7 best graphics cards for graphics and computing

Review of the best according to the editorial board. On the selection criteria. This material is subjective and does not constitute advertising and does not serve as a purchase guide. Before buying, you need to consult with a specialist.

If somewhere it comes to video cards, most ordinary people have associations with either computer games or cryptocurrency mining. But there is also a separate class of video cards that are designed for special and very serious purposes – working with the most complex graphics, which requires the deepest calculations. Often, such video cards are not suitable for either gaming or mining, but at the same time cost much more than the most top gaming 'types'.

Features of special video cards for working with graphics and computing

Unlike gaming, professional video cards are not under the scrutiny of enthusiasts and specialized publications, and there is little information on them. Therefore, today's review from the magazine, compiled together with analysts of the RankQuality.com service, is devoted to these devices. Here, as in the case of gaming video cards, there are very few manufacturers of GPUs in the world (for example, nVidia), and there are a lot of manufacturers of working video cards, they just use chips from major manufacturers, making their own devices that differ in arbitrary parameters – memory, a set of interfaces , software, etc.

Rating of the best video cards for working with graphics and computing

Nomination	a place	Name of product	price
The best low-cost graphics cards for graphics and computing	1	HP Quadro NVS 315 PCI-E 1024Mb 64 bit	RUB 9 652
	2	MSI Radeon RX 5700 1610MHz PCI-E 4.0 8192Mb 14000MHz 256 bit 3xDisplay	RUB 33,474
	3	PNY Quadro P2000 PCI-E 3.0 5120Mb 160bit HDCP	RUB 41,990
	4	PNY Quadro 5000 513Mhz PCI-E 2.0 2560Mb 3000Mhz 320 bit DVI	RUB 38 310
The best premium graphics cards for graphics and computing	1	Sapphire FirePro S9150 PCI-E 3.0 16384Mb 512 bit	RUB 102,450
	2	PNY Tesla K40 745Mhz PCI-E 3.0 12288Mb 6000Mhz 384 bit Silent	RUB 175,300
	3	PNY Quadro RTX 6000 PCI-E 3.0 24576Mb 384 bit (VCQRTX6000-PB)	RUB 515 230

The best low-cost graphics cards for graphics and computing

Let's conditionally divide our small review into two categories according to the price availability of video cards, and start with relatively inexpensive ones. We'll warn you right away that they are really relatively inexpensive, and cost tens of thousands of rubles. However, given the current market conjuncture for 2020, there is no particular choice.

HP Quadro NVS 315 PCI-E 1024Mb 64 bit

Rating: 4.9

First, let's take a look at the most affordable professional graphics card for graphics and computing from Hewlett-Packard. It is really inexpensive when compared with even average gaming ones, but many times more expensive than office 'plugs'. The NVS 315 processor itself was first released by nVidia a relatively long time ago – in 2013, but video cards based on it are still relevant and in demand. The code name of the video processor is GF119, made according to the 40nm process technology.

This particular model and its 'relatives' are used mainly in medium and large enterprises, where production is associated with the operation of complex graphic design systems.

The card is equipped with a DVI video output, supports NVIDIA Mosaic technology, and is completed with specialized software for manipulating desktops. Supports up to two monitors with resolutions up to 2560×1600. Every HP Quadro NVS 315 graphics card undergoes rigorous quality control and is extensively tested in leading commercial applications.

Now let's analyze the most important technical characteristics of a video card. This is a low-profile model, occupies one slot, connects to the motherboard PCI-E 16x version 2.0 connector. The base frequency of the processor is 550MHz. Video memory with a frequency of 875MHz GDDR3 type is installed in the volume of 1GB. RAMDAC frequency – 400MHz. 64-bit memory bus.

In terms of firmware and compatibility with operating systems and applications, this model supports the well-known standards and technologies DirectX 11 and OpenGL 4.1.

The device does not require an additional power supply. Inherent heat dissipation is estimated at TDP 20W. Active cooling – a small fan on a heatsink that occupies almost the entire area of ​​the PCB board.

This model comes with a 36-month manufacturing warranty.

