1 00:00:00,000 --> 00:00:03,740 Oh, Linus Sebastian is remoting in from Texas 2 00:00:03,740 --> 00:00:10,600 to present our next award of the evening. And the least crap GPU of 2025 award goes to AMD 3 00:00:12,160 --> 00:00:16,120 for the Radeon 9070XT. 4 00:00:16,120 --> 00:00:20,960 That's right, folks. After fumbling for years, AMD has finally figured out 5 00:00:20,960 --> 00:00:25,720 how to manufacture and price a GPU that will actually sell. 6 00:00:25,720 --> 00:00:30,240 And all it took for this to happen was for NVIDIA to slip on a banana peel, 7 00:00:30,240 --> 00:00:34,200 fall down an elevator shaft, and then have a piano drop down. 8 00:00:34,200 --> 00:00:39,080 As for the 9070 non-XT, it's also here 9 00:00:39,080 --> 00:00:45,560 in all seriousness, though. At $599 and $550, both of these cards are well-priced, 10 00:00:45,560 --> 00:00:49,680 capable of both 4K gaming and actually good ray tracing, 11 00:00:49,680 --> 00:00:53,240 and even add AI-powered upscaling capabilities. 12 00:00:53,240 --> 00:00:56,320 But as per AMD's long-proud tradition, 13 00:00:56,320 --> 00:01:01,960 they also come with caveats, like performance and productivity and power consumption. 14 00:01:01,960 --> 00:01:07,720 With that said, if they can stay in stock, I think we're gonna be able to look past those little details 15 00:01:07,720 --> 00:01:13,600 toward a brighter future for gamers and a bright segue to our sponsor. 16 00:01:13,600 --> 00:01:17,600 ["Suggestion of the Year"] 17 00:01:22,520 --> 00:01:25,840 AMD says these cards, they're for high-resolution gaming. 18 00:01:25,840 --> 00:01:31,760 So let's talk 1080p later and jump right into 1440p rasterization where, wow, 19 00:01:31,760 --> 00:01:36,480 does the 9070XT ever make NVIDIA look like a bunch of greedy buffoons. 20 00:01:36,480 --> 00:01:41,360 It nearly matches the performance of the RTX 5070 Ti 21 00:01:41,360 --> 00:01:46,360 while coming in $150 below the MSRP of that card, 22 00:01:46,600 --> 00:01:51,160 which of course is an imaginary price that no gamer has ever paid. 23 00:01:51,160 --> 00:01:54,400 Even the lesser non-XT variant looks decent here, 24 00:01:54,400 --> 00:01:59,000 going blow-for-blow with the 5070 across all of our benchmarks. 25 00:01:59,000 --> 00:02:03,320 We even got the occasional clear win for AMD, like in Alan Wake 2, 26 00:02:03,320 --> 00:02:08,320 where the non-XT beats the non-Ti by a whopping 17%. 27 00:02:09,120 --> 00:02:12,160 The only definitive loss for AMD in our suite 28 00:02:12,160 --> 00:02:16,960 was in Blacksmith Wukong, where the Ti beats the XT by around 9%. 29 00:02:16,960 --> 00:02:24,000 But the second you factor in the price, AMD's small L turns into a big W. 30 00:02:24,000 --> 00:02:29,240 In our Vulcan benchmark, Red Dead Redemption 2, we do see some worrying issues in frame-pacing 31 00:02:29,240 --> 00:02:33,640 where both of the new cards drop in ranking because of their poor 1% lows, 32 00:02:33,640 --> 00:02:37,640 despite solid average FPS, so hopefully this is something that AMD's team 33 00:02:37,640 --> 00:02:42,800 can work on post-launch. Overall though, we are off to a great start. 34 00:02:42,800 --> 00:02:47,120 The 9070XT is a clear winner in performance and value, 35 00:02:47,120 --> 00:02:53,080 and as for the 9070, well, it's about as bad a deal as the RTX 5070, 36 00:02:53,080 --> 00:02:57,640 at least on paper, and we'll talk a bit more about our theory as to why later. 37 00:02:57,640 --> 00:03:01,240 First, I wanna talk about 4K, because unlike NVIDIA, 38 00:03:01,240 --> 00:03:06,240 AMD gave their $550 card sufficient VRAM for Ultra HD gaming. 