1 00:00:00,020 --> 00:00:03,100 PC gaming hardware is expensive, right? 2 00:00:03,700 --> 00:00:07,900 Core i7 Extreme Edition, $1,600. 3 00:00:08,660 --> 00:00:13,640 2 terabyte 960 Pro NVMe drive, $1,300. 4 00:00:14,480 --> 00:00:19,320 Two GTX Titan XPPs, we got $2,400. 5 00:00:20,020 --> 00:00:24,800 You add 128 gigs of RAM, as well as a top-of-the-line motherboard, 6 00:00:24,800 --> 00:00:29,500 and yeah, that'll be $6,800, please. 7 00:00:30,020 --> 00:00:33,310 Now, to be clear, I'm not complaining. 8 00:00:33,450 --> 00:00:35,870 This is a spectacularly first-world problem. 9 00:00:36,250 --> 00:00:40,630 But pro video production gear is on a whole freaking other level. 10 00:00:40,970 --> 00:00:42,250 Let me put it in perspective. 11 00:00:42,930 --> 00:00:51,770 This one PCI Express card costs as much as all that stuff I just listed combined. 12 00:00:52,130 --> 00:00:52,910 All of it. 13 00:00:52,910 --> 00:00:58,230 This is the Red Rocket X, a state-of-the-art creation tool 14 00:00:58,230 --> 00:00:59,910 and simultaneously a revolutionary... 15 00:01:00,210 --> 00:01:02,210 ...delic of a bygone era. 16 00:01:02,210 --> 00:01:07,210 A $7,000 co-processor card. 17 00:01:07,210 --> 00:01:10,210 Welcome to holy sh**. 18 00:01:10,210 --> 00:01:24,710 Cooler Master's 25th anniversary edition, Cosmos II, 19 00:01:24,710 --> 00:01:28,710 features a unique dual curved tempered glass side panel. 20 00:01:28,710 --> 00:01:31,710 Check it out now at the link below. 21 00:01:33,030 --> 00:01:35,030 Okay, so as many of you are probably aware, 22 00:01:35,030 --> 00:01:38,030 we moved to RED Digital Cinema cameras earlier this year 23 00:01:38,030 --> 00:01:40,030 in an effort to bump up our production value. 24 00:01:40,030 --> 00:01:42,030 And, I mean, let's be honest, we're enthusiasts. 25 00:01:42,030 --> 00:01:45,030 Give Brandon and Max some new toys to play with. 26 00:01:45,030 --> 00:01:50,060 But have you ever actually tried to work with RED code RAW footage before? 27 00:01:50,060 --> 00:01:52,060 As Taren would put it, 28 00:01:52,060 --> 00:01:54,060 it's a nightmare. 29 00:01:54,060 --> 00:01:59,060 At full 8K on a 10-core Extreme Edition processor 30 00:01:59,060 --> 00:02:04,060 with a $5,000 NVIDIA Quadro P6000 graphics card, 31 00:02:04,060 --> 00:02:08,060 this is what timeline performance looks like. 32 00:02:08,060 --> 00:02:09,060 You see those delays? 33 00:02:09,060 --> 00:02:11,060 As I'm scrubbing around? 34 00:02:11,060 --> 00:02:17,060 And even straight playback performance is abominable. 35 00:02:17,060 --> 00:02:21,060 It's not like it's our 10 gigabit network connection bottlenecking us either. 36 00:02:21,060 --> 00:02:24,160 We're nowhere near saturating it. 37 00:02:24,160 --> 00:02:25,160 No. 38 00:02:25,160 --> 00:02:27,260 It's the CPU. 39 00:02:27,260 --> 00:02:32,260 Even though we're scaling across all 20 of our threads, 40 00:02:32,260 --> 00:02:35,260 we are still running out of processing power. 41 00:02:35,260 --> 00:02:39,260 And we haven't even added any Lumetri colors. 42 00:02:39,260 --> 00:02:41,260 Yet, either. 43 00:02:41,260 --> 00:02:44,260 That's only going to make matters worse. 44 00:02:44,260 --> 00:02:46,260 Throw that into the picture, 45 00:02:46,260 --> 00:02:51,260 and now we are dropping about 90% of our frames 46 00:02:51,260 --> 00:02:56,260 while dealing with a quarter to a third of a second delay 47 00:02:56,260 --> 00:03:00,290 when we're trying to move the playhead around on the timeline. 