NVIDIA launched its new flagship graphics card, the GeForce RTX 5090, on January 6, 2025. Based on the GB202-300-A1 GPU and manufactured using TSMC's 5nm process technology, this monster GPU pushes the boundaries of what's possible in consumer graphics hardware.
The RTX 5090 packs an impressive 21,760 CUDA cores, 176 raster units, and 680 texture units across 170 streaming multiprocessors. For AI and ray tracing workloads, it comes equipped with 680 Tensor cores and 172 dedicated RT cores. The massive 96MB L2 cache further enhances the card's computational capabilities.
Clock speeds have seen a substantial improvement over the previous generation, with a base clock of 2017MHz and a boost clock reaching 2407MHz. The card ships with a whopping 32GB of next-generation GDDR7 memory on a 512-bit bus, delivering memory speeds of 28Gbps and a staggering bandwidth of 1.8TB/s.
All this hardware muscle translates to 104.8 TFLOPS of FP32 performance, making it a computational powerhouse for gaming and professional workloads alike. Unsurprisingly, such performance comes at a cost—both financially and in terms of power draw. The RTX 5090 has a TDP of 575W and carries a suggested retail price of $1,999.
This flagship GPU clearly targets enthusiasts and professionals seeking uncompromising performance, continuing NVIDIA's tradition of pushing the high-end envelope with each new generation.
AMD unveiled its flagship graphics card, the Radeon RX 7900 XTX, on November 3, 2022. Built on TSMC's 5nm manufacturing process, the card is powered by AMD's Navi 31 GPU, representing the pinnacle of the company's RDNA 3 architecture.
The RX 7900 XTX comes equipped with 6,144 stream processors distributed across 96 compute units. The card features 192 raster units and 384 texture units for traditional rendering tasks, alongside 96 dedicated ray tracing accelerators to handle real-time ray tracing workloads.
AMD implemented a multi-level cache hierarchy with the 7900 XTX, combining 6MB of L2 cache with a substantial 96MB of Infinity Cache (L3), designed to reduce memory latency and improve bandwidth efficiency in gaming scenarios.
As AMD's answer to NVIDIA's high-end offerings, the 7900 XTX represents the company's continued push to compete in the premium GPU market segment, delivering impressive performance for demanding games and creative applications.
The NVIDIA RTX 5090 boasts impressive specifications that position it firmly in the high-end graphics card market. Running at a base clock speed of 1855 MHz with boost capabilities up to 2500 MHz, this card delivers substantial processing power for demanding applications.
The RTX 5090 comes equipped with 24GB of GDDR6 memory on a 384-bit memory bus. This configuration, coupled with memory speeds of 20 Gbps, produces a remarkable memory bandwidth of 960GB/s—ensuring smooth performance even in data-intensive scenarios.
Perhaps most notable is the card's exceptional computational capability, with FP32 floating-point performance reaching 61.42 TFLOPS. This level of processing power comes with a power draw of 355W, relatively efficient considering its performance class.
NVIDIA has positioned the RTX 5090 at a recommended retail price of $999, making it a significant investment but competitive within its performance tier when compared to AMD's offerings in the same class.
Test Platform Configuration
The testing was conducted on a high-performance system with the following specifications:
Processor: AMD Ryzen 7 9800X3D
Motherboard: MSI MPG X670E CARBON
Memory: 32GB DDR5 (6000MHz)
Graphics Cards: AMD Radeon RX 7900 XTX and NVIDIA GeForce RTX 5090 (tested separately)
Storage: Dual Samsung 970 EVO 1TB SSDs
Operating System: Windows 11
This configuration ensures that the system performance is balanced across components, minimizing potential bottlenecks and allowing for an accurate comparison between the two graphics cards.
Gaming Performance Comparison
The following section presents benchmark results for seven different games, all tested at 4K resolution settings. In each of the comparative images below, the left side represents performance metrics from the AMD Radeon RX 7900 XTX platform, while the right side displays results from the NVIDIA GeForce RTX 5090 platform.
