WEBVTT 00:00:00.160 --> 00:00:02.320 So I was looking through the 3DMark leaderboards 00:00:02.320 --> 00:00:05.520 the other day because Steve over at Gamers Nexus 00:00:05.520 --> 00:00:09.520 was talking some smack about our benchmarking skills. 00:00:09.520 --> 00:00:12.260 I didn't see it though, I don't watch his videos. 00:00:12.260 --> 00:00:16.660 As usual, Kingpin from EVGA is way up at the top 00:00:16.660 --> 00:00:19.200 holding the record in 3DMark, Time Spy, 00:00:19.200 --> 00:00:20.240 Xtreme and all that stuff. 00:00:20.240 --> 00:00:23.640 But when I looked a little closer, I saw what's this. 00:00:23.640 --> 00:00:28.640 He's only using a 7980XE 18 core processor 00:00:28.640 --> 00:00:31.520 with four GTX 1080Ti graphics cards. 00:00:31.520 --> 00:00:35.400 And I thought, this is my opportunity. 00:00:35.400 --> 00:00:37.980 My one shot, my mom's spaghetti. 00:00:37.980 --> 00:00:41.020 I can't beat him with liquid nitrogen cooling 00:00:41.020 --> 00:00:43.300 and like sawed off graphics cards 00:00:43.300 --> 00:00:44.940 with custom power delivery boards 00:00:44.940 --> 00:00:46.360 soldered to the back of them. 00:00:46.360 --> 00:00:51.040 But what I do have is better hardware than that. 00:00:51.040 --> 00:00:56.040 So my plan is to use 56 CPU cores 00:00:57.280 --> 00:00:59.280 in our ASUS C621E650. 00:00:59.380 --> 00:01:02.480 This is the product I bought and it's my challenge. 00:01:02.480 --> 00:01:04.560 I'm not going to pay you by the way. 00:01:04.560 --> 00:01:06.000 I'm just going to do a quick test. 00:01:06.000 --> 00:01:06.920 I'm going to do a quick test. 00:01:06.920 --> 00:01:08.360 Let's see how far they go. 00:01:08.360 --> 00:01:09.860 We're going to do a little test. 00:01:09.860 --> 00:01:11.560 It's a pretty tight nose test, 00:01:11.560 --> 00:01:16.090 but it's going to be sort of a rough test. 00:01:16.090 --> 00:01:19.230 This is a built-in built-in CPU. 00:01:19.230 --> 00:01:21.290 It's not even a full series. 00:01:21.290 --> 00:01:23.810 So we'll test that out today. 00:01:24.630 --> 00:01:26.690 First we'll test out the chips, 00:01:26.690 --> 00:01:28.710 which are my favorite chips in the market. 00:01:28.710 --> 00:01:38.460 This is what an old chip looks like. 00:01:38.560 --> 00:01:45.040 I'm gonna throw in a quick spoiler alert here. My dastardly plan to troll you guys with this setup 00:01:45.760 --> 00:01:52.640 didn't work out, but it's still worth going through the process because it is a pretty unreal rig. 00:01:53.520 --> 00:02:02.480 So problem number one, SLI support for gaming on Quadros. The GV100 with its high speed 00:02:02.480 --> 00:02:10.800 HBM2 memory and massive 5376 CUDA cores should be a little faster than a Titan V, 00:02:11.360 --> 00:02:18.160 which is faster than a GTX 1080 TI or Titan XP, right out of the box. So even though I can't 00:02:18.160 --> 00:02:23.520 liquid nitrogen cool these things, I should be able to make up a lot of the difference 00:02:23.520 --> 00:02:32.480 in just outright faster hardware, except for one problem. These cards don't support three-way or 00:02:32.480 --> 00:02:32.940 four-way cards. These cards don't support four-way cards. These cards don't support four-way cards. 00:02:32.940 --> 00:02:39.260 SLI. So this right here is called an NVLink connector, and it is only designed for these 00:02:39.260 --> 00:02:47.260 cards anyway to put two of them together at a time. So while you can actually run SLI over NVLink, 00:02:47.900 --> 00:02:53.820 it's not really intended for that. It's more for resource and memory sharing in 3D modeling or 00:02:53.820 --> 00:03:02.460 compute applications. Fortunately though, I do have a way around that. So this is a Quadro Sync 2 board. 