1858:
702:. Like TAA, it uses information from past frames to produce the current frame. Unlike TAA, DLSS does not sample every pixel in every frame. Instead, it samples different pixels in different frames and uses pixels sampled in past frames to fill in the unsampled pixels in the current frame. DLSS uses machine learning to combine samples in the current frame and past frames, and it can be thought of as an advanced and superior TAA implementation made possible by the available tensor cores.
766:(FMA) operations that are used extensively in neural network calculations for applying a large series of multiplications on weights, followed by the addition of a bias. Tensor cores can operate on FP16, INT8, INT4, and INT1 data types. Each core can do 1024 bits of FMA operations per clock, so 1024 INT1, 256 INT4, 128 INT8, and 64 FP16 operations per clock per tensor core, and most Turing GPUs have a few hundred tensor cores.
1512:
aliased inputs and generate high-quality anti-aliased images that match the "perfect frame" as closely as possible. We then repeat the process, but this time we train the model to generate additional pixels rather than applying AA. This has the effect of increasing the resolution of the input. Combining both techniques enables the GPU to render the full monitor resolution at higher frame rates.
33:
502:. The second stage is an image upscaling step which uses the single raw, low-resolution frame to upscale the image to the desired output resolution. Using just a single frame for upscaling means the neural network itself must generate a large amount of new information to produce the high resolution output, this can result in slight
587:. One example of this is neighborhood clamping which forcefully prevents samples collected in previous frames from deviating too much compared to nearby pixels in newer frames. This helps to identify and fix many temporal artifacts, but deliberately removing fine details in this way is analogous to applying a
1166:
As promised, NVIDIA has updated the DLSS network in a new GeForce update that provides better, sharper image quality while still retaining higher framerates in raytraced games. While the feature wasn't used as well in its first iteration, NVIDIA is now confident that they have successfully fixed all
932:
Deep learning super sampling uses artificial intelligence and machine learning to produce an image that looks like a higher-resolution image, without the rendering overhead. Nvidia's algorithm learns from tens of thousands of rendered sequences of images that were created using a supercomputer. That
604:
trained to identify and fix temporal artifacts, instead of manually programmed heuristics as mentioned above. Because of this, DLSS 2.0 can generally resolve detail better than other TAA and TAAU implementations, while also removing most temporal artifacts. This is why DLSS 2.0 can sometimes produce
1045:
Of course, this is to be expected. DLSS was never going to provide the same image quality as native 4K while providing a 37% performance uplift. That would be black magic. But the quality difference comparing the two is almost laughable, in how far away DLSS is from the native presentation in these
608:
Because temporal artifacts occur in most art styles and environments in broadly the same way, the neural network that powers DLSS 2.0 does not need to be retrained when being used in different games. Despite this, Nvidia does frequently ship new minor revisions of DLSS 2.0 with new titles, so this
521:
This first iteration received a mixed response, with many criticizing the often soft appearance and artifacts in certain situations; likely a side effect of the limited data from only using a single frame input to the neural networks which could not be trained to perform optimally in all scenarios
803:
Especially in early versions of DLSS, users reported blurry frames. Andrew
Edelsten, an employee at Nvidia, therefore commented on the problem in a blog post in 2019 and promised that they were working on improving the technology and clarified that the DLSS AI algorithm was mainly trained with 4K
1075:
Recently, two big titles received NVIDIA DLSS support, namely Metro Exodus and
Battlefield V. Both these games come with NVIDIA's DXR (DirectX Raytracing) implementation that at the moment is only supported by the GeForce RTX cards. DLSS makes these games playable at higher resolutions with much
1511:
The DLSS team first extracts many aliased frames from the target game, and then for each one we generate a matching "perfect frame" using either super-sampling or accumulation rendering. These paired frames are fed to NVIDIA's supercomputer. The supercomputer trains the DLSS model to recognize
641:
DLSS 3.0 makes use of a new generation
Optical Flow Accelerator (OFA) included in Ada Lovelace generation RTX GPUs. The new OFA is faster and more accurate than the OFA already available in previous Turing and Ampere RTX GPUs. This results in DLSS 3.0 being exclusive for the RTX 40 Series.
625:
in games will still appear low-resolution when using current TAAU techniques. This is why Nvidia recommends game developers use higher resolution textures than they would normally for a given rendering resolution by applying a mip-map bias when DLSS 2.0 is enabled.
620:
in the same sense as techniques such as ESRGAN or DLSS 1.0, which attempt to create new information from a low-resolution source; instead, TAAU works to recover data from previous frames, rather than creating new data. In practice, this means low resolution
162:), a related and opposite technology where the graphics are rendered at a higher resolution, then downsampled to the native display resolution using an AI-assisted downsampling algorithm to achieve higher image quality than rendering at native resolution.
224:. This time Nvidia said that it used the Tensor Cores again, and that the AI did not need to be trained specifically on each game. Despite sharing the DLSS branding, the two iterations of DLSS differ significantly and are not backwards-compatible.
1247:
The original DLSS required training the AI network for each new game. DLSS 2.0 trains using non-game-specific content, delivering a generalized network that works across games. This means faster game integrations, and ultimately more DLSS
513:
to 64 samples per pixel, as well as the motion vectors for each frame. The data collected must be as comprehensive as possible, including as many levels, times of day, graphical settings, resolutions, etc. as possible. This data is also
134:
for increased performance, and then infer a higher resolution image from this that approximates the same level of detail as if the image had been rendered at this higher resolution. This allows for higher graphical settings and/or
605:
a sharper image than rendering at higher, or even native resolutions using traditional TAA. However, no temporal solution is perfect, and artifacts (ghosting in particular) are still visible in some scenarios when using DLSS 2.0.
653:
DLSS 3.5 adds ray reconstruction, replacing multiple denoising algorithms with a single AI model trained on five times more data than DLSS 3. Ray reconstruction will be available on all RTX GPUs and will first target games with
1076:
better frame rates, although there is a notable decrease in image sharpness. Now, AMD has taken a jab at DLSS, saying that traditional AA methods like SMAA and TAA "offer superior combinations of image quality and performance."
