Main Page / Graphic cards / Comparision NVIDIA GeForce RTX 4090 vs NVIDIA H100 SXM5
NVIDIA GeForce RTX 4090 NVIDIA GeForce RTX 4090
NVIDIA H100 SXM5 NVIDIA H100 SXM5
VS

Comparision NVIDIA GeForce RTX 4090 vs NVIDIA H100 SXM5

NVIDIA GeForce RTX 4090

WINNER
NVIDIA GeForce RTX 4090

Rating: 0 points
NVIDIA H100 SXM5

NVIDIA H100 SXM5

Rating: 0 points
Grade
NVIDIA GeForce RTX 4090
NVIDIA H100 SXM5
Performance
8
8
Memory
2
5
General information
8
8
Functions
8
3
Ports
7
0

Top specs and features

GPU base clock speed

NVIDIA GeForce RTX 4090: 1815 MHz NVIDIA H100 SXM5: 1065 MHz

RAM

NVIDIA GeForce RTX 4090: 24 GB NVIDIA H100 SXM5: 80 GB

Memory bandwidth

NVIDIA GeForce RTX 4090: 1.018 GB/s NVIDIA H100 SXM5: 1.92 GB/s

Gpu memory speed

NVIDIA GeForce RTX 4090: 1325 MHz NVIDIA H100 SXM5: 1500 MHz

FLOPS

NVIDIA GeForce RTX 4090: 70.24 TFLOPS NVIDIA H100 SXM5: 57.68 TFLOPS

Description

The NVIDIA GeForce RTX 4090 video card is based on the Lovelace architecture. NVIDIA H100 SXM5 on the Hopper architecture. The first has There is no data million transistors. The second is 80000 million. NVIDIA GeForce RTX 4090 has a transistor size of 5 nm versus 4.

The base clock speed of the first video card is 1815 MHz versus 1065 MHz for the second.

Let's move on to memory. NVIDIA GeForce RTX 4090 has 24 GB. NVIDIA H100 SXM5 has 24 GB installed. The bandwidth of the first video card is 1.018 Gb/s versus 1.92 Gb/s of the second.

FLOPS of NVIDIA GeForce RTX 4090 is 70.24. At NVIDIA H100 SXM5 57.68.

Goes to tests in benchmarks. In the Passmark benchmark, NVIDIA GeForce RTX 4090 scored There is no data points. And here is the second card There is no data points. In 3DMark, the first model scored There is no data points. Second There is no data points.

In terms of interfaces. The first video card is connected using There is no data. The second is There is no data. Video card NVIDIA GeForce RTX 4090 has Directx version 12.2. Video card NVIDIA H100 SXM5 -- Directx version - There is no data.

Regarding cooling, NVIDIA GeForce RTX 4090 has 450W heat dissipation requirements versus 700W for NVIDIA H100 SXM5.

Why NVIDIA GeForce RTX 4090 is better than NVIDIA H100 SXM5

  • GPU base clock speed 1815 MHz против 1065 MHz, more on 70%
  • FLOPS 70.24 TFLOPS против 57.68 TFLOPS, more on 22%
  • Turbo gpu 1925 MHz против 1780 MHz, more on 8%
  • Power Consumption (TDP) 450 W против 700 W, less by -36%

