When you are configuring a computer, the question you should always ask is, 'What will I be using this computer for?' This question applies to the video card as much as any other part of the system. For many applications - like gaming, 3D design, and animation - video card selection is just as important for performance as picking the right CPU. If you simply need a stable and solid video card for 2D graphics, everyday display and light tasks such as DVD playback then even our most basic models will be sufficient.

When it comes to more intensive tasks, there are generally two types of video cards that you need to look at:

Mainstream / Gaming Cards - Examples include the NVIDIA GeForce line and the AMD Radeon HD line. Among our customers, these are the most common video card choices. These cards are excellent for normal 2D / 3D applications and computer gaming. Most current models will support at least three monitors, and can be paired together to either provide support for more screens or to improve 3D performance in games. While most video cards have fans for cooling there are some which are passively-cooled and therefore silent - they make great choices if you are trying to keep a computer as quiet as possible.

Professional / 3D Design Cards - Examples include the NVIDIA Quadro line and the AMD FirePro line. These cards are particularly useful in CAD, animation and related fields where their custom drivers and enhanced vendor support are major advantages. Professional-grade video cards are more expensive, but they are built specifically to meet the needs of professionals using programs like AutoCAD, Solidworks, 3DS Max, and similar industry-specific applications. There are also some professional-grade cards that are aimed at multiple monitor support rather than advanced 3D performance; if you need to run more than four monitors from a single computer, one of these cards might be appropriate.

Onboard video is our catch-all for various integrated GPU options. If you have further questions about which specific onboard video solution is present on your configuration or quote, please feel free to contact us.

Onboard video is often a good option for systems where 3D graphics performance is not a major concern. Most modern onboard graphics will provide a platform suitable for multimedia and office use, along with reduced power usage, heat, and noise compared to most dedicated graphics cards. If you do take advantage of onboard video, expect it to use a small portion of the system RAM for storing video data.

If you have any questions, be sure to contact our sales consultants at sales@pugetsystems.com. They will be happy to help you find the right video card for your specific needs.

Puget Professional Advice:

William George (Puget Labs Technician) Says: The newest generation of onboard video on today's motherboards is finally reaching a good level of performance and features. In fact, it isn't generally on the motherboard anymore: both Intel and AMD have integrated graphics into parts of their mainstream processor lines! On the Intel side that isn't designed for gaming, but can handle all manner of video acceleration (DVD, Blu-ray, etc) very well and can handle moderate 3D usage in other applications. AMD has done one better, combining Radeon graphics with some of their processors - thus making chips capable of entry-level gaming without need for a dedicated video card. The one exception to this is the add-on video chips found on some server motherboards. Those are extremely basic, designed to simply allow access to a command-line interface for occasional server administration. They are often not even capable of handling a modern graphical user interface well, and should be avoided in any situation where direct access to a system will be commonplace.

The GeForce GT 1030 is an entry-level dedicated video card from NVIDIA. It is generally a step up from integrated graphics, which are built into some CPUs from Intel and AMD, but it is more commonly used for basic video output on systems that don't already have that functionality built-in. It can also be effective for providing two additional monitor outputs, if the applications being run are not dependent on GPU performance.

EVGA's version of this card is passively-cooled, meaning it has no fan... but the heatsink is quite tall, taking up the slot next to the video card as well. As alluded to above it has two video outputs: HDMI and DVI, which can be used at the same time. It is also low-profile, allowing use in some cases which cannot accommodate a full-height card.

The NVIDIA Quadro P620 is an entry-level, Pascal-based professional graphics card. It offers a small performance boost over the earlier P600 model, though the amount of video memory and monitor outputs remained the same. That lets the P620 run up to four displays, though with the limited video memory the resolution of each screen should be kept at a moderate level.