These head-to-head comparisons will provide valuable insights into how these two flagship graphics cards perform under identical gaming conditions, allowing readers to assess their relative merits in practical gaming scenarios. The 4K resolution testing specifically targets high-end gaming performance, which is the primary use case for these premium graphics solutions.
Game 1: Ghostwire: Tokyo
Performance Metrics
AMD Radeon RX 7900 XTX
- Average FPS: 66
- Minimum FPS: 55
- Frame Time: 15.1ms
- CPU Usage: 12%
- CPU Power Consumption: 45W
- CPU Temperature: 44°C
- System Memory Usage: 11.5GB
- GPU Usage: 99%
- VRAM Usage: 10.7GB
- GPU Power Consumption: 346W
- GPU Temperature: 63°C
NVIDIA GeForce RTX 5090
- Average FPS: 117
- Minimum FPS: 97
- Frame Time: 8.0ms
- CPU Usage: 20%
- CPU Power Consumption: 66W
- CPU Temperature: 61°C
- System Memory Usage: 12.1GB
- GPU Usage: 99%
- VRAM Usage: 9.3GB
- GPU Power Consumption: 543W
- GPU Temperature: 56°C
Comparative Analysis
The RTX 5090 delivers substantially higher performance in Ghostwire: Tokyo, achieving an average of 117 FPS compared to the RX 7900 XTX's 66 FPS—representing a remarkable 77% performance advantage. The NVIDIA card also maintains better frame consistency, with minimum frame rates staying above 97 FPS, while the AMD card drops to 55 FPS during demanding sequences.
Frame time measurements further emphasize this difference, with the RTX 5090 processing frames in just 8.0ms versus the RX 7900 XTX's 15.1ms, resulting in noticeably smoother gameplay.
This performance advantage comes with increased power demands, as the RTX 5090 draws 543W compared to the RX 7900 XTX's 346W—a 57% increase in power consumption. Interestingly, despite the higher power draw, the NVIDIA card maintains a lower GPU temperature at 56°C compared to the AMD card's 63°C, suggesting more efficient cooling.
Both cards utilize nearly 100% of their GPU resources, but the RTX 5090 demands slightly more from the system, with higher CPU utilization (20% vs. 12%) and marginally higher system memory usage (12.1GB vs. 11.5GB).
The Second Game: Cyberpunk 2077
On the Radeon RX 7900 XTX platform, we measured an average framerate of 73 FPS with minimum framerates dropping to 53 FPS. Frame time averaged 13.6ms. The CPU utilization hovered around 44% with a power draw of 86W and temperatures of 56°C. System memory usage reached 14.9GB. The GPU itself maintained 99% utilization, consumed 10.4GB of VRAM, drew 343W of power, and operated at 62°C.
The GeForce RTX 5090 delivered significantly better performance with an average of 120 FPS and minimum framerates of 93 FPS. Frame times were reduced to 8.7ms. CPU utilization increased to 58% with power consumption of 102W and temperatures hitting 70°C. System memory usage was 12.3GB. The GPU maintained 98% utilization, used 9.3GB of VRAM, drew a substantial 508W of power, but ran cooler at 53°C.
For this title, the RTX 5090 system showed 14% higher CPU utilization than its competitor but used 2.6GB less system memory. GPU utilization was nearly identical between the two cards, though the RTX 5090 used 1.1GB less VRAM.
In performance metrics, the GeForce RTX 5090 outperformed the Radeon RX 7900 XTX by 47 FPS on average (approximately 1.6x faster), with minimum framerates 40 FPS higher. Frame times were 5ms lower, though power consumption increased by 165W. Interestingly, despite higher power draw, the RTX 5090 ran 9°C cooler than its AMD counterpart.
The Third Game: Silent Hill 2
When testing Silent Hill 2, the Radeon RX 7900 XTX achieved an average framerate of 50 FPS, with minimum framerates dipping to 37 FPS. Frame times averaged 19.9ms. CPU utilization was relatively light at just 10%, drawing 46W and running at 43°C. System memory consumption reached 11.9GB. The GPU was fully maxed out at 99% utilization, using 10.7GB of VRAM, pulling 343W of power, and running at 60°C.
