GeForce RTX 5090
NVIDIA
G3D Mark
38,867
MSRP
$1999
AVG Price
$2700
Performance Spectrum - GPU
About G3D Mark
G3D Mark is a standard benchmark that measures graphics performance in real-world gaming scenarios. It simplifies comparing cards from different brands, where higher scores directly correlate with better fps and smoother gaming experiences.
Overview
The GeForce RTX 5090 is manufactured by NVIDIA. It was released in January 30 2025. It features the Blackwell 2.0 architecture. The core clock ranges from 2017 MHz to 2407 MHz. It has 21760 shading units. The thermal design power (TDP) is 575W. Manufactured using 5 nm process technology. It features 170 dedicated ray tracing cores for enhanced lighting effects. G3D Mark benchmark score: 38,867 points. Launch price was $1,999.
Value Upgrade Path
This is the official ChipVERSUS Value Rating, comparing raw performance (G3D Mark) per dollar. Components placed above yours deliver better value for money.
Avg price is the current average price collected from markets across the web.
Performance Per Dollar GeForce RTX 5090
FPS Benchmarks
Predicted performance for GeForce RTX 5090 across popular games. Tested paired with Ryzen 7 7800X3D to isolate GPU performance.
Counter-Strike 2
Cyberpunk 2077
Delta Force
Dota 2
Honkai: Star Rail
League of Legends
Minecraft
Monster Hunter Wilds
Once Human
Overwatch 2
Technical Analysis
Detailed breakdown of GeForce RTX 5090 specifications and capabilities.
Graphics Performance
The GeForce RTX 5090 scores 38,867 in the G3D Mark benchmark (the standard measure of rasterization GPU performance), placing it in the Enthusiast tier as a Current Gen generation graphics card. It is built on the Blackwell 2.0 (2025−2026) architecture (codename: GB202), manufactured on a 5 nm process— smaller process nodes improve power efficiency and transistor density. It packs 21760 shader units (the primary compute units for rendering pixels), 680 TMUs (texture mapping units), and 176 ROPs (render output units that handle final pixel output). Raw FP32 compute power: 104.8 TFLOPS — this measures the theoretical peak floating-point performance. Boost clock reaches 2407 MHz from a base of 2017 MHz.
| Specification | GeForce RTX 5090 |
|---|---|
| G3D MarkPassMark 3D graphics benchmark score | 38,867 |
| ArchitectureGPU microarchitecture generation | Blackwell 2.0 (2025−2026) |
| CodenameInternal GPU die codename | GB202 |
| Manufacturing ProcessFabrication node — smaller = more efficient | 5 nm |
| Shader UnitsCore compute units for rendering/compute | 21760 |
| TMUsTexture Mapping Units — handle texture sampling | 680 |
| ROPsRender Output Units — final pixel/color blending | 176 |
| Boost ClockMaximum GPU frequency under load | 2407 MHz |
| Base ClockGuaranteed minimum GPU frequency | 2017 MHz |
| Compute (FP32)Peak single-precision floating-point performance | 104.8 TFLOPS |
| Texture Fill RateTextures processed per second | 1,636.8 GTexel/s |
| PCIe InterfaceCPU↔GPU data link generation | PCIe 5.0 x16 |
| Transistor Count | 92,200 million |
| Performance RankPosition in global GPU performance ranking | #3 |
Video Memory (VRAM)
The GeForce RTX 5090 is equipped with 32 GB of GDDR7 video memory connected via a 512-bit memory bus, delivering 1792 GB/s of bandwidth.
| Specification | GeForce RTX 5090 |
|---|---|
| VRAM SizeVideo memory capacity — more = better for high-res textures | 32 GB |
| VRAM TypeMemory technology generation | GDDR7 |
| Memory BusWidth of data path — wider = more bandwidth | 512-bit |
| BandwidthMaximum data throughput between GPU and VRAM | 1792 GB/s |
| L1 Cache | 21.3 MB |
| L2 Cache | 96 MB |
Advanced Features & APIs
The GeForce RTX 5090 supports modern upscaling and frame generation technologies: NVIDIA Reflex, DLSS Frame Generation, DLSS Super Resolution — these can dramatically boost effective FPS by generating intermediate frames or upscaling from lower internal resolution. Graphics API support: DirectX 12.2, Vulkan 1.4, OpenGL 4.6— latest API versions enable the most advanced rendering features like ray tracing and mesh shaders.
