WEBVTT 00:00:00.960 --> 00:00:07.919 look at this guy do you feel old yet because i sure do 00:00:05.600 --> 00:00:12.320 that was me the last time AMD had a truly competitive top tier Radeon 00:00:10.160 --> 00:00:19.039 graphics card that was over seven years ago and this is me the last time AMD sat on 00:00:15.920 --> 00:00:22.880 top as the single GPU king with the 00:00:19.039 --> 00:00:26.480 Radeon hd 7970 and that is why the 00:00:22.880 --> 00:00:28.480 excitement around the rx 6800 series is 00:00:26.480 --> 00:00:33.760 so high and why it would be so devastating if 00:00:30.720 --> 00:00:35.920 AMD let us all down again 00:00:33.760 --> 00:00:39.360 thankfully spoiler alert they didn't 00:00:37.680 --> 00:00:44.000 watch on and watch what apps are stealing your 00:00:41.680 --> 00:00:48.480 bandwidth using our sponsor glasswire it's an easy way to diagnose and fix 00:00:46.480 --> 00:00:55.800 issues like poor video quality in online meetings get 25 off today using offer 00:00:51.200 --> 00:00:55.800 code Linus at the link down below 00:01:02.239 --> 00:01:11.119 during episode two of AMD's where gaming begins event AMD introduced us to their 00:01:07.040 --> 00:01:16.240 three new big navi gpus the rx 6800 rx 00:01:11.119 --> 00:01:18.880 6800 xt and rx 6900 xt nice all of which 00:01:16.240 --> 00:01:24.880 feature ray accelerators in each compute unit for ray tracing support as well as 00:01:21.360 --> 00:01:26.720 16 gigabytes of gddr6 memory and a new 00:01:24.880 --> 00:01:31.119 infinity cache that promises to significantly improve memory performance 00:01:29.040 --> 00:01:35.680 without the additional cost associated with gddr6x 00:01:33.439 --> 00:01:40.880 not only that but AMD also announced a technology that they're calling smart 00:01:37.600 --> 00:01:43.759 access memory which lets compatible cpus 00:01:40.880 --> 00:01:50.079 access the GPU's entire memory space at once rather than in small 256 megabyte 00:01:47.280 --> 00:01:53.280 chunks as before offering a significant performance boost 00:01:51.680 --> 00:01:58.719 it's because of that last feature that we're using a top-end ryzen 5000 series 00:01:56.079 --> 00:02:03.520 CPU for our testing which along with offering leading gaming performance is 00:02:01.040 --> 00:02:06.799 the only CPU platform that currently supports this feature 00:02:05.119 --> 00:02:12.239 before you get too concerned about being left out though both AMD and NVIDIA have 00:02:09.759 --> 00:02:16.400 told us that it's pretty much a fancy marketing name for resizable bar support 00:02:14.959 --> 00:02:21.760 which is something that microsoft is pushing for and that NVIDIA claims will 00:02:18.480 --> 00:02:23.680 be supported on their gpus soon i'd also 00:02:21.760 --> 00:02:28.640 be surprised if we don't see Intel ad platform support in the future but the 00:02:26.480 --> 00:02:32.560 future's the future let's see how things look today shadow of the tomb raider was 00:02:30.879 --> 00:02:37.599 one of the first titles to support ray tracing and with it disabled NVIDIA 00:02:34.959 --> 00:02:43.920 stands no chance against AMD unless we enabled dlss and that's a big 00:02:40.640 --> 00:02:45.599 deal but there is an asterisk here when 00:02:43.920 --> 00:02:51.280 we turn on ray tracing Radeon performance suffers tremendously with 00:02:48.000 --> 00:02:53.360 the 6800 xt sitting 13 frames per second 00:02:51.280 --> 00:02:57.280 below the rtx 3080. we wanted to test the same thing in 00:02:55.040 --> 00:03:01.680 wolfenstein youngblood but unfortunately that game hasn't been updated yet with 00:02:59.280 --> 00:03:06.159 vulcan ray tracing support for AMD so we'll have to settle for a traditional 00:03:03.200 --> 00:03:10.400 rendering where the 6800 xt is once again trailing the rtx 3080 with another 00:03:09.200 --> 00:03:14.800 asterisk with smart access memory enabled AMD 00:03:12.879 --> 00:03:20.080 retakes the lead in this traditional rendering scenario with a single digit 00:03:17.280 --> 00:03:23.680 but measurable performance improvement more good news follows with microsoft 00:03:21.760 --> 00:03:28.560 flight simulator where AMD managed the same or better performance than NVIDIA 00:03:26.159 --> 00:03:34.159 across the board although here we see a scenario where smart access