Cinebench 2026 CPU and GPU Score Tables
We have collected Cinebench 2026 results below so we can compare processors and graphics cards with their test settings in view. Our own CPU results come first, followed by community CPU runs and GPU scores. Higher scores mean faster rendering in the same test.
What Cinebench Means for Everyday Work
Cinebench runs a Redshift rendering workload, which makes it most relevant when we use a PC for 3D scenes, product visualizations or animation rendering. For that kind of work, more rendering throughput can mean shorter waits for a finished frame or more frames completed during an overnight batch.
We can also use repeated CPU runs to see how a system holds up under sustained load. A laptop that starts fast but slows as it heats up may deliver a different result from a desktop with more cooling headroom, even when their processor names look similar.
For a gaming upgrade, we would read these tables alongside actual game tests, average FPS and 1% lows. A higher multi-thread score alone does not establish which processor will run our games faster. Video editing, code compilation and other workloads also need their own application tests.
CPU Scores From Our Test Benches
Results are sorted by multi-thread score. ST means one thread, SC-SMT means one physical core using its SMT threads, and MT means all available threads. The Single Core test is unavailable on processors without SMT.
| Processor | Cores / Threads | Configuration | ST | SC-SMT | MT |
|---|---|---|---|---|---|
| Intel Core Ultra 7 270K Plus | 24C / 24T (8P + 16E) | ASUS AI OC | 606 | N/A | 10356 |
| AMD Ryzen 9 9950X | 16C / 32T | PBO enabled | 583 | 781 | 10233 |
| AMD Ryzen 9 9950X3D2 | 16C / 32T | Stock | 575 | Not measured | 10094 |
| Intel Core Ultra 7 270K Plus | 24C / 24T (8P + 16E) | Intel Performance + 200S BOOST | 584 | N/A | 9757 |
| Intel Core Ultra 7 270K Plus | 24C / 24T (8P + 16E) | ASUS MCE No Limit | 564 | N/A | 9705 |
| Intel Core Ultra 7 270K Plus | 24C / 24T (8P + 16E) | Stock, Intel Performance profile | 582 | N/A | 9646 |
| AMD Ryzen 9 9950X | 16C / 32T | Stock | 571 | 768 | 9612 |
| Intel Core Ultra 9 285K | 24C / 24T (8P + 16E) | Stock | 573 | N/A | 9472 |
| Intel Core Ultra 5 250K Plus | 18C / 18T (6P + 12E) | Stock | 565 | N/A | 7446 |
| AMD Ryzen 7 9800X3D | 8C / 16T | PBO enabled | 552 | 763 | 5816 |
| AMD Ryzen 7 9850X3D | 8C / 16T | PBO enabled | 549 | 786 | 5724 |
| AMD Ryzen 7 9800X3D | 8C / 16T | Stock | 539 | 747 | 5687 |
| AMD Ryzen 7 9850X3D | 8C / 16T | Stock | 574 | 769 | 5562 |
Our stock Ryzen 9 9950X3D2 recorded 10,094 points multi-threaded. The Core Ultra 7 270K Plus moved from 9,646 at stock to 10,356 with ASUS AI OC, a gain of about 7.4%. We keep those profiles in separate rows because tuning changes the comparison.
Reading the CPU Results
- Single Thread (ST): One execution thread. We use this column to compare lightly threaded rendering performance. It gives us one view of CPU capability, but it cannot predict every application’s responsiveness.
- Single Core with SMT (SC-SMT): One physical core using its available SMT threads. Comparing this with ST on the same processor shows the benefit of SMT in this workload. N/A means the test does not apply; it is not a score of zero.
- Multiple Threads (MT): All available CPU threads. This is the column we would prioritize for sustained CPU rendering, while checking whether each result is stock or tuned.
The stock Ryzen 9 9950X3D2 scores 10,094 against 5,687 for the stock Ryzen 7 9800X3D, about 77.5% higher in this test. Their ST scores are much closer, at 575 and 539. The larger gap in MT tells us where the extra rendering throughput appears; it does not promise a matching improvement in every task.
Score gains and time savings are also different percentages. Our Core Ultra 7 270K Plus gains about 7.4% with ASUS AI OC. If a one-hour rendering job scaled in exactly the same proportion, it would take about 56 minutes. That is an illustration of the ratio, not a measured render time.
Community CPU Scores
These results were reported by the named members of the Overclock.net Cinebench 2026 thread. We have retained their stated configurations and credits; we have not independently verified their runs.