811:
It has been criticized that by implementing DLSS in their games, game developers no longer have an incentive to optimize them so that they also run smoothly in native resolution on modern PC hardware. For example, for the game
1106:
The benefit for most people is that, generally, DLSS comes with a sizeable FPS improvement. How much varies from game to game. In Metro Exodus, the FPS jump was barely there and certainly not worth the bizarre hit to image
638:. The DLSS frame generation algorithm takes two rendered frames from the rendering pipeline and generates a new frame that smoothly transitions between them. So for every frame rendered, one additional frame is generated.
612:
The main advancements compared to DLSS 1.0 include: Significantly improved detail retention, a generalized neural network that does not need to be re-trained per-game, and ~2x less overhead (~1-2 ms vs ~2-4 ms).
466:
The linear render scale, compared to the output resolution, that the technology uses to render scenes internally before upsampling. For example, a 1080p scene with a 50% render scale would have an internal resolution of
1701:
552:(TAAU) implementation, using data from previous frames extensively through sub-pixel jittering to resolve fine detail and reduce aliasing. The data DLSS 2.0 collects includes: the raw low-resolution input,
457:
The linear scale factor used for upsampling the input resolution to the output resolution. For example, a scene rendered at 540p with a 2.00x scale factor would have an output resolution of 1080p.
1333:
we developed a new image processing algorithm that approximated our AI research model and fit within our performance budget. This image processing approach to DLSS is integrated into
Control
564:/ brightness information. It can also be used as a simpler TAA implementation where the image is rendered at 100% resolution, rather than being upsampled by DLSS, Nvidia brands this as
1140:
Of course, this isn't the first DLSS implementation we've seen in
Control. The game shipped with a decent enough rendition of the technology that didn't actually use machine learning
2501:
1396:
609:
could suggest some minor training optimizations may be performed as games are released, although Nvidia does not provide changelogs for these minor revisions to confirm this.
142:
As of
September 2022, the 1st and 2nd generation of DLSS are available on all RTX-branded cards from Nvidia in supported titles, while the 3rd generation unveiled at Nvidia's
329:
When using DLSS, depending on the game, users have access to various quality presets in addition to the option to set the internally rendered, upscaled resolution manually:
518:
using common augmentations such as rotations, colour changes, and random noise to help generalize the test data. Training is performed on Nvidia's Saturn V supercomputer.
1230:
1690:
182:) because the algorithm had to be trained specifically on each game on which it was applied and the results were usually not as good as simple resolution upscaling.
448:
The algorithm does not necessarily need to be implemented using these presets; it is possible for the implementer to define custom input and output resolutions.
808:, is due to the fact that the algorithm has far less image information available to calculate an appropriate image compared to higher resolutions like 4K.
307:
DLSS 3.0, augmented with an optical flow frame generation algorithm (only available on RTX 40-series GPUs) to generate frames inbetween rendered frames
1938:
1917:
503:
318:
DLSS 3.5 adds ray reconstruction, replacing multiple denoising algorithms with a single AI model trained on five times more data than DLSS 3.
868:
695:
method. Thus, depending on the game and quality setting used, using DLSS may improve image quality even over native resolution rendering.
915:
933:
trains the algorithm to be able to produce similarly beautiful images, but without requiring the graphics card to work as hard to do it.
1445:
976:
1371:
763:
2843:
76:
1421:
2431:
1890:
1180:
147:
1234:
434:
Rendered resolution dynamically adjusts in real time to achieve user-defined fps targets (e.g., 144 fps on a 144 Hz monitor).
2760:
2361:
826:
enabled, the use of DLSS in
Performance mode is recommended even with current-generation high-end graphics cards such as the
2833:
2666:
2436:
1123:
951:
712:(DLAA). DLAA provides the same AI-driven anti-aliasing DLSS uses, but without any upscaling or downscaling functionality.
1530:
1089:
595:
485:
2401:
2376:
2366:
709:
351:
1582:
1205:
2426:
2421:
2416:
2406:
2396:
2371:
2441:
2411:
2391:
2386:
2381:
2314:
2226:
2096:
1748:
1494:
776:-Level Primitives on 32 parallel threads to take advantage of their parallel architecture. A Warp is a set of 32
749:
1469:
174:
cards when they launched in
September 2018. At that time, the results were limited to a few video games (namely
752:
47:
1669:
2639:
2545:
2332:
827:
784:
494:. The first step is an image enhancement network which uses the current frame and motion vectors to perform
208:
252:
Predominantly spatial image upscaler, required specifically trained for each game integration, included in
2720:
2344:
2139:
1735:
NVIDIA GPUs execute groups of threads known as warps in SIMT (Single
Instruction, Multiple Thread) fashion
823:
699:
546:
293:
2785:
804:
image material. That the use of DLSS leads to particularly blurred images at lower resolutions, such as
580:
499:
192:
484:
The first iteration of DLSS is a predominantly spatial image upscaler with two stages, both relying on
1797:
1772:
509:
The neural networks are trained on a per-game basis by generating a "perfect frame" using traditional
2047:
1883:
1320:
1153:
1032:
635:
278:
187:
1397:"Nvidia announces DLSS 3.5 with ray reconstruction, boosting RT quality with an AI-trained denoiser"
529:
Nvidia also demonstrated the ability for the auto-encoder networks to learn the ability to recreate
2268:
2114:
1295:
221:
100:
2524:
2489:
2119:
2109:
2104:
2085:
2080:
2075:
2070:
2065:
777:
131:
1718:
51:
1558:
2710:
2590:
2158:
2153:
2148:
2129:
2124:
2028:
755:
733:
729:
725:
515:
171:
151:
127:
198:
In April 2020, Nvidia advertised and shipped an improved version of DLSS named DLSS 2.0 with
2770:
2765:
2605:
2290:
2256:
2237:
2209:
2060:
2055:
2038:
2033:
2023:
2018:
1996:
1991:
1345:
838:
495:
1645:
2838:
2580:
2562:
2308:
1876:
1851:
1006:
891:
673:
667:
622:
584:
1857:
1276:
1822:
1595:
276:
DLSS 1.0 adapted for running on the CUDA shader cores instead of tensor cores, used for
2647:
2538:
2533:
2302:
919:
741:
660:
601:
561:
530:
491:
537:, although this functionality has never been included in a publicly released product.