39 00:03:07,120 --> 00:03:10,960 Sure, it's GDDR6 and not GDDR7, 40 00:03:10,960 --> 00:03:15,840 but faster memory is gonna be irrelevant when you don't have the capacity to keep up, 41 00:03:15,840 --> 00:03:21,160 and there is no better illustration of this than the way that the 9070 extends its lead 42 00:03:21,160 --> 00:03:26,440 at higher resolutions, especially when we look at the all-important 1% lows, 43 00:03:26,440 --> 00:03:29,920 which is what determines the smoothness of your gameplay. 44 00:03:29,920 --> 00:03:36,920 Look at last of us part one here. The 9070 leads by a whopping 23% over the RTX 5070. 45 00:03:38,340 --> 00:03:41,560 Why? We'll look at the VRAM usage. 46 00:03:41,560 --> 00:03:47,000 On the 5070, it's full, which results in these gameplay hitches that you see here. 47 00:03:47,000 --> 00:03:52,440 Compare that to the 9070, or to NVIDIA's own cards that do have 16 gigs, 48 00:03:52,440 --> 00:03:57,160 and those hitches disappear. And I mean, sure, not everyone has a 4K monitor, 49 00:03:57,160 --> 00:04:00,280 and these limitations don't rear their heads in every game, 50 00:04:00,280 --> 00:04:04,200 or even most games, but still, it's kind of embarrassing 51 00:04:04,200 --> 00:04:08,000 when you're supposed to be the market leader in gaming GPUs, right? 52 00:04:08,000 --> 00:04:11,120 Overall, the 9000 series does fall a bit short 53 00:04:11,120 --> 00:04:14,680 of true 60 FPS 4K Ultra gaming, 54 00:04:14,680 --> 00:04:19,040 and in our tests, it did not achieve the lofty margins 55 00:04:19,080 --> 00:04:22,520 over the 7900 GRE that AMD promised in their slides, 56 00:04:22,520 --> 00:04:28,720 but we're not that disappointed yet, because AMD included ray tracing in their averages, 57 00:04:28,720 --> 00:04:32,680 and that is a very different story with these new cards. 58 00:04:32,680 --> 00:04:36,520 Before we look at those results though, let's dive into what AMD has done 59 00:04:36,520 --> 00:04:39,800 to bring about this generational leap in performance. 60 00:04:39,800 --> 00:04:44,720 To me, the most impressive thing is shrinkage. The new monolithic die that underpins both the 9070 61 00:04:44,720 --> 00:04:47,800 and 9070 XT should have been codenamed Frightened Turtle. 62 00:04:47,840 --> 00:04:54,240 It's built on TSMC's N4C process node, and it cramps 92% as many transistors as the 7900 XTX 63 00:04:54,240 --> 00:04:59,400 into just two thirds of the die area. This, along with some major improvements in performance 64 00:04:59,400 --> 00:05:02,560 per CU, is what makes these cards such a compelling value. 65 00:05:02,560 --> 00:05:05,760 Diving deeper into the compute engine, we see vastly improved matrix operations 66 00:05:05,760 --> 00:05:08,920 with support for more data types, a new dynamic register allocator, 67 00:05:08,920 --> 00:05:13,120 and improvements to the scheduler. Combine this with dual SIMD 32 vector units, 68 00:05:13,120 --> 00:05:18,440 overhauled AI accelerators, and beefed up ray tracing capabilities, and you get a card that can do a lot of calculations 69 00:05:18,440 --> 00:05:22,200 all at once, especially when it comes to ray tracing and AI. 70 00:05:22,200 --> 00:05:26,160 But wait, there is more. Both cards get an improved media engine, 71 00:05:26,160 --> 00:05:30,080 which provides a considerable improvement for low bit rate encoding during streaming. 72 00:05:30,080 --> 00:05:33,800 YouTube's compression might make this impossible to see, but to the human eye, 73 00:05:33,800 --> 00:05:39,800 these Twitch optimized recordings of Returnal, there's a clear improvement over the 7900 XTX. 74 00:05:39,800 --> 00:05:43,440 And it's even difficult to distinguish AMD from NVIDIA's end banking coding. 