Advantages

• compact, one slot;
• strict quality control;
• reliability;
• does not require additional power supply;
• affordable price.

disadvantages

• the weakest configuration in the collection.

MSI Radeon RX 5700 1610MHz PCI-E 4.0 8192Mb 14000MHz 256 bit 3xDisplay

Rating: 4.8

The second professional graphics card in the selection of relatively inexpensive models is a product of the Taiwanese brand MSI (Micro-Star International Co., Ltd), which is known for its high-quality computer components and not only.

This is already a much more productive video card than the previous one, and several times more expensive. It is larger, more massive, requires more intensive active cooling, therefore it is equipped with two fans on the radiator at once. In addition, it needs additional power beyond what it gets from the PCI-E 16x 4.0 slot. For additional power supply, the board provides 8-pin contacts and a second 8-pin.

As for the scope of application, this model was initially more inclined towards calculations, but the manufacturer's marketers decided not to lose the gaming audience. Therefore, in its description you can often find the word 'Gaming'. It was equally in high demand among cryptocurrency miners at one time due to its strong computing power.

The key technical characteristics of the device are as follows. The system is based on the Navi 10 XL GPU (AMD Radeon RX 5700). Created according to the latest 7nm process technology. Operates at a frequency of 1610MHz.

The device supports simultaneous connection of up to 4 monitors. It can provide a maximum resolution of 7680×4320 thanks to the 8GB video memory – GDDR6 type, 1.4GHz frequency with a 256-bit bus. SLI / CrossFire supported. The math block is characterized by 2304 universal processors, 144 texture units, 64 rasterization units. DirectX 12 and OpenGL 4.6 are supported.

The current connection interfaces are three DisplayPort 1.4, one HDMI 2.0b, HDCP supported. As mentioned above, the card needs additional food. But that's not all, since the power supply of the desktop computer system unit itself must be powerful – at least 750W. The video card has a high TDP of 225W, and this also needs to be borne in mind when thinking about the configuration of the cooling system of the system unit.

Finally, it is worth warning the reader that the device is oversized – 297x140x58mm in size and takes up three slots.

Advantages

• technical process 7nm;
• high performance with this configuration.
• a combination of gaming and computing components;
• HDMI 2.0b;
• fast memory;
• price-performance ratio;
• attractive appearance.

disadvantages

• overall, takes up three slots;
• high TDP.

PNY Quadro P2000 PCI-E 3.0 5120Mb 160bit HDCP

Rating: 4.7

The 'budgetary' part of the review is continued by a video card with a slightly different specialization and focus – less for playing and more for work, although the amount of memory in it is still decent, and it costs a lot. Manufacturer – American company PNY Technologies. Half of the video cards in our review will be of this particular brand, since for the professional environment of designers and designers, as of 2020, they are objectively the best.

The manufacturer does not specify the exact dimensions of the video card, but this is not important, since it is objectively compact and takes up only one slot. Even the cooling system looks modest with just one small fan on the heatsink. All this is a consequence of good optimization, low power consumption and a very low TDP of 75W, which simply cannot be compared with the previous model.

The card is based on the GP106GL graphics chip created according to the 16nm process technology. It is connected to the motherboard via PCI-E 16x 3.0 slot and does not require additional power supply. The volume of the installed video memory is 5GB of GDDR5 type on a 160-bit bus.

The device supports simultaneous connection of up to 4 monitors. Standard interfaces – four DisplayPort 1.4. There is HDCP support. The set includes adapters for other types of interfaces, but professionals confidently criticize them and recommend replacing them with more adequate ones. The maximum supported resolution of the monitor is 5120×2880.

Now, briefly about the capabilities of the device in terms of mathematical calculations. There are 1024 universal processors in the card. 64 texture units and 40 rasterization units. 16x anisotropic filtering. FSAA grade – up to a maximum of 64.

This model supports the following standards and technologies: DirectX 12, OpenGL 4.5, OpenCL 1.2, CUDA 6.1, Vulkan.