2827:
2790:
2624:
2585:
2550:
2296:
2006:
1063:"AMD Thinks NVIDIA DLSS is not Good Enough; Calls TAA & SMAA Better Alternatives"
834:
819:
692:
553:
510:
414:
254:
217:
199:
175:
119:
107:
103:
1296:"Truly Next-Gen: Adding Deep Learning to Games & Graphics (Presented by NVIDIA)"
2795:
2715:
2705:
2619:
2509:
2320:
2203:
2197:
2010:
1372:"Nvidia's new DLSS 3.5 works on all RTX GPUs to improve the quality of ray tracing"
759:
655:
617:
260:
179:
1062:
2595:
2475:
1986:
814:
773:
679:
598:
572:
557:
534:
488:
213:
136:
43:
1619:
575:, however all previous implementations have used some form of manually written
2725:
2514:
2338:
2262:
859:
549:
143:
123:
115:
1971:
1277:"DLSS 2.0 - Image Reconstruction for Real-time Rendering with Deep Learning"
1127:
576:
523:
1446:"DLSS 3 explained: How Nvidia's AI-infused RTX tech turbocharges PC gaming"
1231:"HW News - Crysis Remastered Ray Tracing, NVIDIA DLSS 2, Ryzen 3100 Rumors"
202:
version 445.75. DLSS 2.0 was available for a few existing games including
1966:
1961:
788:
588:
1863:
2775:
2658:
2614:
1912:
1422:"NVIDIA preparing Ultra Quality mode for DLSS, 2.2.9.0 version spotted"
850:
805:
792:
2484:
2326:
1899:
1722:
1534:
1498:
1093:
977:"Nvidia DLDSR tested: better visuals and better performance than DSR"
737:
705:
591:, and thus the final image can appear blurry when using this method.
195:
and an improved version of DLSS, which did not use the Tensor Cores.
111:
1154:"NVIDIA DLSS 2.0 Update Will Fix The GeForce RTX Cards' Big Mistake"
1124:"Remedy's Control vs DLSS 2.0 – AI upscaling reaches the next level"
833:
Furthermore, the use of DLSS frame generation may lead to increased
616:
It should also be noted that forms of TAAU such as DLSS 2.0 are not
17:
2780:
2755:
2629:
2557:
2273:
2248:
2242:
863:
126:. The goal of these technologies is to allow the majority of the
2750:
2232:
872:
770:
565:
2687:
2462:
2179:
1910:
1872:
1181:"NVIDIA DLSS Plugin and Reflex Now Available for Unreal Engine"
1952:
1947:
854:
26:
720:
With the exception of the shader-core version implemented in
139:
for a given output resolution, depending on user preference.
1670:"On Tensors, Tensorflow, And Nvidia's Latest 'Tensor Cores'"
212:, and would later be added to many newly released games and
506:
such as leaves that differ in style to the source content.
1868:
1033:"Battlefield V DLSS Tested: Overpromised, Underdelivered"
58:
1346:"NVIDIA DLSS 2.0 Review with Control – Is This Magic?"
1090:"Nvidia Very Quietly Made DLSS A Hell Of A Lot Better"
780:
which are configured to execute the same instruction.
645:
At release, DLSS 3.0 does not work for VR displays.
2743:
2698:
2656:
2638:
2604:
2573:
2523:
2500:
2473:
2354:
2283:
2219:
2190:
2138:
2094:
2046:
2005:
1937:
1930:
892:"NVIDIA DLSS Technology for Incredible Performance"
413:(since v2.1; only recommended for resolutions from
1391:
1389:
1823:"Nvidia DLSS 3 Revisit: We Try It Out in 9 Games"
1773:"Alan Wake 2 on PC is an embarrassment of riches"
1646:"What Is DLSS and Why Does it Matter for Gaming?"
1559:"What is Nvidia DLAA? An Anti-Aliasing Explainer"
1206:"NVIDIA DLSS Natively Supported in Unity 2021.2"
170:Nvidia advertised DLSS as a key feature of the
1798:"When a high frame rate can lose you the game"
1470:"DLSS: What Does It Mean for Game Developers?"
1366:
1364:
1362:
1027:
1025:
1023:
916:"Nvidia RTX DLSS: Everything you need to know"
748:. Tensor Cores are available since the Nvidia
571:TAA(U) is used in many modern video games and
296:U using Tensor Cores, and trained generically
1884:
1531:"NVIDIA DLSS 2.0: A Big Leap In AI Rendering"
8:
1525:
1523:
1521:
1118:
1116:
154:graphics cards. Nvidia has also introduced
2695:
2684:
2470:
2459:
2187:
2176:
1934:
1927:
1907:
1891:
1877:
1869:
1856:
1001:
999:
997:
762:line of products. They are used for doing
331:
1864:DLSS on official Nvidia developer website
1271:
1269:
1267:
1265:
1263:
1261:
1259:
1257:
862:– AI-augmented upscaling technology from
77:Learn how and when to remove this message
231:
1749:"NVIDIA DLSS: Your Questions, Answered"
1495:"NVIDIA DLSS: Your Questions, Answered"
1007:"Nvidia DLSS in 2020: stunning results"
883:
853:– competing upsampling technology from
740:series of video cards, using dedicated
441:
822:at the highest graphics settings with
579:to prevent temporal artifacts such as
156:Deep learning dynamic super resolution
61:by removing or replacing such wording.