75 00:05:43,440 --> 00:05:47,920 We'll talk more about encoding later, but it's really great to see AMD finally catching up here. 76 00:05:47,920 --> 00:05:51,960 But where AMD is still behind is in 422 hardware encoding and decoding, 77 00:05:51,960 --> 00:05:55,520 which could make this a less desirable option for professional video creators, 78 00:05:55,520 --> 00:06:00,920 but it's unlikely to matter for non-pros. Another slight disappointment compared to the RTX 50 series 79 00:06:00,920 --> 00:06:07,400 is AMD's DisplayPoint 2.1a, which are only UHBR 13.5 rather than UHBR 20. 80 00:06:07,400 --> 00:06:12,720 So these new cards can still do 4K 240 Hertz, but they will rely on display stream compression to do so. 81 00:06:12,720 --> 00:06:16,800 Not a huge deal to me. I find it nigh imperceptible, but your mileage may vary. 82 00:06:16,800 --> 00:06:22,920 Enough specs. Let's talk ray tracing, where both the new cards beat the 7900 XTX, 83 00:06:22,920 --> 00:06:28,720 a card that was well received at a thousand US dollars. Too bad they don't fare quite as well against NVIDIA. 84 00:06:28,720 --> 00:06:32,200 In Alan Wake 2, the 9070 keeps up with the 5070, 85 00:06:32,200 --> 00:06:35,720 showing just how bad the 5070 is, 86 00:06:35,720 --> 00:06:39,240 but the 5070 Ti provides substantially better performance 87 00:06:39,240 --> 00:06:43,040 than the 9070 XT, giving NVIDIA one clear win 88 00:06:43,040 --> 00:06:48,480 for their overpriced 50 series cards, or one win so far. 89 00:06:48,480 --> 00:06:55,640 In the heavily path traced Black Myth Wukong, the 9000 cards handily outclass the 7000 series, 90 00:06:55,640 --> 00:07:01,120 but fall substantially behind NVIDIA's latest, and as for F124, 91 00:07:01,120 --> 00:07:07,080 AMD looks great across the board here, while the 5070 gets beaten by its own predecessor, 92 00:07:07,080 --> 00:07:09,640 the 4070 Super. It... 93 00:07:11,000 --> 00:07:14,560 The clown show. Removing Black Myth Wukong from the equation, 94 00:07:14,560 --> 00:07:17,920 the 9070 family is neck and neck with the competition, 95 00:07:17,920 --> 00:07:21,640 showing that AMD is no longer two generations behind 96 00:07:21,640 --> 00:07:26,760 in ray tracing. With that said, Black Myth Wukong does in fact exist, 97 00:07:26,760 --> 00:07:30,760 and shows that AMD is still one generation behind, 98 00:07:30,760 --> 00:07:35,560 which may matter in future games. And even in AMD's own tech demo, 99 00:07:35,560 --> 00:07:40,680 we can see that their path tracing implementation struggle with boiling artifacts and ghosting, 100 00:07:40,680 --> 00:07:43,920 that reminds me more of last gen's ray reconstruction. 101 00:07:43,920 --> 00:07:47,600 Hey, but here's hoping that AMD can continue to catch up 102 00:07:47,600 --> 00:07:50,840 and maybe even surpass NVIDIA in the future. 103 00:07:50,840 --> 00:07:55,320 Maybe the guy can dream. Hey, did we leave a 1080p raster? 104 00:07:55,320 --> 00:08:00,400 We ran those numbers, so damn it, I want them in the video. Even if the story remains largely the same, 105 00:08:00,400 --> 00:08:07,040 the 5070 Ti is still a pretty bad deal. It does win, but its margin of victory is so small 106 00:08:07,040 --> 00:08:11,040 compared to the pricing chasm between these cards that it's really hard to recommend. 107 00:08:11,040 --> 00:08:16,560 So if you are looking for an overkill 1080p upgrade, the 9070 XT looks like a great way 108 00:08:16,560 --> 00:08:19,800 to push huge FPS numbers in eSports titles. 109 00:08:19,800 --> 00:08:22,800 Or any title, if you don't mind a little AI upscaling. 