memory has 00:03:30.959 --> 00:03:37.680 no tangible benefit so your mileage 00:03:34.159 --> 00:03:40.080 quite literally can vary game to game 00:03:37.680 --> 00:03:44.879 cs go gives us one scenario where NVIDIA pulls ahead in traditional rasterized 00:03:42.000 --> 00:03:49.120 rendering but look at the difference in minimum frame rates 00:03:46.640 --> 00:03:54.879 smart access memory makes a huge difference for AMD here and arguably 00:03:52.560 --> 00:03:58.959 that's even more important for esports titles like this than the average frame 00:03:57.519 --> 00:04:03.920 rate would be finally there's minecraft rtx and it's 00:04:02.000 --> 00:04:10.799 well it's a mess on AMD even without dlss the 6800 xt only gets 00:04:07.760 --> 00:04:15.760 half the average FPS compared to the rtx 00:04:10.799 --> 00:04:19.199 3080 and even the rtx 3070 is 50 faster 00:04:15.760 --> 00:04:21.840 than the 6800 this really highlights the 00:04:19.199 --> 00:04:26.880 weakness of AMD's first generation ray accelerators compared to NVIDIA's more 00:04:24.080 --> 00:04:32.960 mature second gen rt cores a note on dlss before we continue 00:04:29.360 --> 00:04:35.360 dlss is a contentious thing because it's 00:04:32.960 --> 00:04:40.880 an upscaler the scene is rendered at a lower resolution and then intelligently 00:04:37.919 --> 00:04:45.440 scaled up using NVIDIA's tensor course some would call measuring performance 00:04:42.479 --> 00:04:49.759 with dlss cheating but whether it's cheating or whether it's not it looks 00:04:47.520 --> 00:04:53.840 pretty darn good and it's something real gamers are using that AMD doesn't yet 00:04:52.240 --> 00:04:58.639 have an answer for if fidelityfx super resolution which is 00:04:56.720 --> 00:05:03.919 their analogous feature were ready in time for launch then we might have had 00:05:01.440 --> 00:05:09.759 playable frame rates in minecraft with ray tracing enabled but instead we'll 00:05:06.400 --> 00:05:11.440 need to wait for the rx 6900 xt get 00:05:09.759 --> 00:05:15.120 subscribed by the way so you don't miss our coverage of that card when it 00:05:12.800 --> 00:05:20.000 launches in december now productivity used to be an area that Radeon excelled 00:05:17.759 --> 00:05:25.600 but it seems that AMD had to make some sacrifices to catch up in gaming 00:05:22.560 --> 00:05:28.160 in opencl versus cuda Radeon outperforms 00:05:25.600 --> 00:05:31.600 g-force in longer renders while the opposite is true for shorter renders 00:05:30.080 --> 00:05:35.199 like bmw Radeon doesn't currently support 00:05:33.440 --> 00:05:39.280 ray-accelerated rendering which puts NVIDIA light years ahead with their 00:05:37.039 --> 00:05:44.960 optics renderer and it's worth noting that while there is a plug-in for Radeon 00:05:41.919 --> 00:05:47.440 prorender that is accelerated the output 00:05:44.960 --> 00:05:50.960 is rough and looks nothing like what it should 00:05:48.240 --> 00:05:55.120 similarly we couldn't even run luxmark or rather we could but the end result 00:05:53.440 --> 00:06:00.080 was so mangled that it wasn't even recognizable let alone usable 00:05:57.919 --> 00:06:03.919 this led to a problem for us in terms of productivity testing because most 00:06:02.320 --> 00:06:09.360 productivity software that takes advantage of the GPU is optimized for 00:06:06.639 --> 00:06:12.639 cuda and opencl implementations are getting old enough now that they're 00:06:11.120 --> 00:06:18.560 beginning to break we did find a renderer that supports 00:06:14.479 --> 00:06:21.199 both opencl and cuda indigo but 00:06:18.560 --> 00:06:25.520 NVIDIA had the lead in it anyway spec view perf gave us some blow trading 00:06:23.280 --> 00:06:30.160 though with 3ds max performing better on radeons while katia and creo were much 00:06:28.080 --> 00:06:35.520 better on GeForce and the memory intensive energy medical and siemens nx 00:06:33.199 --> 00:06:39.440 workloads heavily favored Radeon thanks to the infinity cache 00:06:37.280 --> 00:06:42.800 mayan solidworks on the other hand once again fell to NVIDIA's faster compute 00:06:41.759 --> 00:06:47.199 engine as AMD predicted when we asked about its 00:06:45.199 --> 00:06:51.120 impact on productivity smart access memory made no difference here just like 00:06:49.680 --> 00:06:55.360 