The P620 is a half-height card, but comes with a full-height bracket for use in normal cases. As alluded to previously, this card has four Mini DisplayPort connections - and adapters to full size DisplayPort are included as well. Adapters to DVI or HDMI can be purchased separately if needed, but please note that if you are using DVI adapters you will be limited to a lower maximum resolution per screen (1920x1200). A wide range of standards are also supported by this card: DirectX 11, OpenGL 4.4, CUDA 5.0, and OpenCL (to name a few).

Puget Professional Advice:

Jeff Stubbers (Technology Consultant Lead) Says: Nvidia lists these (4) Mini DisplayPort 1.4 outputs as supporting 4x 4096x2160 @ 60Hz, or 4x 5120x2880 @ 60Hz. This card comes with (4) Mini DP, to DP adapters.

 Most Popular Part!

NVIDIA's GeForce GTX 1660 6GB is a mid-range GPU, just a little slower than the 1660 Ti. While it is based on Turing technology, the older naming and numbering scheme is still used because this card is missing some of the more advanced features that are found on the RTX 2000-series. That means no hardware ray tracing (RT cores) or machine learning inference (Tensor cores) - but for more traditional applications which don't use those capabilities the GTX 1660 is a fine option.

This particular model from EVGA features a dual-fan cooling layout. This is great for keeping a single card running cool and quiet, but results in most of the heat the card generates being pumped back into the computer. That makes these a poor choice for use in multi-GPU systems, especially in configurations where there will be no space between each card. When installing more than one video card in a system, models with a blower-style fan and rear heat exhaust will be much better.

Finally, in regards to connectivity, the EVGA GTX 1660 supports up to three monitors. Most of the video output standards from the past decade are included, with one connection each for DVI-D, HDMI, and DisplayPort.

Puget Professional Advice:

William George (Puget Labs Technician) Says: For those curious about the details, the GTX 1660 differs from the 1660 Ti in two primary ways: ~9% fewer stream processors (1408 vs 1536) and ~33% slower memory (GDDR5 instead of GDDR6). Both of these are solidly mid-range video cards, for their generation, but there is a measurable difference in performance... and price.

 Most Popular Part!

NVIDIA's GeForce GTX 1660 Ti 6GB is a mid-range GPU, replacing the older GTX 1060. While it is based on Turing technology, the older naming and numbering scheme is still used because this card is missing some of the more advanced features that are found on the RTX 2000-series. That means no hardware ray tracing (RT cores) or machine learning inference (Tensor cores) - but for more traditional applications which don't use those capabilities the GTX 1660 Ti is a strong contender.

This particular model from Asus is a short-depth design, with a single-fan open-air cooling system. That is great for fitting in smaller cases and keeping a single card cool, but results in most of the heat the card generates being pumped back into the computer. That makes this a poor choice for use in multi-GPU systems, especially in configurations where there will be little or no space between each card. In that situation, a video card with a blower-style fan and rear heat exhaust will be much better.

Finally, in regards to connectivity, the GTX 1660 Ti supports up to four monitors. A wide range of outputs are provided: One DVI-D, two HDMI 2.0b, and one DisplayPort 1.4.

The NVIDIA Quadro P1000 is a low-profile, professional-grade video card. It has mid-range 3D performance, and specializes in multi-display situations thanks to four separate outputs. That means you can drive up to four displays by using only one card slot and driver, or stack multiple P1000 cards in a system to expand that to eight or even more outputs. It is perfect for stock/day traders or any other professional users who need multiple monitor support without high-end performance 3D graphics.

As described above, this card has four Mini DisplayPort connections - and includes adapters to full size DisplayPort as well. Adapters to DVI or HDMI can be purchased separately if needed, but please note that if you are using DVI adapters you will be limited to a lower maximum resolution per screen (1920x1200).

Puget Professional Advice:

William George (Puget Labs Technician) Says: This card has the same outputs as the less expensive P600, but since it has double the video memory (4GB instead of 2) it is much better suited to running several monitors at the same time.

