Earlier this week GeForce RTX owners had the chance to put their new and shiny graphics cards to the test with the release of Metro Exodus, which featured both ray tracing and deep learning super-sampling, as well as the Battlefield V update that added deep learning super-sampling to the game (it already had ray tracing).

　　The main purpose of deep learning super-sampling, or DLSS for short, has always been to help GeForce RTX cards deliver playable frame rates at high resolutions when coupled with the taxing ray tracing effects. So far, while there is indeed a performance boost upon activating it in both Metro Exodus and Battlefield V, there's also a significant compromise to image quality which loses sharpness and definition with DLSS enabled.

　　To try and explain what's going on behind the scenes with DLSS, Andrew Edelsten, Technical Director of Deep Learning at NVIDIA, posted a lengthy update on NVIDIA's blog yesterday. Edelsten clarified that DLSS is meant to be used at 60 frames or lower and high resolutions, otherwise it does not necessarily improve performance. He also promised that NVIDIA is constantly working to improve the image quality of deep learning super-sampling, while the neural network itself will yield better results over time. They're also looking into some issues such as those seen in Metro Exodus when both HDR and DLSS are enabled.

　　Q: Where does DLSS provide the biggest benefit? And why isn’t it available for all resolutions?

　　A: The results of DLSS vary a bit, because each game has different characteristics based on the game engine, complexity of content, and the time spent on training. Our supercomputer never sleeps, and we continue to train and improve our deep learning neural network even after a game’s launch. When we have improvements to performance or image quality ready, we provide them to you via NVIDIA software updates.

　　DLSS is designed to boost frame rates at high GPU workloads (i.e. when your framerate is low and your GPU is working to its full capacity without bottlenecks or other limitations). If your game is already running at high frame rates, your GPU’s frame rendering time may be shorter than the DLSS execution time. In this case, DLSS is not available because it would not improve your framerate. However, if your game is heavily utilizing the GPU (e.g. FPS is below ~60), DLSS provides an optimal performance boost. You can crank up your settings to maximize your gains. (Note: 60 FPS is an approximation -- the exact number varies by game and what graphics settings are enabled)

　　To put it a bit more technically, DLSS requires a fixed amount of GPU time per frame to run the deep neural network. Thus, games that run at lower frame rates (proportionally less fixed workload) or higher resolutions (greater pixel shading savings), benefit more from DLSS. For games running at high frame rates or low resolutions, DLSS may not boost performance. When your GPU’s frame rendering time is shorter than what it takes to execute the DLSS model, we don’t enable DLSS. We only enable DLSS for cases where you will receive a performance gain. DLSS availability is game-specific, and depends on your GPU and selected display resolution.

　　Q: Some users mentioned blurry frames. Can you explain?

　　A: DLSS is a new technology and we are working hard to perfect it.

　　We built it to leverage the Turing architecture’s Tensor Cores and to provide the largest benefit when GPU load is high. To this end, we concentrated on high resolutions during development (where GPU load is highest) with 4K (3840x2160) being the most common training target. Running at 4K is beneficial when it comes to image quality as the number of input pixels is high. Typically for 4K DLSS, we have around 3.5-5.5 million pixels from which to generate the final frame, while at 1920x1080 we only have around 1.0-1.5 million pixels. The less source data, the greater the challenge for DLSS to detect features in the input frame and predict the final frame.

　　We have seen the screenshots and are listening to the community’s feedback about DLSS at lower resolutions, and are focusing on it as a top priority. We are adding more training data and some new techniques to improve quality, and will continue to train the deep neural network so that it improves over time.

　　Q: Why don’t I just use upscaled TAA instead?

　　A:  Depending on the resolution, quality settings, and game implementation, some may prefer TAA in one game and DLSS in another.

　　The game industry has used TAA for many years and we know that it can fall down in certain ways. TAA is generated from multiple frames and can suffer from high-motion ghosting and flickering that DLSS tends to handle better.