110 00:08:22,800 --> 00:08:26,120 AMD said at one point that they don't think AI is necessary for upscaling, 111 00:08:26,120 --> 00:08:31,840 and they could get the same results using more traditional means. Ha, spoken like a company that did not have AI figured out 112 00:08:31,840 --> 00:08:37,080 yet. The good news is that AMD has used their slow start to squeeze a lot of image quality out of temporal upscaling, 113 00:08:37,080 --> 00:08:41,600 and now that they've added their proprietary Fp8 machine learning model to improve upscaling even further, 114 00:08:41,600 --> 00:08:46,840 the results are very impressive. Compared to FSR 3.1, we see reductions in artifacting 115 00:08:46,840 --> 00:08:51,680 and improvements in the rendering of fine lines. There are a few instances where it even beats DLSS4, 116 00:08:51,680 --> 00:08:56,640 like with these butterflies in Horizon Forbidden West. On FSR3, there's unsightly trailing and dissolving, 117 00:08:56,640 --> 00:09:02,080 and on DLSS4, they just kind of are blurry. But with FSR4, they appear much crisp and clearer 118 00:09:02,080 --> 00:09:05,800 than either of the other technologies. Sure, when you look closely, the artifacts are still there, 119 00:09:05,800 --> 00:09:09,280 but it's a market improvement. Just like haloing around detailed character models, 120 00:09:09,280 --> 00:09:12,520 it's very heavily reduced. And motion is just sharper in general. 121 00:09:12,520 --> 00:09:17,240 I would say that at least when upscaling to 4K, I found the image quality of FSR4 to be no more distracting 122 00:09:17,240 --> 00:09:20,880 than that of DLSS4. But none of the reduced image quality tradeoffs matter 123 00:09:20,880 --> 00:09:26,520 if we don't get better performance. And compared to DLSS, FSR gives a less substantial performance 124 00:09:26,520 --> 00:09:32,240 uplift at each quality setting than NVIDIA, in both the Last of Us Part 1 and in Horizon Zero Dawn. 125 00:09:32,240 --> 00:09:36,720 But we do see that AMD's Framegen provides a better uplift than NVIDIA's solution, 126 00:09:36,720 --> 00:09:39,880 which indicates that FSR 3.1 Framegen has less overhead 127 00:09:39,880 --> 00:09:43,600 than NVIDIA's Multiframegen. While AMD doesn't quite match NVIDIA yet, 128 00:09:43,600 --> 00:09:47,840 this is the most competitive their upscaler has been ever. 129 00:09:47,840 --> 00:09:52,360 Except support is a problem. Even though you can use the driver to force the updated 130 00:09:52,360 --> 00:09:56,760 FSR4 model in games with FSR3, AMD just has less games that use FSR3. 131 00:09:56,760 --> 00:10:00,160 And unlike DLSS4, older cards are not able to take advantage 132 00:10:00,160 --> 00:10:04,160 of the new machine learning the enhanced upscaling. And AMD still doesn't have a competitor 133 00:10:04,160 --> 00:10:07,760 for Multiframegen yet. However, FSR4 does have Framegen, 134 00:10:07,760 --> 00:10:11,080 but it makes use of FSR3.1's Framegen implementation. 135 00:10:11,080 --> 00:10:15,320 But you can bring Framegen to any game using AMD's driver level fluid motion frames, 136 00:10:15,320 --> 00:10:20,520 now on version 2.1. But even with the additional decimal point, AFMF still sucks. 137 00:10:20,520 --> 00:10:25,400 Without game integration, you get all these weird UI issues and overlays get mangled with motion heavy heck 138 00:10:25,400 --> 00:10:29,200 in even motion light scenes. Let's be real, it's just frame interpolation 139 00:10:29,200 --> 00:10:34,080 and no number of fancy names can change that. And if the use case for NVIDIA's Framegen was already weak, 140 00:10:34,080 --> 00:10:41,560 AFMF feels like it exists only so AMD can say, actually we support Framegen in way more games. 141 00:10:41,560 --> 00:10:45,800 But no, no one wants it. It's, I don't know, who cares? 