Jeff Stubbers (Technology Consultant Lead) Says: This is a great card choice when you need to support many monitors for tasks like stock or commodities trading. This is a moderately powerful graphics card, but if you have higher-end 3D demanding programs that you intend to run such as CAD or 3D modeling, then you would want to choose a more powerful graphics card. When choosing a graphics card, it is always important to make sure the monitor(s) that you intend to use will have compatible connection types with the graphics card outputs. This card can support (4) 4K displays (4096x2160 @ 60Hz) with HDR color using its (4) Mini DisplayPort 1.4 connections. This card comes with (4) Mini DP to DP adapters. Adapters from Mini DP to other types can be purchased separately. Keep in mind that if you use the Mini DP to DVI adapters, the maximum resolution from those DVI adapters will be 1920x1200.

The AMD Radeon Pro WX 5100 is among the first AMD workstation graphics cards to use the Radeon branding name. Compared to its predecessor, the FirePro W5100, it features both more stream processors and higher clock speeds - resulting in significantly more computational performance. While not as fast as the next model up, the WX 7100, this card is less expensive and uses less power. It is equipped with 8GB of video memory and support up to four monitors via DisplayPort outputs.

As with other RadeonPro professional-grade cards, the WX 5100 supports DX12, OpenGL, and OpenCL. Full 10-bit-per-channel color is also available, when paired with supported monitors. In terms of size and power requirements, this card is single-width and does not need a supplemental power connection.

No quick information avaialble. Click to see full part information.

No quick information avaialble. Click to see full part information.

 Most Popular Part!

The NVIDIA Quadro P2000 is a mid-range, Pascal-based professional graphics card. It represents a solid upgrade from the previous M2000 model, with moderate speed increases in most areas and 25% more video memory. A wide range of standards are also supported by this card: DirectX 12, Vulkan, OpenGL 4.5, CUDA, and OpenCL (to name a few).

In addition to solid performance for professional 3D applications, the Quadro P2000 also supports higher resolution monitors than previous generations. It can run up to four 5K screens at the same time via DisplayPort 1.4 outputs. HDMI or DVI monitors can be connected through adapters, available for purchase separately, but when using such adapters the maximum resolution will be reduced substantially if not using an DP to HDMI 2.0 Active adapter.

Puget Professional Advice:

William George (Puget Labs Technician) Says: This is a great mid-range Quadro card, which is ideal for applications which need a "professional grade" video card but don't require cutting-edge 3D performance. It also supports 10-bit color, so it is a great choice for Photoshop users with high-end, 10-bit capable monitors. If you don't fall into one of those categories, though, a GeForce card around the same price will give more raw 3D performance in most situations... or of course a more powerful Quadro model.

Jeff Stubbers (Technology Consultant Lead) Says: This is a great choice if you are looking for a mid-range 30-bit graphics card (10-bit per channel RGB) for most accurate colors when pairing with a 10-bit monitor. The manufacturer lists this card as supporting up to HDR 5120 x 2880 resolution @ 60Hz with each of its (4) DisplayPort 1.4 outputs.

The AMD Radeon Pro WX 7100 is among the first AMD workstation graphics cards to use the Radeon branding name. Compared to its predecessor, the FirePro W7100, it features both more stream processors and higher clock speeds - resulting in significantly more computational performance. Performance-wise it lands between the WX 5100 and WX 9100 cards, while only taking up a single expansion slot. It is equipped with 8GB of video memory and support up to four monitors via DisplayPort outputs.

As with other RadeonPro professional-grade cards, the WX 7100 supports DX12, OpenGL, and OpenCL. Full 10-bit-per-channel color is also available, when paired with supported monitors. One 6-pin PCI-E power connector is required for proper operation.

 Most Popular Part!

