The best graphics cards for 3D graphics, modeling and games
Video cards have two roles in modern computers. In games, they provide beautiful graphics, fast frame rates, explosions of pyrotechnics, and various visual effects that developers have worked on.
And for the CGI industry, including graphic artists, designers, illustrators and 3D professionals working in programs such as Adobe Creative Suite, Photoshop, Blender, Maya, 3DS Max and others, a powerful graphics card can be of great importance for rendering or modeling . Some effects in these programs can no longer even work on the CPU without the help of a video card.
Quadro vs GeForce vs Radeon vs Radeon Pro
Nvidia and AMD produce two types of video cards, which are tentatively designed for games or design. For Nvidia, you probably already know the gaming brand GeForce, while Quadro cards are for professionals, and with AMD, Radeon for games and Radeon Pro for creative software. The bottom line is that professional-grade cards are much more expensive.
At a higher price for Quadros and Radeon Pro, you get basically the same hardware specifications as in much cheaper game cards. They have the same basic design, the same architecture and similar specifications, but with some fundamental differences. Quadro and Radeon cards have certified drivers. This means that they have been tested for compatibility with specific software, offer better performance with design software and are less likely to run into problems. They have ECC memory for added accuracy. And sometimes they operate at lower clock speeds, which means they have lower power requirements and lower thermal requirements.
Look at any review of the video card and it will be full of three-letter abbreviations, which are used to illustrate the expected performance of the software. But it may make you wonder which of these numbers matter to a modern graphics card.
The key features often mentioned in reviews and manufacturers are the amount of memory (capacity, bandwidth, and speed), the number of cores, and the clock speed of the card (in MHz). These specifications differ for different generations of GPUs and for different levels, and the cores in the Nvidia and AMD cards do not match. Nvidia uses the term Cuda core, while AMD refers to its GCN core. The performance between AMD and Nvidia absolutely can not be compared if we assume that the AMD card has more or less cores than the Nvidia card.
Now let's move on to our review of the best graphics cards available now.
The best graphics cards for 3d graphics and games
Nvidia Quadro RTX 8000
The world's first Ray Tracing GPU.
GPU cores: 4.608 | Base frequency: 1,395 MHz | Maximum Clock Frequency: 1.770 MHz | Memory: 48 GB HDDR6 | Memory Bandwidth: 384 GB / s | Maximum supported resolution: 7680x4320
Open a new frontier in professional graphics with unprecedented performance and scalability with 48 GB of high-speed GDDR6 memory and NVIDIA NVLink ™. Designers and artists from various industries can now expand the boundaries of what is possible by working with the largest and most complex workloads of ray tracing, deep learning and visual computing.
With it, you get much more rendering power and 3D modeling than the previous Pascal generation, taking Cuda and OpenCL applications to a new level and leaving any other video card in a relatively weak state.
Nvidia GeForce RTX 2080 Ti
A revolution in game realism and performance.
GPU cores: 4,352 | Base frequency: 1,350 MHz | Maximum Clock Frequency: 1.545 MHz | GFLOPS: 13,448 | Memory: 11 GB GDDR6 | Memory Bandwidth: 616 GB/s
Each generation of Nvidia releases its flagship model, and then the second version with improved features, with a high price, and this is really what PC users will enthusiastically consider the RTX 2080 Ti. This is currently the fastest graphics card on the planet - 4352 Cuda cores, almost double the vanilla RTX 2080 ray tracing hardware, and graphics processing performance nearly doubled.
This card alone is more expensive than a regular mid-range PC, but with some serious hardware it is worth considering investments, including for designers whose workstation is turning into a gaming PC, since the performance of Cuda and OpenCL has grown significantly. gaming performance.
Nvidia Titan RTX
The fastest PC graphics card ever created.
GPU cores: 4.608 | Base frequency: 1,350 MHz | Maximum Clock Frequency: 1.770 MHz | Memory: 24 GB GDDR6 | Memory Bandwidth: 672 GB/s
The most demanding users need the best tools. TITAN RTX is built on the NVIDIA Turing GPU architecture and incorporates the latest Tensor Core and RT Core technologies to accelerate artificial intelligence and ray tracing. It is also supported by NVIDIA and SDK drivers, so developers, researchers and creators can work faster and get better results.