00:03:02.460 --> 00:03:09.420 And you'll probably recognize that if you've watched one of our many K, like 16K gaming videos, 00:03:09.420 --> 00:03:17.020 where we've used it to game on an array of as many as 16 monitors at once. Now this thing's purpose 00:03:17.020 --> 00:03:24.860 is primarily video display walls, like giant digital signage deployments, for example, to keep the monitors 00:03:24.860 --> 00:03:30.800 from getting out of sync, which would cause weird tearing issues. But we already know that it works for gaming. 00:03:30.800 --> 00:03:31.320 But we already know that it works for gaming, as long as you have a very small amount of time on your computer. 00:03:31.320 --> 00:03:31.820 But we already know that it works for gaming, as long as you have a very small amount of time on your computer. 00:03:31.820 --> 00:03:32.320 But we already know that it works for gaming, as long as you have a very small amount of time on your computer. 00:03:32.460 --> 00:03:35.480 long as the resolution we're running at isn't too much 00:03:35.480 --> 00:03:38.600 for the frame buffer of each individual card. 00:03:38.600 --> 00:03:41.320 So that got me thinking, 00:03:41.320 --> 00:03:45.540 what if instead of plugging multiple monitors 00:03:45.540 --> 00:03:46.820 into each one of the cards, 00:03:46.820 --> 00:03:49.340 because remember the more pixels each card drives, 00:03:49.340 --> 00:03:51.620 the harder it is for you to run the game. 00:03:51.620 --> 00:03:53.140 What if instead of that, 00:03:53.140 --> 00:03:57.860 we just ran a single 1080p display out of each one 00:03:57.860 --> 00:04:01.320 in a two by two grid for a total of 4K? 00:04:01.320 --> 00:04:06.320 Well, these are each 32 gig cards, so should be just fine. 00:04:06.700 --> 00:04:09.100 Bringing us to problem number two, 00:04:09.100 --> 00:04:12.140 even if you had a monitor like this one, it's 1080p, 00:04:12.140 --> 00:04:15.940 it's got super thin bezels, it's inexpensive, 00:04:15.940 --> 00:04:17.460 you still have a problem. 00:04:17.460 --> 00:04:21.100 Nobody wants to game where four bezels intersect 00:04:21.100 --> 00:04:23.760 because that's right where your crosshair is. 00:04:23.760 --> 00:04:28.080 This was by far my favorite part of my idea for this video 00:04:28.080 --> 00:04:30.160 and what prompted me to go ahead with it. 00:04:30.160 --> 00:04:31.280 Now I had intended, 00:04:31.280 --> 00:04:33.220 like I said, to troll you guys 00:04:33.220 --> 00:04:36.020 with some unbelievable benchmarks 00:04:36.020 --> 00:04:39.360 with like a test bench under a sheet or something 00:04:39.360 --> 00:04:40.840 to make you guys think 00:04:40.840 --> 00:04:43.220 that I had RTX cards early or something, 00:04:43.220 --> 00:04:45.540 but you'd never buy that 00:04:45.540 --> 00:04:48.140 if I had four of these monitors on an ARM, 00:04:48.140 --> 00:04:51.720 like we did back when we did our first 8K gaming setup. 00:04:51.720 --> 00:04:56.720 So what if we used LG's 43UD79? 00:04:57.780 --> 00:05:00.760 It's 43 inches, 4K, 00:05:00.760 --> 00:05:05.760 and thanks to its four separate inputs, 00:05:07.550 --> 00:05:11.560 it could be split into quadrants. 