Feex·GPU
NVIDIA · GeForce RTX 50 Series

NVIDIA GeForce RTX 5090

The NVIDIA GeForce RTX 5090 is a graphics card released on January 30, 2025, built on the Blackwell architecture using a 4 nm process. It has 21,760 shading units, 32 GB of GDDR7 memory on a 512-bit bus, 1,792.0 GB/s of memory bandwidth, a boost clock of 2.41 GHz and a 575 W board power rating. It scores 38,961 in the PassMark G3D benchmark. It supports DirectX 12 Ultimate, OpenGL 4.6, Vulkan 1.4 and CUDA compute capability 12.0.

BlackwellGB202-3004 nm
10/10 key specs recorded · 117 facts

Depth: 62 recorded fields, 24 features, 5 PassMark workloads

Specification data last verified October 1, 2026 against PassMark Videocard Benchmark, NVIDIA GeForce Specifications, NanoReview (Firecrawl extract), NVIDIA Developer: CUDA GPUs, NVIDIA Developer: Video Encode and Decode GPU Support Matrix and Wikipedia GPU list tables (wikigrid-2026-09). Coverage: 10 of 10 tracked specifications recorded.

Shading units
21,760cores
170 SMs
Memory
32 GB
GDDR7 · 512-bit
Boost clock
2.41 GHz
2.02 GHz base
Board power
575 W
1,000 W PSU advised

Performance

The NVIDIA GeForce RTX 5090 scores 38,961 in the PassMark G3D benchmark, higher than 99.9% of the 2,658 GPUs benchmarked (PassMark rank #2), measured across 10,802 submitted results. Its PassMark G2D score is 1,410.

PassMark G3D Mark
38,961
Rank #2
Faster than 99.9% of 2,658 GPUs
10,802 samples
0relative to the fastest GPU on record (41,928)

PassMark workload tests

DirectX 9 405 fps reference: fastest recorded 526 fps · faster than 99.6% of 1,675 GPUs
DirectX 10 228 fps reference: fastest recorded 231 fps · faster than 99.8% of 1,602 GPUs
DirectX 11 340 fps reference: fastest recorded 362 fps · faster than 99.7% of 1,430 GPUs
DirectX 12 179 fps reference: fastest recorded 213 fps · faster than 99.8% of 1,274 GPUs
GPU Compute 24,392 ops/s reference: fastest recorded 26,591 ops/s · faster than 99.7% of 1,447 GPUs

Averages of PassMark PerformanceTest submissions; frames/sec are PassMark test-scene rates, not game FPS. Tests PassMark does not report for this card are omitted. PassMark listing

In the PassMark workload tests the NVIDIA GeForce RTX 5090 averages DirectX 9 405 fps, DirectX 10 228 fps, DirectX 11 340 fps, DirectX 12 179 fps and GPU Compute 24,392 ops/s, across 10,802 submitted results. Frames per second are PassMark test-scene rates, not game frame rates.

Theoretical throughput

AI performance 3,352 TOPS stated by NVIDIA
Ray tracing 318 TFLOPS stated by NVIDIA

NVIDIA states a peak of AI performance 3,352 TOPS and Ray tracing 318 TFLOPS for the NVIDIA GeForce RTX 5090.

As stated by PassMark

Category Desktop
Bus interface PCIe 5.0 x16
Max memory 32 GB
Core clock 2017 MHz
Memory clock 1750 MHz
DirectX 12_2
OpenGL 4.6
Max TDP 575 W
First benchmarked 2025-01-31
Samples 10,802

PassMark listing (collected 2026-09-30)

PassMark's own listing for the NVIDIA GeForce RTX 5090 states the PCIe 5.0 x16 interface, up to 32 GB of memory, a core clock of 2017 MHz, DirectX 12_2 and OpenGL 4.6, first benchmarked on 2025-01-31. These are PassMark's figures, shown separately from the specification tables above.

Not to be confused with

This page covers the NVIDIA GeForce RTX 5090 specifically. It is a different product from NVIDIA GeForce RTX 5090 D, NVIDIA GeForce RTX 5090 D v2, NVIDIA GeForce RTX 5090 Laptop GPU, which have their own specifications. Check the exact model name before comparing figures.

Graphics processor

Architecture Blackwell
Codename GB202-300
Chip Not recorded
Process node 4 nm
Transistors 92,200 M
Die size 750 mm²

The NVIDIA GeForce RTX 5090 is based on the Blackwell architecture, uses the GB202-300 graphics processor, is manufactured on a 4 nm process, contains 92,200 million transistors and has a die size of 750 mm².

Shader configuration

Shading units 21,760
Compute units Not recorded
Streaming multiprocessors 170
Execution units Not recorded
Texture units (TMUs) 680
Render output units (ROPs) 176
Ray tracing cores 170
Tensor / AI cores 680
XMX engines Not recorded
Infinity cache Not recorded

The NVIDIA GeForce RTX 5090 has 21,760 shading units, 170 streaming multiprocessors, 680 texture mapping units, 176 render output units, 170 ray tracing cores and 680 tensor cores.