NVIDIA's GeForce RTX 2070 is part of the first wave of graphics cards based on the Turing GPU architecture. This generation represents only an incremental improvement in general graphics performance, making the RTX 2070 a little faster than the previous generation GTX 1080 (and slightly more expensive as well). The big thing that this new GPU architecture brings to the table is forward-looking features not found on previous GTX cards: dedicated RT cores for ray tracing and Tensor cores for machine learning inference. The RT cores are brand-new, while the Tensor cores have existed on previous Titan and Quadro cards - but this is the first time they've come to the mainstream GeForce series. This specialized hardware will allow for new functionality in games and GPU-accelerated applications, but will require such programs to be written specifically to utilize them in order to see much benefit.

This version of the RTX 2070 features a dual-fan cooling layout. This is great for keeping a single card running cool and quiet, but results in most of the heat the card generates being pumped back into the computer. That makes these a poor choice for use in multi-GPU systems, especially in configurations where there will be no space between each card. When installing more than one video card in a system, models with a blower-style fan and rear heat exhaust will be much better.

Finally, in regards to connectivity, the RTX 2070 supports up to four monitors. Unlike the higher-end RTX 2080 and 2080 Ti, it still has one dual-link DVI output - along with one HDMI 2.0b and two DisplayPort 1.4. VirtualLink - a specialized Type C port combining DisplayPort, USB 3.1, and power - is also included, for single-cable connectivity to next-generation VR headsets.

NVIDIA's GeForce RTX 2070 is part of the first wave of graphics cards based on the Turing GPU architecture. This generation represents only an incremental improvement in general graphics performance, making the RTX 2070 a little faster than the previous generation GTX 1080 (and slightly more expensive as well). The big thing that this new GPU architecture brings to the table is forward-looking features not found on previous GTX cards: dedicated RT cores for ray tracing and Tensor cores for machine learning inference. The RT cores are brand-new, while the Tensor cores have existed on previous Titan and Quadro cards - but this is the first time they've come to the mainstream GeForce series. This specialized hardware will allow for new functionality in games and GPU-accelerated applications, but will require such programs to be written specifically to utilize them in order to see much benefit.

This particular variant has a single-fan "blower" style cooler. This won't be as quiet under load as a dual-fan design, when comparing individual cards, but since it exhausts most of the card's heat out the back of the computer it is much better for use in multi-GPU configurations.

Finally, in regards to connectivity, this variant has outputs for three monitors: one HDMI 2.0b and two DisplayPort 1.4. VirtualLink - a specialized Type C port combining DisplayPort, USB 3.1, and power - is also included, for single-cable connectivity to next-generation VR headsets.

 Most Popular Part!

NVIDIA's GeForce RTX 2070 is part of the first wave of graphics cards based on the Turing GPU architecture. This generation represents only an incremental improvement in general graphics performance, making the RTX 2070 a little faster than the previous generation GTX 1080 (and slightly more expensive as well). The big thing that this new GPU architecture brings to the table is forward-looking features not found on previous GTX cards: dedicated RT cores for ray tracing and Tensor cores for machine learning inference. The RT cores are brand-new, while the Tensor cores have existed on previous Titan and Quadro cards - but this is the first time they've come to the mainstream GeForce series. This specialized hardware will allow for new functionality in games and GPU-accelerated applications, but will require such programs to be written specifically to utilize them in order to see much benefit.

This particular variant, from EVGA, features a dual-fan cooling layout. This is great for keeping a single card running cool and quiet, but results in most of the heat the card generates being pumped back into the computer. That makes these a poor choice for use in multi-GPU systems, especially in configurations where there will be no space between each card. In those cases, a video card with a blower-style fan and rear heat exhaust will be much better.

Finally, in regards to connectivity, the RTX 2070 supports up to four monitors. Unlike many other high-end RTX cards, it still has one dual-link DVI output- along with one HDMI 2.0b and two DisplayPort 1.4 outputs. VirtualLink is also included: a specialized Type C port combining DisplayPort, USB 3.1, and power for single-cable connectivity to next-generation VR headsets.