Summing up on this model, we can say that users notice a noticeable increase in performance under Solidworks and stability when switching to work with this video card. Also, users involved in design activities are almost unanimous in the fact that graphics and previews are drawn much faster in CAD applications. Photo artists will appreciate the quality of work with raster graphics in 4K resolution.

Advantages

• the technical process of creating a processor – 16nm;
• compact;
• moderate power consumption;
• does not require additional power supply, in addition to that obtained from the motherboard slot;
• high performance and stability for Solidworks;
• little heat dissipation;
• overall performance to price ratio.

disadvantages

• there are minor driver failures (rarely and do not affect the overall performance);
• there are complaints about poor-quality complete adapters for monitors.

PNY Quadro 5000 513Mhz PCI-E 2.0 2560Mb 3000Mhz 320 bit DVI

Rating: 4.7

And the first part of the review will end with a relatively inexpensive professional video card, also made by PNY, only slightly different and less pretentious.

The secret here is that the capabilities of the card itself do not have to be something limiting for professional activities. Graphics cards are often used in tandem with standalone accelerators, which are not graphics cards themselves, but when combined with them for complex graphics and scientific computing can potentiate the effect many times over. We will consider an example of such a discrete accelerator in the second part of the article.

So, PNY Quadro 5000 is based on the GPU of the same name from nVidia codenamed GF100 (clock frequency – 513MHz). Produced according to the 40nm process technology. Plugs into PCI-E 16x 2.0 motherboard slot. Although the technical process here is a generation older than some neighboring models, the chip heats up moderately – at the TDP level of 152W. The card needs additional power, for this a 6pin connector is provided.

The video card supports up to two monitors. Standard interfaces – two DisplayPort, one DVI-I, and even VESA Stereo. The maximum possible monitor resolution is 2560×1600. The video memory in the card is 2GB. Memory type – GDDR5, runs at 3GHz. The memory bus is 320-bit. RAMDAC frequency – 400MHz. Full support for SLI / CrossFire is built in.

A few words about the math block. In this model, we have 352 arithmetic logic units, the fifth version of shaders, 16x anisotropic filtering. The FSAA limit is 64 times.

Like all modern graphics cards, PNY Quadro 5000 has built-in support for OpenGL 4.0 and DirectX 11.

We emphasize once again that against the background of the models described above, the characteristics of the PNY Quadro 5000 may look 'pale', but, again, in combination with a discrete accelerator, this model is capable of a lot, both in terms of creating the most complex polygonal 3D graphics, and for extremely complex calculations.

Advantages

• reliability and susceptibility to many systems;
• efficient work in tandem with discrete accelerators;
• relatively low power consumption;
• moderate TDP;
• relatively affordable price.

disadvantages

• not the highest maximum resolution;
• relatively limited performance due to obsolescence.

The best premium graphics cards for graphics and computing

The magazine's second selection of the best graphics cards for graphics and computing presents the brightest models that can impress even the layman. Here our experts no longer made price restrictions, but considered only top-end video cards available on the Russian market.

Sapphire FirePro S9150 PCI-E 3.0 16384Mb 512 bit

Rating: 4.9

Let's start the premium part of the review of the best video cards according to the version for working with graphics and computing, a model that is highly valued by professionals. This device has an even narrower specialization than all the others – this is the so-called server video card, and it has a serious emphasis on computing capabilities.

At the heart of this graphics accelerator is the powerful FirePro S9150 processor of the same name from AMD, codenamed Hawaii XT. Created according to the 28nm process technology. Able to generate up to 5.07 teraflops of computing power. In terms of one watt of power consumption, the video card produces 21.6 gigaflops, which is a record value in this class.

Consider the rest of the technical characteristics of the device. Installed amount of RAM – 16384MB, memory type – GDDR5 on 512-bit bus. The memory is 'fast' – operates at 5GHz. Bandwidth – 320GB / sec.

The mathematical module of the video card is characterized by 2816 universal processors, 176 texture units, 64 rasterization units, and the fifth version of shaders. OpenGL 4.4 and DirectX 12 are supported.

Although this device is characterized by a relatively moderate power consumption, it still does not have enough power from the slot for operation. Therefore, to connect additional electricity, it has two 8-pin and 6-pin sockets.