1553:
1551:
1416:
1414:
1290:
1288:
1286:
1284:
869:PlayStation Spectral Super Resolution
698:It operates on similar principles to
7:
1057:
1055:
946:
944:
942:
658:(aka "full ray tracing"), including
89:Image upscaling technology by Nvidia
1319:Edelsten, Andrew (30 August 2019).
634:Augments DLSS 2.0 by making use of
1719:"Using CUDA Warp-Level Primitives"
1691:"TENSOR CORE DL PERFORMANCE GUIDE"
1596:"NVIDIA DLSS DLL (2.3.7) Download"
691:DLSS requires and applies its own
25:
1321:"NVIDIA DLSS: Control and Beyond"
975:Archer, James (17 January 2022).
1707:from the original on 2020-11-11.
1644:Smith, Matthew S. (2023-12-28).
31:
568:(deep learning anti-aliasing).
146:2022 event is exclusive to the
122:, and available in a number of
2341:(framebuffer in system memory)
1672:. tomshardware.com. 2017-04-11
758:, which was first used on the
1:
1348:. techquila.co.in. 2020-04-05
1156:. techquila.co.in. 2020-03-24
1065:. techquila.co.in. 2019-02-15
787:, Microsoft Windows provided
1233:. 2020-04-19. Archived from
830:in order to achieve 60 fps.
724:, DLSS is only available on
93:Deep learning super sampling
1399:. EuroGamer. 23 August 2023
1374:. The Verge. 22 August 2023
952:"Introducing NVIDIA DLSS 3"
851:FidelityFX Super Resolution
710:deep learning anti-aliasing
2860:
1035:. techspot.com. 2019-02-19
1009:. techspot.com. 2020-02-26
871:– similar technology from
292:An AI accelerated form of
114:that are exclusive to its
110:technologies developed by
2694:
2683:
2546:RSX 'Reality Synthesizer'
2469:
2458:
2186:
2181:Software and technologies
2175:
1926:
1906:
1620:"NVIDIA Optical Flow SDK"
795:to support Tensor Cores.
433:
240:
237:
234:
2844:Anti-aliasing algorithms
1167:the issues it had before
185:In 2019, the video game
2333:Scalable Link Interface
2299:(variable refresh rate)
2191:Multimedia acceleration
828:Nvidia GeForce RTX 4080
785:Windows 10 version 1903
270:"1.9" (unofficial name)
209:Wolfenstein: Youngblood
2355:GPU microarchitectures
2347:(live video upscaling)
2345:Video Super Resolution
2265:(ray tracing platform)
2095:Unified shaders &
547:temporal anti-aliasing
106:image enhancement and
42:contains wording that
2786:Mellanox Technologies
1474:NVIDIA Developer Blog
1210:NVIDIA Developer Blog
1185:NVIDIA Developer Blog
769:The Tensor Cores use
500:spatial anti-aliasing
334:Standard DLSS presets
193:real-time ray tracing
2834:3D computer graphics
1931:Fixed pixel pipeline
636:motion interpolation
46:the subject through
2269:Nvidia System Tools
1275:Edward Liu, NVIDIA
336:
2276:(video decode API)
2229:(shading language)
2142:& tensor cores
981:Rock Paper Shotgun
764:fused multiply-add
332:
130:to run at a lower
52:unnoteworthy facts
2821:
2820:
2817:
2816:
2813:
2812:
2711:Chris Malachowsky
2679:
2678:
2675:
2674:
2591:Shield Android TV
2454:
2453:
2450:
2449:
2245:(ray tracing API)
2171:
2170:
2167:
2166:
1981:
1980:
1921:
1583:Temporal AA small
756:microarchitecture
438:
437:
411:Ultra Performance
322:
321:
172:GeForce 20 series
128:graphics pipeline
99:) is a family of
87:
86:
79:
16:(Redirected from
2851:
2771:Cumulus Networks
2766:Bright Computing
2751:3dfx Interactive
2696:
2685:
2566:
2554:
2542:
2471:
2460:
2291:Nvidia 3D Vision
2257:Nvidia Omniverse
2238:Nvidia GameWorks
2212:(video decoding)
2206:(video decoding)
2200:(video encoding)
2188:
2177:
1935:
1928:
1915:
1908:
1893:
1886:
1879:
1870:
1860:
1855:
1854:
1852:Official website
1838:
1837:
1835:
1834:
1819:
1813:
1812:
1810:
1809:
1794:
1788:
1787:
1785:
1784:
1769:
1763:
1762:
1760:
1759:
1745:
1739:
1738:
1731:
1730:
1715:
1709:
1708:
1706:
1695:
1687:
1681:
1680:
1678:
1677:
1666:
1660:
1659:
1657:
1656:
1641:
1635:
1634:
1632:
1631:
1624:NVIDIA Developer
1616:
1610:
1609:
1607:
1606:
1592:
1586:
1580:
1574:
1573:
1571:
1570:
1555:
1546:
1545:
1543:
1542:
1527:
1516:
1515:
1507:
1506:
1491:
1485:
1484:
1482:
1481:
1466:
1460:
1459:
1457:
1456:
1442:
1436:
1435:
1433:
1432:
1418:
1409:
1408:
1406:
1404:
1393:
1384:
1383:
1381:
1379:
1368:
1357:
1356:
1354:
1353:
1342:
1336:
1335:
1330:
1328:
1316:
1310:
1309:
1307:
1306:
1300:www.gdcvault.com
1292:
1279:
1273:
1252:
1251:
1243:
1242:
1227:
1221:
1220:
1218:
1217:
1202:
1196:
1195:
1193:
1192:
1177:
1171:
1170:
1162:
1161:
1150:
1144:
1143:
1136:
1135:
1120:
1111:
1110:
1102:
1101:
1086:
1080:
1079:
1071:
1070:
1059:
1050:
1049:
1046:stressful areas.
