AMD Ryzen Threadripper PRO 3995WX vs AMD Ryzen Threadripper PRO 9975WX Comparison

The AMD Ryzen Threadripper PRO 9975WX and the AMD Ryzen Threadripper PRO 3995WX represent two distinct approaches to high-end workstation computing, separated by five years of architectural evolution. The 9975WX, built on the Zen 5 architecture, wins 13 of the 17 head-to-head comparisons, while the older 3995WX, with its Zen 2 design, retains victories in four specific workloads. This analysis breaks down where each processor excels, what the benchmark data reveals about their respective strengths, and how their fundamental specifications dictate their performance profiles.

Where Each One Wins

The 9975WX is the clear winner in nearly every throughput and latency-sensitive category. Its dominance is most pronounced in Cinebench workloads, where it leads by roughly 31% across all R15, R20, and R23 multi-core and single-core tests. The data shows a consistent 31.1% to 31.3% advantage in these rendering benchmarks, indicating that the newer architecture delivers substantial generational improvements regardless of thread count or workload intensity. The 9975WX also wins in PassMark’s multi-thread test (25% ahead), physics simulation (32.6% ahead), and extended instructions (29.4% ahead). Its single-thread performance is extraordinary, with a 69.9% lead over the 3995WX in PassMark’s single-thread test, a margin that reflects the massive IPC gains from Zen 5.

Conversely, the 3995WX retains wins in four specific PassMark tests: data compression (10.7% ahead), data encryption (31.7% ahead), integer math (5.8% ahead), and random string sorting (0.3% ahead). These victories are notable because they occur despite the older chip’s significantly lower clock speeds and older architecture. The 3995WX’s advantage in these areas is likely attributable to its 64 physical cores and 128 threads—double the 9975WX’s 32 cores and 64 threads—which provide raw parallel throughput that can overcome architectural disadvantages in workloads that scale almost perfectly with core count. The encryption result is particularly striking, with the 3995WX outperforming the 9975WX by 31.7%, suggesting that certain cryptographic workloads remain heavily core-count dependent rather than clock-speed dependent.

FAQ

Q: Which processor has higher single-core performance?

A: The AMD Ryzen Threadripper PRO 9975WX is dramatically faster in single-threaded workloads. In PassMark’s single-thread test, it scores 4408 compared to the 3995WX’s 2595, a 69.9% advantage. Cinebench R23 single-core results show a similar trend: 13207 versus 10071, a 31.1% lead.

Q: Does the older 3995WX win any benchmarks?

A: Yes, the 3995WX wins four of the 17 head-to-head comparisons. It leads in PassMark data compression (1841292 vs 1644573), data encryption (124433 vs 85035), integer math (490116 vs 461724), and random string sorting (188574 vs 188014). These wins are concentrated in workloads that benefit from its higher core count.

Q: How do the two chips compare in multi-threaded rendering?

A: The 9975WX is consistently about 31% faster in Cinebench multi-core tests. In Cinebench R23 multi-core, it scores 93553 versus 71338 for the 3995WX. The PassMark multi-thread test shows a 25% advantage for the 9975WX (104902 vs 83928).

Q: What is the core and thread count difference?

A: The 3995WX has 64 cores and 128 threads, while the 9975WX has 32 cores and 64 threads. Despite having half the cores, the 9975WX wins the majority of benchmarks due to its superior architecture and higher clock speeds.

Q: Which processor has a higher boost clock?

A: The 9975WX has a boost clock of 5.40 GHz, compared to 4.20 GHz for the 3995WX. Its base clock is also higher at 4.00 GHz versus 2.70 GHz.

Q: How does memory bandwidth differ between the two?

A: The 9975WX supports DDR5 memory with a bandwidth of 409.6 GB/s, exactly double the 204.8 GB/s of the 3995WX, which uses DDR4. Both utilize an eight-channel memory bus.

Head-to-Head Benchmarks

The most decisive victories for the 9975WX come in single-threaded performance tests. PassMark’s single-thread test shows the 9975WX scoring 4408 against 2595 for the 3995WX, a 69.9% delta that indicates the Zen 5 architecture’s IPC improvements are nothing short of transformative. Cinebench R15 single-core tells a similar story: 1331 vs 1014, a 31.3% lead. This pattern repeats across every Cinebench iteration, with the 9975WX achieving precisely 31.1% to 31.3% advantages in both single-core and multi-core tests. The consistency of these margins suggests that the performance gap is architectural rather than workload-dependent, with the newer chip simply executing instructions more efficiently per clock.

