NVIDIA GeForce RTX 3090

NVIDIA GeForce RTX 3090

NVIDIA GeForce RTX 3090 in 2025: Power for Gamers and Professionals

Updated review of the flagship graphics card


Ampere Architecture: Innovations and Technologies

The NVIDIA GeForce RTX 3090, released in 2020, remains relevant even in 2025 due to its Ampere architecture. It is built on Samsung's 8nm process technology, which, despite the emergence of thinner nodes (such as 5nm in the RTX 40 series), provides a high transistor density of 28 billion.

Key Features:

- 2nd Generation RT Cores for ray tracing, accelerating lighting and shadow calculations in real-time.

- 3rd Generation Tensor Cores, enhancing the performance of DLSS (Deep Learning Super Sampling) and other AI-based algorithms.

- Support for DirectX 12 Ultimate, including technologies like Variable Rate Shading (VRS) and Mesh Shading.

- PCIe 4.0 interface, offering bandwidth up to 64 GB/s (though by 2025, PCIe 5.0 is already dominant).

DLSS 3.0 is the RTX 3090’s key advantage. Even five years post-release, this technology allows for gaming in 4K with high FPS, boosting performance by 40-70% through frame generation via neural networks.


24 GB GDDR6X: Memory for Any Task

The RTX 3090 is equipped with 24 GB of GDDR6X memory with a bandwidth of 936 GB/s. This solution continues to impress:

- For Gaming: Allows for working with 8K textures and complex scenes without data loading. For instance, in Microsoft Flight Simulator 2024 at max settings, the card has no issues even at 4K.

- For Professionals: 24 GB serves as a safety buffer for rendering, 3D modeling, and neural network applications. In Blender, rendering a scene with 10 million polygons takes 20% less time than competitors with 16 GB.

Con: GDDR6X heats up significantly, so effective cooling is a must.


4K Gaming and Beyond: Performance in 2025

Despite its age, the RTX 3090 handles modern titles effectively. Here are some examples of FPS (Ultra settings, without DLSS):

- Cyberpunk 2077 (2023):

- 4K + RT Ultra: 45-50 FPS → with DLSS 3.0 Quality: 65-70 FPS.

- 1440p + RT Ultra: 75-80 FPS.

- Starfield (2025):

- 4K: 60 FPS (with DLSS 3.0).

- Call of Duty: Modern Warfare V:

- 4K: 90-100 FPS.

Ray tracing still reduces FPS by 30-40%, but DLSS helps compensate for the losses. In games that support FidelityFX Super Resolution (AMD), the RTX 3090 shows commendable results, though it falls behind the Radeon RX 7900 XTX in some scenarios.


Not Just Gaming: Professional Tasks

The RTX 3090 is a hybrid of gaming and professional-grade graphics card.

- Video Editing: In Adobe Premiere Pro, rendering an 8K video lasting 10 minutes takes about 8-9 minutes thanks to its 10,496 CUDA cores.

- 3D Rendering: In Blender (Cycles engine), the card performs at the level of the NVIDIA RTX A5000 but costs 2-3 times less (by 2025, a new RTX 3090 is priced around $1000-1200).

- Scientific Calculations: Support for CUDA and OpenCL makes it popular for machine learning and simulations. For example, training a neural network based on TensorFlow is accelerated 3-4 times compared to CPU.

Limitation: For specific tasks (like rendering in AutoCAD), certified Quadro cards are more suitable.


Power Consumption and Cooling: What You Need to Know

The TDP of the RTX 3090 is 350 W, which requires a thoughtful power supply and cooling system:

- Power Supply: Minimum 750 W (850 W recommended) with an 80+ Gold certification.

- Cooling:

- An optimal option is a case with good ventilation (3-4 fans).

- For compact builds, hybrid liquid cooling systems (e.g., from EVGA or ASUS ROG Strix LC) work well.

- Operating temperature under load: 70-80°C (with good airflow).

Advice: Avoid cases shorter than 320 mm — the card takes 3 slots and is 313 mm long.


Competitors in 2025: Who is the RTX 3090 Compared To?

- AMD Radeon RX 7900 XTX (24 GB GDDR6):

- Cheaper ($900-1000) but weaker in ray tracing and lacks an equivalent to DLSS 3.0.