our new keyboard shirts won't make you a better typist they just make you look 00:06:53.039 --> 00:06:59.440 cooler lttestore.com it looks like AMD's expanded vapor 00:06:57.280 --> 00:07:04.160 chamber cooler design though is roughly comparable to NVIDIA's with both the rx 00:07:01.680 --> 00:07:09.120 6800 gpus closely matching their team green counterparts with the curious 00:07:06.319 --> 00:07:14.479 exception of the tail end of the test where specvuperf's medical and siemens 00:07:11.840 --> 00:07:19.520 nx benchmarks happen to fall as you might recall that's one of the areas 00:07:16.240 --> 00:07:21.599 where AMD absolutely spanked NVIDIA 00:07:19.520 --> 00:07:26.880 as far as clock stability goes it looks like AMD is slightly less consistently 00:07:24.160 --> 00:07:31.039 running at maximum boost clock but it's pretty similar across the board 00:07:29.199 --> 00:07:35.199 you might expect that with clocks going as low as they do at idle power 00:07:33.280 --> 00:07:39.039 consumption would be lower than NVIDIA's when you're not doing anything 00:07:36.639 --> 00:07:43.599 unfortunately that's not so still power consumption on the whole 00:07:40.960 --> 00:07:47.680 appears lower with some exceptions so you can again see that dip for NVIDIA 00:07:45.759 --> 00:07:52.319 here suggesting a bottleneck with medical and siemens nx that AMD isn't 00:07:50.319 --> 00:07:56.879 experiencing you can't see it on this graph unfortunately but we recorded a 00:07:54.319 --> 00:08:03.120 spike as high as 270 watts for the rx 6800 and an eye watering 411 watts for 00:08:01.120 --> 00:08:06.879 the 6800 xt thankfully that's not all the time so 00:08:04.879 --> 00:08:10.720 it's not really a problem in practice that is as long as you've got a quality 00:08:08.720 --> 00:08:14.000 power supply that won't trip its over current protection under a momentary 00:08:12.400 --> 00:08:18.960 spike like that the main trouble with the rx 6800 series 00:08:16.639 --> 00:08:22.400 aside from the last gen ray tracing performance which is becoming more 00:08:20.720 --> 00:08:26.960 important day by day is its lack of truly useful features in 00:08:25.039 --> 00:08:30.800 comparison to NVIDIA not only do they not yet have a dlss 00:08:29.120 --> 00:08:35.039 equivalent that's ready to help out their weaker ray accelerators they're 00:08:32.719 --> 00:08:38.959 not even apologetic about it here's what they told us when we asked about the ray 00:08:36.800 --> 00:08:43.279 tracing performance and why they weren't comparing with NVIDIA even for press 00:08:40.800 --> 00:08:48.080 briefings super resolution techniques like dlss do not produce the same visual 00:08:46.000 --> 00:08:50.399 quality as native rendering especially with motion 00:08:49.200 --> 00:08:56.399 i mean yeah that's true but it's gotten pretty 00:08:53.440 --> 00:09:01.440 good and that fact alone doesn't change that NVIDIA has it and AMD doesn't and 00:08:59.920 --> 00:09:07.200 that was literally the difference between playable ray tracing and sub 60 00:09:04.399 --> 00:09:10.399 frame rates in our testing also AMD if you really thought it was that garbage 00:09:08.959 --> 00:09:15.200 why are you working on an analogous feature like we weren't expecting AMD to blow us 00:09:13.600 --> 00:09:20.160 away with their first gen ray tracing support it's just that this fox and the 00:09:17.200 --> 00:09:23.920 grapes response is a little bit tone deaf 00:09:21.279 --> 00:09:29.040 of course AMD has a point not everyone cares about upscaling or ray tracing but 00:09:26.240 --> 00:09:33.279 then AMD's hardware video encoder is still frankly rubbish with what looks 00:09:30.880 --> 00:09:36.560 like significant chroma subsampling and macroblock artifacts compared to NVIDIA 00:09:35.200 --> 00:09:40.720 at the same resolution in bitrate and x264 00:09:38.399 --> 00:09:45.760 it's basically unwatchable and on the subject of media NVIDIA is also rolling 00:09:43.279 --> 00:09:51.440 features like rtx voice background noise cancellation ai camera background 00:09:48.560 --> 00:09:56.320 removal and more that are accelerated by their tensor cores 00:09:52.959 --> 00:09:58.480 AMD currently has no equivalent machine 00:09:56.320 --> 00:10:03.600 learning hardware in the Radeon 6000 series this puts streamers and