1 00:00:00,160 --> 00:00:07,359 here it is NVIDIA's new RTX 580 compared 2 00:00:04,880 --> 00:00:13,599 to its predecessor the 4080 super it is around 15% faster in games that use its 3 00:00:10,759 --> 00:00:17,720 new features it's faster in AI as long as you don't need more than 16 gigs of 4 00:00:15,440 --> 00:00:22,439 RAM for your model it's more efficient even though it draws more power overall 5 00:00:19,880 --> 00:00:27,279 and it's got a great cooler design as long as you don't mind heating up your 6 00:00:24,400 --> 00:00:32,759 CPU and that's it that's the most positive spin we can put on this thing 7 00:00:29,519 --> 00:00:35,480 why the do we need all of these caveats 8 00:00:32,759 --> 00:00:39,399 when we say the new one is better than the old one I mean isn't the point of 9 00:00:37,719 --> 00:00:44,200 the new one to be clearly and substantially better well it is better 10 00:00:41,960 --> 00:00:49,320 in cranked out cyberpunk at 4K it not only beats the 480 super but it even 11 00:00:46,719 --> 00:00:52,680 gives the 490 a run for its money but it's such a mixed bag that in other 12 00:00:50,879 --> 00:00:58,280 situations it barely out competes the 7900 XTX the flagship GPU that made AMD 13 00:00:55,640 --> 00:01:03,600 think you know maybe we just shouldn't do flagships anymore hey class half full 14 00:01:01,800 --> 00:01:08,080 at least they didn't raise the price and we did find some situations where the 80 15 00:01:05,760 --> 00:01:11,600 offers up pretty compelling use case and if you don't mind the weirdness of 16 00:01:09,200 --> 00:01:14,479 generated frames maybe NVIDIA is onto something with the neural rendering 17 00:01:13,000 --> 00:01:19,280 stuff we're going to tell you more about it after I nearly render a better way to 18 00:01:16,600 --> 00:01:19,280 sell your used 19 00:01:20,960 --> 00:01:28,520 Hardware is it available 20 00:01:24,079 --> 00:01:30,079 yeah is it available yeah again is it 21 00:01:28,520 --> 00:01:34,119 available just let her sponsor Jawa buy your old 22 00:01:31,759 --> 00:01:36,759 gpus no need for more stupid questions yeah so are you 23 00:01:44,479 --> 00:01:49,159 interested before we explore the wildlands of AI assisted rendering let's 24 00:01:48,040 --> 00:01:54,479 get our bearings with some good oldfashioned 1440p results all raster no 25 00:01:52,079 --> 00:01:59,119 DLS upscaling or rate tracing in cyberpunk the 580 makes a great first 26 00:01:56,600 --> 00:02:03,200 impression with a 20% lead over the 480 super then in Allen wake the 580 brings 27 00:02:01,159 --> 00:02:08,800 a decent generation uplift that brings 1% low FPS into the triple digits but 28 00:02:06,000 --> 00:02:13,640 also it uh only barely outpaces the 7900 XTX here in f124 things look even less 29 00:02:11,720 --> 00:02:19,160 exciting with the 580 failing to overtake AMD's now 2-year-old Flagship 30 00:02:16,599 --> 00:02:25,239 and only offering a measly 7ish per uplift over the 480 super does it get 31 00:02:21,239 --> 00:02:27,879 any better no no again and triple no why 32 00:02:25,239 --> 00:02:31,080 does the last gen 4070 TI super make everything else here look like an 33 00:02:29,239 --> 00:02:34,720 overpriced joke at least Blackmouth Wukong provides a reason for the 580 to 34 00:02:32,840 --> 00:02:38,879 exist but overall I'm just kind of getting sad reading you all these charts 35 00:02:36,400 --> 00:02:44,720 and you can expect a roughly 10% uplift over the 480 super and just 7% over the 36 00:02:41,800 --> 00:02:48,200 7900 XTX if you're curious about the 480 nons super there's a reason we called it 37 00:02:46,280 --> 00:02:51,720 the 480 cheaper it performs basically identically and you can't buy it anymore 38 00:02:50,159 --> 00:02:56,720 so we opted to use our limited graph space for older cards that might actually be worth upgrading from all 39 00:02:54,720 --> 00:03:01,680 right well maybe the story changes at 4K oh yeah okay so it's a bit Happ 40 00:02:59,840 --> 00:03:07,280 happier but this feels a little bit like going from Bridge to Terabithia to Dead 41 00:03:04,080 --> 00:03:10,000 Poet Society on average the 80 takes a 42 00:03:07,280 --> 00:03:14,440 commanding 18% lead over its predecessor which is good but that parade gets 43 00:03:11,879 --> 00:03:19,680 somewhat rained on by its mere 10% lead over the 7900 