- Better at handling 4K gaming without RT (e.g., +10% in Horizon Forbidden West).

- NVIDIA RTX 4070 Ti (16 GB GDDR6X):

- New Ada Lovelace architecture, DLSS 3.5, but only has 16 GB of memory. Price — $800-900.

- Lags behind the RTX 3090 in professional tasks.

Conclusion: The RTX 3090 wins against competitors due to its memory capacity and versatility.


Practical Tips for Owners

1. Power Supply: Don’t skimp — Corsair RM850x or Be Quiet! Straight Power 11.

2. Compatibility:

- Motherboard with PCIe 4.0 (although the card works with PCIe 3.0 with minimal losses).

- Update BIOS to support Resizable BAR (increases performance by 5-10%).

3. Drivers: Regularly update via GeForce Experience. For professional tasks, use Studio Drivers.


Pros and Cons of the RTX 3090 in 2025

Pros:

- 24 GB of memory for gaming and work.

- Excellent 4K performance and with RT.

- Support for DLSS 3.0.

- Versatility (gaming + professional use).

Cons:

- High power consumption.

- Size (won't fit in small cases).

- Price: $1000-1200 compared to $900 for the RX 7900 XTX.


Final Verdict: Who Is the RTX 3090 For?

This graphics card is the choice for those who are not willing to compromise:

- Gamers wanting to play in 4K with maximum quality and RT.

- Professionals needing a versatile tool for rendering, editing, and ML.

- Enthusiasts building PCs with future-proof capability.

If you’re seeking a card for "all occasions" and are willing to tolerate its power appetite — the RTX 3090 is still relevant. However, for pure gaming in 2025, there are newer and more energy-efficient options available (like the RTX 4070 Ti or RX 7900 XTX).

Basic

Label Name
NVIDIA
Platform
Desktop
Launch Date
September 2020
Model Name
GeForce RTX 3090
Generation
GeForce 30
Base Clock
1395MHz
Boost Clock
1695MHz
Bus Interface
PCIe 4.0 x16
Transistors
28,300 million
RT Cores
82
Tensor Cores
?
Tensor Cores are specialized processing units designed specifically for deep learning, providing higher training and inference performance compared to FP32 training. They enable rapid computations in areas such as computer vision, natural language processing, speech recognition, text-to-speech conversion, and personalized recommendations. The two most notable applications of Tensor Cores are DLSS (Deep Learning Super Sampling) and AI Denoiser for noise reduction.
328
TMUs
?
Texture Mapping Units (TMUs) serve as components of the GPU, which are capable of rotating, scaling, and distorting binary images, and then placing them as textures onto any plane of a given 3D model. This process is called texture mapping.
328
Foundry
Samsung
Process Size
8 nm
Architecture
Ampere

Memory Specifications

Memory Size
24GB
Memory Type
GDDR6X
Memory Bus
?
The memory bus width refers to the number of bits of data that the video memory can transfer within a single clock cycle. The larger the bus width, the greater the amount of data that can be transmitted instantaneously, making it one of the crucial parameters of video memory. The memory bandwidth is calculated as: Memory Bandwidth = Memory Frequency x Memory Bus Width / 8. Therefore, when the memory frequencies are similar, the memory bus width will determine the size of the memory bandwidth.
384bit
Memory Clock
1219MHz
Bandwidth
?
Memory bandwidth refers to the data transfer rate between the graphics chip and the video memory. It is measured in bytes per second, and the formula to calculate it is: memory bandwidth = working frequency × memory bus width / 8 bits.
936.2 GB/s

Theoretical Performance

Pixel Rate
?
Pixel fill rate refers to the number of pixels a graphics processing unit (GPU) can render per second, measured in MPixels/s (million pixels per second) or GPixels/s (billion pixels per second). It is the most commonly used metric to evaluate the pixel processing performance of a graphics card.
189.8 GPixel/s
Texture Rate
?
Texture fill rate refers to the number of texture map elements (texels) that a GPU can map to pixels in a single second.
556.0 GTexel/s
FP16 (half)
?
An important metric for measuring GPU performance is floating-point computing capability. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable. Single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks, while double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy.
35.58 TFLOPS
FP64 (double)
?
An important metric for measuring GPU performance is floating-point computing capability. Double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy, while single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable.
556.0 GFLOPS
FP32 (float)
?
An important metric for measuring GPU performance is floating-point computing capability. Single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks, while double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable.
34.868 TFLOPS