00:05:11.560 --> 00:05:15.600 So let's get it set up, shall we? 00:05:15.600 --> 00:05:17.160 One thing I should do while we're setting this up 00:05:17.160 --> 00:05:20.460 is give a massive shout out to Benny over at Memex 00:05:20.460 --> 00:05:22.280 who did a personal favor, 00:05:22.280 --> 00:05:27.280 hooking me up very last minute with one of these guys, 00:05:27.760 --> 00:05:32.760 actually two of them, DisplayPort to HDMI 2.0, Active. 00:05:32.780 --> 00:05:34.020 Adapters. 00:05:34.020 --> 00:05:37.260 The Active part's important because, 00:05:37.260 --> 00:05:38.820 well, for one thing, 00:05:38.820 --> 00:05:41.200 these Quadros don't have HDMI, 00:05:41.200 --> 00:05:43.140 so you have to run DisplayPort, 00:05:43.140 --> 00:05:47.240 and we've only got two DisplayPort inputs on this monitor, 00:05:47.240 --> 00:05:50.580 only one of which is actually usable because it's a Type-C, 00:05:50.580 --> 00:05:53.880 and you can't adapt this to that. 00:05:53.880 --> 00:05:56.780 There are cables, but they are directional cables. 00:05:56.780 --> 00:05:57.960 They don't go both ways. 00:05:57.960 --> 00:06:00.960 The Quadro Sync 2 card here only works 00:06:00.960 --> 00:06:01.960 if every one of these cards, 00:06:01.960 --> 00:06:05.540 is hooked up to the same type of display. 00:06:05.540 --> 00:06:07.940 The card has to think it's DisplayPort, 00:06:07.940 --> 00:06:10.480 so that's why the Active signal adapter is important, 00:06:10.480 --> 00:06:14.070 so every card thinks it's hooked up to DisplayPort. 00:06:14.070 --> 00:06:16.430 Just the amount of hassle that I went through 00:06:16.430 --> 00:06:19.930 to get this thing working, just, you know. 00:06:19.930 --> 00:06:22.850 But it was totally worth it, sort of. 00:06:22.850 --> 00:06:25.450 Here's an example of one of those directional cables. 00:06:25.450 --> 00:06:28.630 They only go from a Type-C output, 00:06:28.630 --> 00:06:30.990 like on your laptop or something, 00:06:30.990 --> 00:06:32.310 to a DisplayPort monitor. 00:06:32.310 --> 00:06:34.190 They don't go the other way. 00:06:34.190 --> 00:06:36.150 Oh, it totally makes sense. 00:06:36.150 --> 00:06:38.310 It's just not something that I thought about 00:06:38.310 --> 00:06:43.500 when this video was in the early conceptualization stages. 00:06:43.620 --> 00:06:46.920 Can we all just stop and appreciate for a moment 00:06:46.920 --> 00:06:52.440 that this is $36,000 worth of graphics cards here? 00:06:54.010 --> 00:06:57.470 I guess, you know, props to NVIDIA for lending them to us, 00:06:57.470 --> 00:06:59.250 even though we're totally not doing 00:06:59.250 --> 00:07:02.520 what we said we were gonna do with them. 00:07:02.520 --> 00:07:03.940 NVIDIA gets real cagey about, you know, this. 00:07:03.940 --> 00:07:03.980 NVIDIA gets real cagey about, you know, this. 00:07:03.980 --> 00:07:05.580 NVIDIA gets real cagey about gaming on Quadros. 00:07:05.580 --> 00:07:09.370 They just, they don't like people talking about it. 00:07:09.370 --> 00:07:14.370 Okay, so, HDMI 4, HDMI 3, that's right. 00:07:16.090 --> 00:07:24.930 Boom, boom, done, finish. 00:07:24.950 --> 00:07:28.750 And you can see how seamless it is as you move between them. 00:07:28.750 --> 00:07:30.110 See, you'll get a little bit of tearing. 00:07:30.110 --> 00:07:32.070 We got a little bit vertically there. 00:07:32.070 --> 00:07:33.470 Generally, actually, that's probably 00:07:33.470 --> 00:07:35.010 about the worst I've seen. 00:07:35.010 --> 00:07:37.850 But then you can see here, moving between these two, 00:07:37.850 --> 00:07:41.240 we get nothing. 