| Specification | GeForce RTX 5090 |
|---|---|
| NVIDIA Reflex | ✅ Supported |
| DLSS Frame Generation | ✅ Supported |
| DLSS Super Resolution | ✅ Supported |
| DirectXMicrosoft graphics API version | 12.2 |
| VulkanCross-platform graphics API | 1.4 |
| OpenGLLegacy graphics API | 4.6 |
| Compute APIGPGPU framework (CUDA/ROCm/OpenCL) | CUDA 13.0 |
| Multi-GPUSLI/CrossFire/NVLink support | None |
Display & Media
Display outputs: 3x DP, 1x HDMI — supports up to 4 simultaneous displays. Hardware encoder: 9th Gen NVENC (2x), decoder: 6th Gen NVDEC— dedicated hardware for video encoding/decoding accelerates streaming, recording, and video playback. Supported codecs: MPEG-2,H.264,HEVC,VP9,AV1.
| Specification | GeForce RTX 5090 |
|---|---|
| DP PortsNumber of DP display connectors | 3x |
| HDMI PortsNumber of HDMI display connectors | 1x |
| Max DisplaysMaximum simultaneously connected monitors | 4 |
| Hardware EncoderVideo encoding engine version | 9th Gen NVENC (2x) |
| Hardware DecoderVideo decoding engine version | 6th Gen NVDEC |
| CodecsSupported video compression formats | MPEG-2,H.264,HEVC,VP9,AV1 |
Power & Dimensions
The GeForce RTX 5090 draws 575 Watt under load (TDP) — a 1000W PSU minimum is recommended. Power connector: 16-pin (12VHPWR). Physical size: 304mm long, 137mm tall, occupying 2 expansion slots— verify your case has clearance before purchasing. Typical temperatures: 35°C at idle, 85°C under load.
| Specification | GeForce RTX 5090 |
|---|---|
| TDPPower consumption under load | 575 Watt |
| Recommended PSUMinimum power supply wattage | 1000W |
| Power ConnectorRequired PCIe power cable type | 16-pin (12VHPWR) |
| Card LengthGPU card length — check case clearance | 304mm |
| Card Height | 137mm |
| Slot WidthExpansion slots occupied | 2-slot |
| Idle TemperatureTypical temperature at desktop idle | 35°C |
| Load TemperatureTypical temperature under gaming load | 85°C |
Value Analysis
At an average price of $2,700 (MSRP: $1999), the GeForce RTX 5090 delivers 14 G3D Mark points per dollar. Overall performance ranking position: #3 among all indexed GPUs.
| Specification | GeForce RTX 5090 |
|---|---|
| MSRPManufacturer's Suggested Retail Price | $1999 |
| Avg Price (30d)Average market price over the last 30 days | $2700 |
| Performance RankPosition in global GPU performance ranking | #3 |
| Release Date | January 30 2025 |
| Release Year | 2025 |
Specifications Reference
Raw spec sheet for quick lookup. The analysis cards above explain what these fields mean in practice.
2 categories
Technical Specifications
22 specs
Technical Specifications
22 specs
| ROPs | 176 |
| TMUs | 680 |
| Vram | 32 GB |
| L1 Cache | 21.3 MB |
| L2 Cache | 96 MB |
| Core Clock | 2017 MHz |
| Boost Clock | 2407 MHz |
| Memory Type | GDDR7 |
| Tensor Cores | 680 |
| Shading Units | 21760 |
| Upscaling Tech | DLSS 4 Super Resolution |
| PCIe Generation | PCIe 5.0 x16 |
| Memory Bandwidth | 1792 GB/s |
| Memory Bus Width | 512-bit |
| Transistor Count | 92,200 million |
| Frame Generation | NVIDIA Reflex: Yes | DLSS Frame Generation: Yes | DLSS Super Resolution: Yes |
| Ray Tracing Cores | 170 |
| Texture Fill Rate | 1,636.8 |
| Manufacturing Process | 5 nm |
| Frame Generation Tech | DLSS 4 Multi Frame Generation |
| Power Consumption (TDP) | 575 Watt |
| Floating Point Performance | 104.8 TFLOPS |
General Information
5 specs
General Information
5 specs
| Codename | GB202 |
| Architecture | Blackwell 2.0 (2025−2026) |
| Launch Price | $1,999 |
| Release Date | January 30 2025 |
| Performance Ranking Position | 3 |
Similar Performance
Graphics cards with benchmark scores closest to the current hardware, grouped by manufacturer.
NVIDIA (12)
Our Recommendation for GeForce RTX 5090
Suggested pairings based on performance balance
Minimum (Avoid Bottleneck) (12)
Recommended (Balanced) (12)
Popular Comparisons
Compare GeForce RTX 5090 with other popular GPUs
Compare GeForce RTX 5090
See how this GPU stacks up against similar alternatives
Better Alternatives (3)
Similar or Lower Performance (8)