 Most Popular Part!

The NVIDIA Quadro P4000 is a high-end "professional grade" video card, based on the Pascal GPU architecture and offering solid performance for 3D applications. Because of the Pascal architecture it is more efficient than previous Quadro cards of this caliber, with very good single-precision floating point performance as well. This card is not very good at double-precision calculations, though, being handicapped in that regard compared to more specialized models like the GP100. It has 1792 stream processors, a 256-bit memory bus, and 8GB of memory - giving it overall roughly half of the capability of the P6000... but in a single slot card, rather than the dual slot size of the P5000 and P6000.

In addition to good performance for professional 3D applications, the Quadro P4000 also supports several high-res monitors. It can run up to four 5K screens at the same time, thanks to the use of DisplayPort 1.4 outputs - which can be adapted to older video standards as well, if needed (only one DVI adapter is included). A wide range of standards are supported by this card: DirectX 12, OpenGL 4.5, CUDA, and OpenCL (to name a few).

Puget Professional Advice:

William George (Puget Labs Technician) Says: This is the fastest single-slot video card of its generation, and a good choice for professional applications which need a moderate amount of 3D performance across multiple monitors.

Jeff Stubbers (Technology Consultant Lead) Says: This graphics card offer (4) DP 1.4 outputs. Nvidia lists this card as supporting (4) 4096x2160 @ 120Hz, or (4) 5120x2880 @ 60Hz.

NVIDIA's GeForce RTX 2080 is one of the first graphics cards based on the Turing GPU architecture. This generation represents an incremental improvement in general graphics performance, as usual, along with the addition of two major features: dedicated RT cores for ray tracing and Tensor cores for machine learning inference. The RT cores are brand-new, while the Tensor cores have existed on previous Titan and Quadro cards - but this is the first time they've come to the mainstream GeForce series. This specialized hardware will allow for new functionality in games and GPU-accelerated applications, but will require such programs to be written specifically to utilize them in order to see much benefit. General performance should be about 20-30% faster than the previous GTX 1080, in situations where the video card is the limiting factor.

This particular variant has a single-fan "blower" style cooler. This won't be as quiet under load as a dual-fan design, when comparing individual cards, but since it exhausts most of the card's heat out the back of the computer it is much better for use in multi-GPU configurations.

Finally, in regards to connectivity, the RTX 2080 has dropped DVI output entirely. There are still enough outputs for three monitors, though: one HDMI 2.0b and two DisplayPort 1.4. VirtualLink - a specialized Type C port combining DisplayPort, USB 3.1, and power - is also included, for single-cable connectivity to next-generation VR headsets.

 Most Popular Part!

NVIDIA's GeForce RTX 2080 is one of the first graphics cards based on the Turing GPU architecture. This generation represents an incremental improvement in general graphics performance, as usual, along with the addition of two major features: dedicated RT cores for ray tracing and Tensor cores for machine learning inference. The RT cores are brand-new, while the Tensor cores have existed on previous Titan and Quadro cards - but this is the first time they've come to the mainstream GeForce series. This specialized hardware will allow for new functionality in games and GPU-accelerated applications, but will require such programs to be written specifically to utilize them in order to see much benefit. General performance should be about 20-30% faster than the previous GTX 1080, in situations where the video card is the limiting factor.

This variant has a dual-fan cooling layout. That is great for keeping a single card running cool and quiet, but results in most of the heat the card generates being pumped back into the computer. That makes these a poor choice for use in multi-GPU systems, especially in configurations where there will be no space between each card. In those cases, a video card with a blower-style fan and rear heat exhaust will be much better.

Finally, in regards to connectivity, the RTX 2080 has dropped DVI output entirely. There are still enough outputs for four monitors, though: one HDMI 2.0b and three DisplayPort 1.4. VirtualLink - a specialized Type C port combining DisplayPort, USB 3.1, and power - is also included, for single-cable connectivity to next-generation VR headsets.

