Intel Xeon 6741P vs Intel Xeon 6980P Comparison

Head-to-Head Benchmarks

The benchmark data presents a fascinating split between these two Granite Rapids server processors. The Intel Xeon 6741P, despite having fewer cores, dominates the Cinebench suite with a consistent 26.2% advantage across every single test. In Cinebench R23 multicore, the 6741P scores 85,561 against the 6980P's 63,175, while in single-core R23 it posts 12,079 versus 8,918. This same margin repeats in R15 and R20, suggesting a fundamental per-core performance advantage for the smaller chip rather than a workload-specific quirk.

The PassMark results tell a different story entirely. The 128-core 6980P wins six of the eleven PassMark workloads, with particularly strong showings in data encryption (125,246 versus 89,746, a 39.6% advantage), extended instructions (214,794 versus 142,682, a 50.5% edge), and floating-point math (501,720 versus 358,423, a 40% lead). Integer math also favors the larger chip at 637,476 versus 458,058, a 39.2% gap. Data compression and random string sorting round out the 6980P's wins with margins of 30.2% and 35.8% respectively.

Yet the 6741P counters decisively in several PassMark workloads. Its passmark_physics score of 13,890 crushes the 6980P's 3,350, a 75.9% advantage that stands as the largest delta in the entire comparison. Prime number finding also strongly favors the 6741P (1,242 versus 555, a 55.3% gap), and single-thread performance shows a 47.4% lead (3,195 versus 1,681). The multithread PassMark score goes to the 6741P as well: 100,660 versus 74,324, matching the 26.2% Cinebench margin.

What emerges from the data is a processor that wins on raw throughput in encryption, compression, and math-heavy workloads, and another that wins in rendering, physics simulation, single-thread tasks, and the overall PassMark multithread score. The 6980P's average benchmark score of 251,516 places it 3% ahead of the AMD EPYC 9634 and 5.6% ahead of the Intel Xeon 6747P, while trailing the AMD EPYC 9684X by 5.8% and the AMD Ryzen Threadripper 9970X by 10.1%. The 6741P, with an average score of 194,901, sits 0.3% ahead of the AMD EPYC 9335 and 0.7% ahead of the Intel Xeon 678X, with larger margins over the Intel Xeon 6740E (3.8%) and AMD EPYC 8534P (5.3%). Both chips rank in the 99th percentile of all CPUs in the database.

The Verdict

The data makes the choice workload-dependent rather than absolute. The Intel Xeon 6980P is the pick for data-centric server tasks. Its 30.2% lead in data compression, 39.6% in encryption, and 50.5% in extended instructions point directly at database, security, and analytics workloads where these operations dominate. The 40% floating-point advantage and 39.2% integer math edge further cement its position for scientific computing and financial modeling.

The Intel Xeon 6741P is the choice for rendering and general-purpose compute. Its 26.2% advantage across every Cinebench test, both multicore and single-core, makes it the stronger option for 3D rendering, animation, and CPU-based content creation. The 75.9% physics score gap is particularly striking and suggests simulation workloads will run dramatically better on the smaller chip. The 47.4% single-thread advantage also matters for legacy applications that scale poorly across cores.

For organizations running mixed workloads, the 6741P's broader wins (11 versus 6) and its superiority in the widely-used Cinebench suite may make it the more versatile daily driver. The 6980P is more specialized, excelling precisely where massive core counts and large shared caches pay off. The 6980P's 504 MB of L3 cache versus 288 MB on the 6741P likely explains much of its throughput advantage in cache-sensitive workloads like compression and encryption.

Architecture Differences

Both processors share the Granite Rapids architecture and Intel's 5 nm process node, but they diverge significantly in configuration. The 6980P packs 128 cores and 256 threads, while the 6741P offers 48 cores and 96 threads. Both use the same per-core cache layout: 112 KB of L1 and 2 MB of L2 per core. The shared L3 cache differs substantially, with the 6980P carrying 504 MB against the 6741P's 288 MB.