　　Q: When’s the next DLSS update for Battlefield V and Metro Exodus?

　　A: We are constantly working to improve image quality. Recently we updated the core of DLSS so that you get the latest model updates the moment you launch your game. So make sure you have our latest Game Ready Driver (418.91 or higher) installed.

　　For Battlefield V, we think it delivers a big improvement in 4K and 2560x1440 performance -- up to 40% -- for the corresponding quality, but also hear the community.  For the next push, we are focusing our testing and training to improve the image quality at 1920x1080 and also for ultrawide monitors (e.g. 3440x1440). The current experience at these resolutions is not where we want them.

　　For Metro Exodus, we’ve got an update coming that improves DLSS sharpness and overall image quality across all resolutions that didn’t make it into day of launch. We’re also training DLSS on a larger cross section of the game, and once these updates are ready you will see another increase in quality. Lastly, we are looking into a few other reported issues, such as with HDR, and will update as soon as we have fixes.

　　What was your experience with DLSS so far? Vote in the poll below, and don't forget to send your feedback to NVIDIA on both RTX and DLSS through this form.

　　.democracy:after{content:'';display:table;clear:both}.democracy ul li,.democracy ul{background:none;padding:0;margin:0}.democracy ul li:before{display:none}.democracy input[type='radio'],.democracy input[type='checkbox']{margin:0;vertical-align:middle}.democracy input[type='radio']:focus,.democracy input[type='checkbox']:focus{outline:0}.democracy ul{list-style:none !important;border:0 !important;padding-left:0 !important}.democracy ul li{list-style:none !important}input[type="submit"].dem-button,a.dem-button,.dem-button{position:relative;display:inline-block;text-decoration:none;-webkit-user-select:none;-moz-user-select:none;-ms-user-select:none;user-select:none;line-height:1;border:0;margin:0;padding:0;width:100px;font-size:14px;font-weight:normal;float:right}input[type="submit"].dem-button:focus,a.dem-button:focus,.dem-button:focus{outline:0}input[type="submit"].dem-button:disabled,a.dem-button:disabled,.dem-button:disabled{opacity:.6;cursor:not-allowed}.dem-button:hover{cursor:pointer}a.dem-button:hover{text-decoration:none !important}.democracy{position:relative;margin:0 auto}.democracy input[type='radio'],.democracy input[type='checkbox']{margin-right:.2em}.dem-screen{position:relative;line-height:1.3;padding:20px}ul.dem-vote{margin-bottom:1em}ul.dem-vote li{padding-bottom:0;margin-bottom:0}ul.dem-vote li>*{margin-bottom:.6em}ul.dem-answers li{padding-bottom:1em}.dem-answers .dem-label{margin-bottom:.2em;line-height:1.2}.dem-graph{position:relative;color:#fff;box-sizing:content-box;width:100%;display:table;height:1.2em;line-height:1.2em}.dem-fill{position:absolute;top:0;left:0;height:100%;background-color:#7cb4dd}.dem-voted-this .dem-fill{background-color:#3498db}.dem-votes-txt,.dem-percent-txt{position:relative;display:table-cell;padding-left:.3em;vertical-align:middle;font-size:90%}.dem-poll-info{float:left}.dem-poll-info:after{content:'';display:table;clear:both}.dem-poll-info>*{font-size:85%;display:block;clear:both;opacity:.7;line-height:1.3}.dem-vote label{float:none;display:block}.dem-results-link{display:inline-block;line-height:1;margin:.5em 0;float:left}.dem-vote .dem-disabled{opacity:.5}.democracy .dem-bottom{border-top:1px solid rgba(0,0,0,.1);margin-left:-20px;margin-right:-20px;padding:19px 20px 0 20px}.dem-bottom:after{content:'';display:table;clear:both}.dem-vote-button{float:right}.dem-poll-title{display:block;margin-bottom:1.5em;margin-top:1.2em;font-size:120%}.dem-cache-notice{z-index:10;position:absolute;top:0;border-radius:2px;width:100%;padding:1.5em 