00:07:41.240 --> 00:07:42.080 Oh, and now. 00:07:42.080 --> 00:07:42.920 You can see its behavior. 00:07:42.920 --> 00:07:44.380 Oh, yeah. 00:07:44.380 --> 00:07:46.720 So, it's not perfect, 00:07:46.720 --> 00:07:51.300 but I figured it would be good enough to trick you guys. 00:07:51.300 --> 00:07:53.200 Bring up some 3D Mark here. 00:07:53.200 --> 00:07:56.300 And there it is, Time Spy Extreme. 00:07:56.300 --> 00:08:03.230 I think this would have been believable. 00:08:03.230 --> 00:08:04.590 Yeah, you can see. 00:08:04.590 --> 00:08:06.790 The character's walking right through the intersection 00:08:06.790 --> 00:08:13.130 of all four of them, and you can't see any tearing. 00:08:13.130 --> 00:08:17.770 Pretty mean. 00:08:17.770 --> 00:08:21.320 Yeah. 00:08:21.320 --> 00:08:26.070 So, that's pretty underwhelming. 00:08:26.070 --> 00:08:29.400 But, perhaps, but, perhaps, 00:08:29.400 --> 00:08:31.540 if we weren't so CPU bottlenecked. 00:08:31.540 --> 00:08:34.820 So, while our CPU score is actually really good, 00:08:34.820 --> 00:08:37.040 like 10,000, 00:08:37.040 --> 00:08:42.040 that's just because we got a great score in the CPU tests, 00:08:42.300 --> 00:08:47.120 not because that CPU is actually not causing a bottleneck 00:08:47.120 --> 00:08:48.960 for our overpowered graphics setup. 00:08:48.960 --> 00:08:53.900 So, maybe what we can do is transfer the whole thing 00:08:53.900 --> 00:08:57.710 onto a bench with a 7980 XE. 00:08:57.710 --> 00:08:59.750 So, why don't we do that? 00:09:00.110 --> 00:09:03.230 Maybe we can alleviate the CPU bottleneck a little bit 00:09:03.230 --> 00:09:05.590 because we've got lots of cores here, 00:09:05.590 --> 00:09:08.470 but they're not particularly fast. 00:09:08.470 --> 00:09:11.890 You know, you do enough bench swaps, 00:09:11.890 --> 00:09:14.360 you get pretty fast at it. 00:09:14.360 --> 00:09:19.360 And, as expected, this is really not a lot better. 00:09:20.240 --> 00:09:22.160 Also, I'm having another weird issue 00:09:22.160 --> 00:09:23.800 where the color is a little bit off 00:09:23.800 --> 00:09:26.100 on the bottom left quadrant for some reason. 00:09:26.100 --> 00:09:28.880 I've actually run this before on both of these setups. 00:09:28.880 --> 00:09:30.220 I have no idea what it's doing. 00:09:30.220 --> 00:09:31.580 it's doing this time. 00:09:31.580 --> 00:09:33.760 As you guys can imagine, by this point 00:09:33.760 --> 00:09:36.880 in this whole process, I was feeling pretty discouraged 00:09:36.880 --> 00:09:41.840 because we are a far cry from the results 00:09:41.840 --> 00:09:43.760 that I was expecting. 00:09:43.760 --> 00:09:47.440 I mean, it's not like the entire thing is terrible. 00:09:47.440 --> 00:09:49.700 Like here, we can fire up a game here 00:09:49.700 --> 00:09:52.460 and see that actually like in-game performance 00:09:52.460 --> 00:09:53.860 looks pretty darn good. 00:09:53.860 --> 00:09:55.600 We're just gonna ignore this weirdness 00:09:55.600 --> 00:09:58.960 because this is the first time I've seen that. 00:09:58.960 --> 00:10:03.720 But this is not the benchmark breaker 00:10:03.720 --> 00:10:06.630 that I was really hoping that it would be. 00:10:06.630 --> 00:10:09.990 Oh, I know what it is. 00:10:09.990 --> 00:10:15.290 It's the whole zero to 255 versus 16 00:10:15.290 --> 00:10:20.300 to whatever black depth thing. 00:10:20.300 --> 00:10:22.440 How do you adjust that? 00:10:22.440 --> 00:10:25.260 You know what, I bet I have to break down the mosaic. 