GeForce RTX 5090 vs Radeon RX 6950 XT
Aggregate performance score
We've compared Radeon RX 6950 XT and GeForce RTX 5090, covering specs and all relevant benchmarks.
RTX 5090 outperforms RX 6950 XT by a substantial 37% based on our aggregate benchmark results.
Primary details
GPU architecture, market segment, value for money and other general parameters compared.
| Place in the ranking | 19 | 1 |
| Place by popularity | not in top-100 | 57 |
| Cost-effectiveness evaluation | 27.27 | 10.80 |
| Power efficiency | 14.95 | 11.93 |
| Architecture | RDNA 2.0 (2020−2024) | Blackwell 2.0 (2025) |
| GPU code name | Navi 21 | GB202 |
| Market segment | Desktop | Desktop |
| Release date | 10 May 2022 (2 years ago) | 30 January 2025 (recently) |
| Launch price (MSRP) | $1,099 | $1,999 |
Cost-effectiveness evaluation
The higher the performance-to-price ratio, the better. We use the manufacturer's recommended prices for comparison.
RX 6950 XT has 153% better value for money than RTX 5090.
Detailed specifications
General parameters such as number of shaders, GPU core base clock and boost clock speeds, manufacturing process, texturing and calculation speed. Note that power consumption of some graphics cards can well exceed their nominal TDP, especially when overclocked.
| Pipelines / CUDA cores | 5120 | 21760 |
| Core clock speed | 1925 MHz | 2017 MHz |
| Boost clock speed | 2324 MHz | 2407 MHz |
| Number of transistors | 26,800 million | 92,200 million |
| Manufacturing process technology | 7 nm | 5 nm |
| Power consumption (TDP) | 335 Watt | 575 Watt |
| Texture fill rate | 743.7 | 1,637 |
| Floating-point processing power | 23.8 TFLOPS | 104.8 TFLOPS |
| ROPs | 128 | 176 |
| TMUs | 320 | 680 |
| Tensor Cores | no data | 680 |
| Ray Tracing Cores | 80 | 170 |
Form factor & compatibility
Information on compatibility with other computer components. Useful when choosing a future computer configuration or upgrading an existing one. For desktop graphics cards it's interface and bus (motherboard compatibility), additional power connectors (power supply compatibility).
| Interface | PCIe 4.0 x16 | PCIe 5.0 x16 |
| Length | 267 mm | 304 mm |
| Width | 3-slot | 2-slot |
| Supplementary power connectors | 2x 8-pin | 1x 16-pin |
VRAM capacity and type
Parameters of VRAM installed: its type, size, bus, clock and resulting bandwidth. Integrated GPUs have no dedicated video RAM and use a shared part of system RAM.
| Memory type | GDDR6 | GDDR7 |
| Maximum RAM amount | 16 GB | 32 GB |
| Memory bus width | 256 Bit | 512 Bit |
| Memory clock speed | 2250 MHz | 1750 MHz |
| Memory bandwidth | 576.0 GB/s | 1.79 TB/s |
| Shared memory | - | - |
Connectivity and outputs
Types and number of video connectors present on the reviewed GPUs. As a rule, data in this section is precise only for desktop reference ones (so-called Founders Edition for NVIDIA chips). OEM manufacturers may change the number and type of output ports, while for notebook cards availability of certain video outputs ports depends on the laptop model rather than on the card itself.
| Display Connectors | 1x HDMI 2.1, 2x DisplayPort 1.4a | 1x HDMI 2.1b, 3x DisplayPort 2.1b |
| HDMI | + | + |
API and SDK compatibility
List of supported 3D and general-purpose computing APIs, including their specific versions.
| DirectX | 12 Ultimate (12_2) | 12 Ultimate (12_2) |
| Shader Model | 6.5 | 6.8 |
| OpenGL | 4.6 | 4.6 |
| OpenCL | 2.1 | 3.0 |
| Vulkan | 1.3 | 1.4 |
| CUDA | - | 10.1 |
| DLSS | - | + |
Synthetic benchmark performance
Non-gaming benchmark results comparison. The combined score is measured on a 0-100 point scale.
Combined synthetic benchmark score
This is our combined benchmark score.
