NVIDIA announced their next-gen RTX 3000 video cards yesterday

Yesterday NVIDIA announced the release of their next generation of RTX video cards, the 3000 series. NVIDIA is one of the two primary makers of mainstream video cards. The other being AMD. As GIS folks, you may ask why is this important to me? That is a great question. So let me answer that.

As we all know GIS is graphics intensive. Our maps, graphs, tables, and reports display a lot of data in multiple colors, layers, and transparency. That is just the traditional uses and outputs of GIS. Now we also have lidar point clouds, multispectral raster, and 3D scenes. All of these are putting even more strain on our computer systems to render these on our screens.

New GIS applications such as ArcGIS Pro are now designed to leverage the capabilities of a dedicated graphics card to not only render these but to also use the GPU to assist with some analysis as well. This expanded use of graphics cards will only continue to increase as more computers include them and the card's capabilities go up. 

So without getting too technical, what benefits can you expect from these new video/graphics cards from NVIDIA.

• GPUs build on new Ampere architecture which provides more and enhanced tensor cores and more CUDA cores for much better performance (up to 2 times the current RTX 2000 series cards)
• An increased amount of VRAM using GDDR6 and GDDR6X
• Much faster GDDR6X memory in higher-end models. DDR6X is reported to double the data rate per clock over the more traditional GDDR6. This should make it much faster when supporting 3D renderings, multispectral imagery, and AI. 
• Support for HDMI 2.1 which allows for up to 8K graphics
• Improved thermal solutions to reduce throttling caused by high temps and allow for better overclocking
• Support for PCIe 4.0 which allows for faster and large data transfer rates
• Increase per watt of power performance (reportedly 1.9x the current 2000 series)
• Higher frame rates with full-motion video

Unlike previously released NVIDIA cards, the prices for these new 3000 series are much more affordable. Still not cheap but certainly better. Below are the MSRPs NVIDIA announced

RTX 3070

CUDA Cores - 5.888

VRAM - 8GB GDDR6

Price Starting at $499

Release Date - October

RTX 3080

CUDA Cores - 8,704

VRAM - 10GB GDDR6X

Price Starting at $699

Release Date - September 17

RTX 3090

CUDA Cores - 10,496

VRAM - 24GB GDDR6X

Price Starting at $1,499

Release Date - September 24

As you can see these cards are not cheap. But if you compare them to the cost of the current series of RTX cards that they are designed to replace, you do get a lot more performance for the dollars you are spending.

• RTX 2070 Super - $499 (same price as RTX 3070. RTX 3070 is reported to outperform the RTX 2080 TI which is a tier higher)
• RTX 2080 TI - $1,199 (almost twice as much as RTX 3080)
• RTX Titan - $2,500 ($1000 more than the RTX 3090)

These prices are direct from NVIDIA's website as of September 9, 2020. For more information about these cards and what they can do you can click on the video below: 

The other good news I expect from this announcement is the current RTX 2000 series cards should start going down in price both on the retail and used markets. These are still powerful cards and will work well for those that are working with GIS in smaller datasets, not working in 3D, or needing to process large rasters. I am currently running an NVIDIA RTX 2070 Super and is works well for 90% of my GIS work. So this might be a good time for you to upgrade your video card. 

I should probably add that to get the most from these new video cards from NVIDIA, you will most likely need to upgrade your other components such as motherboard and CPU. Only the X570 and B550 motherboards for AMD processors and the Z490 motherboards for Intel processors will support PCIe 4.0 for example. These cards will run in the older PCIe 3.0 slots but the bandwidth will be limited to half what a PCIe 4.0 can support. These new cards also have much higher power requirements. So you will need to have a power supply that can support them as well as the rest of your system. This means you will most likely need at least a 650-watt power supply to successfully run these cards.

I hope this is useful information. We will know more once the hardware is actually available to test and benchmark.