XTX there are a couple of 44 00:03:17,920 --> 00:03:23,599 standouts that Merit specific mention here black myth Wukong has the 580 45 00:03:21,879 --> 00:03:29,080 taking a price appropriate lead over the 7900 XTX with a worthwhile uplift over 46 00:03:26,599 --> 00:03:33,400 the 480 super and that's also true in cyberpunk where 47 00:03:30,360 --> 00:03:35,480 again the 80 nearly catches the 4090 48 00:03:33,400 --> 00:03:41,120 losing only in the less important average FPS but that one result there 49 00:03:39,040 --> 00:03:45,799 it's less of a trend and more of just an outlier in other titles it seems like 50 00:03:43,439 --> 00:03:51,799 the 80 is content to just be its disappointing self better but not better 51 00:03:49,200 --> 00:03:56,360 enough of course NVIDIA doesn't think that traditional raster rendering is the 52 00:03:53,400 --> 00:04:00,280 way forward no their way forward is with the AI GPU Hardware that made them the 53 00:03:58,480 --> 00:04:04,280 most valuable company in the world and any conflicting data would almost 54 00:04:01,959 --> 00:04:09,360 certainly mean a cratering stock price so with the AI boom in mind NVIDIA 55 00:04:07,640 --> 00:04:12,640 focused most of their architectural upgrades on that if you want to learn 56 00:04:11,360 --> 00:04:16,440 more about the architectural and Technology upgrades we highly recommend 57 00:04:14,519 --> 00:04:20,320 you check out our 5090 review where we spend a bit more time explaining what 58 00:04:18,040 --> 00:04:25,400 everything means what really matters for gamers is how this car differs from the 59 00:04:22,120 --> 00:04:28,240 4080 and 4080 super it's not much we're 60 00:04:25,400 --> 00:04:32,199 still on tsmc's 4N process node but at the same MSRP we get get four more 61 00:04:30,160 --> 00:04:38,080 streaming multiprocessors in the 480 super along with a clock speed bump and 62 00:04:34,160 --> 00:04:40,680 a memory swap from 16 GB of gdr 6X to 16 63 00:04:38,080 --> 00:04:44,240 GB of gdr 7 that all explains why raw gaming performance isn't improved much 64 00:04:42,680 --> 00:04:49,280 the bigger improvements come in the form of Next Generation NVIDIA Ray tracing RT 65 00:04:46,759 --> 00:04:53,120 cores and AI accelerating tensor cores the tensor cores also now have Hardware 66 00:04:51,160 --> 00:04:56,840 support for fp4 calculations for improved performance and memory usage 67 00:04:54,880 --> 00:05:00,360 with lightweight AI models and are more tightly integrated with Cuda course to 68 00:04:58,560 --> 00:05:04,039 make way for the Sci-Fi sound in neural rendering for hypothetical future Gamers 69 00:05:02,680 --> 00:05:08,520 these neural rendering techniques actually sound really cool and they might turn out great but Gamers 70 00:05:06,960 --> 00:05:11,960 shouldn't feel pressure to upgrade until a must playay title that uses them hits 71 00:05:10,199 --> 00:05:15,840 the steam store instead feel pressured to buy our Precision multi-bit 72 00:05:13,440 --> 00:05:18,960 screwdriver bundle from LTT Store.com anywh who those working with high-end 73 00:05:17,320 --> 00:05:23,560 video might upgrade for the New Media engine on the 5000 series that provides 74 00:05:21,120 --> 00:05:27,039 support for 422 chroma sub sampling in a nutshell this means much higher quality 75 00:05:25,240 --> 00:05:30,240 Hardware accelerated video encoding which will be a boon for professionals 76 00:05:28,360 --> 00:05:34,560 and anyone who likes streaming games to a hand old or lightweight decoding box 77 00:05:32,039 --> 00:05:40,120 we're also stoked about DisplayPort 2.1 uhb 20 the 580 can drive 4K 240 HZ with 78 00:05:38,319 --> 00:05:44,720 no display stream compression and even go higher with DSC which doesn't impact 79 00:05:43,039 --> 00:05:49,039 image quality as far as our eyes can perceive that means no compromises 80 00:05:46,800 --> 00:05:53,479 between image quality and latency and responsiveness let's move on to Ray 81 00:05:50,919 --> 00:05:56,960 tracing where yeah NVIDIA is really just competing against themselves here but is 82 00:05:55,080 --> 00:06:00,440 more big titles list Ray tracing as a minimum requirement it's becoming 83 00:05:58,720 --> 00:06:05,360 something that just can can't be brushed off which is a problem for AMD at 1440p 84 00:06:03,960 --> 00:06:08,639 while AMD can survive