00:10:25.260 --> 00:10:27.460 You can see not in mosaic, I'm not having the issue. 00:10:27.460 --> 00:10:29.840 So I think it's just a matter of like, 00:10:29.840 --> 00:10:31.420 it just went wonko. 00:10:31.460 --> 00:10:33.500 For a second there. 00:10:33.500 --> 00:10:36.080 Yeah, see, it works perfectly this time. 00:10:36.080 --> 00:10:39.850 So let's run a game or two on this thing. 00:10:39.850 --> 00:10:42.350 Not bad. 00:10:42.350 --> 00:10:47.110 So what I can hope for here is that 00:10:47.110 --> 00:10:51.190 we will get some pretty awesome frame rates in GTA V. 00:10:51.190 --> 00:10:56.360 Why is this game so bad at running in full screen? 00:10:56.360 --> 00:10:58.500 Hey, Anthony, can I borrow you for a minute? 00:10:58.500 --> 00:11:01.780 No, I don't wanna, like, can I just change more things? 00:11:01.780 --> 00:11:02.750 GTA V is so horrible. 00:11:02.750 --> 00:11:06.400 There's more things than one at a time, like. 00:11:06.400 --> 00:11:14.050 Okay, how do you crank it? 00:11:14.050 --> 00:11:16.050 All the settings first before resetting it? 00:11:16.050 --> 00:11:16.890 You can't. 00:11:16.890 --> 00:11:17.710 You can't. 00:11:17.710 --> 00:11:20.010 You can reset it if you say no. 00:11:20.010 --> 00:11:22.430 So the good news is throughout the process 00:11:22.430 --> 00:11:25.130 of getting all of our GTA V settings cranked, 00:11:25.130 --> 00:11:26.710 this tearing line down the middle is gone. 00:11:26.710 --> 00:11:28.410 So that's pretty cool. 00:11:28.410 --> 00:11:33.820 Now we can see if our 4K performance is impressive 00:11:33.880 --> 00:11:39.130 because we have got this thing dialed in to hurt 00:11:39.410 --> 00:11:43.630 anything other than a godly graphics card setup. 00:11:43.630 --> 00:11:45.780 Why is that off? 00:11:45.780 --> 00:11:46.900 Did it even, oh, it did. 00:11:46.900 --> 00:11:49.140 It registered everything else, just not the MSAA. 00:11:49.140 --> 00:11:51.510 So that's neat. 00:11:51.510 --> 00:11:54.840 Multiple reboots later, we've got all that on. 00:11:54.840 --> 00:11:56.500 Yes, I wanna keep the new settings. 00:11:56.500 --> 00:11:58.260 Yes, yes, yes. 00:11:58.260 --> 00:11:59.620 And then we should be able to do 00:11:59.620 --> 00:12:01.900 extended shadow distances now. 00:12:01.900 --> 00:12:05.720 Now let's find out if in a situation 00:12:05.720 --> 00:12:10.220 where we're not quite so CPU bottlenecked, this thing can. 00:12:10.300 --> 00:12:13.320 Knock our socks off. 00:12:13.320 --> 00:12:15.100 Anthony, are you ready? 00:12:15.100 --> 00:12:15.940 Am I ready? 00:12:15.940 --> 00:12:17.130 Yeah. 00:12:17.130 --> 00:12:18.050 Are you ready? 00:12:18.050 --> 00:12:18.870 Are you ready? 00:12:18.870 --> 00:12:19.690 I'm ready. 00:12:19.690 --> 00:12:20.820 I'm ready. 00:12:20.820 --> 00:12:21.880 It's loading still. 00:12:21.880 --> 00:12:23.540 So what would you expect? 00:12:23.540 --> 00:12:25.560 I would say this will do. 00:12:25.560 --> 00:12:26.400 You think so? 00:12:26.400 --> 00:12:27.520 Yeah. 00:12:27.520 --> 00:12:30.390 Yeah, that's stuttering. 00:12:30.390 --> 00:12:32.260 It's suffering. 00:12:32.260 --> 00:12:36.700 So I still maintain this was a pretty cool idea. 00:12:36.700 --> 00:12:38.860 Perhaps works in GTA V, DX11, yeah. 00:12:38.860 --> 00:12:40.370 Yes. 00:12:40.370 --> 00:12:42.310 Let's just throw it on here. 00:12:42.310 --> 00:12:43.150 Okay. 00:12:43.150 --> 00:12:44.350 So how impressed are you? 00:12:44.350 --> 00:12:50.720 How impressed are you for a $36,000? 