1041:
1040:
1029:
1018:
1017:
1015:
1014:
1003:
992:
991:
989:
987:
972:
966:
965:
963:
962:
948:
937:
936:
928:
927:
912:
906:
905:
903:
902:
888:
839:visual artifacts
594:DLSS 2.0 uses a
496:edge enhancement
468:
464:
458:
455:
449:
446:
337:
232:
82:
75:
71:
68:
62:
35:
34:
27:
21:
2859:
2858:
2854:
2853:
2852:
2850:
2849:
2848:
2824:
2823:
2822:
2809:
2739:
2733:Ranga Jayaraman
2730:Debora Shoquist
2690:
2671:
2652:
2634:
2600:
2581:Shield Portable
2569:
2563:Nintendo Switch
2560:
2548:
2536:
2519:
2496:
2465:
2446:
2350:
2323:(module/socket)
2317:(module/socket)
2311:(multi-monitor)
2309:Nvidia Surround
2305:(GPU switching)
2279:
2215:
2182:
2163:
2134:
2090:
2048:Unified shaders
2042:
2001:
1977:
1976:
1957:
1922:
1902:
1897:
1850:
1849:
1846:
1841:
1832:
1830:
1821:
1820:
1816:
1807:
1805:
1796:
1795:
1791:
1782:
1780:
1771:
1770:
1766:
1757:
1755:
1747:
1746:
1742:
1728:
1726:
1717:
1716:
1712:
1704:
1693:
1689:
1688:
1684:
1675:
1673:
1668:
1667:
1663:
1654:
1652:
1643:
1642:
1638:
1629:
1627:
1618:
1617:
1613:
1604:
1602:
1594:
1593:
1589:
1581:
1577:
1568:
1566:
1557:
1556:
1549:
1540:
1538:
1529:
1528:
1519:
1504:
1502:
1493:
1492:
1488:
1479:
1477:
1468:
1467:
1463:
1454:
1452:
1444:
1443:
1439:
1430:
1428:
1420:
1419:
1412:
1402:
1400:
1395:
1394:
1387:
1377:
1375:
1370:
1369:
1360:
1351:
1349:
1344:
1343:
1339:
1326:
1324:
1318:
1317:
1313:
1304:
1302:
1294:
1293:
1282:
1274:
1255:
1240:
1238:
1229:
1228:
1224:
1215:
1213:
1204:
1203:
1199:
1190:
1188:
1179:
1178:
1174:
1159:
1157:
1152:
1151:
1147:
1133:
1131:
1122:
1121:
1114:
1099:
1097:
1088:
1087:
1083:
1068:
1066:
1061:
1060:
1053:
1038:
1036:
1031:
1030:
1021:
1012:
1010:
1005:
1004:
995:
985:
983:
974:
973:
969:
960:
958:
950:
949:
940:
925:
923:
914:
913:
909:
900:
898:
890:
889:
885:
881:
847:
801:
791:as one part of
742:AI accelerators
718:
689:
674:Portal with RTX
668:Phantom Liberty
651:
632:
543:
492:neural networks
482:
477:
472:
471:
465:
461:
456:
452:
447:
443:
420:
417:
367:
340:Quality preset
327:
325:Quality presets
315:September 2023
304:September 2022
230:
228:Release history
168:
90:
83:
72:
66:
63:
59:help improve it
56:
36:
32:
23:
22:
15:
12:
11:
5:
2857:
2855:
2847:
2846:
2841:
2836:
2826:
2825:
2819:
2818:
2815:
2814:
2811:
2810:
2808:
2807:
2804:
2801:
2798:
2793:
2788:
2783:
2778:
2773:
2768:
2763:
2758:
2753:
2747:
2745:
2741:
2740:
2738:
2737:
2736:Jonah M. Alben
2734:
2731:
2728:
2723:
2718:
2713:
2708:
2706:Jen-Hsun Huang
2702:
2700:
2692:
2691:
2688:
2681:
2680:
2677:
2676:
2673:
2672:
2670:
2669:
2663:
2661:
2654:
2653:
2651:
2650:
2648:Project Denver
2644:
2642:
2636:
2635:
2633:
2632:
2627:
2622:
2617:
2611:
2609:
2602:
2601:
2599:
2598:
2593:
2588:
2583:
2577:
2575:
2571:
2570:
2568:
2567:
2555:
2543:
2530:
2528:
2521:
2520:
2518:
2517:
2512:
2506:
2504:
2498:
2497:
2495:
2494:
2493:
2492:
2481:
2479:
2467:
2466:
2464:Other products
2463:
2456:
2455:
2452:
2451:
2448:
2447:
2445:
2444:
2439:
2434:
2429:
2424:
2419:
2414:
2409:
2404:
2399:
2394:
2389:
2384:
2379:
2374:
2369:
2364:
2358:
2356:
2352:
2351:
2349:
2348:
2342:
2336:
2330:
2324:
2318:
2312:
2306:
2303:Nvidia Optimus
2300:
2294:
2287:
2285:
2281:
2280:
2278:
2277:
2271:
2266:
2260:
2254:
2253:
2252:
2246:
2235:
2230:
2223:
2221:
2217:
2216:
2214:
2213:
2207:
2201:
2194:
2192:
2184:
2183:
2180:
2173:
2172:
2169:
2168:
2165:
2164:
2162:
2161:
2156:
2151:
2145:
2143:
2136:
2135:
2133:
2132:
2127:
2122:
2117:
2112:
2107:
2101:
2099:
2092:
2091:
2089:
2088:
2083:
2078:
2073:
2068:
2063:
2058:
2052:
2050:
2044:
2043:
2041:
2036:
2031:
2026:
2021:
2016:
2014:
2003:
2002:
2000:
1999:
1994:
1989:
1982:
1979:
1978:
1975:
1974:
1969:
1964:
1958:
1956:
1955:
1950:
1944:
1943:
1941:
1932:
1924:
1923:
1911:
1904:
1903:
1898:
1896:
1895:
1888:
1881:
1873:
1867:
1866:
1861:
1845:
1844:External links
1842:
1840:
1839:
1814:
1802:Digital Trends
1789:
1777:Digital Trends
1764:
1753:www.nvidia.com
1740:
1710:
1682:
1661:
1636:
1611:
1587:
1575:
1563:Digital Trends
1547:
1517:
1486:
1461:
1437:
1426:VideoCardz.com
1410:
1385:
1358:
1337:
1311:
1280:
1253:
1222:
1197:
1172:
1145:
1112:
1081:
1051:
1019:
993:
967:
938:
920:Digital Trends
907:
882:
880:
877:
876:
875:
866:
857:
846:
843:
800:
797:
734:GeForce RTX 40
730:GeForce RTX 30
726:GeForce RTX 20
717:
714:
688:
685:
661:Cyberpunk 2077
650:
647:
631:
628:
602:neural network
554:motion vectors
545:DLSS 2.0 is a
542:
539:
531:depth-of-field
504:hallucinations
481:
478:
476:
475:Implementation
473:
470:
469:
459:
450:
440:
439:
436:
435:
432:
428:
427:
424:
421:
418:
412:
408:
407:
404:
401:
397:
396:
393:
390:
386:
385:
382:
379:
375:
374:
371:
368:
365:
361:
360:
357:
354:
348:
347:
344:
341:
326:
323:
320:
319:
316:
313:
309:
308:
305:
302:
298:
297:
290:
287:
283:
282:
274:
271:
267:
266:
250:
247:
243:
242:
239:
236:
229:
226:
167:
164:
120:graphics cards
88:
85:
84:
39:
37:
30:
24:
14:
13:
10:
9:
6:
4:
3:
2:
2856:
2845:
2842:
2840:
2837:
2835:
2832:
2831:
2829:
2805:
2802:
2799:
2797:
2794:
2792:
2791:Mental Images
2789:
2787:
2784:
2782:
2779:
2777:
2774:
2772:
2769:
2767:
2764:
2762:
2759:
2757:
2754:
2752:
2749:
2748:
2746:
2742:
2735:
2732:
2729:
2727:
2724:
2722:
2719:
2717:
2714:
2712:
2709:
2707:
2704:
2703:
2701:
2697:
2693:
2686:
2682:
2668:
2665:
2664:
2662:
2660:
2655:
2649:
2646:
2645:
2643:
2641:
2637:
2631:
2628:
2626:
2623:
2621:
2618:
2616:
2613:
2612:
2610:
2607:
2603:
2597:
2594:
2592:
2589:
2587:
2586:Shield Tablet
2584:
2582:
2579:
2578:
2576:
2574:Nvidia Shield
2572:
2564:
2559:
2556:
2552:
2551:PlayStation 3
2547:
2544:
2540:
2535:
2532:
2531:
2529:
2526:
2522:
2516:
2513:
2511:
2508:
2507:
2505:
2503:
2499:
2491:
2488:
2487:
2486:
2485:Nvidia Quadro
2483:
2482:
2480:
2477:
2472:
2468:
2461:
2457:
2443:
2440:
2438:
2435:
2433:
2430:
2428:
2425:
2423:
2420:
2418:
2415:
2413:
2410:
2408:
2405:
2403:
2400:
2398:
2395:
2393:
2390:
2388:
2385:
2383:
2380:
2378:
2375:
2373:
2370:
2368:
2365:
2363:
2360:
2359:
2357:
2353:
2346:
2343:
2340:
2337:
2334:
2331:
2328:
2325:
2322:
2319:
2316:
2313:
2310:
2307:
2304:
2301:
2298:
2297:Nvidia G-Sync
2295:
2292:
2289:
2288:
2286:
2282:
2275:
2272:
2270:
2267:
2264:
2261:
2259:(3D graphics)
2258:
2255:
2251:(physics SDK)
2250:
2247:
2244:
2241:
2240:
2239:
2236:
2234:
2231:
2228:
2225:
2224:
2222:
2218:
2211:
2208:
2205:
2202:
2199:
2196:
2195:
2193:
2189:
2185:
2178:
2174:
2160:
2157:
2155:
2152:
2150:
2147:
2146:
2144:
2141:
2137:
2131:
2128:
2126:
2123:
2121:
2118:
2116:
2113:
2111:
2108:
2106:
2103:
2102:
2100:
2098:
2093:
2087:
2084:
2082:
2079:
2077:
2074:
2072:
2069:
2067:
2064:
2062:
2059:
2057:
2054:
2053:
2051:
2049:
2045:
2040:
2037:
2035:
2032:
2030:
2027:
2025:
2022:
2020:
2017:
2015:
2012:
2008:
2004:
1998:
1995:
1993:
1990:
1988:
1984:
1983:
1973:
1970:
1968:
1965:
1963:
1960:
1959:
1954:
1951:
1949:
1946:
1945:
1942:
1940:
1936:
1933:
1929:
1925:
1919:
1914:
1909:
1905:
1901:
1894:
1889:
1887:
1882:
1880:
1875:
1874:
1871:
1865:
1862:
1859:
1853:
1848:
1847:
1843:
1828:
1824:
1818:
1815:
1803:
1799:
1793:
1790:
1778:
1774:
1768:
1765:
1754:
1750:
1744:
1741:
1737:
1736:
1724:
1720:
1714:
1711:
1703:
1699:
1692:
1686:
1683:
1671:
1665:
1662:
1651:
1647:
1640:
1637:
1625:
1621:
1615:
1612:
1601:
1597:
1591:
1588:
1584:
1579:
1576:
1564:
1560:
1554:
1552:
1548:
1536:
1532:
1526:
1524:
1522:
1518:
1514:
1513:
1500:
1496:
1490:
1487:
1475:
1471:
1465:
1462:
1451:
1447:
1441:
1438:
1427:
1423:
1417:
1415:
1411:
1398:
1392:
1390:
1386:
1373:
1367:
1365:
1363:
1359:
1347:
1341:
1338:
1334:
1323:. nividia.com
1322:
1315:
1312:
1301:
1297:
1291:
1289:
1287:
1285:
1281:
1278:
1272:
1270:
1268:
1266:
1264:
1262:
1260:
1258:
1254:
1250:
1249:
1237:on 2020-09-26
1236:
1232:
1226:
1223:
1211:
1207:
1201:
1198:
1186:
1182:
1176:
1173:
1169:
1168:
1155:
1149:
1146:
1142:
1141:
1129:
1125:
1119:
1117:
1113:
1109:
1108:
1095:
1091:
1085:
1082:
1078:
1077:
1064:
1058:
1056:
1052:
1048:
1047:
1034:
1028:
1026:
1024:
1020:
1008:
1002:
1000:
998:
994:
982:
978:
971:
968:
957:
953:
947:
945:
943:
939:
935:
934:
921:
917:
911:
908:
897:
893:
887:
884:
878:
874:
870:
867:
865:
861:
858:
856:
852:
849:
848:
844:
842:
840:
836:
835:input latency
831:
829:
825:
821:
820:4K resolution
817:
816:
809:
807:
798:
796:
794:
790:
786:
781:
779:
775:
772:
767:
765:
761:
757:
754:
751:
747:
743:
739:
735:
731:
727:
723:
715:
713:
711:
707:
703:
701:
696:
694:
693:anti-aliasing
687:Anti-aliasing
686:
684:
682:
681:
676:
675:
670:
669:
664:
662:
657:
648:
646:
643:
639:
637:
629:
627:
624:
619:
614:
610:
606:
603:
600:
597:
596:convolutional
592:
590:
586:
582:
578:
574:
569:
567:
563:
559:
558:depth buffers
555:
551:
548:
540:
538:
536:
532:
527:
525:
519:
517:
512:
511:supersampling
507:
505:
501:
497:
493:
490:
487:
486:convolutional
479:
474:
463:
460:
454:
451:
445:
442:
430:
429:
425:
422:
416:
410:
409:
405:
402:
399:
398:
394:
391:
388:
387:
383:
380:
377:
376:
372:
369:
364:Ultra Quality
363:
362:
358:
355:
353:
350:
349:
346:Render scale
345:
343:Scale factor
342:
339:
338:
335:
330:
324:
317:
314:
311:
310:
306:
303:
300:
299:
295:
291:
288:
285:
284:
281:
280:
275:
272:
269:
268:
265:among others
264:
262:
257:
256:
255:Battlefield V
251:
249:February 2019
248:
245:
244:
238:Release date
233:
227:
225:
223:
219:
218:Unreal Engine
215:
211:
210:
205:
201:
196:
194:
191:shipped with
190:
189:
183:
181:
177:
176:Battlefield V
173:
165:
163:
161:
157:
153:
152:RTX 40 series
149:
145:
140:
138:
133:
129:
125:
121:
117:
113:
109:
105:
104:deep learning
102:
98:
94:
81:
78:
70:
60:
54:
53:
49:
45:
40:This article
38:
29:
28:
19:
2796:PortalPlayer
2744:Acquisitions
2716:Curtis Priem
2608:and embedded
2558:Tegra NX-SoC
2510:Nvidia Tesla
2432:Ada Lovelace
2284:Technologies
1918:List of GPUs
1831:. Retrieved
1829:. 2023-03-08
1826:
1817:
1806:. Retrieved
1804:. 2023-11-21
1801:
1792:
1781:. Retrieved
1779:. 2023-10-26
1776:
1767:
1756:. Retrieved
1752:
1743:
1734:
1733:
1727:. Retrieved
1725:. 2018-01-15
1713:
1697:
1685:
1674:. Retrieved
1664:
1653:. Retrieved
1649:
1639:
1628:. Retrieved
1626:. 2018-11-29
1623:
1614:
1603:. Retrieved
1599:
1590:
1578:
1567:. Retrieved
1565:. 2021-09-28
1562:
1539:. Retrieved
1537:. 2020-03-23
1510:
1509:
1503:. Retrieved
1501:. 2019-02-15
1489:
1478:. Retrieved
1476:. 2018-09-19
1473:
1464:
1453:. Retrieved
1449:
1440:
1429:. Retrieved
1425:
1401:. Retrieved
1376:. Retrieved
1350:. Retrieved
1340:
1332:
1325:. Retrieved
1314:
1303:. Retrieved
1299:
1246:
1245:
1239:. Retrieved
1235:the original
1225:
1214:. Retrieved
1212:. 2021-04-14
1209:
1200:
1189:. Retrieved
1187:. 2021-02-11
1184:
1175:
1165:
1164:
1158:. Retrieved
1148:
1139:
1138:
1132:. Retrieved
1130:. 2020-04-04
1105:
1104:
1098:. Retrieved
1096:. 2020-02-22
1084:
1074:
1073:
1067:. Retrieved
1044:
1043:
1037:. Retrieved
1011:. Retrieved
984:. Retrieved
980:
970:
959:. Retrieved
955:
931:
930:
924:. Retrieved
922:. 2020-02-14
910:
899:. Retrieved
895:
886:
832:
813:
810:
802:
782:
768:
746:Tensor Cores
745:
721:
719:
716:Architecture
708:also offers
704:
697:
690:
678:
672:
666:
659:
656:path tracing
652:
644:
640:
633:
615:
611:
607:
599:auto-encoder
593:
573:game engines
570:
544:
528:
520:
508:
489:auto-encoder
483:
462:
453:
444:
400:Performance
333:
328:
277:
261:Metro Exodus
259:
253:
214:game engines
207:
203:
197:
186:
184:
180:Metro Exodus
169:
159:
155:
148:Ada Lovelace
141:
96:
92:
91:
73:
64:
48:exaggeration
41:
2596:GeForce Now
2490:Quadro Plex
2476:Workstation
2335:(multi-GPU)
2293:(stereo 3D)
2140:Ray tracing
2105:GeForce 600
1987:GeForce 256