lightly rate traced games the path tracing found in 85 00:06:06,960 --> 00:06:12,639 Allen wake and Blackmouth Wukong make AMD look like an 8-year-old kid made it 86 00:06:10,520 --> 00:06:17,199 onto the field at an NFL game hopefully that will change with AMD's 9000 series 87 00:06:14,639 --> 00:06:20,280 of gpus but we don't know for sure today across our suite of games the 80 88 00:06:18,960 --> 00:06:26,880 maintains the same performance margin over the 480 super that we saw in our raster tests but fails to scale as well 89 00:06:24,160 --> 00:06:31,280 as the 490 does into 4K maybe the 490 is extra RT cors give it the edge or maybe 90 00:06:29,039 --> 00:06:36,599 it's the vram check out this result in Allan Wake 2 why did so many cards fail 91 00:06:33,880 --> 00:06:41,560 miserably guess now did you say lack of vram well you're right and if you didn't 92 00:06:38,680 --> 00:06:46,960 say lack of vram pretend that you did 4K path traced Allen wake asks for over 15 93 00:06:44,160 --> 00:06:52,520 GB of vram on 50 Series cards and over 16 on AMD cards anything less and it's 94 00:06:50,080 --> 00:06:57,080 wholly unplayable well we can't say that it causes a performance hit on the 580 95 00:06:54,759 --> 00:07:02,240 it highlights just how skimpy 16 GB of vram feels on a $1,000 96 00:06:59,840 --> 00:07:05,560 card and how close we are to a future where it's necessary for high-end gaming 97 00:07:04,319 --> 00:07:08,960 and the more frustrating thing is that if you want more vram from Team Green 98 00:07:07,560 --> 00:07:15,000 you literally have to spend twice as much for a 5090 or go back a generation 99 00:07:11,400 --> 00:07:17,039 or two I digress the rest of the 4K RT 100 00:07:15,000 --> 00:07:21,800 results show that the 580 is still a mean green ray tracing machine it's just 101 00:07:19,280 --> 00:07:25,919 not that much meaner than the 4080 super another disappointing result for the 102 00:07:23,360 --> 00:07:29,800 5,000 series for productivity the story is more mixed while it doesn't do much 103 00:07:27,879 --> 00:07:34,000 to impress in our blender render in Premiere Pro we can see that the updated 104 00:07:31,720 --> 00:07:39,240 media engine on the new 50 Series puts in the work beating the 4090 by 5% with 105 00:07:37,199 --> 00:07:42,000 a similar lead in Da Vinci resolve we'd love to talk about improvements to 106 00:07:40,560 --> 00:07:45,840 encoding but we're working on our benchmarks for that let us cook but who 107 00:07:44,240 --> 00:07:50,759 needs to shoot and edit videos when you could just generate garbage while 108 00:07:47,400 --> 00:07:53,360 wasting precious Earthly resources in AI 109 00:07:50,759 --> 00:07:57,639 the 5080 manages a 20% lead over the 4080 super in our Tex generation 110 00:07:55,599 --> 00:08:02,159 benchmarks with the only Hiccup being a smaller lead in our 53.5 111 00:08:00,159 --> 00:08:07,159 in our stable diffusion image generation benchmarks we see 16 is% improvement 112 00:08:04,879 --> 00:08:12,280 over the 4080 super which is cool I guess sadly a lack of vram will limit 113 00:08:09,879 --> 00:08:16,000 the models you can play with wamp wamp but this is positioned as a gaming card 114 00:08:14,199 --> 00:08:21,639 so let's talk about the AI that makes games run good D lss4 is a massive 115 00:08:18,840 --> 00:08:25,120 overhaul from dss3 it switches from a CNN model to a Transformer model if you 116 00:08:23,879 --> 00:08:29,159 want to learn more about the theory you should check out our 5090 review what it 117 00:08:27,039 --> 00:08:33,080 means in practice is that dss4 allows for better image quality at the same 118 00:08:30,919 --> 00:08:35,919 settings with some caveats of course it's the 50 Series there's going to be 119 00:08:34,320 --> 00:08:38,880 caveats there's a performance hit using the new Transformer model while the 120 00:08:37,560 --> 00:08:42,360 effect is mild for things like super resolution digital Foundry saw that 121 00:08:40,800 --> 00:08:47,360 older cards can take a real hit from enabling certain features with the new model but it scales better which means 122 00:08:45,800 --> 00:08:52,360 that the 50 Series and all of its tensor core huta generates up to three frames 123 00:08:50,200 --> 00:08:56,640 for every rendered frame essentially quadrupling your FPS multiframe gen is 124 00:08:54,959 --> 00:09:00,560 impressive and despite generating so many frames the tech does not impact 125 00:08:58,360 --> 00:09:04,480 latency substantially at least over top of your base frame rate sadly more 126 00:09:02,560 --> 00:09:08,399 frames is not more performance if you were to run a game natively at 30 FPS 127 00:09:06,519 --> 00:09:13,360 each frame would have 33 milliseconds of latency running at 60 FPS would half 128 00:09:10,760 --> 00:09:13,360 your latency to 129 00:09:20,120 --> 00:09:27,560 16.66% 33 milliseconds of latency which is 25 milliseconds worse than rendering 130 00:09:24,519 --> 00:09:29,760 120 FPS natively a large disparity 131 00:09:27,560 --> 00:09:35,079 meanwhile if you quadruple your 12 FPS game to 480 FPS well that only results 132 00:09:33,279 --> 00:09:39,120 in about a 6 millisecond difference in latency Which is far less noticeable but 133 00:09:36,959 --> 00:09:43,399 here's the kicker MFG has a performance overhead that will reduce your base 134 00:09:41,320 --> 00:09:47,040 frame rate and the more frames you gen the more overhead you need to contend 135 00:09:44,839 --> 00:09:51,519 with that's why we don't see a literal quadrupling of frames this means that 136 00:09:49,680 --> 00:09:55,920 using it to bump up an already low frame rate will make it feel even worse to 137 00:09:53,560 --> 00:09:59,480 play as it adds more latency there's a ton of nuance to this deck and we would 138 00:09:57,560 --> 00:10:02,880 recommend checking out Hardware unbox excellent video on it the good news is 139 00:10:01,279 --> 00:10:06,399 that multiframe gen scales about as well in the 5080 as it did on the 5090 140 00:10:04,959 --> 00:10:11,279 hopefully this trend carries forward to the lower 50 Series because it did not 141 00:10:09,240 --> 00:10:16,320 carry over to the lower end of the 40 series sheesh so tldr the higher your 142 00:10:14,600 --> 00:10:20,839 base frame rate the better your frame gen will feel but the less you need it 143 00:10:19,360 --> 00:10:25,880 NVIDIA likes to pretend that frame gen is comparable to Native rendering in their marketing but it's just not should 144 00:10:24,120 --> 00:10:30,279 you forgo MFG and just stick with the new dss4 Transformer model on the 80 145 00:10:28,320 --> 00:10:34,640 you'll see about a 5% % lower frame rate for the improved image quality and we 146 00:10:33,079 --> 00:10:39,959 think it's worth the tradeoff now all of this talk of frame rates and latency might have you wondering how does the 147 00:10:37,480 --> 00:10:43,240 580 fair in Esports well it does a dang good job we didn't test the rest of the 148 00:10:41,639 --> 00:10:48,560 lineup today because frankly any of the high-end cards will give you a great experience in Esports titles but this is 149 00:10:46,680 --> 00:10:52,920 a reminder that despite our Dow and disappointed tone it's still a very 150 00:10:50,440 --> 00:10:57,360 high-end very performant card and that also means it has big Power draw the 151 00:10:55,120 --> 00:11:01,680 5080 gets the same sleek and Innovative double flowr cooler as the 5090 allowing 152 00:10:59,800 --> 00:11:06,480 it to stay at a reasonable 70° in combuster despite pulling as much as 403 153 00:11:04,639 --> 00:11:09,839 Watts if you're going to small form factor you're far less likely to bump up 154 00:11:08,320 --> 00:11:14,440 against the thermal limits of this car than with the 5090 when gaming the 155 00:11:11,720 --> 00:11:20,240 average atts drop further to 65° thanks to the lower power budget in f124 we see 156 00:11:16,959 --> 00:11:23,279 the 580 pulling an average of 337 Watts 157 00:11:20,240 --> 00:11:25,480 nearly 15% more than the 480 super this 158 00:11:23,279 --> 00:11:29,880 Nets around 14% more frames which indicates no efficiency gains just like 159 00:11:27,680 --> 00:11:34,600 we saw in the 5090 by feeling deflated me too taking 160 00:11:32,560 --> 00:11:39,399 everything into account we have a Next Generation card that outperforms its 161 00:11:36,600 --> 00:11:45,000 last gen counterpart by 10 to 20% using roughly 15% more power I mean at least 162 00:11:42,959 --> 00:11:49,120 the price is the same that's that's good even then it feels like it could have 163 00:11:46,360 --> 00:11:53,000 been a midcycle refresh or a 4080 TI given it doesn't catch the 490 and raw 164 00:11:51,160 --> 00:11:57,440 performance something we really hoped it would do at best if you're considering 165 00:11:54,959 --> 00:12:01,839 the 480 super the 80 might be enough to push you over the edge to buy at least 166 00:11:59,320 --> 00:12:05,839 at MSRP but that's the hardest working s in the world because last gen cards are 167 00:12:04,040 --> 00:12:09,519 nearly impossible to find at NVIDIA's suggested price and if the 5090 is 168 00:12:07,920 --> 00:12:12,519 anything to go by this generation will be the same or even 169 00:12:12,560 --> 00:12:18,920 worse it feels like NVIDIA just doesn't care much about Gamers and that's 170 00:12:16,160 --> 00:12:21,320 because they don't they care about AI That's What made them one of the most 171 00:12:20,000 --> 00:12:27,279 valuable companies in the world and they're trying to use their large market share to move the market in their 172 00:12:25,079 --> 00:12:32,639 Direction they're saying no no no no no you don't want more raster what the 173 00:12:29,399 --> 00:12:34,440 wants nay what it needs is AI and AI 174 00:12:32,639 --> 00:12:38,560 just so happens to need the proprietary features enabled by the chips that we 175 00:12:36,199 --> 00:12:42,720 produce I get it if they were to admit that AI Hardware wasn't the solution to 176 00:12:40,800 --> 00:12:46,760 every conceivable problem then it would crater their stock price a thing that 177 00:12:44,800 --> 00:12:51,320 actually just happened on Monday the moment a Chinese AI company with limited 178 00:12:48,959 --> 00:12:57,040 access to NVIDIA's best AI Hardware supposedly overtook chat GPT team 179 00:12:54,199 --> 00:13:01,800 Green's stock dropped 17 points if the hype for AI and more specifically to 180 00:12:58,920 --> 00:13:05,399 them AI Hardware deflates Jensen might have to trade his crocodile leather 181 00:13:03,199 --> 00:13:09,320 jacket for a regular leather jacket but with all of that said without a clear 182 00:13:07,240 --> 00:13:12,839 competitor in the GPU space and video will still sell a buttload of these to 183 00:13:11,079 --> 00:13:19,440 Gamers and will continue to drive the industry in the direction of they're choosing so until AMD or Intel brings 184 00:13:16,440 --> 00:13:22,000 the Heat this is what we got really kind 185 00:13:19,440 --> 00:13:26,959 of feels like the Intel uh hey that's a nice quad core you got there how about 186 00:13:23,720 --> 00:13:28,760 this shiny new quad core days all I can 187 00:13:26,959 --> 00:13:33,120 do is hope at this point then that NVIDIA can nearly render some more 188 00:13:30,839 --> 00:13:37,639 innovative ideas before they suffer the same eventual fate that Intel did just 189 00:13:35,920 --> 00:13:41,880 like I'm going to nearly render this segue to our 190 00:13:39,279 --> 00:13:46,519 sponsor what you expected an AI to come up with a new segue I mean it was 191 00:13:43,399 --> 00:13:48,880 trained on my segue to our sponsor 192 00:13:46,519 --> 00:13:53,519 Bessie hiding a stash in your shoes is nothing special keeping it dry while 193 00:13:51,079 --> 00:13:59,839 stepping through water that's what vessie can offer still Dy now not your 194 00:13:57,600 --> 00:14:03,519 typical clunky robber rain F they kind of leave your feet looking and smelling 195 00:14:01,720 --> 00:14:08,120 like blue cheese not only they keep your s dry but also stay breathable 196 00:14:05,519 --> 00:14:11,759 lightweight and stretchy plus they look way more stylish check out the Ulta high 197 00:14:10,199 --> 00:14:16,160 top it's got a thicker snow gripping also a warm liner to keep your feet cozy 198 00:14:14,519 --> 00:14:20,560 and comfy whether you are chasing your New Year's goals or running away from 199 00:14:18,000 --> 00:14:23,800 your responsibilities in the RAM vessie got you covered we have been working 200 00:14:21,839 --> 00:14:28,320 with vasie for years and so many of our team members actually wear vasy every 201 00:14:26,040 --> 00:14:33,880 single day including myself check out v.com / LTT and get automatic 15% off 202 00:14:32,120 --> 00:14:38,480 your first order at checkout thanks for watching guys if you liked this video 203 00:14:35,399 --> 00:14:40,600 check out our 90 review or check out LTT 204 00:14:38,480 --> 00:14:45,120 Labs.com for even more info on these new cards the site is always getting updated