48 00:03:00,290 --> 00:03:01,290 Ridiculous. 49 00:03:01,290 --> 00:03:05,290 And both our CPU and GPU are basically maxed. 50 00:03:05,290 --> 00:03:07,290 You actually need to drop your preview quality 51 00:03:07,290 --> 00:03:09,290 all the way down 52 00:03:09,290 --> 00:03:10,290 to 1 eighth 53 00:03:10,290 --> 00:03:13,290 in order to get what I would consider to be 54 00:03:13,290 --> 00:03:15,290 acceptable performance 55 00:03:15,290 --> 00:03:17,290 with much lower usage. 56 00:03:17,290 --> 00:03:19,350 True to form though, 57 00:03:19,350 --> 00:03:22,350 RED Digital Cinema has a solution for you. 58 00:03:22,350 --> 00:03:25,350 Are you unhappy with your editing experience 59 00:03:25,350 --> 00:03:28,350 of the footage from your $50,000 camera? 60 00:03:28,350 --> 00:03:31,350 Buy this expansion card 61 00:03:31,350 --> 00:03:32,350 to accelerate it. 62 00:03:32,350 --> 00:03:35,350 Available now for 7,000 easy payments 63 00:03:35,350 --> 00:03:37,350 of only $1. 64 00:03:37,350 --> 00:03:39,350 Call now and we'll throw in this nothing. 65 00:03:39,350 --> 00:03:42,350 Absolutely free because we're RED 66 00:03:42,350 --> 00:03:45,350 and your wallet is more than just our job. 67 00:03:45,350 --> 00:03:46,350 It's our passion. 68 00:03:46,350 --> 00:03:49,350 But what is this thing? 69 00:03:49,350 --> 00:03:51,350 What does it even do? 70 00:03:51,350 --> 00:03:54,350 Well, we can figure out a little bit on our own. 71 00:03:54,350 --> 00:03:57,350 So there's some kind of processor here 72 00:03:57,350 --> 00:04:00,350 that's under 150 watts of total power consumption 73 00:04:00,350 --> 00:04:02,350 based on the single slot cooler 74 00:04:02,350 --> 00:04:05,350 and the single six pin power connector back here. 75 00:04:05,350 --> 00:04:07,350 There's some memory on board 76 00:04:07,350 --> 00:04:08,350 so we can see those things. 77 00:04:08,350 --> 00:04:11,350 So we can see those chips arranged around the chip here 78 00:04:11,350 --> 00:04:14,350 and the dual DisplayPort ports 79 00:04:14,350 --> 00:04:17,350 are so it can actually function as a video output device. 80 00:04:17,350 --> 00:04:21,350 Though our workload necessitates a powerful GPU anyway 81 00:04:21,350 --> 00:04:24,350 so we won't be using it that way. 82 00:04:24,350 --> 00:04:27,350 And the main purpose of this thing anyway 83 00:04:27,350 --> 00:04:32,350 is to accelerate decoding, scaling, and debayering 84 00:04:32,350 --> 00:04:34,350 of our 3D files. 85 00:04:34,350 --> 00:04:36,350 The first two I think are fairly self-explanatory 86 00:04:36,350 --> 00:04:37,350 but the last one 87 00:04:37,350 --> 00:04:42,350 is the process of reconstructing a full color image 88 00:04:42,350 --> 00:04:44,350 from incomplete color samples 89 00:04:44,350 --> 00:04:46,350 that are captured by the sensor 90 00:04:46,350 --> 00:04:48,350 in a raw recording format. 91 00:04:48,350 --> 00:04:53,350 But they give us no actual details about the hardware 92 00:04:53,350 --> 00:04:58,350 other than some nebulous 5X performance claim 93 00:04:58,350 --> 00:05:02,350 relative to their own first generation RED rocket. 