WEBVTT 00:00:00.000 --> 00:00:05.759 it's obvious that NVIDIA gpus are just 00:00:03.360 --> 00:00:10.320 playing better I mean look how few people are buying Radeon cards and if 00:00:08.400 --> 00:00:14.040 everyone's doing it it must be right right wrong 00:00:14.280 --> 00:00:22.320 actually when you look at the price and performance of AMD gpus right now on 00:00:19.260 --> 00:00:26.519 paper they're the obvious choice you can 00:00:22.320 --> 00:00:29.580 get a 6700 XT which delivers near RTX 00:00:26.519 --> 00:00:34.200 3070 performance for just about the 00:00:29.580 --> 00:00:36.600 price of an RTX 3060 and we just got our 00:00:34.200 --> 00:00:44.280 hands on a one-click tuning software from monasmus the creator of clock tuner 00:00:39.180 --> 00:00:48.600 Verizon that he claims will put a 6800 00:00:44.280 --> 00:00:52.920 XT this isn't it a 6800 XT in the same 00:00:48.600 --> 00:00:55.800 performance class as a 3090 TI a GPU 00:00:52.920 --> 00:01:01.379 that costs twice as much now obviously a claim like that needs to 00:00:58.500 --> 00:01:05.220 be taken with a grain of salt but if there is even any free performance on 00:01:03.359 --> 00:01:08.580 the table it's worth investigating to make sure that you don't get taken for a 00:01:07.619 --> 00:01:13.580 ride a ride to this message from our sponsor 00:01:11.000 --> 00:01:18.420 enlisted enlisted is a free-to-play multiplayer World War II shooter jump 00:01:16.380 --> 00:01:21.659 into large-scale battles with authentic moments in history like the invasion of 00:01:20.040 --> 00:01:25.619 Normandy check it out by following the link in the description the truth is I'm 00:01:24.060 --> 00:01:31.380 at least partly to blame for the current situation after some very bad 00:01:28.500 --> 00:01:36.240 experiences with AMD's GPU drivers that delayed our video production almost 00:01:33.420 --> 00:01:41.460 every time we tried to use them we switched to almost exclusively using 00:01:38.579 --> 00:01:46.439 NVIDIA gpus in our showcase builds and in our test rigs I mean during a review 00:01:44.220 --> 00:01:52.020 for a new card from AMD we would often find that it was pretty okay some 00:01:49.200 --> 00:01:55.680 instability or bugs notwithstanding but then it would invariably sit on the 00:01:53.880 --> 00:02:00.840 Shelf until it was needed as a point of comparison for some other review for me 00:01:58.860 --> 00:02:06.420 the experience that turned me off of AMD was a bit more personal back in 2015 I 00:02:03.659 --> 00:02:10.679 was swayed by the please by this person to vote for competition with your wall 00:02:08.520 --> 00:02:15.900 call it and I use my incredibly hard earned cash I was a student at the time 00:02:12.360 --> 00:02:17.700 to buy a Radeon R9 380. my favorite game 00:02:15.900 --> 00:02:21.300 at the time was dirt rally and it had an AMD splash screen at the start so it 00:02:19.680 --> 00:02:26.459 should be fine right yeah you would think so 00:02:23.300 --> 00:02:28.860 turns out Not only was AMD's performance 00:02:26.459 --> 00:02:33.480 subpart but due to driver issues the AMD gpus couldn't properly render foliage 00:02:31.500 --> 00:02:40.379 and shadows in effect making my experience barely better than the anemic 00:02:36.620 --> 00:02:43.920 gt740 I had before that DDR3 RAM in it 00:02:40.379 --> 00:02:46.080 in a game with AMD's logo on the splash 00:02:43.920 --> 00:02:50.459 screen it was a similar story many of my favorite games too you can tell how over 00:02:48.000 --> 00:02:55.260 it he is by how red his face is right now however there is good reason to 00:02:53.340 --> 00:03:01.459 think that things have changed back in July of this year AMD completely rewrote 00:02:58.739 --> 00:03:06.599 their DirectX 11 drivers apparently dramatically increasing stability and 00:03:04.080 --> 00:03:12.599 performance at the same time so to unscientifically test this Harrison and 00:03:10.319 --> 00:03:17.220 Alex from the writing team switched to team red for two