142 00:10:45,800 --> 00:10:49,840 Despite the swath of shareholder approved AI junk 143 00:10:49,840 --> 00:10:55,120 in their latest driver software, AMD falls pretty well short of the mark here. 144 00:10:55,120 --> 00:11:01,400 There are some great gen over gen improvements in computer vision, but even the 4070 Super 145 00:11:01,400 --> 00:11:05,200 manages a sizable lead over the 9070XT. 146 00:11:05,200 --> 00:11:10,960 The generational gains are even more apparent in stable diffusion, but the story remains the same. 147 00:11:10,960 --> 00:11:17,920 Finally, there's AI text generation, and I bet AMD wishes an AI could rewrite these benchmark results 148 00:11:17,920 --> 00:11:21,120 over the new cards failed to improve substantially 149 00:11:21,120 --> 00:11:24,200 over last gen and get absolutely dunked on by NVIDIA, 150 00:11:24,200 --> 00:11:30,520 regardless of which model you're using. The one silver lining is that either of these cards 151 00:11:30,520 --> 00:11:34,160 can run all the same models that the 5080 can, 152 00:11:34,160 --> 00:11:37,640 and can run some that the 5070 cannot. 153 00:11:37,640 --> 00:11:41,960 Thanks to AMD's generous 16 gigs of VRAM. 154 00:11:41,960 --> 00:11:46,000 I mean, it's only generous compared to NVIDIA, but thank you guys. 155 00:11:46,000 --> 00:11:49,920 Don't be too thankful though. AMD still needs to impress in content creation 156 00:11:49,920 --> 00:11:53,280 where they do okay for video editing, 157 00:11:53,280 --> 00:11:58,600 but the 9070 and 70XT provide good, but not exceptional performance in Premiere Pro 158 00:11:58,600 --> 00:12:03,960 and DaVinci Resolve. And in Blender, the 9070 and 70XT perform about on par 159 00:12:03,960 --> 00:12:07,040 with the flagships from AMD's last generation, 160 00:12:07,040 --> 00:12:12,080 but they still just don't really get anywhere close to NVIDIA thanks to optics rendering. 161 00:12:12,080 --> 00:12:17,080 Yeah, there was a transition that I was written that made more sense before we had to record this. 162 00:12:17,080 --> 00:12:22,680 Speaking of updates, we have some encoder testing for you. While AMD has made strides to improve their encoder, 163 00:12:22,680 --> 00:12:25,960 they still fall behind in quality compared to NVIDIA and Intel, 164 00:12:25,960 --> 00:12:31,160 especially at lower bit rates in H.264. For this encoding test, we only set the bit rate 165 00:12:31,160 --> 00:12:34,320 and no other parameters. Depending on what encoder tweaks are used, 166 00:12:34,320 --> 00:12:38,240 any of these cards could have improved quality compared to our tests. 167 00:12:38,240 --> 00:12:41,480 But the simplest way to improve quality is use AV1. 168 00:12:41,480 --> 00:12:46,080 But sadly, we've excluded the 7,000 series from our results due to a hardware bug. 169 00:12:46,080 --> 00:12:49,240 The good news is everyone outputs better quality in AV1, 170 00:12:49,240 --> 00:12:51,960 and AMD is sadly still in last place. 171 00:12:52,880 --> 00:12:55,880 We are almost there. Let's talk power and thermals. 172 00:12:55,880 --> 00:13:00,200 Over the past several generations, AMD has made some substantial improvements to efficiency, 173 00:13:00,200 --> 00:13:04,400 and in their announcement, they didn't really talk much about it. I guess they were being modest. 174 00:13:04,400 --> 00:13:10,360 In F124, the 9070 pulls lower power on average than the 5070, but with less stable power delivery, 175 00:13:10,360 --> 00:13:14,480 having transient spikes as high as 321 watts. 176 00:13:14,480 --> 00:13:20,440 Speaking of which, the 9070 XT has a massive spike to 426 watts, and its average is 309 watts, 177 00:13:20,440 --> 00:13:24,480 which is already higher than its rated TBP of 305 watts. 