94 00:05:02,350 --> 00:05:04,350 But why would this older chip 95 00:05:04,350 --> 00:05:07,350 based on older process nodes 96 00:05:07,350 --> 00:05:09,350 be better at decoding technology 97 00:05:09,350 --> 00:05:12,350 compared to a GPU running CUDA 98 00:05:12,350 --> 00:05:15,350 be better at any of that stuff? 99 00:05:15,350 --> 00:05:19,350 I mean RED even supports CUDA acceleration 100 00:05:19,350 --> 00:05:22,350 in their own Cinex desktop application. 101 00:05:22,350 --> 00:05:24,350 Is this thing even relevant today? 102 00:05:24,350 --> 00:05:25,350 So with it installed 103 00:05:25,350 --> 00:05:27,350 we throw on the latest drivers 104 00:05:27,350 --> 00:05:29,860 and I don't know 105 00:05:29,860 --> 00:05:31,860 it's kind of weird. 106 00:05:31,860 --> 00:05:33,860 There's no configuration utility 107 00:05:33,860 --> 00:05:34,860 or anything like that 108 00:05:34,860 --> 00:05:36,860 but if you go into device manager 109 00:05:36,860 --> 00:05:39,860 it's working properly 110 00:05:39,860 --> 00:05:41,860 and everything seems fine. 111 00:05:41,860 --> 00:05:43,860 So all that's left to do then 112 00:05:43,860 --> 00:05:46,860 is go into our video editing software 113 00:05:46,860 --> 00:05:48,860 and press enable. 114 00:05:48,860 --> 00:05:50,860 So this is the same project 115 00:05:50,860 --> 00:05:52,860 that we were just looking at 116 00:05:52,860 --> 00:05:53,860 and we're going to change this 117 00:05:53,860 --> 00:05:57,860 use rocket drop down to all available. 118 00:05:57,860 --> 00:06:00,120 At full quality 119 00:06:00,120 --> 00:06:03,120 my CPU is still pinned 120 00:06:03,120 --> 00:06:06,120 and now I'm seeing 90% GPU usage 121 00:06:06,120 --> 00:06:11,120 even with lumetri color disabled. 122 00:06:11,120 --> 00:06:13,310 What? 123 00:06:13,310 --> 00:06:16,820 Okay, well now 124 00:06:16,820 --> 00:06:19,910 hold on just a minute here. 125 00:06:19,910 --> 00:06:22,910 Surely there must be something wrong. 126 00:06:22,910 --> 00:06:25,910 Let's try disabling the rocket 127 00:06:25,910 --> 00:06:28,100 just to see if we can get back 128 00:06:28,100 --> 00:06:30,100 to where we started. 129 00:06:30,100 --> 00:06:31,100 Now hold on 130 00:06:31,100 --> 00:06:33,100 this doesn't make sense either. 131 00:06:33,100 --> 00:06:34,100 With it disabled 132 00:06:34,100 --> 00:06:36,100 our timeline scrubbing 133 00:06:36,100 --> 00:06:38,100 looks better than ever. 134 00:06:38,100 --> 00:06:41,330 Our CPU usage is way down 135 00:06:41,330 --> 00:06:44,330 and even our GPU usage 136 00:06:44,330 --> 00:06:47,100 is under control now. 137 00:06:47,100 --> 00:06:49,100 But we are still dropping some frames 138 00:06:49,100 --> 00:06:51,100 far fewer though. 139 00:06:51,100 --> 00:06:53,100 Well RED only advertises 140 00:06:53,100 --> 00:06:55,100 the Rocket X as being capable 141 00:06:55,100 --> 00:06:57,100 of handling up to 6K footage. 142 00:06:57,100 --> 00:06:59,100 So maybe what we're looking at here 143 00:06:59,100 --> 00:07:00,100 makes sense 144 00:07:00,100 --> 00:07:02,100 if somehow it's enabled. 145 00:07:02,100 --> 00:07:04,100 Dropping then down to 146 00:07:04,100 --> 00:07:07,100 half quality in the preview window 147 00:07:07,100 --> 00:07:09,100 yields perfect playback 148 00:07:09,100 --> 00:07:11,100 without lumetri color. 