread PassMark measurement.

Where Each One Wins

The Intel Xeon 678X is the clear winner for compute-heavy, parallel workloads. Its 57.6% lead in floating point math makes it the obvious choice for scientific simulations, rendering, and any workload that relies heavily on FPU throughput. The 26.4% encryption win and 25.7% compression win point to advantages in data-intensive server tasks, database operations, and content delivery. The 30% lead in extended instructions suggests better support for advanced instruction sets, which matters for code that uses SIMD or specialized operations. The 34.3% win in prime number finding indicates strong integer throughput in algorithmic workloads. With 48 cores and 96 threads, the Xeon is positioned for heavily threaded environments like large-scale virtualization, batch processing, and multi-tenant server deployments. Its 192 MB of L3 cache likely helps in workloads with large working sets, such as in-memory databases or big-data analytics.

The AMD Ryzen Threadripper PRO 9965WX wins in single-thread responsiveness. Its 17.4% lead in PassMark single-thread tests, combined with a 5.40 GHz boost clock, makes it attractive for applications that are poorly parallelized or latency-sensitive. This includes certain legacy software, interactive workstation tasks, and code that cannot take advantage of many cores. The Threadripper also has a lower launch MSRP, though the data does not quantify the performance-per-dollar relationship. For users who prioritize the fastest possible response in single-threaded applications and are willing to accept lower multi-threaded scores, the Threadripper PRO 9965WX is the data-backed choice. For everyone else, the Xeon 678X's 14 wins out of 16 benchmarks make it the dominant pick. The aggregate average score of 193477 versus 147009 reinforces that conclusion: the Xeon is the higher-performing processor overall, and the Threadripper's single-thread niche does not offset its losses in the broader benchmark suite.