The multi-core Cinebench results deserve particular attention because they show the 9975WX winning despite a 2:1 core deficit. In Cinebench R23 multi-core, the 9975WX scores 93553 while the 3995WX manages 71338. The 9975WX’s 31.1% advantage here means that its 32 cores deliver more aggregate rendering performance than the 3995WX’s 64 cores. This is a direct consequence of the 5.40 GHz boost clock versus 4.20 GHz, combined with the IPC gains of Zen 5 over Zen 2. The PassMark physics test shows the largest relative margin for the 9975WX at 32.6% (7288 vs 5498), indicating that simulation workloads heavily favor the newer architecture.

The 3995WX’s wins are equally instructive. Its 31.7% advantage in data encryption (124433 vs 85035) is the largest margin for either processor in any test. Data compression shows a 10.7% lead (1841292 vs 1644573), while integer math and random string sorting show narrower 5.8% and 0.3% edges respectively. These four workloads share a common characteristic: they are memory-latency sensitive and scale exceptionally well with raw core count. The 3995WX’s 256 MB of L3 cache—double the 9975WX’s 128 MB—likely plays a role in its compression and string sorting performance, providing a larger working set for data manipulation tasks.

Specification Differences

The most fundamental difference is core count: the 3995WX has 64 cores and 128 threads, while the 9975WX has 32 cores and 64 threads. This is inverted from what one might expect from a newer product, but the 9975WX compensates with significantly higher clocks. Its base clock is 4.00 GHz versus 2.70 GHz, and its boost clock reaches 5.40 GHz compared to 4.20 GHz. The 9975WX also has a higher TDP at 350 W versus 280 W for the 3995WX.

Memory support represents a generational leap. The 9975WX uses DDR5 with 409.6 GB/s of bandwidth, while the 3995WX uses DDR4 with 204.8 GB/s. Both feature eight-channel memory buses and ECC support. PCIe capabilities also differ: the 9975WX offers Gen 5 with 128 lanes (CPU only), while the 3995WX offers Gen 4 with the same 128 lanes. The 9975WX has an unlocked multiplier, whereas the 3995WX does not. The 9975WX’s launch MSRP is $4099, while the 3995WX launched at $5489.

Cache hierarchies differ notably. Both have 64 KB of L1 cache per core, but the 9975WX has 1 MB of L2 per core versus 512 KB for the 3995WX. L3 cache totals 128 MB for the 9975WX and 256 MB for the 3995WX, giving the older chip a 2:1 advantage in last-level cache. The 3995WX uses eight 74 mm² dies, while the 9975WX uses four 70.6 mm² dies.

Architecture Differences

The 9975WX is built on the Zen 5 architecture, codenamed Shimada Peak, while the 3995WX uses Zen 2, codenamed Castle Peak. This architectural gap spans three generations and explains the benchmark disparities. The 9975WX is manufactured on a 4 nm process at TSMC, compared to the 7 nm process for the 3995WX. The newer process node enables higher clock speeds and better power efficiency, contributing to the 9975WX’s ability to reach 5.40 GHz despite having fewer cores.

Transistor counts are similar—33,260 million for the 9975WX versus 30,400 million for the 3995WX—but the 9975WX packs these into four dies instead of eight. This die configuration, combined with the denser 4 nm process, allows for more efficient communication between cores and reduced latency. The 9975WX belongs to the 9000 series and the Ryzen Threadripper (Zen 5 Shimada Peak) generation, while the 3995WX is from the 3000 series and the Ryzen Threadripper (Zen 2 Castle Peak) generation.

The memory architecture shift from DDR4 to DDR5 doubles available bandwidth, which is critical for the 9975WX’s performance in bandwidth-intensive workloads. The PCIe upgrade from Gen 4 to Gen 5 similarly doubles I/O throughput potential. The 9975WX is classified for the Server/Workstation market segment, while the 3995WX is listed as Desktop, though both target professional workloads. The 9975WX’s L2 cache is double per core (1 MB vs 512 KB), but the 3995WX’s L3 cache is double in total (256 MB vs 128 MB). These architectural choices reflect different design philosophies: the 9975WX prioritizes per-core efficiency and low latency, while the 3995WX maximizes aggregate cache capacity and raw thread count.