The beginning of 2025 from Nvidia is marked with the launch of the GeForce RTX 5090, the world’s most powerful consumer graphics card. The card took the crown from GeForce RTX 4090 with an impressive margin. The card packs enormous processing power and is also quite power-hungry. It is the ultimate graphics card for 4K gaming, and no other card can come even close to it when it comes to raw performance. However, if you wish to find out how the flagship card from AMD, the Radeon RX 7900 XTX, and Nvidia’s previous flagship RTX 4090 compare against the mighty RTX 5090, please go through the comparison below.
RTX 5090 vs RX 7900 XTX vs RTX 4090 Comparison
A quick comparison between RTX 5090, RTX 4090, and RX 7900 XTX graphics cards based on their specifications, gaming performance, pricing, power consumption, and features.
GPU Architecture
GeForce RTX 5090 is built on the newer Blackwell GPU architecture on a 5nm manufacturing process. It uses GB202 GPU chip and has 21760 CUDA Cores, 170 RT Cores (4th Generation), and 680 Tensor Cores (5th Generation). On the other hand, RTX 4090 is built on the Ada Lovelace GPU architecture with a 5nm manufacturing process. It uses AD102 GPU chip and has 16384 CUDA Cores, 128 RT Cores (3rd Generation), and 512 Tensor Cores (4th Generation)
Radeon RX 7900 XTX is based on RDNA 3 GPU architecture. It uses a Chiplet design where the GPU package consists of a Graphics Compute Die (GCD) and multiple Memory Cache Dies (MCD) surrounding the GCD. The GCD is built on a 5nm fabrication process, and the MCD is built on a 6nm manufacturing process. The card uses Navi 31 GPU chip and has 6144 Stream Processors, 96 Ray Accelerators, and 192 AI Accelerators.
Must Read: CUDA Cores vs Stream Processors Difference
| RTX 5090 | RX 7900 XTX | RTX 4090 | |
| GPU Chip | GB202 | Navi 31 | AD102 |
| GPU Architecture | Blackwell | RDNA 3 | Ada Lovelace |
| Fabrication Process | 5nm | 5nm | 5nm |
| CUDA Cores/Stream Processors | 21760 CUDA Cores | 6144 Stream Processors | 16384 CUDA Cores |
| RT Cores | 170 (4th Generation) | 96 | 128 (3rd Generation) |
| Tensor Cores / | 680 (5th Generation) | 192 | 512 (4th Generation) |
Video RAM [VRAM]
RTX 5090 has the latest, largest, and fastest video memory of all cards. It comes with 32 GB GDDR7 memory having a 512-bit interface. The memory runs at a speed of 28 Gbps and delivers a bandwidth of 1.79 GB/s, which is quite huge. RX 7900 XT comes with 24 GB GDDR6 memory with a 384-bit interface and runs at 20 Gbps, delivering a bandwidth of 960 GB/s. RX 7900 XTX also comes equipped with 96MB Infinity Cache, an ultra-fast cache that can boost peak memory bandwidth of up to 3500 GB/s.
RTX 4090 has 24 GB GDDR6X memory, a 384-bit interface, and a 21 Gbps speed. The memory bandwidth generated is 1.01 GB/s, which is also on the higher side but significantly lower than that of RTX 5090.
| RTX 5090 | RX 7900 XTX | RTX 4090 | |
| Memory Size | 32 GB GDDR7 | 24 GB GDDR6 | 24 GB GDDR6X |
| Memory Interface | 512-bit | 384-bit | 384-bit |
| Memory Speed | 28 Gbps | 20 Gbps | 21 Gbps |
| Memory Bandwidth | 1.79 TB/s | 960 GB/s | 1.01 GB/s |
| Infinity Cache | NA | 96 MB | NA |
Features
All three VR Ready graphics cards support DirectX 12, OpenGL 4.6, Vulkan, AV1 Encode/Decode, Real-Time Ray Tracing, and variable refresh rate technologies G-Sync/FreeSync. RTX 5090 supports PCIe 5.0 while RTX 4090 and RX 7900 XTX support PCIe 4.0 interface. Regarding image upscaling technologies for increasing frame rate in games, RTX 5090 supports DLSS 4, RTX 4090 supports DLSS 3, while RX 7900 XTX supports AMD FidelityFX Super Resolution (FSR 3). When it comes to video output ports, RTX 5090 supports DisplayPort 2.1b and HDMI 2.1b, RX 7900 XTX supports DisplayPort 2.1a and HDMI 2.1a, while RTX 4090 supports HDMI 2.1a but does not support DisplayPort 2.1.