The GeForce RTX 5090 demonstrated much stronger performance, delivering an average of 104 FPS and minimum framerates of 81 FPS. Frame times were significantly reduced to 9.2ms. CPU utilization increased to 18% with power consumption of 59W and temperatures of 49°C. System memory usage was nearly identical at 11.8GB. The GPU ran at 97% utilization, consumed 9.1GB of VRAM, drew 421W of power, and operated at a cooler 49°C.
In this title, the RTX 5090 system showed 8% higher CPU utilization than its competitor, with essentially equal system memory usage. GPU utilization was actually 2% lower on the NVIDIA card, and it used 1.6GB less VRAM.
Performance-wise, the difference was dramatic - the GeForce RTX 5090 outperformed the Radeon RX 7900 XTX by 54 FPS on average (approximately 2.1x faster), with minimum framerates 44 FPS higher. Frame times were cut in half, being 10ms lower on the RTX 5090. Power consumption increased by 78W compared to the AMD card, but despite this, the NVIDIA card ran 11°C cooler.
The Fourth Game: Stalker 2
For Stalker 2, the Radeon RX 7900 XTX delivered an average framerate of 59 FPS with minimum framerates of 46 FPS. Frame times averaged 14.7ms. CPU utilization reached 39% with power consumption of 73W and temperatures of 53°C. System memory usage was particularly high at 20.6GB. The GPU maintained 98% utilization, consumed 9.9GB of VRAM, drew 346W of power, and ran at 63°C.
The GeForce RTX 5090 once again showed its performance advantage with an average of 96 FPS and minimum framerates of 60 FPS. Frame times were reduced to 8.1ms. CPU utilization increased to 53% with power consumption of 88W and temperatures of 64°C. System memory usage was lower at 17.3GB. The GPU ran at 95% utilization, used almost the same amount of VRAM at 9.8GB, drew 360W of power, and operated at a much cooler 46°C.
In this demanding title, the RTX 5090 system pushed the CPU 14% harder than the AMD setup, while using 3.3GB less system memory. GPU utilization was 3% lower on the NVIDIA card, while VRAM usage was virtually identical between the two cards.
Looking at performance metrics, the GeForce RTX 5090 outpaced the Radeon RX 7900 XTX by 37 FPS on average (approximately 1.6x faster), with minimum framerates 14 FPS higher. Frame times were 6ms lower, while power consumption was only slightly higher at 14W more than the AMD card. Most impressively, despite the higher performance, the RTX 5090 ran a substantial 17°C cooler during this test.
The Fifth Game: Star Wars Outlaws
Star Wars Outlaws presented a significant challenge for the Radeon RX 7900 XTX, which managed only 42 FPS on average, with minimum framerates of 33 FPS. Frame times were quite high at 24.3ms. CPU utilization was relatively low at 19%, drawing 57W and operating at 47°C. System memory usage settled at 15.6GB. The GPU was pushed to its absolute limit with 100% utilization, consuming a hefty 14.5GB of VRAM, drawing 347W of power, and running at 61°C.
The GeForce RTX 5090 demonstrated exceptional performance in this title, delivering a smooth 102 FPS on average with minimum framerates of 87 FPS. Frame times were dramatically reduced to 9.8ms. CPU utilization increased significantly to 44% with power consumption of 80W and temperatures of 63°C. System memory usage was nearly identical at 15.4GB. The GPU ran at 97% utilization, used 13.7GB of VRAM, drew a substantial 541W of power, yet maintained a temperature of 57°C.
In this graphically demanding game, the RTX 5090 system pushed the CPU utilization 25% higher than the AMD setup, while system memory usage was practically identical (just 0.2GB less). GPU utilization was 3% lower on the NVIDIA card, and it used 0.8GB less VRAM.