1939:Pre-GeForce
1600:TechPowerUp
1585:Cloud Front
1403:6 September
1378:6 September
986:23 February
824:ray tracing
815:Alan Wake 2
680:Alan Wake 2
589:blur filter
535:motion blur
273:August 2019
241:Highlights
150:generation
137:frame rates
124:video games
2828:Categories
2726:Bill Dally
2721:David Kirk
2699:Key people
2527:components
2362:Fahrenheit
2339:TurboCache
2329:(protocol)
2263:Nvidia RTX
2149:GeForce 20
1833:2024-07-09
1808:2024-07-09
1783:2024-07-09
1758:2024-07-09
1729:2020-04-08
1676:2020-04-08
1655:2024-06-13
1630:2022-09-20
1605:2022-02-10
1569:2022-02-10
1541:2020-04-07
1505:2020-04-19
1480:2022-02-07
1455:2024-06-08
1431:2021-07-06
1352:2020-04-06
1305:2022-02-07
1241:2020-04-19
1216:2022-02-07
1191:2022-02-07
1160:2020-04-06
1134:2020-04-05
1100:2020-04-06
1069:2020-04-06
1039:2020-04-06
1013:2020-04-05
961:2022-09-20
926:2020-04-05
901:2022-02-07
879:References
860:Intel XeSS
760:Tesla V100
738:Quadro RTX
585:flickering
577:heuristics
550:upsampling
524:edge-cases
289:April 2020
132:resolution
67:March 2024
2657:Computer
2474:Graphics
2437:Blackwell
2210:PureVideo
2056:GeForce 8
2019:GeForce 3
1327:11 August
1128:Eurogamer
618:upscalers
516:augmented
389:Balanced
108:upscaling
101:real-time
2659:chipsets
2220:Software
1967:RIVA TNT
1962:RIVA 128
1827:TechSpot
1702:Archived
1107:quality.
845:See also
789:DirectML
649:DLSS 3.5
630:DLSS 3.0
623:textures
581:ghosting
562:exposure
541:DLSS 2.0
480:DLSS 1.0
378:Quality
366:(unused)
235:Release
216:such as
118:line of
44:promotes
2776:DeepMap
2689:Company
2615:GoForce
2525:Console
2402:Maxwell
2377:Rankine
2367:Celsius
2013:shaders
1913:GeForce
1450:PCWorld
806:Full HD
793:DirectX
778:threads
744:called
722:Control
279:Control
204:Control
188:Control
166:History
57:Please
2839:Nvidia
2806:Stexar
2803:MediaQ
2800:Exluna
2667:nForce
2625:Jetson
2427:Hopper
2422:Ampere
2417:Turing
2407:Pascal
2397:Kepler
2372:Kelvin
2327:NVLink
2007:Vertex
1985:
1900:Nvidia
1723:Nvidia
1698:Nvidia
1535:Nvidia
1499:Nvidia
1248:games.
1094:Kotaku
956:NVIDIA
896:NVIDIA
799:Issues
783:Since
736:, and
706:Nvidia
677:, and
560:, and
498:, and
426:33.3%
423:3.00x
406:50.0%
403:2.00x
395:58.0%
392:1.72x
384:66.7%
381:1.50x
373:76.0%
370:1.32x
200:driver
112:Nvidia
2781:Icera
2756:Ageia
2630:Tegra
2620:Drive
2502:GPGPU
2478:cards
2442:Rubin
2412:Volta
2392:Fermi
2387:Tesla
2382:Curie
2274:VDPAU
2249:PhysX
2243:OptiX
2204:NVDEC
2198:NVENC
2011:pixel
1705:(PDF)
1694:(PDF)
864:Intel
750:Volta
671:DLC,
467:540p.
431:Auto
359:100%
222:Unity
160:DLDSR
2640:CPUs
2606:SoCs
2539:Xbox
2534:NV2A
2233:CUDA
2115:800M
2097:NUMA
2024:4 Ti
2009:and
1997:4 MX
1972:TNT2
1405:2023
1380:2023
1329:2020
988:2022
873:Sony
837:and
774:Warp
771:CUDA
583:and
566:DLAA
533:and
522:and
352:DLAA
312:3.5
301:3.0
258:and
220:and
206:and
178:and
97:DLSS
18:DLSS
2761:ULi
2515:DGX
2321:SXM
2315:MXM
2120:900
2110:700
2086:500
2081:400
2076:300
2071:200
2066:100
1953:NV2
1948:NV1
1650:IGN
855:AMD
818:in
753:GPU
700:TAA
356:1x
294:TAA
286:2.0
246:1.0
144:GTC
116:RTX
50:of
2830::
2227:Cg
2159:40
2154:30
2130:16
2125:10
2029:FX
1825:.
1800:.
1775:.
1751:.
1732:.
1721:.
1700:.
1696:.
1648:.
1622:.
1598:.
1561:.
1550:^
1533:.
1520:^
1508:.
1497:.
1472:.
1448:.
1424:.
1413:^
1388:^
1361:^
1331:.
1298:.
1283:^
1256:^
1244:.
1208:.
1183:.
1163:.
1137:.
1126:.
1115:^
1103:.
1092:.
1072:.
1054:^
1042:.
1022:^
996:^
979:.
954:.
941:^
929:.
918:.
894:.
841:.
732:,
728:,
683:.
665:s
556:,
526:.
415:8K
2565:)
2561:(
2553:)
2549:(
2541:)
2537:(
2061:9
2039:7
2034:6
1992:2
1920:)
1916:(
1892:e
1885:t
1878:v
1836:.
1811:.
1786:.
1761:.
1679:.
1658:.
1633:.
1608:.
1572:.
1544:.
1483:.
1458:.
1434:.
1407:.
1382:.
1355:.
1308:.
1219:.
1194:.
1016:.
990:.
964:.
904:.
663:'
419:)
263:,
158:(
95:(
80:)
74:(
69:)
65:(
55:.
20:)
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.