The NVIDIA GeForce RTX 5090 does not use XMX engines; those are specific to Intel Arc graphics. Infinity Cache is an AMD architectural feature and does not apply to the NVIDIA GeForce RTX 5090.

Clock speeds

Base clock 2.02 GHz
Boost clock 2.41 GHz
Game clock Not recorded
Memory clock 1.75 GHz
Effective memory clock Not recorded

The NVIDIA GeForce RTX 5090 runs at a base clock of 2.02 GHz and boosts to 2.41 GHz.

Memory

Size 32 GB
Type GDDR7
Bus width 512-bit
Bandwidth 1,792.0 GB/s

The NVIDIA GeForce RTX 5090 is equipped with 32 GB of GDDR7 memory, uses a 512-bit memory bus and delivers 1,792.0 GB/s of memory bandwidth.

Board and power

The NVIDIA GeForce RTX 5090 has a board power rating of 575 W, a total board power of 575 W, a power supply of at least 1,000 W is recommended, power is supplied through 4x PCIe 8-pin cables (adapter in box) OR 1x 600 W PCIe Gen 5 cable, the card occupies 2-slot, it is 304 mm long, its width is 137 mm, the maximum operating temperature is 90 °C and it connects over PCIe 5.0.

Board power (TDP) 575 W
Total board power 575 W
Suggested PSU 1,000 W
Power connectors 4x PCIe 8-pin cables (adapter in box) OR 1x 600 W PCIe Gen 5 cable
Slot width 2-slot
Length 304 mm
Width 137 mm
Height Not recorded
Cooler Not recorded
Max temperature 90 °C
Bus interface PCIe 5.0

Market

Launched January 30, 2025
Launch price $1,999
Status Not recorded
Form factor Not recorded
Segment Desktop
Series GeForce RTX 50 Series

The NVIDIA GeForce RTX 5090 was released on January 30, 2025 and launched at a price of $1,999.

Benchmarks

PassMark G3D 38,961
PassMark G2D 1,410
Rank #2
Samples 10,802

Graphics and compute APIs

The NVIDIA GeForce RTX 5090 supports DirectX 12 Ultimate, OpenGL 4.6, Vulkan 1.4 and CUDA compute capability 12.0.

DirectX 12 Ultimate
OpenGL 4.6
Vulkan 1.4
CUDA compute capability 12.0

Feex has no recorded OpenCL version for the NVIDIA GeForce RTX 5090.

Features

The NVIDIA GeForce RTX 5090 has 15 supported features recorded outside the media engine. A feature that no source mentions is shown as not listed, never as unsupported.

Rendering

  • Supported: DirectX 12 Ultimate
    Microsoft's top DirectX 12 feature tier, covering hardware ray tracing, mesh shaders, variable-rate shading and sampler feedback.
  • Supported: NVIDIA DLSS DLSS 4.5
    NVIDIA's AI-based upscaling and frame-generation technology that can raise frame rates in supported games.
    Modes: Super Resolution, DLAA, Ray Reconstruction, Frame Generation, Multi Frame Generation, Dynamic Multi Frame Generation
  • Supported: Hardware ray tracing 4th Generation
    Dedicated GPU hardware that speeds up ray-traced lighting, reflections and shadows in supported games and apps.
    Ray tracing units: 170
Not listed (1)

No source Feex holds states whether the NVIDIA GeForce RTX 5090 supports Vulkan ray tracing. That is silence, not a statement that they are unsupported.

The NVIDIA GeForce RTX 5090 supports DirectX 12 Ultimate, NVIDIA DLSS (DLSS 4.5) and Hardware ray tracing (4th Generation) under Rendering.

Compute and AI

  • Supported: CUDA Compute capability 12.0
    Support for NVIDIA's CUDA computing platform, used by many creative, scientific and AI applications.
  • Supported: Tensor cores 5th Generation
    NVIDIA's dedicated matrix-math units used for AI workloads such as DLSS and machine-learning inference.
    Tensor cores: 680
Not listed (1)

No source Feex holds states whether the NVIDIA GeForce RTX 5090 supports NVLink / SLI. That is silence, not a statement that they are unsupported.

The NVIDIA GeForce RTX 5090 supports CUDA (Compute capability 12.0) and Tensor cores (5th Generation) under Compute and AI.

Display features

  • Supported: NVIDIA G-SYNC
    Support for NVIDIA's variable refresh rate technology, which syncs the display to the GPU to reduce tearing and stutter.
  • Supported: HDCP content protection HDCP 2.3
    Support for HDCP, the copy-protection standard that some protected 4K streaming and disc video requires.
  • Supported: VR ready
    The vendor states the GPU meets the requirements for running mainstream PC virtual reality headsets.

The NVIDIA GeForce RTX 5090 supports NVIDIA G-SYNC, HDCP content protection (HDCP 2.3) and VR ready under Display features.

Platform

  • Supported: Small-form-factor ready
    The vendor lists the GPU as suitable for compact small-form-factor PC builds.
Not listed (1)

No source Feex holds states whether the NVIDIA GeForce RTX 5090 supports Resizable BAR. That is silence, not a statement that they are unsupported.

The NVIDIA GeForce RTX 5090 supports Small-form-factor ready under Platform.