| RTX 5090 | RX 7900 XTX | RTX 4090 | |
| Bus Interface | PCI Express 5.0 | PCI Express 4.0 | PCI Express 4.0 |
| DirectX | 12 Ultimate | 12 Ultimate | 12 Ultimate |
| OpenGL | 4.6 | 4.6 | 4.6 |
| Vulkan | 1.4 | 1.3 | 1.3 |
| SLI / CrossFire | NA | NA | NA |
| VR Ready | Yes | Yes | Yes |
| G-Sync/FreeSync | Yes | Yes | Yes |
| HDMI 2.1 | Yes (HDMI 2.1b) | Yes (HDMI 2.1a) | Yes (HDMI 2.1a) |
| DisplayPort 2.1 | Yes (DisplayPort 2.1b) | Yes (DisplayPort 2.1a) | No |
| AV1 Support (Encode/Decode) | Yes | Yes | Yes |
| Real-Time Ray Tracing | Yes | Yes | Yes |
| Deep Learning Super Sampling (DLSS) | Yes (DLSS 4) | NA | Yes (DLSS 3) |
| FidelityFX Super Resolution (FSR) | NA | FSR 2 | NA |
Gaming Performance
Here are the gaming benchmarks of these cards at 1440p and 4K resolution in various latest AAA games.
1440p Gaming Benchmarks
4K Gaming Benchmarks
From the above gaming benchmarks, we can see that RTX 5090 is, on average, around 16 percent faster than RTX 4090 and 40 percent faster than RX 7900 XTX.
Ray Tracing Performance
In Ray Tracing, RTX 5090 is around 20 percent faster than RTX 4090 and 60 percent faster than RX 7900 XTX.
Power Consumption
RTX 5090 has the highest power consumption of all cards. Its maximum power consumption is 575W, and it needs a 950W or higher PSU. RTX 4090 power consumption is at 450W and needs an 850W power supply. RX 7900 XTX consumes 355W of power at maximum load and requires an 800W or better power supply. RTX 5090 and RTX 4090 use a single 16-pin 12VHPWR connector for power, while the RX 7900 XTX uses two 8-pin PCIe connectors.
| RTX 5090 | RX 7900 XTX | RTX 4090 | |
| Power Connector | 16-pin 12V-2×6 HPWR | 8-pin PCIe (x 2) | 16-pin 12VHPWR |
| Power Consumption | 575W | 355W | 450W |
| Recommended PSU | 950W | 800W | 850W |
Pricing & Availability
RTX 5090 is the most expensive GeForce card to date. The official MSRP for the Founders edition is 1999 USD. The MSRP of RTX 4090 is 1599 USD, while the MSRP of RX 7900 XTX is 999 USD. The custom models of these cards cost more, and you can check the latest prices from the links below.
Check RTX 5090 Price on Amazon
Check RX 7900 XTX Price on Amazon
Check RTX 4090 Price on Amazon
RTX 5090 vs RX 7900 XTX vs RTX 4090 Specifications
| RTX 5090 | RX 7900 XTX | RTX 4090 | |
| GPU Chip | GB202 | Navi 31 | AD102 |
| GPU Architecture | Blackwell | RDNA 3 | Ada Lovelace |
| Fabrication Process | 5nm | 5nm | 5nm |
| CUDA Cores/Stream Processors | 21760 CUDA Cores | 6144 Stream Processors | 16384 CUDA Cores |
| Ray Tracing Cores | 170 (4th Generation) | 96 | 128 (3rd Generation) |
| Tensor Cores | 680 (5th Generation) | 192 | 512 (4th Generation) |
| Memory Size | 32 GB GDDR7 | 24 GB GDDR6 | 24 GB GDDR6X |
| Memory Interface | 512-bit | 384-bit | 384-bit |
| Memory Speed | 28 Gbps | 20 Gbps | 21 Gbps |
| Memory Bandwidth | 1.79 TB/s | 960 GB/s | 1.01 GB/s |
| Bus Interface | PCI Express 5.0 | PCI Express 4.0 | PCI Express 4.0 |
| DirectX | 12 Ultimate | 12 Ultimate | 12 Ultimate |
| OpenGL | 4.6 | 4.6 | 4.6 |
| Vulkan | 1.4 | 1.3 | 1.3 |
| SLI / CrossFire | NA | NA | NA |
| VR Ready | Yes | Yes | Yes |
| G-Sync/FreeSync | Yes | Yes | Yes |
| Power Consumption | 575W | 355W | 450W |
| Recommended PSU | 950W | 800W | 850W |
Final Words
Well, there is no real competition to the RX 5900 when it comes to performance. It overpowers its competitors by a significant margin. Also, RX 5900 is a card that not everyone can afford or even need for their requirements. It is a card meant for professionals and hardcore enthusiasts where money is not a serious issue. If you want to share your thoughts on this comparison, please let me know in the comment section below.
(*This post may contain affiliate links, which means I may receive a small commission if you choose to purchase through the links I provide (at no extra cost to you). Thank you for supporting the work I put into this site!)
![Best GTX 1630 Cards for Budget and SFF PC [Low-power models]](https://graphicscardhub.com/wp-content/uploads/2022/07/Gigabyte-GeForce-GTX-1630-OC-Low-Profile-4G-211x150.jpg "Best GTX 1630 Cards for Budget and SFF PC [Low-power models]")