NVIDIA GeForce RTX 4090 vs NVIDIA H100 SXM5: highlights

NVIDIA GeForce RTX 4090
NVIDIA GeForce RTX 4090
NVIDIA H100 SXM5
NVIDIA H100 SXM5
Performance
GPU base clock speed
The graphics processing unit (GPU) has a high clock speed.
1815 MHz
max 2457
Average: 1124.9 MHz
1065 MHz
max 2457
Average: 1124.9 MHz
Gpu memory speed
This is an important aspect for calculating memory bandwidth.
1325 MHz
max 16000
Average: 1468 MHz
1500 MHz
max 16000
Average: 1468 MHz
FLOPS
Measuring the processing power of a processor is called FLOPS.
70.24 TFLOPS
max 1142.32
Average: 53 TFLOPS
57.68 TFLOPS
max 1142.32
Average: 53 TFLOPS
RAM
RAM in video cards (also known as video memory or VRAM) is a special type of memory used by a video card to store graphics data. It serves as a temporary buffer for textures, shaders, geometry, and other graphics resources that are needed to display images on the screen. More RAM allows the graphics card to work with more data and handle more complex graphic scenes with high resolution and detail. Show more
24 GB
max 128
Average: 4.6 GB
80 GB
max 128
Average: 4.6 GB
Number of threads
The more threads a video card has, the more processing power it can provide.
17408
max 18432
Average: 1326.3
16896
max 18432
Average: 1326.3
Number of PCIe lanes
The number of PCIe lanes in video cards determines the speed and bandwidth of data transfer between the video card and other computer components through the PCIe interface. The more PCIe lanes a video card has, the more bandwidth and ability to communicate with other computer components. Show more
16
max 16
Average:
16
max 16
Average:
Pixel rendering speed
The higher the pixel rendering speed, the smoother and more realistic the display of graphics and the movement of objects on the screen will be.
370 GTexel/s    
max 563
Average: 94.3 GTexel/s    
43 GTexel/s    
max 563
Average: 94.3 GTexel/s    
TMUs
Responsible for texturing objects in 3D graphics. TMU provides textures to the surfaces of objects, which gives them a realistic look and detail. The number of TMUs in a video card determines its ability to process textures. The more TMUs, the more textures can be processed at the same time, which contributes to better texturing of objects and increases the realism of graphics. Show more
544
max 880
Average: 140.1
528
max 880
Average: 140.1
ROPs
Responsible for the final processing of pixels and their display on the screen. ROPs perform various operations on pixels, such as blending colors, applying transparency, and writing to the framebuffer. The number of ROPs in a video card affects its ability to process and display graphics. The more ROPs, the more pixels and image fragments can be processed and displayed on the screen at the same time. A higher number of ROPs generally results in faster and more efficient graphics rendering and better performance in games and graphics applications. Show more
192
max 256
Average: 56.8
24
max 256
Average: 56.8
Number of shader blocks
The number of shader units in video cards refers to the number of parallel processors that perform computational operations in the GPU. The more shader units in the video card, the more computing resources are available for processing graphics tasks. Show more
17408
max 17408
Average:
16896
max 17408
Average:
L2 cache size
Used to temporarily store data and instructions used by the graphics card when performing graphics calculations. A larger L2 cache allows the graphics card to store more data and instructions, which helps speed up the processing of graphics operations. Show more
96000
50000
Turbo gpu
If the GPU speed has dropped below its limit, then to improve performance, it can go to a high clock speed.
1925 MHz
max 2903
Average: 1514 MHz
1780 MHz
max 2903
Average: 1514 MHz
architecture name
Lovelace
Hopper
GPU name
AD102
GH100
Memory
Memory bandwidth
This is the rate at which the device stores or reads information.
1.018 GB/s
max 2656
Average: 257.8 GB/s
1.92 GB/s
max 2656
Average: 257.8 GB/s
RAM
RAM in video cards (also known as video memory or VRAM) is a special type of memory used by a video card to store graphics data. It serves as a temporary buffer for textures, shaders, geometry, and other graphics resources that are needed to display images on the screen. More RAM allows the graphics card to work with more data and handle more complex graphic scenes with high resolution and detail. Show more
24 GB
max 128
Average: 4.6 GB
80 GB
max 128
Average: 4.6 GB
GDDR memory versions
Latest versions of GDDR memory provide high data transfer rates to improve overall performance
6
max 6
Average: 4.9
3
max 6
Average: 4.9
Memory bus width
A wide memory bus means that it can transfer more information in one cycle. This property affects memory performance as well as the overall performance of the device's graphics card. Show more
384 bit
max 8192
Average: 283.9 bit
5120 bit
max 8192
Average: 283.9 bit
General information
Crystal size
The physical dimensions of the chip on which the transistors, microcircuits and other components necessary for the operation of the video card are located. The larger the die size, the more space the GPU takes up on the graphics card. Larger die sizes can provide more computing resources such as CUDA cores or tensor cores, which can result in increased performance and graphics processing capabilities. Show more
611
max 826
Average: 356.7
814
max 826
Average: 356.7
Length
334
max 524
Average: 250.2
max 524
Average: 250.2
Generation
A new generation of graphics card usually includes improved architecture, higher performance, more efficient use of power, improved graphics capabilities, and new features. Show more
GeForce 40
Tesla
Manufacturer
TSMC
TSMC
Power supply power
When choosing a power supply for a video card, you must take into account the power requirements of the video card manufacturer, as well as other computer components. Show more
850
max 1300
Average:
1100
max 1300
Average:
Year of issue
2022
max 2023
Average:
2022
max 2023
Average:
Power Consumption (TDP)
Heat Dissipation Requirements (TDP) is the maximum possible amount of energy dissipated by the cooling system. The lower the TDP, the less power will be consumed Show more
450 W
Average: 160 W
700 W
Average: 160 W
Technological process
The small size of the semiconductors means this is a new generation chip.
5 nm
Average: 34.7 nm
4 nm
Average: 34.7 nm
PCIe connection interface
A considerable speed of the expansion card used to connect the computer to the peripherals is provided. The updated versions offer impressive bandwidth and high performance. Show more
4
max 4
Average: 3
4
max 4
Average: 3
Width
142 mm
max 421.7
Average: 192.1 mm
mm
max 421.7
Average: 192.1 mm
Height
63 mm
max 620
Average: 89.6 mm
mm
max 620
Average: 89.6 mm
Purpose
Desktop
Desktop
Functions
OpenGL Version
OpenGL provides access to the graphics card's hardware capabilities for displaying 2D and 3D graphics objects. New versions of OpenGL may include support for new graphical effects, performance optimizations, bug fixes, and other improvements. Show more
4.6
max 4.6
Average:
max 4.6
Average:
DirectX
Used in demanding games, providing improved graphics
12.2
max 12.2
Average: 11.4
max 12.2
Average: 11.4
Shader model version
The higher the version of the shader model in the video card, the more functions and possibilities are available for programming graphic effects.
6.6
max 6.7
Average: 5.9
max 6.7
Average: 5.9
CUDA Version
Allows you to use the compute cores of your graphics card to perform parallel computing, which can be useful in areas such as scientific research, deep learning, image processing, and other computationally intensive tasks. Show more
9
max 9
Average:
9
max 9
Average:
Ports
Number of connectors 16-pin
1
max 1
Average: 1
max 1
Average: 1
Has hdmi output
HDMI output allows you to connect devices with HDMI or mini HDMI ports. They can send video and audio to the display.
Available
There is no data
HDMI version
The latest version provides a wide signal transmission channel due to the increased number of audio channels, frames per second, etc.
2.1
max 2.1
Average: 1.9
max 2.1
Average: 1.9
Number of HDMI connectors
The more their number, the more devices can be connected at the same time (for example, game / TV set-top boxes)
1
max 3
Average: 1.1
max 3
Average: 1.1
HDMI
A digital interface that is used to transmit high-resolution audio and video signals.
Available
There is no data