The physical packages are distinct as well. The 6980P uses a 3x 598 mm² die configuration on Intel Socket 7529, while the 6741P uses a 2x 598 mm² setup on Intel Socket 4710. Power envelopes differ, with the 6980P rated at 500 W and the 6741P at 300 W. Clock speeds favor the smaller chip: the 6741P runs at 2.50 GHz base and 3.80 GHz boost, versus 2.00 GHz base and 3.90 GHz boost on the 6980P. The higher base clock on the 6741P helps explain its single-core and lightly-threaded advantages.

Memory subsystems also diverge. The 6980P supports twelve-channel DDR5 with 614.4 GB/s of bandwidth, while the 6741P uses eight-channel DDR5 at 409.6 GB/s. Both support ECC memory. PCIe connectivity runs in the opposite direction: the 6741P offers 136 Gen 5 lanes, while the 6980P provides 96 lanes. Both lack integrated graphics and have locked multipliers. The 6980P launched on 2024-09-23, while the 6741P followed on 2025-02-23. The 6980P carries a launch MSRP of $12460; the 6741P carries a launch MSRP of $4421.

FAQ

Q: Which processor has more cores and threads?

A: The Intel Xeon 6980P has 128 cores and 256 threads, while the Intel Xeon 6741P has 48 cores and 96 threads.

Q: Why does the 6741P win Cinebench tests despite having fewer cores?

A: The 6741P has a higher base clock of 2.50 GHz versus 2.00 GHz on the 6980P, and the benchmark results show a consistent 26.2% advantage across all Cinebench tests, indicating superior per-core performance.

Q: Which processor has more L3 cache?

A: The 6980P has 504 MB of shared L3 cache, while the 6741P has 288 MB.

Q: What is the memory bandwidth difference?

A: The 6980P supports twelve-channel DDR5 with 614.4 GB/s bandwidth, while the 6741P supports eight-channel DDR5 with 409.6 GB/s.

Q: Which processor ranks higher in the database's average benchmark score?

A: The 6980P has an average benchmark score of 251,516, compared to 194,901 for the 6741P. The 6980P sits 3% above the AMD EPYC 9634, while the 6741P sits 0.3% above the AMD EPYC 9335.

Q: Do both processors use the same socket?

A: No. The 6980P uses Intel Socket 7529, while the 6741P uses Intel Socket 4710.

Where Each One Wins

The 6980P wins in data-heavy and throughput-oriented workloads. Data compression performance is 30.2% higher, making it the better fit for backup systems, log processing, and database storage engines. Encryption workloads see a 39.6% improvement, which matters for secure communications, VPN gateways, and encrypted storage. Extended instruction workloads, where the 6980P leads by 50.5%, point to specialized scientific and enterprise applications that leverage AVX-style instruction sets. The 40% floating-point math advantage suits computational fluid dynamics, weather modeling, and machine learning inference. Integer math at 39.2% higher helps with financial risk calculations and large-scale data analytics. Random string sorting, 35.8% faster, benefits text processing, search indexing, and data transformation pipelines.

The 6741P wins in rendering, simulation, and general compute. Its 26.2% lead across all Cinebench versions makes it the stronger choice for 3D rendering, animation studios, and architectural visualization. The 75.9% physics advantage points to engineering simulation, game physics, and scientific modeling that rely on rigid body dynamics. Prime number finding at 55.3% higher suits cryptography research, number theory computation, and certain optimization algorithms. The 47.4% single-thread advantage helps legacy enterprise applications, database query processing that runs single-threaded, and software development workloads. The PassMark multithread score, 26.2% higher, suggests the 6741P delivers better aggregate performance for mixed server workloads and virtualization hosts running diverse application stacks.

The socket and platform differences reinforce these divisions. The 6980P's twelve-channel memory and 504 MB L3 cache make it a data platform, while the 6741P's higher clocks and 136 PCIe lanes position it as a flexible compute node for high-throughput I/O and per-core performance. Neither chip is universally better; the recorded data shows two different optimization philosophies within the same architecture family.