weeks and played video 00:03:14.819 --> 00:03:21.180 games like they normally would to see if anything wonky happened I wasn't 00:03:18.959 --> 00:03:26.580 informed people were just taking gpus home but okay you're in Hawaii and it 00:03:24.120 --> 00:03:30.420 was fine the point is that after I made the switch and installed AMD's drivers I 00:03:28.800 --> 00:03:33.540 honestly just completely forgot that I'd swapped the GPU until this morning when 00:03:32.040 --> 00:03:38.099 I looked at the script again I didn't even have to adjust any of my 00:03:35.819 --> 00:03:44.159 in-game settings going from an RTX 3080 to an RX 6800 XT I could have easily 00:03:41.159 --> 00:03:45.840 just saved like 200 by going team red of 00:03:44.159 --> 00:03:49.379 course you guys deserve a more scientific approach to testing so we 00:03:48.060 --> 00:03:53.640 handed the cards that we'll be looking at today over to our Labs team to whip 00:03:51.840 --> 00:03:58.500 up some charts and graphs that I assume you're pouring over in detail right now 00:03:56.159 --> 00:04:03.900 if you're not super detail-oriented don't worry about that the bottom line 00:04:00.480 --> 00:04:05.940 is this AMD is at or near the 00:04:03.900 --> 00:04:13.260 performance of NVIDIA in traditionally rendered so not Ray traced games at 00:04:09.360 --> 00:04:16.019 anywhere from 100 to 200 dollars cheaper 00:04:13.260 --> 00:04:21.660 and that's before we pull out our secret weapon when Asma sent us an early 00:04:18.479 --> 00:04:23.699 preview of what he's calling Hydra 1.2 a 00:04:21.660 --> 00:04:28.199 user-friendly tool for rdna 2 architecture Radeon cards that can be 00:04:25.919 --> 00:04:33.360 used to easily improve both performance and power consumption Alex is going to 00:04:30.840 --> 00:04:37.919 show us how it works to get Hydra right now now you have to subscribe to one 00:04:35.100 --> 00:04:43.020 usmas's patreon but this tool will be free once it's you know not broken and 00:04:40.680 --> 00:04:48.900 stuff speaking of which if you have an Intel CPU it won't work unless you have 00:04:46.080 --> 00:04:53.759 an F model so right now the igpu will take priority when you boot it up and it 00:04:52.560 --> 00:04:59.100 just doesn't work I'm sure that's very easy to fix but uh we have a I9 12900kf 00:04:57.000 --> 00:05:04.259 in here so that solved that problem first of all a warning you can destroy 00:05:02.160 --> 00:05:08.040 your processors by using the software if you do something really really bad it's 00:05:06.600 --> 00:05:11.880 probably fine most of the time you'll just crash your stuff but you know you 00:05:10.259 --> 00:05:17.220 need the warning right there so this is your basic overclocking 00:05:14.400 --> 00:05:22.800 dashboard your sort of main stuff is all in here so we have like this eraser 00:05:20.280 --> 00:05:26.400 deletes all of your stuff and restores it all back to your factory defaults you 00:05:24.840 --> 00:05:31.380 can also come back in here so that you change your horror frequency you can 00:05:28.500 --> 00:05:37.259 save it right there boom it's saved now that save button only applies it to you 00:05:34.620 --> 00:05:43.560 know this it doesn't actually apply it to the GPU so what this is basically 00:05:40.080 --> 00:05:46.680 doing is changing stuff within the GPU 00:05:43.560 --> 00:05:48.180 drivers itself the power play table and 00:05:46.680 --> 00:05:53.699 that's what's doing your overclocking you do that with this little 00:05:50.639 --> 00:05:55.680 lightning bolt then driver resets and 00:05:53.699 --> 00:05:59.460 you're good to go now all this stuff right here is just your basic manual 00:05:57.360 --> 00:06:02.580 overclocking but why would you overclock something yourself when you can have 00:06:00.660 --> 00:06:07.860 just the robots do it for you let's go over the diagnostic window GPU there we 00:06:05.460 --> 00:06:13.139 go this allows us to change what kind of a test we're doing so we can change it 00:06:10.560 --> 