NVIDIA is working on a new generation of graphics cards that will take over the GeForce GTX 10 series. We have already talked to you about the GTX 10 series on articles like these where we reviewed it with all the information that was available at the time. Today, we will focus on the GeForce GTX 2080 and the GeForce GTX 2070, two high-end graphics cards that will take over the GeForce GTX 1080 and the GeForce GTX 2070.
Before actually reviewing both graphics cards, we want to point out something about the name that the new NVIDIA lineup could have. Nothing is set in stone just yet, but there are two possible names: GeForce GTX 11 and GeForce GTX 20. The former sounds more likely but suggests a minor evolution compared to the latter name, so I personally think that the brand will go for GeForce GTX 20.
Having said this, let’s begin. We hope you like this article and, as usual, we invite you to leave your questions in the comments section.
GeForce GTX 2080 and GeForce GTX 2070: new architecture
With these graphics cards, NVIDIA will switch to a new architecture that will be a successor to Pascal, which is currently being used on the GeForce GTX 10 and is based on 14 nm and 16 nm processes.
According to the latest news, NVIDIA will use the Turing architecture on its new graphics cards. This will be a “minor” evolution compared to the Volta architecture used on the GeForce GTX TITAN V and on the Tesla and Quadro series.
The GV100 GPU is fabricated on a 14 nm process, and it has 5,120 shading units, 320 TMUs and 120 ROPs. The version found on the GTX TITAN V has 640 tensor cores that are used for artificial intelligence and deep learning.
It is possible for Turing to take the same route, meaning that the GT100 GPU should have specifications that are very similar to the GV100’s. However, the GT100 will be fabricated on a 12 nm process and its architecture will be improved, which will have an effect performance and power consumption.
This means that this new generation will not only offer a better performance from raw power (more shading units, TMUs and ROPs) but also an improvement in terms of silicon and efficiency. Additionally, the GT100 will use a GDDR6 memory to get a better bandwidth (only seen on high-end models) and, according to some rumors, it is possible that NVIDIA uses a specific amount of tensor cores to power its ray tracing technology for gaming.
GeForce GTX 2080, a thorough look
This graphics card (also known as GTX 1180) will be NVIDIA’s new flagship. It will be based on the GT104 GPU. According to some reports that we were able to see, it will have:
- GV104 GPU (Volta architecture) fabricated on a 12 nm process
- 3.584 shading units at 1,405-1,582 MHz in normal and turbo mode, respectively
- 224 TMUs
- 64 ROPs
- A 256-bit bus
- A 16 GB GDDR6 memory with 12 GHz of effective data-rate
- A 200 W TDP
- Possible price point: $699
- You would need 1×6-pin and 1×8-pin connectors.
In terms of performance, it should top the GeForce GTX 1080 Ti by an average of 30-40%, which is really good if we take into account that the GTX 1080 Ti has a 250 W TDP and a $940 price point.
The improvement in terms of power and efficiency are clear, but what about the tensor cores? Although there is no official information yet, if NVIDIA decides to keep the amount of tensor cores to power the GTX 2080’s ray tracing, I think it could have a total of 480 (the GTX 2080 Ti would be the only option for consumers that would have 640 tensor cores).
As for the price, $699 sounds like a grounded number. However, given the rollercoaster that the market is right now due to the high demand caused by cryptocurrency mining, it is possible to see a higher price point (some rumors put it at $799).
GeForce GTX 2070, a high-end graphics card at a reasonable price
We now switch to what could be NVIDIA’s best-selling graphics card within its generation. The GeForce 2070 will also be based on a smaller version of the GT104 GPU. These could be its specifications:
- GT104 GPU (Turing architecture, which is derived from Volta) fabricated on a 12 nm process
- 2,688 shading units at 1,500 MHz in normal mode and 1,800 MHz in turbo mode
- 168 TMUs
- 64 ROPs
- A 256-bit bus
- A GDDR6 memory of around 8-16 GB at 12 GHz
- A 140-160 W TDP
- Price: $499
- You would need an 8-pin connector.
In terms of raw performance, it should be quite on par with the GTX 1080 Ti. Like above, this is a positive aspect because it shows a major improvement on raw performance when compared to the GTX 1070 and a much better efficiency, as it reduces the TDP by about 100 W.
The GTX 2070’s price would be incredible, as it stands at almost half the price of a GTX 1080 Ti, and both cards virtually perform the same. However, the GTX 2070 might notably outshine the 1080 Ti in terms of ray-tracing gaming due to the amount of tensor cores. The GeForce GTX 2070 should have a total of 320 tensor cores.
Again, we can expect the same thing that we discussed concerning the GeForce GTX 2080: the price could end up being higher if the demand from the cryptocurrency mining sector rises again.
The GeForce GTX 2080 will be a high-performance graphics card focused on full 4K gaming. It will also be NVIDIA’s response to ray-tracing technology. This graphics card will certainly be an improvement over the GTX 1080 Ti, prompting the arrival of subpar units onto the market.
As for the GeForce GTX 2070, it will also be a high-performance graphics card for a seamless 4K gaming experience. However, the difference in terms of specifications in comparison to the GTX 2080 might be felt when maxing out settings and using the ray-tracing technology. The GTX 2070 will be the best option for those who have a tight budget but still want to get a powerful last-gen unit. It will also prompt the GTX 1080 Ti to be seen on thrift stores.
As we have told you in previous posts, if you are thinking about upgrading your graphics card, the best thing for you to do is to wait for NVIDIA’s newest graphics card because this will mean that there will be new products on the market and on thrift shops.