In an interview with PCWorld, AMD reaffirmed its plans to launch the FidelityFX Super Resolution feature for RDNA 2 graphics cards within this year to address NVIDIA’s DLSS. The information comes from the latest Full Nerd Special podcast with Scott Herkelman of AMD as one of the show’s guests.
AMD’s FidelityFX Super Resolution ‘FSR’ for RDNA 2 Radeon RX GPUs Compete with NVIDIA’s DLSS for GeForce RTX GPUs This Year
It seems the rumor mill was a bit optimistic when it was announced that the FidelityFX Super Resolution feature for AMD Radeon RX GPUs based on the RDNA 2 architecture will be released in the spring of 2021. Scott Herkelman (CVP & GM at AMD Radeon) has confirmed that their NVIDIA The DLSS competitor is still up and running this year. Scott also confirmed that the technology will first come to gaming computers with Radeon RX GPUs, but will later be extended to other platforms, such as game consoles that are also powered by RDNA 2 architecture.
NVIDIA started head on with its AI-based supersampling technology back in 2018 when the Turing-based GeForce RTX 20 series was introduced. DLSS 1.0 had a rather rough start and there were not enough games that used the feature, and although users could see an impressive performance increase, it also had a loss of image quality which was often too vague compared to the play at native resolution with standard AA methods. That changed over time and DLSS 2.0 showed the true form of the feature with the still impressive gains, while the image quality was almost the same as the original resolution. From now on, the feature has seen major updates with The Medium and Cyberpunk 2077, which are the latest highlight titles and made possible on 8K on a single graphics card.
The difference DLSS makes makes the NVIDIA GeForce RTX 30 Series graphics cards a league ahead of AMD’s Radeon RX 6000 series.
It is progressing well internally in our laboratory, but it is our commitment to the game community that it should be open, that it should work in all respects and that game developers should accept it. Although progressing well, we still have more work to do and not only internally but also with our game developer partners. We want to introduce it this year. We believe we can do it this year, but at the same time we are working much harder. We need to make sure the image quality is there. We need to make sure that it can scale according to different resolutions. And at the same time that our game developers are happy with what we produce.
– Scott Herkelman (AMD)
We also have an official abbreviation for the AMD FiedilityFX Super Resolution, which is FSR for short, as confirmed by Scott below:
This is probably one of the biggest software initiatives we have internally, because we know how important it is if you want to turn on jet tracking that you do not just want that competitive hit or your GPU gets so hard. The FSR (which will be called the acronym) is something important for us to introduce this year, but it’s going to have a little more time. We are making good progress, but we still have to work.
– Scott Herkelman (AMD)
Scott also mentions that they do not have pure machine learning for FSR, but in fact go with developers and the gaming community to find out which approach is best. NVIDIA’s DLSS relies solely on AI – assisted machine learning powered by their Tensor core GPU architecture, while AMD seems to be more in line with the Microsoft DirectML approach, from standard hardware and non-specialized AI cores can not be turned off.
You do not need machine learning to do this; you can do it in different ways and we evaluate many different ways. What’s most important to us is what game developers want to use, because if we’re just for ourselves at the end of the day, forcing our people to do it is not a good result. We would rather say: the game community, which one of these techniques do you want us to see implemented so that it can be spread across the industry immediately and hopefully on a platform.
– Scott Herkelman (AMD)
AMD says it will not limit the mining rate of cryptocurrency on its Radeon RX GPUs
In another interview posted on PCGamer, an AMD representative confirmed that the red team does not intend to limit the mining rate of cryptocurrency for its Radeon RX GPUs, unlike NVIDIA which tries to limit it by software -hacks, but then leaked their own developer drivers. full mining hash rate for its GeForce RTX 3060 graphics card earlier this week.
“The short answer is no,” said Nish Neelalojanan, an AMD product manager, about a potential mining constraint during a pre-Radeon RX 6700 XT briefing. “We will not block any workload, but also not just mining.
“There are a few things though. RDNA was designed for games from the beginning and RDNA 2 doubles it. And what I mean by that is that Infinity Cache and a smaller bus width have been carefully selected to give a very specific hit rate. However, mining specifically enjoys greater bandwidth and bus width, so that from an architectural level there are limitations to the mining itself. ‘
‘All our optimization, as always, is going to play first, and we’ve optimized everything for play. It is clear that players will benefit greatly from this, and it will not be ideal for the workload for mining. That being said, in this market it is always a nice thing to look at. ‘
– PCGamer
Unlike NVIDIA, AMD does not offer a mining-specific range of GPUs, but there are rumors that this is possible. For cryptocurrency mining itself, AMD GPUs are not as good compared to NVIDIA’s Ampere offerings, due to the large reliance on raw bandwidth required by the mining algorithms to deliver higher hash rates. AMD’s best dog, the RX 6900 XT, offers around 60-70 MH / s in Ethereum, while NVIDIA’s GeForce RTX 3090 can push up to 125 MH / s if set.
This makes the Ampere cards a more profitable investment for miners, but at the same time, due to the shortage of GPUs and the increased payout that the current GPUs offer, it can be prevented from enlarging even RX 6000 cards.
News source: Videocardz