FAQ

How does the NVIDIA GeForce RTX 4090 processor perform in benchmarks?

Passmark NVIDIA GeForce RTX 4090 scored There is no data points. The second video card scored There is no data points in Passmark.

What FLOPS do video cards have?

FLOPS NVIDIA GeForce RTX 4090 is 70.24 TFLOPS. But the second video card has FLOPS equal to 57.68 TFLOPS.

What power consumption?

NVIDIA GeForce RTX 4090 450 Watt. NVIDIA H100 SXM5 700 Watt.

How fast are NVIDIA GeForce RTX 4090 and NVIDIA H100 SXM5?

NVIDIA GeForce RTX 4090 operates at 1815 MHz. In this case, the maximum frequency reaches 1925 MHz. The clock base frequency of NVIDIA H100 SXM5 reaches 1065 MHz. In turbo mode it reaches 1780 MHz.

What kind of memory do graphics cards have?

NVIDIA GeForce RTX 4090 supports GDDR6. Installed 24 GB of RAM. Throughput reaches 1.018 GB/s. NVIDIA H100 SXM5 works with GDDR3. The second one has 80 GB of RAM installed. Its bandwidth is 1.018 GB/s.

How many HDMI connectors do they have?

NVIDIA GeForce RTX 4090 has 1 HDMI outputs. NVIDIA H100 SXM5 is equipped with There is no data HDMI outputs.

What power connectors are used?

NVIDIA GeForce RTX 4090 uses There is no data. NVIDIA H100 SXM5 is equipped with There is no data HDMI outputs.

What architecture are video cards based on?

NVIDIA GeForce RTX 4090 is built on Lovelace. NVIDIA H100 SXM5 uses the Hopper architecture.

What graphics processor is being used?

NVIDIA GeForce RTX 4090 is equipped with AD102. NVIDIA H100 SXM5 is set to GH100.

How many PCIe lanes

The first graphics card has 16 PCIe lanes. And the PCIe version is 4. NVIDIA H100 SXM5 16 PCIe lanes. PCIe version 4.

How many transistors?

NVIDIA GeForce RTX 4090 has There is no data million transistors. NVIDIA H100 SXM5 has 80000 million transistors

Graphic cards