Doom

The AMD Ryzen 9 9950X3D and NVIDIA GeForce RTX 5070 Ti combination delivers exceptional performance in Doom, with the data showing a clear dominance at lower resolutions and a predictable shift toward GPU limitation at 4K. The measured frame rates indicate this pairing is capable of exceeding 380 FPS at 1080p on Low settings, while still maintaining playable performance at 4K Ultra. This analysis breaks down the specific roles of the CPU and GPU, the measured FPS across all settings, and how resolution scaling affects the bottleneck profile.

FAQ

Q: What is the average FPS at 1080p with Ultra settings?

A: The measured average frame rate is 229.7 FPS at 1920x1080 resolution with Ultra settings. This is the most demanding preset tested at this resolution and still results in a highly playable experience.

Q: How does the RTX 5070 Ti compare to its nearest rival in average benchmark score?

A: The RTX 5070 Ti's average benchmark score is 52086, which is exactly 0% different from the NVIDIA GeForce RTX 4080's score of 52085. The data shows a 2.5% lead over the NVIDIA RTX A1000 and a 0.8% deficit against the AMD Radeon RX 7700S.

Q: What is the FPS difference between High and Ultra settings at 1440p?

A: At 2560x1440, the High preset delivers 185.3 FPS, while the Ultra preset delivers 150.3 FPS. This represents a drop of 35 FPS when moving from High to Ultra at this resolution.

Q: Does the Ryzen 9 9950X3D have a higher single-thread or multi-thread benchmark score?

A: The multi-thread score is significantly higher. The Cinebench R23 multi-core score is 59712, while the single-core score is 8430. This indicates the 16-core, 32-thread processor is heavily optimized for parallel workloads.

Q: What is the lowest average FPS recorded across all tested configurations?

A: The lowest average FPS is 78.4, which occurs at 3840x2160 resolution with Ultra settings. This is the only configuration tested that drops below 96.9 FPS.

Q: What is the GPU's memory bandwidth and how does it relate to performance?

A: The RTX 5070 Ti has a memory bandwidth of 896.0 GB/s, paired with 16 GB of GDDR7 memory on a 256-bit bus. This high bandwidth is a key factor in maintaining high frame rates, especially at 4K resolution.

GPU Role

The NVIDIA GeForce RTX 5070 Ti is built on the Blackwell 2.0 architecture with a GB203 chip, fabricated on a 5 nm process at TSMC. The GPU contains 45,600 million transistors on a 378 mm² die, resulting in a transistor density of 120.6M per mm². The core configuration includes 8960 shading units, 280 texture mapping units, 96 ROPs, 70 RT cores, and 280 tensor cores.

Clock speeds for the RTX 5070 Ti are set at a base of 2295 MHz and a boost of 2452 MHz. The memory operates at 1750 MHz, which translates to 28 Gbps effective, delivering the 896.0 GB/s bandwidth. Compute performance is rated at 43.94 TFLOPS for both FP32 and FP16 (1:1), with pixel rate at 235.4 GPixel/s and texture rate at 686.6 GTexel/s.

In Doom's measured results, the GPU's role becomes more pronounced as resolution increases. At 1080p, the RTX 5070 Ti's raw throughput is sufficient to push frame rates beyond 383 FPS on Low settings, indicating the GPU is not the primary constraint at this resolution. However, at 4K, the GPU's workload scales significantly, with the average FPS dropping from 383.5 at 1080p Low to 138.1 at 4K Low. The gap between Low and Ultra settings at 4K is substantial, with a 59.7 FPS difference (138.1 vs 78.4), showing how the GPU struggles with the higher preset's demands at that resolution.

The GPU's benchmark scores provide context for its performance tier. The Passmark G3D score is 32974, and the Geekbench OpenCL score is 231226. The GPU achieves a percentile rank of 87 among all GPUs, with an average benchmark score of 52086. This places it statistically equivalent to the RTX 4080, which scores 52085 with a deltaPct of 0%. The data shows a 2.5% advantage over the RTX A1000 and a 2.5% deficit against the Radeon Pro W6600M.

CPU Role

The AMD Ryzen 9 9950X3D is a 16-core, 32-thread processor from the 9000 series, based on the Zen 5 architecture (Granite Ridge) and manufactured on a 4 nm process at TSMC. It operates with a base clock of 4.30 GHz and a boost clock of 5.70 GHz, with a TDP of 170 W. The CPU supports DDR5 memory via a dual-channel interface, providing a memory bandwidth of 89.6 GB/s, and uses the AMD Socket AM5 platform with PCIe Gen 5 (24 lanes).