2em;text-align:center;background:rgba(247,241,212,.8);color:#6d6214}.dem-notice-close{position:absolute;top:0;right:0;padding:5px;cursor:pointer;line-height:.6;font-size:150%}.dem-notice-close:hover{color:#d26616}.dem-star{font-size:90%;vertical-align:baseline;padding-left:.3em;color:#ff4e00}.dem-poll-note{font-size:90%;padding:.5em;opacity:.8;line-height:1.3}.democracy .dem-copyright{position:absolute;bottom:-1em;right:0;text-decoration:none;border-bottom:0;color:#b2b2b2;opacity:.5;line-height:1}.democracy .dem-copyright:hover{opacity:1}.dem-add-answer{position:relative}.dem-add-answer>*{display:block}.dem-add-answer a{display:inline-block}input.dem-add-answer-txt{width:100%;box-sizing:border-box}.dem-add-answer-close{position:absolute;right:0;padding:0 .7em;cursor:pointer;color:#333}.dem-add-answer-close:hover{color:#ff2700}.dem-edit-link{display:block;position:absolute;top:0;right:0;line-height:1;text-decoration:none !important;border:0 !important}.dem-edit-link svg{width:1.2em !important;fill:#5a5a5a;fill:rgba(0,0,0,.6)}.dem-edit-link:hover svg{fill:#35a91d}.dem-loader{display:none;position:absolute;top:0;left:0;width:100%;height:100%}.dem-loader>*{display:table-cell;vertical-align:middle;text-align:center}.dem-loader svg{width:20%;max-width:100px;min-width:80px;margin-bottom:15%}.dem-loader [class^="dem-"]{margin-bottom:15% !important}.dem-archives .democracy{margin-bottom:2em;padding-bottom:20px;border-bottom:1px dashed #ccc}.dem-archives .dem-archive-link{display:none}.democracy{background:#f6f6f6;margin-bottom:1em;border:1px solid rgba(0,0,0,.1)}.dem-poll-title{margin:0;font-size:16px;padding:20px;border-bottom:1px solid rgba(0,0,0,.1)}.dem-percent-txt,.dem-label-percent-txt{display:none}.dem-votes-txt-percent:before{content:'- ';display:inline}.dem-revote-button:before{content:'◂ '}.dem-graph{background-color:#ddd}.dem-results-link{margin-bottom:0}.dem__checkbox,.dem__radio{display:none !important}.dem__checkbox_label,.dem__radio_label{position:relative;box-sizing:content-box}.dem__checkbox_label .dem__spot,.dem__radio_label .dem__spot{position:relative;display:inline-block;vertical-align:baseline;top:.1em;cursor:pointer;width:12px;height:12px;margin-right:.3em;background:#777}.dem__checkbox_label .dem__spot{border-radius:3px}.dem__radio_label .dem__spot{border-radius:50%}.dem__checkbox_label .dem__spot:after,.dem__radio_label .dem__spot:after{content:'';position:absolute;opacity:0}.dem__checkbox_label .dem__spot:after{width:11px;height:4px;top:0;left:2px;border:2px solid #fff;border-top:0;border-right:0;background:transparent;-ms-transform:rotate(-45deg);-webkit-transform:rotate(-45deg);transform:rotate(-45deg);box-shadow:-1px 1px 1px #555}.dem__radio_label .dem__spot:after{width:6px;height:6px;border-radius:50%;top:50%;left:50%;background:#fff;-ms-transform:translate(-50%,-50%);-webkit-transform:translate(-50%,-50%);transform:translate(-50%,-50%);box-shadow:0 0 0 1px #666,inset -1px -1px 1px #ccc}.dem__checkbox_label:hover .dem__spot::after,.dem__radio_label:hover .dem__spot::after{opacity:.4}.dem__checkbox:checked+.dem__spot,.dem__radio:checked+.dem__spot{border-color:#999}.dem__checkbox:checked+.dem__spot:after,.dem__radio:checked+.dem__spot:after{opacity:1}.dem__checkbox:disabled+.dem__spot,.dem__radio:disabled+.dem__spot{opacity:.5}.dem__checkbox:not(:checked):disabled+.dem__spot:after,.dem__radio:not(:checked):disabled+.dem__spot:after{opacity:0}input[type="submit"].dem-button,a.dem-button,.dem-button{font-weight:700;color:#fff;padding:10px 30px;border-radius:3px;background:#eb0254;transition:background .2s}input[type="submit"].dem-button:hover,a.dem-button:hover,.dem-button:hover{background:#eb0254}input[type="submit"].dem-button:active,a.dem-button:active,.dem-button:active{background:#21935a;box-shadow:0 3px #21935a inset}a.dem-button:visited{color:#fff}What was your experience with NVIDIA DLSS so far?Disappointing and I wish NVIDIA didn't waste silicon on it 44%, 725 votes