178 00:13:24,480 --> 00:13:28,000 And that propensity for pulling profuse power can be found in combustor, 179 00:13:28,000 --> 00:13:32,940 where both AMD cards also use more than their rated total board power on average. 180 00:13:32,940 --> 00:13:36,720 But AMD is allowing some partner cards to use a higher power budget, 181 00:13:36,720 --> 00:13:41,120 so I guess this isn't fully out of spec, but I would definitely recommend you stick 182 00:13:41,120 --> 00:13:45,960 with the manufacturer recommended power supply capacity. And I recommend you check out PSU circuit 183 00:13:45,960 --> 00:13:49,720 if you need to make an upgrade decision. Well, that's a good info there. Thankfully, despite the power draw, 184 00:13:49,720 --> 00:13:53,040 thermals seem to be under control on our provided sapphire pulse samples. 185 00:13:53,040 --> 00:13:58,160 I'm sure in part to their use of PTM7950, which you can buy for yourself over at LTTstore.com. 186 00:13:58,160 --> 00:14:03,840 That's a double plug, double whammy. Self promo, baby! AMD doesn't feel the need to hide GPU hotspot metrics 187 00:14:03,840 --> 00:14:07,600 like NVIDIA does, and in combustor, we see that sapphire's coolers provide 188 00:14:07,600 --> 00:14:10,600 ample thermal headroom, which is good. And they're a little big. 189 00:14:10,600 --> 00:14:14,640 They're not obnoxious, but you know, come on. It does allow the cards to stay much cooler 190 00:14:14,640 --> 00:14:18,760 than say the 5070 or the 6700 XT in F124. 191 00:14:18,760 --> 00:14:22,040 So if you're coming from an older card, you can confidently upgrade to either the 9070 192 00:14:22,040 --> 00:14:27,200 or the 9070 XT, and know you'll get a solid improvement in raster performance and ray tracing, 193 00:14:27,200 --> 00:14:31,480 which is a big plus for folks who are on older ray tracing cards like the 2000 series, 194 00:14:31,480 --> 00:14:36,240 or if you're coming from the 6000 series on AMD. It's just, it's good, it's good, it's cool, it's nice. 195 00:14:36,240 --> 00:14:39,240 We like it, we're happy. They could be cheaper. 196 00:14:39,240 --> 00:14:42,560 They could be, they could be cheaper. There's kind of two conclusions here. 197 00:14:42,560 --> 00:14:48,440 A short one and a long one. The short one is that the 9070 XT is a winner. 198 00:14:48,440 --> 00:14:53,400 If AMD can keep this thing in stock and you have $600 to spend on a gaming GPU, 199 00:14:53,400 --> 00:14:57,440 you are gonna love this thing, and NVIDIA needs to respond now, 200 00:14:57,440 --> 00:15:01,020 or AMD might actually take some real market share for a change. 201 00:15:01,020 --> 00:15:06,140 The long conclusion is that it seems like AMD is trying to eat their cake and have it too here. 202 00:15:06,140 --> 00:15:09,880 See, the XT's 599 price point is giving real 203 00:15:09,880 --> 00:15:14,080 good guy AMD vibes, but the 9070 non-XT's price 204 00:15:14,080 --> 00:15:17,160 is giving maximized margins while GPUs 205 00:15:17,160 --> 00:15:23,240 are in short supply vibes. By matching the 5070 in both price and performance, 206 00:15:23,240 --> 00:15:26,720 unfortunately, you've also matched it in value, 207 00:15:26,720 --> 00:15:31,200 and let's be real. The 5070 is not a great value. 208 00:15:31,200 --> 00:15:34,440 Once nobody buys the non-XT because the XT 209 00:15:34,440 --> 00:15:40,200 is so much better for just $50 more, you're gonna end up dropping the price on this thing 210 00:15:40,200 --> 00:15:44,120 after the reputational damage has already been done to it. 211 00:15:44,120 --> 00:15:48,920 Did you guys learn nothing from the terrible initial reception to the 7900XT? 212 00:15:48,920 --> 00:15:52,280 And it's not like you have just reputation 213 00:15:52,280 --> 00:15:56,400 for days to give up. You have nothing to compete in the high end. 214 00:15:56,400 --> 00:16:00,840 And sure, 85% of gamers do buy cards under $700. 