The cache hierarchy is extensive: 80 KB of L1 per core, 1 MB of L2 per core, and a substantial 128 MB of shared L3 cache. This large L3 cache is critical for gaming performance, as it reduces latency for frequently accessed data. The CPU contains 16,630 million transistors across a dual-die design with each die measuring 70.6 mm².

Benchmark results highlight the CPU's strengths. In Cinebench R23, the multi-core score is 59712, while the single-core score is 8430. The Geekbench multi-core score is 22515, and the single-core score is 3401. The 3DMark results show scaling from 1292 (single-thread) to 17184 (max threads), demonstrating excellent thread scaling. The Passmark multithread score is 70255, with a single-thread score of 4737.

The Ryzen 9 9950X3D achieves a percentile rank of 96 among all CPUs, with an average benchmark score of 71660. Its nearest rival is the Intel Xeon Platinum 8270, which scores 71056, a deltaPct of 0.9% slower. The AMD Ryzen 9 9950X scores 70420 (1.8% slower), while the Intel Xeon 6517P scores 73050 (1.9% faster). The Intel Core i7-14700KF scores 70104, which is 2.2% slower.

In Doom, the CPU's influence is most evident at 1080p, where frame rates are exceptionally high. At 1080p Low, the average FPS is 383.5, which suggests the CPU is keeping up with the GPU's output. The CPU's high single-thread performance, evidenced by the Cinebench R23 single-core score of 8430, is crucial for Doom's game logic and draw calls, especially at lower resolutions where the GPU has more headroom.

Measured FPS Breakdown

The measured FPS data covers three resolutions and four settings presets, providing a comprehensive view of performance. At 1920x1080, the average frame rates are 383.5 FPS (Low), 333.8 FPS (Medium), 285.1 FPS (High), and 229.7 FPS (Ultra). The spread from Low to Ultra is 153.8 FPS, indicating that the GPU scales well with settings at this resolution.

Moving to 2560x1440, the frame rates remain high but show a noticeable decrease. The measured averages are 251.6 FPS (Low), 215.9 FPS (Medium), 185.3 FPS (High), and 150.3 FPS (Ultra). The Low to Ultra spread narrows to 101.3 FPS, showing that the GPU is becoming more of a limiting factor.

At 3840x2160, the performance drops significantly. The averages are 138.1 FPS (Low), 117.9 FPS (Medium), 96.9 FPS (High), and 78.4 FPS (Ultra). The spread between Low and Ultra is 59.7 FPS, the smallest of all resolutions, which indicates the GPU is heavily loaded even at Low settings.

The combination's overall rank in Doom is 132 out of 1681 tested configurations, placing it in the top 8% of all possible pairings. The data is marked as measured, not estimated, ensuring accuracy.

Resolution Scaling

The FPS drop from 1080p to 4K reveals the shifting bottleneck profile. At 1080p Low, the system achieves 383.5 FPS. Moving to 1440p Low, the FPS drops to 251.6, a reduction of 131.9 FPS (34.4%). At 4K Low, the FPS falls to 138.1, a further drop of 113.5 FPS from 1440p (45.1%). The total drop from 1080p to 4K is 245.4 FPS, or 64% of the original performance.

For Ultra settings, the scaling is similar but with lower absolute values. The 1080p Ultra score is 229.7 FPS, dropping to 150.3 FPS at 1440p (a 34.6% decrease), and then to 78.4 FPS at 4K (a 47.8% decrease from 1440p). The total drop from 1080p to 4K Ultra is 151.3 FPS, or 65.9%.

The data indicates that the CPU is the primary limiting factor at 1080p, as the GPU is capable of higher frame rates but is constrained by the Ryzen 9 9950X3D's ability to feed it. At 4K, the GPU becomes the bottleneck, as evidenced by the relatively smaller FPS spread between Low and Ultra settings (59.7 FPS) compared to 1080p (153.8 FPS). This suggests that at 4K, the GPU is saturated and the CPU has headroom, while at 1080p, the CPU is the constraint and the GPU has unused capacity. The 128 MB L3 cache of the CPU helps mitigate CPU-bound scenarios, but the fundamental physics of frame rate scaling remain evident in the data.