149 00:07:11,100 --> 00:07:14,660 Let's check that out. 150 00:07:14,660 --> 00:07:15,660 Okay 151 00:07:15,660 --> 00:07:16,700 and 152 00:07:16,700 --> 00:07:17,700 all right 153 00:07:17,700 --> 00:07:18,700 turning it on 154 00:07:18,700 --> 00:07:19,700 well 155 00:07:19,700 --> 00:07:21,700 it looks like we are still dropping 156 00:07:21,700 --> 00:07:22,700 some frames 157 00:07:22,700 --> 00:07:25,020 in that case. 158 00:07:25,020 --> 00:07:27,020 But that's probably caused then 159 00:07:27,020 --> 00:07:28,020 by a GPU bottleneck 160 00:07:28,020 --> 00:07:29,020 because you can see 161 00:07:29,020 --> 00:07:31,020 our Quadro is sitting at around 162 00:07:31,020 --> 00:07:34,020 80 to 85% usage. 163 00:07:34,020 --> 00:07:36,020 So then finally as a last step 164 00:07:36,020 --> 00:07:39,020 we drop down to one quarter quality 165 00:07:39,020 --> 00:07:40,720 and 166 00:07:40,720 --> 00:07:42,720 that's looking great. 167 00:07:42,720 --> 00:07:44,720 Acceptable CPU 168 00:07:44,720 --> 00:07:46,720 and GPU usage 169 00:07:46,720 --> 00:07:49,720 and a much nicer looking timeline 170 00:07:49,720 --> 00:07:51,720 than what we were able to achieve 171 00:07:51,720 --> 00:07:55,840 without the RED Rocket X. 172 00:07:55,840 --> 00:07:57,840 But in a lot of ways 173 00:07:57,840 --> 00:08:01,290 this actually raises 174 00:08:01,290 --> 00:08:04,290 a lot more questions than it answers. 175 00:08:04,290 --> 00:08:07,290 Why would disabling our add-in card 176 00:08:07,290 --> 00:08:10,290 give me the best possible editing experience? 177 00:08:10,290 --> 00:08:11,290 Is it just 178 00:08:11,290 --> 00:08:13,290 a snake oil hardware key 179 00:08:13,290 --> 00:08:14,290 to run an improved 180 00:08:14,290 --> 00:08:17,640 GPU acceleration algorithm? 181 00:08:17,640 --> 00:08:18,640 So I actually spent 182 00:08:18,640 --> 00:08:19,640 a couple of weeks 183 00:08:19,640 --> 00:08:21,640 investigating this 184 00:08:21,640 --> 00:08:22,640 because timeline performance 185 00:08:22,640 --> 00:08:25,640 has genuinely been an issue for us. 186 00:08:25,640 --> 00:08:27,640 But investing in 187 00:08:27,640 --> 00:08:28,640 one of these 188 00:08:28,640 --> 00:08:29,640 for each of our editors 189 00:08:29,640 --> 00:08:30,640 would be a whopping 190 00:08:30,640 --> 00:08:33,640 $35,000 total. 191 00:08:33,640 --> 00:08:35,770 And here's what I found. 192 00:08:35,770 --> 00:08:36,900 First 193 00:08:36,900 --> 00:08:37,900 some backup 194 00:08:37,900 --> 00:08:38,900 for what we already knew. 195 00:08:38,900 --> 00:08:40,900 Without the RED Rocket X 196 00:08:40,900 --> 00:08:42,900 REDCODE RAW footage 197 00:08:42,900 --> 00:08:43,900 loves 198 00:08:43,900 --> 00:08:44,900 CPU cores. 199 00:08:44,900 --> 00:08:47,900 Knocking my Extreme Edition down to 6 cores 200 00:08:47,900 --> 00:08:49,900 yielded dropped frames 201 00:08:49,900 --> 00:08:52,900 even at only one quarter quality. 202 00:08:52,900 --> 00:08:55,900 But more cores only carries you so far. 203 00:08:55,900 --> 00:08:58,900 A 22 core Xeon like this 204 00:08:58,900 --> 00:08:59,900 costs more 205 00:08:59,900 --> 00:09:02,900 significantly slashes single core performance 206 00:09:02,900 --> 00:09:05,900 and doesn't even approach the performance improvement 207 00:09:05,900 --> 00:09:07,900 from the Rocket X installed 208 00:09:07,900 --> 00:09:09,900 but disabled in Premiere. 