 Most Popular Part!

The NVIDIA Quadro RTX 4000 is a professional Turing-based graphics card, offering good overall performance along with specialized ray tracing and machine learning inference hardware (RT cores and Tensor cores, respectively). While not sporting as much onboard memory as higher-end Quadro models, the RTX 4000's 8GB of VRAM is on par with similarly priced GeForce models. A rarity among modern video cards, the RTX 4000 uses a single-width cooler - allowing it to take up less space in systems which have a lot of other expansion cards or fit easily into smaller form factor computers.

To compliment its focus on professional 3D applications, the Quadro RTX 4000 supports up to four monitors at the same time. Three can be connected via DisplayPort outputs, and the fourth through VirtualLink. That is a newer connector - a specialized Type C port combining DisplayPort, USB 3.1, and power. It was conceived as a single-cable connectivity to next-generation VR headsets, but can also be used (with the included DisplayPort adapter) to connect a normal monitor.

NVIDIA's GeForce RTX 2080 Ti is the highest-end GeForce model based on the Turing GPU architecture. This generation represents an incremental improvement in general graphics performance, as usual, along with the addition of two major features: dedicated RT cores for ray tracing and Tensor cores for machine learning inference. The RT cores are brand-new, while the Tensor cores have existed on previous Titan and Quadro cards - but this is the first time they've come to the mainstream GeForce series. This specialized hardware will allow for new functionality in games and GPU-accelerated applications, but will require such programs to be written specifically to utilize them in order to see much benefit. The 2080 Ti should be about 20-30% faster than the previous GTX 1080 Ti in normal GPU-limited situations, and even higher in applications that utilize RT and/or Tensor cores.

This particular variant has a single-fan "blower" style cooler. This won't be as quiet under load as a dual-fan design, when comparing individual cards, but since it exhausts most of the card's heat out the back of the computer it is much better for use in multi-GPU configurations.

Finally, in regards to connectivity, the RTX 2080 Ti has dropped DVI output entirely. There are still enough outputs for three monitors, though: one HDMI 2.0b and two DisplayPort 1.4. VirtualLink - a specialized Type C port combining DisplayPort, USB 3.1, and power - is also included, for single-cable connectivity to next-generation VR headsets.

 Most Popular Part!

NVIDIA's GeForce RTX 2080 Ti is the highest-end GeForce model based on the Turing GPU architecture. This generation represents an incremental improvement in general graphics performance, as usual, along with the addition of two major features: dedicated RT cores for ray tracing and Tensor cores for machine learning inference. The RT cores are brand-new, while the Tensor cores have existed on previous Titan and Quadro cards - but this is the first time they've come to the mainstream GeForce series. This specialized hardware will allow for new functionality in games and GPU-accelerated applications, but will require such programs to be written specifically to utilize them in order to see much benefit. The 2080 Ti should be about 20-30% faster than the previous GTX 1080 Ti in normal GPU-limited situations, and even higher in applications that utilize RT and/or Tensor cores.

This version of the RTX 2080 Ti features a dual-fan cooling layout. This is great for keeping a single card running cool and quiet, but results in most of the heat the card generates being pumped back into the computer. That makes these a poor choice for use in multi-GPU systems, especially in configurations where there will be no space between each card. When installing more than one video card in a system, models with a blower-style fan and rear heat exhaust will be much better.

Finally, in regards to connectivity, the RTX 2080 Ti has dropped DVI output entirely. There are still enough outputs for four monitors, though: one HDMI 2.0b and three DisplayPort 1.4. VirtualLink - a specialized Type C port combining DisplayPort, USB 3.1, and power - is also included, for single-cable connectivity to next-generation VR headsets.