| Processor | Configuration | ST | SC-SMT | MT | GPU | Reported by |
|---|---|---|---|---|---|---|
| AMD Ryzen 9 9950X3D2 | 16C / 32T, 5715 MHz, custom water, ASUS ROG CROSSHAIR X870E APEX, 2x16GB G.SKILL DDR5, BenchMate 14.1.0 | Not stated | 841 | 12244 | Not stated | domdtxdissar |
| AMD Ryzen 9 9950X3D2 | 16C / 32T at 4.35 GHz, PBO with BCLK adjustment, no delid, Radeon RX 9070 XT | Not stated | Not stated | 11734 | Not stated | ChrisG82 |
| Intel Core Ultra 7 270K Plus | P-core 5600 MHz at 1.30V, E-core 5150 MHz, 322W package, ASUS ROG MAXIMUS Z890 APEX, 2x16GB Team Group UD5-7200 Gear 2, BenchMate 14.1.0 | Not stated | Not stated | 11515 | Not stated | Digg de |
| AMD Ryzen 9 9950X3D | 16C / 32T at 4.3 GHz, delidded, TG Mycro Direct-Die Pro, ASUS Strix X870E-E, 96GB DDR5-6000 CL26, RTX 5090 | 593 | 803 | 11278 | 181623 | Asmodian |
| Intel Core Ultra 9 285K | Tuned, not delidded | Not stated | Not stated | 11201 | Not stated | MrTOOSHORT |
| Intel Core Ultra 9 285K | Tuned, earlier run | Not stated | Not stated | 11077 | Not stated | MrTOOSHORT |
| Intel Core Ultra 7 265K | P 8×52 / 6×55, E47, R39, DDR5-7200 CL34, RTX 5060 Ti 16GB. Power draw recorded at 150W package multi-thread, 15W single-thread, 75W on the GPU test | 578 | Not stated | 8666 | 61838 | persizi |
| Intel Core Ultra 7 265K | 24/7 profile, P-core 5200 MHz, E-core 4600 MHz, 147W package, MSI PRO Z890-A WIFI, 64GB Corsair DDR5 Gear 2 CL34, RTX 5060 Ti, BenchMate 14.1.0 | 578 | Not stated | 8580 | Not stated | persizi |
| AMD Ryzen 7 9700X | Windows 10, multi-core run only | Not stated | Not stated | 5913 | N/A | aEm |
| Intel Core Ultra 5 225F | 10C / 10T (6P + 4E), P-core 5057 MHz, E-core 4734 MHz, 84W package, ASUS ROG MAXIMUS Z890 APEX, 2x16GB DDR5 Gear 2 CL38, BenchMate 14.1.0 | Not stated | N/A | 4661 | Not stated | igor_kavinski |
GPU Scores
Our GPU reference table combines Cinebench’s built-in comparison entries with attributed Overclock.net runs. Driver versions and tuning settings are shown where available. Multi-GPU results represent the combined configuration.
| Graphics Card | VRAM | Driver | Configuration | GPU Score | Source |
|---|---|---|---|---|---|
| RTX 5090 + RTX 4090 + RTX 3090 | Multi-GPU | Not stated | Three cards combined, 5090 on stock vBIOS | 417997 | J7SC (Overclock.net) |
| NVIDIA GeForce RTX 5090 | 32GB | Not stated | Maxon reference entry | 182454 | Cinebench built-in ranking |
| NVIDIA GeForce RTX 5090 | 32GB | 591.86 | MSI Vanguard, stock | 181623 | Asmodian (Overclock.net) |
| NVIDIA GeForce RTX 5070 | 12GB | 595.97 | Core +330, memory +3000, power limit 110 percent | 87278 | ultraex2003_9978 (Overclock.net) |
| NVIDIA GeForce RTX 5070 | 12GB | Earlier driver | Core +340, memory +3000 | 84642 | ultraex2003_9978 (Overclock.net) |
| Apple M3 Ultra | Unified | Metal | Maxon reference entry | 83865 | Cinebench built-in ranking |
| NVIDIA GeForce RTX 5060 Ti | 16GB | Not stated | 2880 MHz at 0.900V, 75W measured | 61838 | persizi (Overclock.net) |
We would use the GPU column when our renderer actually runs on the graphics card. A high CPU MT result does not automatically improve a GPU-bound render, and a Cinebench GPU score does not tell us how fast a game runs. The three-card result above also cannot be treated as the performance of one RTX 5090.
Memory capacity matters alongside throughput. Maxon requires at least 8GB of GPU memory for this test, or 16GB unified memory on Apple silicon. Completing the benchmark does not guarantee that a larger production scene will fit. We would check the memory needs of our own projects before choosing a card.
How to Judge Your Own Result
- Match the test: Use Cinebench 2026 and compare ST with ST, SC-SMT with SC-SMT, MT with MT, or GPU with GPU.
- Find the matching configuration: Start with the same processor or graphics card at similar settings. A stock result should not be expected to match an overclocked community run.
- Repeat under consistent conditions: Close heavy background applications and keep the benchmark duration and power mode the same. Repeated runs help us distinguish a persistent gap from a one-off result.
- Investigate a persistent shortfall: Check temperatures, sustained clocks, power limits and memory configuration. For GPU testing, also record the driver version. A low score is a reason to inspect the system, not proof that a component is faulty.
There is no single passing score for every PC. We want our machine to perform consistently near comparable configurations, then confirm that it handles the applications and games we actually use. The tables give us a reference point for that comparison.
Our AM5 platform uses an ASUS ROG STRIX X870E-E GAMING WIFI7 NEO, 32GB DDR5 and an ARCTIC Liquid Freezer III PRO 420mm cooler. Our LGA1851 platform uses an ASUS ROG MAXIMUS Z890 HERO and 32GB G.SKILL Trident Z5 CK DDR5-8200 C40. Room temperature is approximately 28°C.
We can download the free benchmark from Maxon’s Cinebench page. Our Cinebench results guide explains the tests in more detail. For community comparisons, visit the Overclock.net Cinebench 2026 thread.