00:12:50.720 --> 00:12:53.300 All right, let's see what it's actually running at here. 00:12:53.300 --> 00:12:56.160 Yeah, no, no, it's just only running at like 30 FPS. 00:12:56.160 --> 00:12:56.980 Okay. 00:12:56.980 --> 00:13:00.850 What is a Peltzer V8 something or other? 00:13:00.850 --> 00:13:02.470 Oh, 15 FPS. 00:13:02.470 --> 00:13:03.510 Oh, oh, oh, okay. 00:13:03.510 --> 00:13:06.210 Well, I mean, that was clearly the frame rate drops. 00:13:06.210 --> 00:13:09.310 This is very realistic car physics we've got going on here. 00:13:12.680 --> 00:13:16.080 I'm saying that they probably wanted this car in their pool 00:13:16.080 --> 00:13:18.800 and I'm doing them a favor. 00:13:18.800 --> 00:13:21.620 So basically between the amount of GPU horsepower, 00:13:21.620 --> 00:13:25.060 we have and the difficulty of synchronizing them 00:13:25.060 --> 00:13:27.380 and managing resources between them, 00:13:27.380 --> 00:13:28.840 there are many, many bottlenecks 00:13:28.840 --> 00:13:31.140 that prevent this theoretically cool setup 00:13:31.140 --> 00:13:36.150 from actually performing well in the real world. 00:13:36.150 --> 00:13:38.030 Speaking of performing well in the real world, 00:13:38.030 --> 00:13:42.850 Corsair's Strafe MK2 or Mark II mechanical RGB keyboard 00:13:42.850 --> 00:13:45.010 is an update from the previous fan favorite 00:13:45.010 --> 00:13:47.690 and now has all the features you've been asking for. 00:13:47.690 --> 00:13:50.690 It comes in either Cherry MX Red or Cherry MX Silent switches 00:13:50.690 --> 00:13:52.190 with reds offering smooth and linear actuation, 00:13:52.190 --> 00:13:53.010 and it's a great feature. 00:13:53.010 --> 00:13:55.770 It also comes with anti-ghosting with silent dampening the sound by 30%. 00:13:55.770 --> 00:13:59.290 It's got 100% anti-ghosting with 104 key rollover. 00:13:59.290 --> 00:14:03.210 So no matter how fast all of your keystrokes are on the keyboard, 00:14:03.210 --> 00:14:04.730 it will register all of them. 00:14:04.730 --> 00:14:06.130 It's got eight megs of built-in storage 00:14:06.130 --> 00:14:08.130 so you can take all your RGB configurations 00:14:08.130 --> 00:14:09.910 and macro customization with you. 00:14:09.910 --> 00:14:12.450 It's got easy access multimedia audio controls 00:14:12.450 --> 00:14:14.030 allowing you to make on the fly adjustments. 00:14:14.030 --> 00:14:17.370 And the top of the keyboard is a built-in USB pass-through, 00:14:17.370 --> 00:14:19.290 which allows you to plug in a mouse, 00:14:19.290 --> 00:14:21.590 smartphone, flash drive, et cetera. 00:14:21.590 --> 00:14:23.010 So check it out at the link. 00:14:23.010 --> 00:14:25.420 In the video description. 00:14:25.420 --> 00:14:26.260 Thanks for watching, guys. 00:14:26.260 --> 00:14:28.300 If you just liked this video, you can hit that button. 00:14:28.300 --> 00:14:30.080 But if you liked it, hit like, get subscribed, 00:14:30.080 --> 00:14:31.020 or maybe consider checking out 00:14:31.020 --> 00:14:32.500 where to buy the stuff we featured, 00:14:32.500 --> 00:14:34.800 assuming that you have a real professional task 00:14:34.800 --> 00:14:37.340 to use them for at the link in the video description. 00:14:37.340 --> 00:14:39.760 Also down there is our merch store, which has cool shirts, 00:14:39.760 --> 00:14:42.020 not like this one, they're dead, 00:14:42.020 --> 00:14:44.360 which has cool shirts and our community forum, 00:14:44.360 --> 00:14:45.760 which you should totally join.