The AMD Radeon Pro WX 9100 is among the first AMD workstation graphics cards to use the Radeon branding name. Compared to its predecessor, the FirePro W9100, it features both more stream processors and higher clock speeds - resulting in significantly more computational performance. It is equipped with 16GB of HBM2 memory and support up to six monitors via Mini DisplayPort outputs.

As with other RadeonPro professional-grade cards, the WX 9100 supports DX12, OpenGL, and OpenCL. Full 10-bit-per-channel color is also available, when paired with supported monitors.

The NVIDIA Quadro P5000 is a high-end "professional grade" video card, based on the Pascal GPU architecture and offering great performance for 3D applications. Because of the Pascal architecture it is more efficient than previous Quadro cards of this caliber, with very good single-precision floating point performance as well. This architecture is not very good at double-precision calculations, though, falling behind some older generations of Quadro cards in that area. It has 2560 stream processors, a 256-bit memory bus, and 16GB of ECC memory - giving it overall about 2/3rds of the capability of the P6000.

In addition to great performance for professional 3D applications, the Quadro P5000 also supportsup to four 4K monitors at the same time. That connectivity is provided by four DisplayPort outputs, and there is a dual-link DVI output as well for legacy monitors. A wide range of standards are supported by this card: DirectX 12, OpenGL 4.5, CUDA, and OpenCL (to name a few).

The NVIDIA Quadro RTX 5000 is a professional Turing-based graphics card, offering great overall performance along with specialized ray tracing and machine learning inference hardware (RT cores and Tensor cores, respectively). While not sporting as much onboard memory as higher-end Quadro models, the RTX 5000's 16GB of VRAM is still more than any contemporary GeForce model. As with most Quadros, the single-fan cooler on this card is ideally suited for use in multi-GPU systems.

In addition to great performance for professional 3D applications, the Quadro RTX 5000 supports up to four monitors at the same time via DisplayPort outputs. It also has a VirtualLink connector - a specialized Type C port combining DisplayPort, USB 3.1, and power for single-cable connectivity to next-generation VR headsets. Using that counts toward the card's limit of four simultaneous outputs.

NVIDIA's Titan RTX is the Turing-generation entry in the Titan video card line, which straddles the fence between consumer and workstation graphics segments. It looks just like the GeForce RTX 2080 / 2080 Ti Founders Edition cards, except with a gold finish instead of silver. Performance-wise it is a hair faster than the 2080 Ti, and comes equipped with more than double the VRAM: 24GB of fast GDDR6 memory.

Like the GeForce and Quadro cards in this generation, the Titan RTX features both dedicated RT cores for ray tracing and Tensor cores for machine learning inference. The RT cores are brand-new with the Turing architecture, while the Tensor cores debuted on the recent Titan V and Quadro V100 cards. This specialized hardware will allow for new functionality in games and GPU-accelerated applications, but will require such programs to be written specifically to utilize them in order to see much benefit.

In a first for the Titan series, this card features dual cooling fans - again, like the GeForce 2000-series Founders Edition models. This is great for keeping a single card running cool and quiet, but results in most of the heat it generates being pumped back into the computer. That makes these a poor choice for use in multi-GPU systems, especially in configurations where there will be little or no space between each card. When installing more than GPU in a system, cards with a blower-style fan and rear heat exhaust will be much better.

Finally, in regards to connectivity, the Titan RTX has four monitor outputs: one HDMI 2.0b and three DisplayPort 1.4. VirtualLink - a specialized Type C port combining DisplayPort, USB 3.1, and power - is also included, for single-cable connectivity to next-generation VR headsets.

The NVIDIA Quadro RTX 6000 is the first Turing-based Quadro card, offering the best single card performance available in this generation. Because of the Turing architecture it also gains specialized ray tracing and machine learning inference hardware as well: RT cores and Tensor cores, respectively. The more expensive Quadro RTX 8000 is effectively the same card but with double the already massive 24GB of VRAM this model has. As with most Quadros, the single-fan cooler on this card is ideally suited for use in multi-GPU systems.