Passmark
This is the most ubiquitous GPU benchmark. It gives the graphics card a thorough evaluation under various types of load, providing four separate benchmarks for Direct3D versions 9, 10, 11 and 12 (the last being done in 4K resolution if possible), and few more tests engaging DirectCompute capabilities.
Gaming performance
Let's see how good the compared graphics cards are for gaming. Particular gaming benchmark results are measured in FPS.
Average FPS across all PC games
Here are the average frames per second in a large set of popular games across different resolutions:
| Full HD | 220
+8.4%
| 203
−8.4%
|
| 1440p | 135
−37.8%
| 186
+37.8%
|
| 4K | 84
−76.2%
| 148
+76.2%
|
Cost per frame, $
| 1080p | 5.00
+97.1%
| 9.85
−97.1%
|
| 1440p | 8.14
+32%
| 10.75
−32%
|
| 4K | 13.08
+3.2%
| 13.51
−3.2%
|
- RX 6950 XT has 97% lower cost per frame in 1080p
- RX 6950 XT has 32% lower cost per frame in 1440p
- RX 6950 XT and RTX 5090 have nearly equal cost per frame in 4K
FPS performance in popular games
Full HD
Low Preset
| Atomic Heart | 347
+38.2%
|
250−260
−38.2%
|
| Counter-Strike 2 | 204
−4.4%
|
210−220
+4.4%
|
| Cyberpunk 2077 | 161
−49.1%
|
240−250
+49.1%
|
Full HD
Medium Preset
| Atomic Heart | 257
+2.4%
|
250−260
−2.4%
|
| Battlefield 5 | 170−180
−10.7%
|
190−200
+10.7%
|
| Counter-Strike 2 | 198
−7.6%
|
210−220
+7.6%
|
| Cyberpunk 2077 | 143
−67.8%
|
240−250
+67.8%
|
| Far Cry 5 | 181
−19.9%
|
210−220
+19.9%
|
| Fortnite | 300−350
+0%
|
300−350
+0%
|
| Forza Horizon 4 | 270−280
−27.4%
|
300−350
+27.4%
|
| Forza Horizon 5 | 237
−5.5%
|
250−260
+5.5%
|
| PLAYERUNKNOWN'S BATTLEGROUNDS | 170−180
+0%
|
170−180
+0%
|
| Valorant | 350−400
−75.3%
|
650−700
+75.3%
|
Full HD
High Preset
| Atomic Heart | 159
−57.9%
|
250−260
+57.9%
|
| Battlefield 5 | 170−180
−10.7%
|
190−200
+10.7%
|
| Counter-Strike 2 | 168
−26.8%
|
210−220
+26.8%
|
| Counter-Strike: Global Offensive | 270−280
+0%
|
270−280
+0%
|
| Cyberpunk 2077 | 128
−87.5%
|
240−250
+87.5%
|
| Dota 2 | 199
−35.7%
|
270−280
+35.7%
|
| Far Cry 5 | 173
−25.4%
|
210−220
+25.4%
|
| Fortnite | 300−350
+0%
|
300−350
+0%
|
| Forza Horizon 4 | 270−280
−27.4%
|
300−350
+27.4%
|
| Forza Horizon 5 | 229
−9.2%
|
250−260
+9.2%
|
| Grand Theft Auto V | 172
−1.2%
|
170−180
+1.2%
|
| Metro Exodus | 189
+174%
|
69
−174%
|
| PLAYERUNKNOWN'S BATTLEGROUNDS | 170−180
+0%
|
170−180
+0%
|
| The Witcher 3: Wild Hunt | 376
−16%
|
400−450
+16%
|
| Valorant | 350−400
−75.3%
|
650−700
+75.3%
|
Full HD
Ultra Preset
| Battlefield 5 | 170−180
−10.7%
|
190−200
+10.7%
|
| Counter-Strike 2 | 153
−32%
|
202
+32%
|
| Cyberpunk 2077 | 122
−96.7%
|
240−250
+96.7%
|
| Dota 2 | 167
−31.7%
|
220−230
+31.7%
|
| Far Cry 5 | 164
−32.3%
|
210−220
+32.3%
|
| Forza Horizon 4 | 270−280
−27.4%
|
300−350
+27.4%
|
| PLAYERUNKNOWN'S BATTLEGROUNDS | 170−180
+0%
|