TITAN RTX, built with precision tensor cores, delivers unmatched FP32, FP16, INT8, and INT4 performance, enabling faster training and output of neural networks. With twice the memory capacity of previous generation TITAN GPUs and NVIDIA NVLink ™ GPUs, TITAN RTX allows data scientists and data scientists to experiment with larger neural networks and datasets than ever before, all in the GPU's memory.
Nvidia Quadro RTX 4000
Professional-grade basic graphics card.
GPU cores: 2,304 | Base frequency: 1.005 MHz | Maximum Clock Frequency: 1.545 MHz | GFLOPS: 7,100 | Memory: 8 GB GDDR6 | Memory Bandwidth: 416 GB/s
This is our main recommendation for workstation-class graphics cards at an affordable price with excellent performance in design applications. It comes in an elegant single-slot design that helps fit into small cases and requires less power than a larger GeForce card .
In particular, OpenCL and Cuda applications make full use of the new Turing architecture, so the RTX 4000 will be of great importance when working with all kinds of creative software, plugins and filters, providing excellent performance when rendering images, 3D and video.
Flagship graphics card
GPU cores: 5.120 | Base frequency: 1200 MHz | Maximum Clock Frequency: 1447 MHz | Memory: 32 GB GDDR6 | Memory Bandwidth: 870 GB/s | Maximum supported resolution: 7680x4320
NVIDIA® Quadro® GV100 reinvents the workstation to meet the demands of advanced design, simulation and 3d visualization workflows. It is powered by NVIDIA Volta technology, providing incredible memory capacity, scalability and performance that designers, architects and scientists must create, create and solve the impossible.
In terms of computing performance, NVIDIA highlights real-time ray tracing. At the Games Developer Conference a few days ago, NVIDIA introduced a developer interface called RTX, which will soon be supported in games. However, the first time ray tracing will be used only for some scene effects. In a professional environment, ray tracing is usually performed for the entire scene. And here the Quadro GV100 should show its strengths.
The novelty can rightfully be considered the most productive professional graphics card in the world. It is reported that the peak performance in double-precision computing is 7.4 TFlops, in single-precision problems - 14.8 TFlops, and in problems related to deep learning - 118.5 TFlops. The last indicator is precisely what the "tensor kernels" provide.
AMD Radeon VII
AMD's new high-performance graphics card is fierce competition for Nvidia
GPU cores: 3840 | Base frequency: 1400 | Maximum Clock Frequency: 1800 MHz | GFLOPS: 13,824 | Memory: 16 GB HBM2 | Memory Bandwidth: 1028 GB/s
The video card (codenamed Vega 20) is aimed at the high-end GPU market, breaking AMD’s tradition of simply offering better results than Nvidia’s offerings. The Radeon VII is the GeForce RTX 2070 for gaming, and it is especially good for creative work with high resolution thanks to the huge 16 GB ultra-fast HBM2 memory and amazing OpenCL performance.
This new generation of AMD cards costs a little more, and, most importantly, it is a bit noisier and hotter than Nvidia's offerings. There is also no hardware support for ray tracing that RTX cards have, which can be a decisive factor when making a purchase decision if you are not shaken by the additional amount of Radeon VII memory.
Nvidia GeForce GTX 1660 Ti
A modern card that replaces the GTX 1060
GPU cores: 1,536 | Base frequency: 1500 | Maximum Clock Frequency: 1770 MHz | GFLOPS: 4.608 | Memory: 6 GB GDDR6 | Memory Bandwidth: 288 GB/s
With a price and spec that puts the Geforce GTX 1660 Ti. This new Nvidia card will no doubt find its way into more affordable off-the-shelf PCs than the expensive high-end RTX series, with features roughly located between the (still impressive) GTX 1070 and GTX 1060.