In addition to extreme performance for professional 3D applications, the Quadro RTX 6000 also supports up to four monitors at the same time. It sports four DisplayPort connections, along with VirtualLink - a specialized Type C port combining DisplayPort, USB 3.1, and power for single-cable connectivity to next-generation VR headsets. Using that counts toward the limit of four simultaneous outputs.

Puget Professional Advice:

William George (Puget Labs Technician) Says: This graphics card is very similar to the NVIDIA Titan RTX in specs, with the same CUDA and RT core counts as well as onboard memory capacity (24GB). The RTX 6000 does have ECC protection on the memory, which is handy, and being in the Quadro line means it may have more specialized drivers and better support for certain high-end applications. The biggest practical difference, though, is the cooling: this card has a fan layout that works great with multiple GPUs in a single workstation or rackmount chassis, while the Titan RTX has a dual-fan design that limits it to use as an individual card.

The NVIDIA Quadro P6000 is the highest-end Pascal-based Quadro card, offering the best NVIDIA single card performance available in this generation. Because of the Pascal architecture it is more efficient than previous Quadro cards of this caliber. General graphics performance and single-precision floating point performance are higher than any Quadro card that came before it, though unlike some older Quadro cards its double-precision floating point performance is limited. It has 3840 stream processors, a 384-bit memory bus, and 24GB of memory with error correction for improved reliability.

In addition to extreme performance for professional 3D applications, the Quadro P6000 also supports up to four monitors at the same time. In terms of connectivity it has one dual-link DVI port and four DisplayPort outputs. A wide range of standards are supported by this card: DirectX 12, OpenGL 4.5, CUDA, and OpenCL (to name a few).

Puget Professional Advice:

William George (Puget Labs Technician) Says: This is very similar to the NVIDIA Titan X (Pascal version), including its handicap in double-precision floating point performance, but with slightly more CUDA cores, twice the memory, and ECC protection. That means it is a very powerful GPU, and as part of the Quadro line it may have more specialized drivers and better support for certain high-end applications. For many users seeking high performance, though, the Titan X will offer similar capabilities for a lower price.

The NVIDIA Quadro RTX 8000 is the highest-end Turing-based Quadro card, offering the best single card performance available in this generation. In fact it has the same GPU as the RTX 6000, but double the amount of onboard memory: a tremendous 48GB! Because of the Turing architecture it also gains specialized ray tracing and machine learning inference hardware as well: RT cores and Tensor cores, respectively. As with most Quadros, the single-fan cooler on this card is ideally suited for use in multi-GPU systems.

In addition to extreme performance for professional 3D applications and a massive amount of VRAM, the Quadro RTX 8000 also supports up to four monitors at the same time. It sports four DisplayPort connections, along with VirtualLink - a specialized Type C port combining DisplayPort, USB 3.1, and power for single-cable connectivity to next-generation VR headsets. Using that counts toward the limit of four simultaneous outputs.

Puget Professional Advice:

William George (Puget Labs Technician) Says: This is effectively a RTX 6000 with double the memory (48 vs 24GB) and a mirror finish on the heatsink shroud. If you need a ton of VRAM, this is the most I've ever seen on a card from NVIDIA (as of early 2019, at least).

The NVIDIA Quadro GV100 is effectively a Tesla V100 but with graphics output capabilities included. That is why it is included under the Quadro banner, but even the model number conveys its relation to the Tesla line. Because of its heritage, the GV100 has unhindered double-precision floating point performance (FP64) - a big advantage over most Quadro and GeForce cards, albeit in a niche area.

The GV100 has 16GB of high bandwidth memory (HBM), and includes NVIDIA's NVLINK technology so that if you use two of these cards they can communicate directly with each other in a much faster way than PCI-Express or SLI would allow. For video output, the GV100 has four DisplayPort connections - all of which can be used at the same time, supporting up to four high-res monitors.