209 00:09:09,900 --> 00:09:10,900 So then 210 00:09:10,900 --> 00:09:11,900 the ideal 211 00:09:11,900 --> 00:09:14,900 8K video editing configuration is 212 00:09:14,900 --> 00:09:15,900 a 10 core 213 00:09:15,900 --> 00:09:16,900 with a Quadro 214 00:09:16,900 --> 00:09:18,900 and a RED Rocket X 215 00:09:18,900 --> 00:09:19,900 disabled, right? 216 00:09:19,900 --> 00:09:22,410 Maybe not. 217 00:09:22,410 --> 00:09:23,410 Is the Quadro 218 00:09:23,410 --> 00:09:25,410 at over four times the price 219 00:09:25,410 --> 00:09:27,410 really accelerating this process 220 00:09:27,410 --> 00:09:28,410 any better 221 00:09:28,410 --> 00:09:30,410 than a high-end GeForce card would? 222 00:09:30,410 --> 00:09:33,600 We'll be moving to an HDR workflow 223 00:09:33,600 --> 00:09:34,600 sometime in Q4. 224 00:09:34,600 --> 00:09:35,600 Expect an update on that. 225 00:09:35,600 --> 00:09:37,600 So we need 226 00:09:37,600 --> 00:09:39,600 10-bit color over DisplayPort. 227 00:09:39,600 --> 00:09:41,600 A traditional Quadro feature. 228 00:09:41,600 --> 00:09:44,600 But that was actually added to GeForce 229 00:09:44,600 --> 00:09:45,600 a while back. 230 00:09:45,600 --> 00:09:48,790 And the results here are fascinating. 231 00:09:48,790 --> 00:09:50,790 Our Titan X Pascal 232 00:09:50,790 --> 00:09:53,790 despite its nearly identical 233 00:09:53,790 --> 00:09:54,790 on-paper specs 234 00:09:54,790 --> 00:09:56,790 to our Quadro P6000 235 00:09:56,790 --> 00:09:59,790 is able to use its sky-high 236 00:09:59,790 --> 00:10:02,790 nearly 1800 MHz boost clocks 237 00:10:02,790 --> 00:10:04,790 to reduce GPU usage 238 00:10:04,790 --> 00:10:06,790 lower our dropped frames 239 00:10:06,790 --> 00:10:08,790 and even drop CPU usage 240 00:10:08,790 --> 00:10:09,790 a little bit 241 00:10:09,790 --> 00:10:10,790 at our target 242 00:10:10,790 --> 00:10:12,790 one-quarter playback quality. 243 00:10:12,790 --> 00:10:14,790 It even almost managed 244 00:10:14,790 --> 00:10:15,790 to pull off 245 00:10:15,790 --> 00:10:17,790 half playback quality. 246 00:10:17,790 --> 00:10:19,790 Both GPUs by the way were run 247 00:10:19,790 --> 00:10:21,790 with their power sliders cranked 248 00:10:21,790 --> 00:10:23,790 and power saving mode disabled. 249 00:10:23,790 --> 00:10:25,400 So then 250 00:10:25,400 --> 00:10:26,400 having finished running 251 00:10:26,400 --> 00:10:27,400 all of these numbers 252 00:10:27,400 --> 00:10:28,400 I came to the same conclusion 253 00:10:28,400 --> 00:10:29,400 that RED support 254 00:10:29,400 --> 00:10:31,400 did get back and give me. 255 00:10:31,400 --> 00:10:33,400 Premiere is just being Premiere. 256 00:10:33,400 --> 00:10:35,400 Disabled is enabled 257 00:10:35,400 --> 00:10:38,400 and enabled is borked. 258 00:10:38,400 --> 00:10:39,400 And as for 259 00:10:39,400 --> 00:10:41,400 an optimal workstation configuration 260 00:10:41,400 --> 00:10:42,400 well it turns out 261 00:10:42,400 --> 00:10:45,400 the Rocket X does do something 262 00:10:45,400 --> 00:10:47,400 for users with 8K footage 263 00:10:47,400 --> 00:10:49,400 even though it's not rated for it. 264 00:10:49,400 --> 00:10:51,400 But for the difference that it provides 265 00:10:51,400 --> 00:10:53,400 I'm not going to invest that much 266 00:10:53,400 --> 00:10:55,400 just to get a bump 267 00:10:55,400 --> 00:10:57,400 in preview quality. 268 00:10:57,400 --> 00:11:00,100 Do you ever find yourself thinking about 269 00:11:00,100 --> 00:11:02,100 the privacy of your information 270 00:11:02,100 --> 00:11:04,100 when you're making purchases online? 