Miscellaneous

SM Count
?
Multiple Streaming Processors (SPs), along with other resources, form a Streaming Multiprocessor (SM), which is also referred to as a GPU's major core. These additional resources include components such as warp schedulers, registers, and shared memory. The SM can be considered the heart of the GPU, similar to a CPU core, with registers and shared memory being scarce resources within the SM.
82
Shading Units
?
The most fundamental processing unit is the Streaming Processor (SP), where specific instructions and tasks are executed. GPUs perform parallel computing, which means multiple SPs work simultaneously to process tasks.
10496
L1 Cache
128 KB (per SM)
L2 Cache
6MB
TDP
350W
Vulkan Version
?
Vulkan is a cross-platform graphics and compute API by Khronos Group, offering high performance and low CPU overhead. It lets developers control the GPU directly, reduces rendering overhead, and supports multi-threading and multi-core processors.
1.3
OpenCL Version
3.0
OpenGL
4.6
DirectX
12 Ultimate (12_2)
CUDA
8.6
Power Connectors
1x 12-pin
Shader Model
6.6
ROPs
?
The Raster Operations Pipeline (ROPs) is primarily responsible for handling lighting and reflection calculations in games, as well as managing effects like anti-aliasing (AA), high resolution, smoke, and fire. The more demanding the anti-aliasing and lighting effects in a game, the higher the performance requirements for the ROPs; otherwise, it may result in a sharp drop in frame rate.
112
Suggested PSU
750W

Benchmarks

Shadow of the Tomb Raider 2160p
Score
95 fps
Shadow of the Tomb Raider 1440p
Score
157 fps
Shadow of the Tomb Raider 1080p
Score
196 fps
Cyberpunk 2077 2160p
Score
65 fps
Cyberpunk 2077 1440p
Score
81 fps
Cyberpunk 2077 1080p
Score
116 fps
Battlefield 5 2160p
Score
118 fps
Battlefield 5 1440p
Score
185 fps
Battlefield 5 1080p
Score
195 fps
GTA 5 2160p
Score
129 fps
GTA 5 1440p
Score
186 fps
GTA 5 1080p
Score
213 fps
FP32 (float)
Score
34.868 TFLOPS
3DMark Time Spy
Score
19416
Blender
Score
5266.54
Vulkan
Score
170158
OpenCL
Score
194529
Hashcat
Score
1198430 H/s

Compared to Other GPU

Shadow of the Tomb Raider 2160p / fps
193 +103.2%
45 -52.6%
34 -64.2%
24 -74.7%
Shadow of the Tomb Raider 1440p / fps
292 +86%
67 -57.3%
49 -68.8%
Shadow of the Tomb Raider 1080p / fps
310 +58.2%
101 -48.5%
72 -63.3%
Cyberpunk 2077 2160p / fps
90 +38.5%
24 -63.1%
Cyberpunk 2077 1440p / fps
185 +128.4%
35 -56.8%
Cyberpunk 2077 1080p / fps
203 +75%
48 -58.6%
Battlefield 5 2160p / fps
194 +64.4%
56 -52.5%
Battlefield 5 1440p / fps
203 +9.7%
Battlefield 5 1080p / fps
213 +9.2%
169 -13.3%
139 -28.7%
122 -37.4%
GTA 5 2160p / fps
174 +34.9%
GTA 5 1440p / fps
73 -60.8%
GTA 5 1080p / fps
156 -26.8%
141 -33.8%
86 -59.6%
FP32 (float) / TFLOPS
44.257 +26.9%
39.2 +12.4%
31.615 -9.3%
28.567 -18.1%
3DMark Time Spy
36233 +86.6%
9097 -53.1%
Blender
15026.3 +185.3%
2020.49 -61.6%
1064 -79.8%
552 -89.5%
Vulkan
382809 +125%
91662 -46.1%
61331 -64%
34688 -79.6%
OpenCL
385013 +97.9%
109617 -43.7%
75816 -61%
57474 -70.5%
Hashcat / H/s
2602280 +117.1%
971947 -18.9%
883336 -26.3%
881523 -26.4%