Software and drivers

  • Supported: NVIDIA Game Ready drivers
    Driver branch tuned and released alongside new games.
  • Supported: NVIDIA Broadcast
    NVIDIA software that applies AI noise removal and camera effects for calls and streams.
  • Supported: NVIDIA Reflex Reflex 2
    NVIDIA's latency-reduction technology that lowers the delay between input and what you see in supported games.
  • Supported: RTX HDR
    NVIDIA feature that adds HDR to games and video that were made without it.
  • Supported: RTX Remix
    NVIDIA tooling for modding older games with ray tracing and updated assets.
  • Supported: NVIDIA Studio drivers
    Driver branch tuned and tested for creative applications.
Not listed (3)

No source Feex holds states whether the NVIDIA GeForce RTX 5090 supports NVIDIA Freestyle, NVIDIA Omniverse and NVIDIA ShadowPlay. That is silence, not a statement that they are unsupported.

The NVIDIA GeForce RTX 5090 supports NVIDIA Game Ready drivers, NVIDIA Broadcast, NVIDIA Reflex (Reflex 2), RTX HDR, RTX Remix and NVIDIA Studio drivers under Software and drivers.

Media engine

The NVIDIA GeForce RTX 5090 provides hardware encode for H.264 (AVC), HEVC (H.265) and AV1 and hardware decode for H.264 (AVC), HEVC (H.265), AV1 and VP9.

NVIDIA NVENC encoder 9th Generation Encoder engines: 3NVIDIA's dedicated video-encoding engine, used by streaming and capture software; newer generations encode more efficiently.
NVIDIA NVDEC decoder 6th Generation Decoder engines: 2NVIDIA's dedicated video-decoding engine; newer generations support more codecs and formats.
Hardware video codec support of the NVIDIA GeForce RTX 5090
CodecEncodeDecode
H.264 (AVC) Supported4:4:4: YesSupported
HEVC (H.265) Supported4:2:2: Yes; 4:4:4: Yes; 10-bit: Yes; B-frames: Yes; 8K: YesSupported4:4:4: Yes
AV1 SupportedSupported
VP9 Not stated by the vendorSupported

supported   explicitly not supported   not stated by the vendor

Media engines recorded for the NVIDIA GeForce RTX 5090: NVIDIA NVENC encoder (9th Generation, encoder engines: 3) and NVIDIA NVDEC decoder (6th Generation, decoder engines: 2).

Display outputs

The NVIDIA GeForce RTX 5090 offers outputs 3x DisplayPort, 1x HDMI, up to 4 simultaneous displays, a maximum resolution of 4K at 480Hz or 8K at 165Hz with DSC and HDCP 2.3. It is stated to be VR ready.

Outputs 3x DisplayPort, 1x HDMI
Max simultaneous displays 4
Max resolution 4K at 480Hz or 8K at 165Hz with DSC
HDCP 2.3
VR ready Yes

Frequently asked questions

How much VRAM does the NVIDIA GeForce RTX 5090 have?

The NVIDIA GeForce RTX 5090 has 32 GB of GDDR7 memory on a 512-bit bus. Its memory bandwidth is 1,792.0 GB/s.

How much power does the NVIDIA GeForce RTX 5090 use and what power supply does it need?

The NVIDIA GeForce RTX 5090 has a board power rating of 575 W. Feex records a suggested power supply of at least 1,000 W. Power connectors: 4x PCIe 8-pin cables (adapter in box) OR 1x 600 W PCIe Gen 5 cable.

How fast is the NVIDIA GeForce RTX 5090?

The NVIDIA GeForce RTX 5090 scores 38,961 in the PassMark G3D benchmark, higher than 99.9% of the 2,658 GPUs benchmarked. Its PassMark rank is #2. In the PassMark DirectX 12 test it averages 179 fps.

Does the NVIDIA GeForce RTX 5090 support ray tracing?

Yes. The NVIDIA GeForce RTX 5090 has hardware ray tracing (4th Generation) with 170 ray tracing cores.

Does the NVIDIA GeForce RTX 5090 support AV1 encoding?

Yes. The NVIDIA GeForce RTX 5090 supports AV1 hardware encoding and AV1 hardware decoding.

Does the NVIDIA GeForce RTX 5090 support DirectX 12 Ultimate?

Yes. The NVIDIA GeForce RTX 5090 supports DirectX 12 Ultimate.

What is the CUDA compute capability of the NVIDIA GeForce RTX 5090?

The NVIDIA GeForce RTX 5090 has CUDA compute capability 12.0.

What bus interface does the NVIDIA GeForce RTX 5090 use?

The NVIDIA GeForce RTX 5090 connects over PCIe 5.0.

When was the NVIDIA GeForce RTX 5090 released and how much did it cost at launch?

The NVIDIA GeForce RTX 5090 was released on January 30, 2025 and launched at a price of $1,999.

Corrections to this record

After an audit, 1 value on the NVIDIA GeForce RTX 5090 record was corrected. Each change is listed with its old value, new value, rule and sources.

The audit log is public: all changes for this GPU, the corrections overview and the method.

Similar performance