170−180
+0%
|
| The Witcher 3: Wild Hunt | 216
−62%
|
350
+62%
|
| Valorant | 350−400
−75.3%
|
650−700
+75.3%
|
Full HD
Epic Preset
| Fortnite | 300−350
+0%
|
300−350
+0%
|
1440p
High Preset
| Counter-Strike 2 | 70−75
−158%
|
183
+158%
|
| Counter-Strike: Global Offensive | 500−550
+0%
|
500−550
+0%
|
| Grand Theft Auto V | 153
−10.5%
|
160−170
+10.5%
|
| Metro Exodus | 120
−68.3%
|
202
+68.3%
|
| PLAYERUNKNOWN'S BATTLEGROUNDS | 170−180
+0%
|
170−180
+0%
|
| Valorant | 450−500
−0.2%
|
450−500
+0.2%
|
1440p
Ultra Preset
| Battlefield 5 | 170−180
−13.3%
|
190−200
+13.3%
|
| Cyberpunk 2077 | 93
−67.7%
|
150−160
+67.7%
|
| Far Cry 5 | 163
−27%
|
200−210
+27%
|
| Forza Horizon 4 | 230−240
−30.2%
|
300−350
+30.2%
|
| The Witcher 3: Wild Hunt | 160−170
−96.9%
|
321
+96.9%
|
1440p
Epic Preset
| Fortnite | 150−160
+0%
|
150−160
+0%
|
4K
High Preset
| Atomic Heart | 60−65
−100%
|
128
+100%
|
| Counter-Strike 2 | 50−55
−248%
|
174
+248%
|
| Grand Theft Auto V | 174
−7.5%
|
180−190
+7.5%
|
| Metro Exodus | 77
−117%
|
167
+117%
|
| The Witcher 3: Wild Hunt | 144
−163%
|
378
+163%
|
| Valorant | 300−350
+0%
|
300−350
+0%
|
4K
Ultra Preset
| Battlefield 5 | 120−130
−6.3%
|
130−140
+6.3%
|
| Counter-Strike 2 | 20
−175%
|
55
+175%
|
| Cyberpunk 2077 | 46
−69.6%
|
75−80
+69.6%
|
| Dota 2 | 141
−34.8%
|
190−200
+34.8%
|
| Far Cry 5 | 124
−37.1%
|
170−180
+37.1%
|
| Forza Horizon 4 | 180−190
−61.4%
|
300−350
+61.4%
|
| PLAYERUNKNOWN'S BATTLEGROUNDS | 95−100
+0%
|
95−100
+0%
|
4K
Epic Preset
| Fortnite | 75−80
+0%
|
75−80
+0%
|
This is how RX 6950 XT and RTX 5090 compete in popular games:
- RX 6950 XT is 8% faster in 1080p
- RTX 5090 is 38% faster in 1440p
- RTX 5090 is 76% faster in 4K
Here's the range of performance differences observed across popular games:
- in Metro Exodus, with 1080p resolution and the High Preset, the RX 6950 XT is 174% faster.
- in Counter-Strike 2, with 4K resolution and the High Preset, the RTX 5090 is 248% faster.
All in all, in popular games:
- RX 6950 XT is ahead in 3 tests (5%)
- RTX 5090 is ahead in 45 tests (74%)
- there's a draw in 13 tests (21%)
Pros & cons summary
| Performance score | 73.02 | 100.00 |
| Recency | 10 May 2022 | 30 January 2025 |
| Maximum RAM amount | 16 GB | 32 GB |
| Chip lithography | 7 nm | 5 nm |
| Power consumption (TDP) | 335 Watt | 575 Watt |
RX 6950 XT has 71.6% lower power consumption.
RTX 5090, on the other hand, has a 36.9% higher aggregate performance score, an age advantage of 2 years, a 100% higher maximum VRAM amount, and a 40% more advanced lithography process.
The GeForce RTX 5090 is our recommended choice as it beats the Radeon RX 6950 XT in performance tests.
Other comparisons
We selected several comparisons of graphics cards with performance close to those reviewed, providing you with more options to consider.