215 00:16:00,840 --> 00:16:04,280 I'm sure that's true, but without a high end card at all, 216 00:16:04,280 --> 00:16:09,080 you're giving up precious mind share. I mean, we didn't even bother to compare 217 00:16:09,080 --> 00:16:12,480 the 9000 series against NVIDIA's last gen flagship, 218 00:16:12,480 --> 00:16:15,440 the 4090, let alone their new flagship. 219 00:16:15,480 --> 00:16:19,560 And yes, 16 gigs of VRAM at 550 bucks is nice, 220 00:16:19,560 --> 00:16:23,440 but it's a bummer that the cheapest way to get more VRAM in your lineup 221 00:16:23,440 --> 00:16:27,520 is still a last generation 7900XT. 222 00:16:27,520 --> 00:16:32,920 If you have no plans to compete in the high end, we need you guys dominating the mid-range 223 00:16:32,920 --> 00:16:36,960 and enthusiast segments, especially because your non-gaming performance 224 00:16:36,960 --> 00:16:40,940 is still kind of lacking. Another thing that we need to acknowledge 225 00:16:40,940 --> 00:16:44,920 in our longer conclusion is the PlayStation-shaped elephant in the room. 226 00:16:44,920 --> 00:16:49,160 As impressive as these cards are, there is still a strong argument to be made 227 00:16:49,160 --> 00:16:52,600 that PC gaming has just plain gotten too expensive 228 00:16:52,600 --> 00:16:59,960 when you can pick up an all-in-one gaming box that will already run at 4K for less than the MSRP of a 9070 229 00:16:59,960 --> 00:17:03,340 and even lesser if you buy it secondhand. 230 00:17:03,340 --> 00:17:06,800 Sure, you'll pay more for games and online services in the long run, 231 00:17:06,800 --> 00:17:12,080 but boy, is it ever a reasonable upfront cost. But hey, maybe the best is yet to come 232 00:17:12,080 --> 00:17:17,260 on the discrete GPU side. AMD has announced that 9060 cards will be coming in Q2 233 00:17:17,260 --> 00:17:20,260 and those could provide even better value, 234 00:17:20,260 --> 00:17:25,220 assuming of course that AMD doesn't launch them at a high price in order to maximize margin for those ones, 235 00:17:25,220 --> 00:17:29,900 which grosses me out just thinking about it. What I'm not grossed out by is ending this video 236 00:17:29,900 --> 00:17:33,540 saying that this is truly probably the best GPU launch 237 00:17:33,540 --> 00:17:39,780 of the year and safe for Intel's B580, perhaps the best launch of the past several years 238 00:17:39,780 --> 00:17:44,540 if they can keep it in stock. But as per usual, for you, the consumer, 239 00:17:44,540 --> 00:17:47,580 there is no single right answer, just what's right for you. 240 00:17:47,580 --> 00:17:51,380 So whether you wanna go with these new cards or pay extra for Team Green 241 00:17:51,380 --> 00:17:54,500 or go for a console, all the power to you. 242 00:17:54,500 --> 00:17:57,580 And all the power to our sponsor! 243 00:17:57,580 --> 00:18:01,380 Thanks for watching, guys. We are pretty tired from doing back to back to back 244 00:18:01,380 --> 00:18:05,620 to back GPU launches. Massive shout out to you for watching 245 00:18:05,620 --> 00:18:11,100 and of course to our Labs team, our editors, our camera team, our writers, everyone for being part of it. 246 00:18:11,100 --> 00:18:15,220 I think there's a little time to breathe. Wait, 90, 60 and Q2? Okay, okay, no. 247 00:18:15,220 --> 00:18:18,700 Oh, and 50, 60 I guess is coming. Well, whatever, if you liked this video, 248 00:18:18,700 --> 00:18:22,220 check out, I don't know, 50, 80 review.