　　725 votes 44%

　　725 votes - 44% of all votes

　　Underwhelming, but I believe in future improvements 38%, 624 votes

　　624 votes 38%

　　624 votes - 38% of all votes

　　Acceptable, as the performance boost is worth the image quality tradeoff for me 18%, 302 votes

　　302 votes 18%

　　302 votes - 18% of all votes

　　Total Votes: 1651

　　February 16, 2019 - February 23, 2019

　　Voting is closed

　　Poll Options are limited because JavaScript is disabled in your browser. .dem-roller { margin: 0 auto; width: 40px; height: 40px; position: relative; } .con1 > div, .con2 > div, .con3 > div { width: 10px; height: 10px; background-color: #2686cc; border-radius: 100%; position: absolute; -webkit-animation: bouncedelay 1.2s infinite ease-in-out; animation: bouncedelay 1.2s infinite ease-in-out; /* Prevent first frame from flickering when animation starts */ -webkit-animation-fill-mode: both; animation-fill-mode: both; } .dem-roller .demrcont { position: absolute; width: 100%; height: 100%; } .con2 { -webkit-transform: rotateZ(45deg); transform: rotateZ(45deg); } .con3 { -webkit-transform: rotateZ(90deg); transform: rotateZ(90deg); } .circle1 { top: 0; left: 0; } .circle2 { top: 0; right: 0; } .circle3 { right: 0; bottom: 0; } .circle4 { left: 0; bottom: 0; } .con2 .circle1 { -webkit-animation-delay: -1.1s; animation-delay: -1.1s; } .con3 .circle1 { -webkit-animation-delay: -1.0s; animation-delay: -1.0s; } .con1 .circle2 { -webkit-animation-delay: -0.9s; animation-delay: -0.9s; } .con2 .circle2 { -webkit-animation-delay: -0.8s; animation-delay: -0.8s; } .con3 .circle2 { -webkit-animation-delay: -0.7s; animation-delay: -0.7s; } .con1 .circle3 { -webkit-animation-delay: -0.6s; animation-delay: -0.6s; } .con2 .circle3 { -webkit-animation-delay: -0.5s; animation-delay: -0.5s; } .con3 .circle3 { -webkit-animation-delay: -0.4s; animation-delay: -0.4s; } .con1 .circle4 { -webkit-animation-delay: -0.3s; animation-delay: -0.3s; } .con2 .circle4 { -webkit-animation-delay: -0.2s; animation-delay: -0.2s; } .con3 .circle4 { -webkit-animation-delay: -0.1s; animation-delay: -0.1s; } @-webkit-keyframes bouncedelay { 0%, 80%, 100% { -webkit-transform: scale(0.0) } 40% { -webkit-transform: scale(1.0) } } @keyframes bouncedelay { 0%, 80%, 100% { transform: scale(0.0); -webkit-transform: scale(0.0); } 40% { transform: scale(1.0); -webkit-transform: scale(1.0); } }