The performance gap was at its most dramatic in this title - the GeForce RTX 5090 outperformed the Radeon RX 7900 XTX by a whopping 60 FPS on average (approximately 2.4x faster), with minimum framerates 54 FPS higher. Frame times were 15ms lower on the RTX 5090. Power consumption increased by 194W over the AMD card, but surprisingly, the NVIDIA GPU still ran 4°C cooler.
The Sixth Game: Black Myth: Wukong
The Chinese action RPG Black Myth: Wukong proved challenging for both cards. The Radeon RX 7900 XTX struggled somewhat, delivering an average framerate of just 36 FPS with minimum framerates of 30 FPS. Frame times were quite high at 28.1ms. CPU utilization was minimal at 7%, drawing 44W and running at 44°C. System memory usage was 11.4GB. The GPU was pushed near its limits with 99% utilization, consuming 11.2GB of VRAM, drawing 349W of power, and reaching 64°C.
The GeForce RTX 5090 handled the game significantly better, achieving 69 FPS on average with minimum framerates of 58 FPS. Frame times were cut in half to 14.3ms. CPU utilization remained low but doubled to 15%, with power consumption of 52W and temperatures of 53°C. System memory usage was slightly higher at 11.6GB. The GPU maintained 98% utilization, used 9.1GB of VRAM, drew 484W of power, and operated at 55°C.
In this visually stunning title, the RTX 5090 system exhibited 8% higher CPU utilization and marginally higher system memory usage (0.2GB more). GPU utilization was nearly identical between the two cards, though the RTX 5090 used 2.1GB less VRAM.
Performance-wise, the GeForce RTX 5090 outperformed the Radeon RX 7900 XTX by 33 FPS on average (approximately 1.9x faster), with minimum framerates 28 FPS higher. Frame times were 14ms lower on the RTX 5090. Power consumption increased by 135W compared to the AMD card, but the NVIDIA GPU still managed to run 9°C cooler.
The Seventh Game: Starfield
Bethesda's space epic Starfield showed interesting performance characteristics across both platforms. The Radeon RX 7900 XTX achieved an average framerate of 64 FPS with minimum framerates of 51 FPS. Frame times averaged 15.8ms. CPU utilization reached 35% with power consumption of 69W and temperatures of 49°C. System memory usage was 13.5GB. The GPU maintained 98% utilization, consumed 8.6GB of VRAM, drew 346W of power, and ran at 62°C.
The GeForce RTX 5090 delivered substantially better performance with an average of 100 FPS and minimum framerates of 91 FPS. Frame times were reduced to 10.1ms. CPU utilization increased to 50% with power consumption of 89W and temperatures of 62°C. System memory usage was slightly lower at 13.2GB. The GPU ran at 97% utilization, used 7.7GB of VRAM, drew 432W of power, and operated at a much cooler 46°C.
For Starfield, the RTX 5090 system pushed the CPU 15% harder than the AMD configuration, while using marginally less system memory (0.3GB less). GPU utilization was nearly identical between the two cards, with the RTX 5090 using 0.9GB less VRAM.
Looking at performance metrics, the GeForce RTX 5090 outperformed the Radeon RX 7900 XTX by 36 FPS on average (approximately 1.6x faster), with minimum framerates 40 FPS higher. Frame times were 5ms lower on the RTX 5090. Power consumption increased by 86W over the AMD card, yet the NVIDIA card still ran a remarkable 16°C cooler.
Comparison Conclusion
The results of our testing couldn't be clearer: the GeForce RTX 5090 platform holds an enormous performance advantage over its competitor, demonstrating performance that's several tiers above the Radeon RX 7900 XTX.
When we aggregate the data across all our test titles, the GeForce RTX 5090 performs at approximately 1.9 (or more precisely, 1.85) times the performance level of the Radeon RX 7900 XTX—nearly doubling its framerate in most scenarios. This dramatic performance gap represents our final verdict on these two high-end graphics solutions.
These findings should provide clear guidance for enthusiasts weighing their options between these flagship cards, though individual priorities regarding power consumption, temperatures, and value will naturally factor into purchasing decisions.
0 Comments