In operation under significant load, the video card heats up significantly. Its TDP is estimated at 235W. But this is mostly due to the passive cooling system. For the server version, the developers decided to implement just this approach, hoping that the heat removal will be carried out by a more powerful cooling system of the server itself.

Advantages

• record computing power in this class;
• reliable and durable;
• server room;
• relatively moderate energy consumption;
• more or less affordable price for such computing power.

disadvantages

• very hot due to passive cooling.

PNY Tesla K40 745Mhz PCI-E 3.0 12288Mb 6000Mhz 384 bit Silent

Rating: 4.8

The next striking model in our review is the PNY Tesla K40. Its price is capable of impressing even sophisticated gamers, but when it comes to working with really serious calculations, then such an investment will definitely pay off, especially if we are talking not about an individual user, but about a company.

And let's say right away that this is not quite a video card, but an accelerator, which was discussed above. Simply put, it is the math center for a graphics card. Such a device needs to be paired with a video card (Tesla works best with professional video cards of the Quadro family), and then the graphics stream will go through the video card with all the expected amazing effect in terms of performance, speed, stability. So you can get a 'monstrous', in a good way, computing power of 3-3.5 teraflops at home. Let us emphasize – at home, and not in a data center, where the capacities are even higher. Of course, such solutions are simply blasphemous to use for computer games. The lot of PNY Tesla K40 is the processing of high-quality video material in 4K resolution or even higher, but to a greater extent – the most complex calculations within the framework of deep learning. For rendering complex 3D graphics in Maya, 3ds Max and similar tasks, it is better to look at another model.

Let's go directly to the technical parameters of the card. The basis of the video card is the eponymous Tesla K40 graphics processor codenamed GK110B from nVidia. Manufactured according to the 28nm process technology. Operates at 745MHz. The processor architecture contains 2880 arithmetic-logical units.

The graphics accelerator supports DirectX 11.2 technology. It has 12GB of GDDR5 video memory. The memory is 'fast' – it operates at 6GHz, the bus is 384-bit. SLI / CrossFire support has been implemented.

Advantages

• power;
• high professional class;
• exceptional performance and impact in deep learning;
• reliability and stability.

disadvantages

• less suitable for graphics than all previous models.

PNY Quadro RTX 6000 PCI-E 3.0 24576Mb 384 bit (VCQRTX6000-PB)

Rating: 4.8

And the review of the best, according to the magazine, professional video cards for working with graphics and computing is completed by the most powerful and productive model on the market in 2020, which even some people are scared to take in their hands because of the insane high cost. Indeed, the price tag goes over 400 thousand rubles and reaches half a million at some points of sale. But, this is the objective price for a high professional level.

Let's write down the main technical characteristics of the video card. The processor is built on the Turing architecture, its code name is TU102. Created on a super-thin 12nm process technology. Operates at a standard clock frequency of 1.44GHz. The mode of the standard acceleration of the processor within the Boost technology is supported up to 1.77 GHz. As much as 24GB of RAM is installed in it. The most advanced memory type to date is GDDR6. The memory bus is 384-bit. The memory operates at an unprecedented speed of 14 GHz and delivers 672 GB per second of bandwidth.

Translating all these parameters into a more applied language, we can say that the video card in combination with all the rest of the hardware in the computer with adequate power is capable of providing texturing speeds of up to 509.8 gigatexels per second.

Physically, the video card is 267mm long and connects to a PCIe 3.0 x16 motherboard slot. Such exorbitant power, of course, requires additional power. For this, the card itself has two slots – 6-pin and 8-pin.

To connect monitors, the card provides 4 DisplayPort interfaces, and in addition there is a USB Type-C port.

Advantages

• exceptional quality of production;
• 12nm;
• exclusive technologies;
• 24GB of RAM;
• the highest memory speed;
• CPU acceleration is supported;
• high efficiency in deep learning.

disadvantages

• many experts consider the price to be very high.

---

Attention! This rating is subjective and does not constitute an advertisement and does not serve as a purchase guide. Before buying, you need to consult with a specialist.