00:06:17.580 from this right here is opengl you can go to whatever you want and it's fine 00:06:14.580 --> 00:06:19.320 the duration is how long it's going to 00:06:17.580 --> 00:06:24.060 test for stability right now it's 90 seconds that's okay but if this was my 00:06:22.620 --> 00:06:28.919 personal rig and I was going to use the GPU like day in day out I would increase 00:06:26.639 --> 00:06:31.680 this by quite a bit it's the sort of thing you don't want your computer to 00:06:30.180 --> 00:06:37.560 crash in the middle of A ranked game below that we have the do overclocking 00:06:34.560 --> 00:06:40.440 window this right here is the real meat 00:06:37.560 --> 00:06:44.960 and potatoes of hydro 1.2 basically what this does is move your whole voltage 00:06:42.180 --> 00:06:44.960 frequency curve 00:06:45.600 --> 00:06:53.160 so your GPU is going to have something that looks a little bit like this given 00:06:50.880 --> 00:06:58.259 a frequency you will have a certain voltage to make sure that it doesn't 00:06:54.479 --> 00:07:00.419 crash there now AMD doesn't want to have 00:06:58.259 --> 00:07:04.319 this like right on the edge of crashing all the time so they're a little bit 00:07:02.400 --> 00:07:10.199 conservative with their values what Hydra is doing is just moving this curve 00:07:06.720 --> 00:07:12.300 over a tiny bit seeing if it crashes if 00:07:10.199 --> 00:07:16.919 it's all good and it just moves over a little bit more once it gets instability 00:07:14.419 --> 00:07:19.560 pushes it back a little bit you're good to go 00:07:18.120 --> 00:07:23.520 oh and that is exactly what we're doing 00:07:21.539 --> 00:07:29.340 right here so you can see our start offset that is negative 00:07:26.479 --> 00:07:35.039 43.75 millivolts so we'll be starting with the Curve moved over by that much 00:07:31.800 --> 00:07:36.599 and then the step offset is 6.25 so 00:07:35.039 --> 00:07:43.259 every time that it does an iteration it's going to be decreasing the voltage 00:07:39.060 --> 00:07:45.660 by another 6.25 millivolts so that's how 00:07:43.259 --> 00:07:49.740 you can see for the same amount of power you're going to be getting it going 00:07:47.400 --> 00:07:54.180 faster because let's say default we have this amount of voltage 00:07:51.960 --> 00:07:58.620 this frequency yeah you get it you get it GP goes 00:07:56.340 --> 00:08:02.580 faster the last field here is the power offset this is if you want to increase 00:08:00.660 --> 00:08:07.979 your power so that you have a custom water cooler you can just crank your 00:08:05.460 --> 00:08:13.800 power right up so say right now it's 15 that increased our TDP from 255 watts to 00:08:12.120 --> 00:08:18.780 292. I'm just going to set that back to zero 00:08:15.599 --> 00:08:21.000 additional power tune up here is where 00:08:18.780 --> 00:08:25.620 some things get a little bit strange so all the do overclocking is concerned 00:08:22.919 --> 00:08:30.180 about your GPU core whereas additional power tuning is focused a little bit 00:08:27.419 --> 00:08:34.919 more on the memory on some gpus if you get really good RAM you can actually 00:08:32.219 --> 00:08:39.300 drop your thermals by a lot while also overclocking the GPU core if you're able 00:08:37.320 --> 00:08:42.779 to drop the voltage of your memory chips yeah but some of the other ones it 00:08:40.680 --> 00:08:48.380 doesn't work at all just hope you get lucky well Brandon should we run it 00:08:45.120 --> 00:08:48.380 to do let's go 00:08:50.940 --> 00:08:58.620 here we have MSI combustor it's not crashing so that's good and now 00:08:56.100 --> 00:09:03.860 we wait okay the diagnostic process successfully completed so we have a core 00:09:01.200 --> 00:09:09.240 curve offset of negative 68.75 millivolts 00:09:06.959 --> 00:09:14.339 should have been .025 more we'll fix that later anyway our total board power 00:09:11.640 --> 00:09:21.060 has stayed exactly the same our core frequency has only increased what is 00:09:17.220 --> 00:09:24.360 