content 00:10:01.120 --> 00:10:08.000 creators and frankly even work from home professionals who like to game in their 00:10:05.040 --> 00:10:11.040 off hours at a disadvantage with Radeon compared to if they had bought GeForce 00:10:09.760 --> 00:10:14.800 instead speaking personally those features 00:10:12.959 --> 00:10:20.240 really are enough to compel me to go GeForce even if i didn't give two hoots 00:10:17.680 --> 00:10:24.560 about ray tracing AMD is trying to make up for this somewhat though by focusing 00:10:22.079 --> 00:10:28.320 more on game enhancements via variable rate shading mesh shading and fidelity 00:10:27.279 --> 00:10:34.160 effects but these are up to the developer to include you can't just turn on the 00:10:32.399 --> 00:10:39.120 ambient occlusion feature on any game that you want this combined with enhanced media 00:10:37.120 --> 00:10:44.720 decoding support and support for full fat HDMI 2.1 with freesync 00:10:41.920 --> 00:10:49.120 well they merely bring AMD up to date rather than allowing them to leap over 00:10:46.880 --> 00:10:52.399 NVIDIA who already has mesh shading variable rate shading ambient occlusion 00:10:50.800 --> 00:10:55.760 sharpening on top of all the other things that AMD's missing out on here 00:10:54.880 --> 00:11:01.680 so i'm not disappointed by the rx 6800 00:10:58.800 --> 00:11:05.200 series AMD delivered on everything they promised in their keynote but if you 00:11:03.760 --> 00:11:08.480 were hoping that there was going to be some extra dazzle you know one more 00:11:07.519 --> 00:11:14.480 thing then you might feel a bit let down 00:11:11.440 --> 00:11:16.480 what you see is what you get and at 00:11:14.480 --> 00:11:22.480 these prices they're just a little tough to recommend 00:11:19.360 --> 00:11:24.880 the rx 6800 xt is just fifty dollars 00:11:22.480 --> 00:11:28.880 less than the rtx 3080 and gets you roughly similar performance in 00:11:26.480 --> 00:11:35.279 traditional gaming with a big l in almost every other area the rx 6800 well 00:11:32.640 --> 00:11:40.320 honestly it's even tougher at 80 dollars more than the competing rtx 3070 well it 00:11:38.000 --> 00:11:43.120 competes well until you turn on ray tracing 00:11:41.360 --> 00:11:48.480 or start streaming but it's also only 70 dollars more for 00:11:45.680 --> 00:11:52.399 the significantly faster xt model which doesn't seem like a lot more by the time 00:11:50.480 --> 00:11:56.959 you're already spending 600 on a freaking graphics card 00:11:53.920 --> 00:11:59.440 so maybe the 6800 exists mostly as like 00:11:56.959 --> 00:12:04.640 a kicker sku and we'll see some crazy promotions on it or something 00:12:01.279 --> 00:12:06.800 all in all a plus for effort but anyone 00:12:04.640 --> 00:12:12.160 who wanted to see NVIDIA get absolutely embarrassed this generation is just 00:12:09.600 --> 00:12:15.839 gonna have to wait for AMD's next crack at top tier gaming performance 00:12:14.240 --> 00:12:19.600 though AMD's probably still gonna move a lot of these given that you can't 00:12:17.279 --> 00:12:23.120 actually buy an rtx card at the moment now that i think about it 00:12:21.200 --> 00:12:27.839 just like i always think about our segways get the best prices and best 00:12:25.600 --> 00:12:32.720 selection on pc hardware and technology at any of microcenter's 25 locations 00:12:30.399 --> 00:12:37.680 across the united states micro centers got new prices on the 9th generation 8 00:12:35.200 --> 00:12:43.360 core cpus from Intel with the 9700k at just 199 and 9900k at 299 and you can 00:12:41.600 --> 00:12:46.880 save an additional 20 bucks when you bundle with a compatible eligible 00:12:44.880 --> 00:12:51.279 motherboard and you can follow the link in the description for a free 32gb flash 00:12:48.880 --> 00:12:54.480 drive and 32gb microsd card valid in store only no purchase necessary really 00:12:53.040 --> 00:12:57.920 no purchase necessary how the hell does that work well thanks for watching guys 00:12:56.079 --> 00:13:01.760 go check out our coverage of AMD's where gaming begins to get a little more 00:12:59.440 --> 00:13:05.680 background on today's launch we decided to focus this video more on just like 00:13:03.600 --> 00:13:09.440 the performance and feature analysis let us know if you like that approach