It has 6 GB of GDDR6 memory and a modest 1,536 Cuda cores, and is based on the newer 12nm Turning RTX card architecture, but without ray tracing hardware.
he GTX 1660 Ti, which provides excellent gaming performance in 1080p and 1440p, as well as many features for accelerating plug-ins and filters in creative software, is offered at a very good price and is offered by some manufacturers (for example, PNY) in a very short form, this can squeeze her into tiny pc.
AMD Radeon RX 590
Middle class card for a great price
GPU cores: 2,304 | Base frequency: 1469 MHz | Maximum Clock Frequency: 1.545 MHz | GFLOPS: 7,100 | Memory: 8 GB GDDR5 | Memory Bandwidth: 256 GB/s
AMD now has a new flagship top-end processor Radeon VII, the older RX590 video card has fallen to about the price point, where it directly competes with the GeForce GTX 1060 from Nvidia. But in many situations, the RX590 is a better buy, outperforming the GTX 1060 in some games and OpenCL processing. It has more memory, bandwidth and processing power. Given its low price, it’s worth it to think about it.
AMD Radeon Pro WX8200
Unlimited freedom for the embodiment of creative ideas.
GPU cores: 3.384 | Base frequency: 1,200 MHz | Maximum Clock Frequency: 1,500 MHz | GFLOPS: 10,075 | Memory: 8 GB HBM2 | Memory Bandwidth: 512 GB/s
AMD has released an excellent professional-grade graphics card, the WX8200. It is still based on the current Vega architecture, but has 3,584 cores and 8 GB of ultra-fast HBM2 memory, which provides excellent performance in specific tests. Equipped with advanced features, Radeon Pro WX 8200 graphics accelerators are professional-oriented and certified for use with many of today's most popular applications. They are ideal solutions for those who are looking for highly professional graphics for areas such as design, 3d modeling, manufacturing, design work, planning, construction or multimedia and entertainment applications.
Depending on the tools you use, AMD cards may offer better performance. For example, 3D animators who rely solely on Maya or Blender may benefit from choosing AMD instead of Nvidia, although we recommend some additional research and review your workflow before investing. OpenCL applications work great with AMD graphics cards, and the WX8200 is the best choice for AMD professional-level graphics cards.
How to choose a graphics card
There are a few basic things to consider when buying. The higher resolution you work with (or play with) requires more memory. If you intend to work with 4K resolution on the screen or with larger textures, you need a graphics card with a large amount of memory. 8 GB or more is now common on higher level cards.
The number of cores really determines the overall rendering power of the card. They vary significantly in different levels of price and performance: from entry-level cards $ 100 to hippos $ 1000 +.
The clock speed of the graphics card is indicated as a base digit and “GPU boost”. Like Turbo mode on Intel processors, when the video card is under heavy load, it will operate at a higher clock frequency to increase performance until it reaches a predetermined maximum that is set to avoid overheating.
Be sure to also consider the display (s) you are working with and the output of the video card you are buying. All modern video cards use only digital video outputs, either HDMI or DisplayPort (it can be either a small square miniDP connector or a large D-shaped connector).
For 4K or 5K displays, all graphics cards now support at least DisplayPort 1.4 and HDMI 2.0 standards, which offer the bandwidth needed for a 60 Hz refresh rate - which was a serious problem on older graphics cards when higher resolution displays were first introduced became a popular offer for some many years ago. When 8K displays finally become more affordable, this problem will once again be on its feet.
And finally, the single biggest difference in graphics card performance that may be obvious to some readers is the hardware generation of a series of cards, always encoded for reference. Nvidia names its cards after scientists like Pascal, Turing, etc., while AMD is a bit more obscure, with Polaris and the newer Vega architecture that is now on the market.
Nvidia and AMD release a new series of video cards approximately every two years, and when a new generation comes out, this means raising the bar in all technical areas - more cores, more memory, more bandwidth and more functions, often compressing the card to the same power and thermal requirements previous generation.
For best performance and a better perspective, look only towards the new graphics cards.