271 00:11:04,100 --> 00:11:06,100 Well privacy.com is a great way 272 00:11:06,100 --> 00:11:07,100 to control and protect 273 00:11:07,100 --> 00:11:08,100 your online spending. 274 00:11:08,100 --> 00:11:09,100 They offer the ability 275 00:11:09,100 --> 00:11:12,100 to set up virtual burner cards 276 00:11:12,100 --> 00:11:13,100 which allow you to hide 277 00:11:13,100 --> 00:11:15,100 sensitive information 278 00:11:15,100 --> 00:11:17,100 and cut off specific points 279 00:11:17,100 --> 00:11:19,100 of spending immediately 280 00:11:19,100 --> 00:11:20,100 without interfering with the places 281 00:11:20,100 --> 00:11:23,100 that you make online purchases. 282 00:11:23,100 --> 00:11:24,100 Two of the best features 283 00:11:24,100 --> 00:11:25,100 of privacy.com 284 00:11:25,100 --> 00:11:27,100 are the ability to set charge limits 285 00:11:27,100 --> 00:11:31,100 per transaction, per day, and per year 286 00:11:31,100 --> 00:11:34,100 and you can close any card 287 00:11:34,100 --> 00:11:36,100 with just a click. 288 00:11:36,100 --> 00:11:37,100 They've also got their 289 00:11:37,100 --> 00:11:38,100 private payments feature 290 00:11:38,100 --> 00:11:39,100 which masks your spending 291 00:11:39,100 --> 00:11:41,100 because it's nobody's business 292 00:11:41,100 --> 00:11:42,100 what you're buying 293 00:11:42,100 --> 00:11:44,100 and their browser extension 294 00:11:44,100 --> 00:11:45,100 automatically detects 295 00:11:45,100 --> 00:11:47,100 checkout fields on a webpage 296 00:11:47,100 --> 00:11:48,100 so with just two clicks 297 00:11:48,100 --> 00:11:51,100 you can create and autofill 298 00:11:51,100 --> 00:11:54,100 a dynamic number into the checkout form. 299 00:11:54,100 --> 00:11:55,100 So head over to 300 00:11:55,100 --> 00:11:57,100 privacy.com forward slash Linus 301 00:11:57,100 --> 00:11:58,100 or click the link 302 00:11:58,100 --> 00:11:59,100 in the video description. 303 00:11:59,100 --> 00:12:00,100 As a special offer 304 00:12:00,100 --> 00:12:02,100 privacy will be randomly selecting 305 00:12:02,100 --> 00:12:04,100 50 of those who sign up 306 00:12:04,100 --> 00:12:06,100 via privacy.com slash Linus 307 00:12:06,100 --> 00:12:07,100 and crediting 50 bucks 308 00:12:07,100 --> 00:12:09,100 to their privacy.com accounts 309 00:12:09,100 --> 00:12:11,100 to spend whenever. 310 00:12:11,100 --> 00:12:13,100 That's like a rival 300 mouse 311 00:12:13,100 --> 00:12:15,100 or a cheap mechanical keyboard 312 00:12:15,100 --> 00:12:16,100 for nothing. 313 00:12:16,100 --> 00:12:17,460 So thanks for watching guys 314 00:12:17,460 --> 00:12:18,460 if you disliked this video 315 00:12:18,460 --> 00:12:19,460 you can hit that button 316 00:12:19,460 --> 00:12:20,460 but if you liked it 317 00:12:20,460 --> 00:12:21,460 hit the like button 318 00:12:21,460 --> 00:12:22,460 get subscribed 319 00:12:22,460 --> 00:12:23,460 maybe consider checking out 320 00:12:23,460 --> 00:12:24,460 where to buy the stuff we featured 321 00:12:24,460 --> 00:12:27,060 at the link in the video description. 322 00:12:27,060 --> 00:12:28,060 Also down there 323 00:12:28,060 --> 00:12:29,060 is our community forum 324 00:12:29,060 --> 00:12:30,060 which you should totally join 325 00:12:30,060 --> 00:12:31,060 as well as our merch store 326 00:12:31,060 --> 00:12:33,060 which has cool shirts like this one.