that uh 72 megahertz so that's not a 00:09:21.060 --> 00:09:26.700 whole lot oh wait uh average FPS in MSI 00:09:24.360 --> 00:09:31.920 combustor went up by one so it's not incredible that didn't do too much but 00:09:29.700 --> 00:09:37.620 what we did see is about a three to five percent increase in games after we had 00:09:34.680 --> 00:09:42.000 done this so not too bad to be clear what we're doing isn't black magic 00:09:39.300 --> 00:09:46.260 NVIDIA and AMD for that matter are running these kinds of optimizations 00:09:43.880 --> 00:09:51.660 constantly on their gpus to try to squeeze as much performance out as 00:09:48.600 --> 00:09:53.959 possible we're just using the excellent 00:09:51.660 --> 00:10:00.660 access that AMD provides to overclockers to fly a little closer to the Sun to be 00:09:58.019 --> 00:10:04.620 clear three to five percent it's not a ton of extra performance but as I like 00:10:03.000 --> 00:10:08.700 to say it's better than a kick in the teeth and given that it didn't cost us 00:10:06.240 --> 00:10:13.140 anything but 15 to 20 minutes of our time and it didn't even increase power 00:10:11.580 --> 00:10:17.339 consumption or heat I'd say it's kind of a no-brainer 00:10:15.720 --> 00:10:21.660 not everyone has 15 minutes though thankfully there's an easier way 00:10:19.279 --> 00:10:26.640 whenasmus created a one button overclocking tool that he calls monster 00:10:24.000 --> 00:10:33.060 profiles and it's with one of these that he claims we should be able to beat a 00:10:29.820 --> 00:10:36.839 3090 TI for half the price 00:10:33.060 --> 00:10:40.680 let's give it a shot for the RX 6800 XT 00:10:36.839 --> 00:10:43.560 and RX 6700xt their Radeon monster 00:10:40.680 --> 00:10:49.320 profiles so all you have to do is click this profile button right here so we 00:10:46.320 --> 00:10:50.760 have 355 good bin and bad bin those are 00:10:49.320 --> 00:10:54.779 coming soon we don't actually have those and given this cooler that would not 00:10:53.339 --> 00:11:01.380 work anyway we also have our 305 watt good bin and 00:10:58.100 --> 00:11:03.480 255 so that's stock power good bid we're 00:11:01.380 --> 00:11:09.140 going to try the 305 watt good bin use it 00:11:04.500 --> 00:11:11.640 apply Radeon monster profile enabled 00:11:09.140 --> 00:11:19.820 this is where we should be able to beat a 3090 TI this guy right here in time 00:11:15.839 --> 00:11:23.700 spy got us a score of 21 00:11:19.820 --> 00:11:25.019 425. can we beat it well I'm not gonna 00:11:23.700 --> 00:11:28.019 let the whole thing run because I already ran it and I'm just gonna spoil 00:11:26.880 --> 00:11:37.740 it for you we did not beat the 3090 TI the score of 00:11:32.640 --> 00:11:41.279 our 6800 XT was 21 255 just missed it 00:11:37.740 --> 00:11:43.680 under 200 points oh but that still is 00:11:41.279 --> 00:11:46.800 damn close for a GPU that's half as much money 00:11:44.760 --> 00:11:52.200 this thing works really quite well although honestly in real games outside 00:11:49.200 --> 00:11:54.839 of synthetic benchmarks it's not quite 00:11:52.200 --> 00:11:59.640 as impressive but at the same time it's free performance you can just put it on 00:11:57.120 --> 00:12:05.579 here your car goes faster the monster profile doesn't always work 00:12:01.980 --> 00:12:08.100 out great though our 6700 XT this one 00:12:05.579 --> 00:12:13.440 right here absolutely refused to work with any of them even the bad bin 00:12:11.100 --> 00:12:18.180 profile caused it to immediately crash the good news though is that hydra's 00:12:16.260 --> 00:12:22.200 automated tuning process you know the one that takes about 15-20 minutes did 00:12:20.640 --> 00:12:25.860 work for this card and we actually managed a similar performance 00:12:24.000 --> 00:12:31.200 Improvement by fine-tuning the voltage frequency curve just like on our 6800 XT 00:12:28.740 --> 00:12:37.500 which puts this card comfortably in competition with NVIDIA's 3070 and 3070 00:12:34.320 --> 00:12:40.320 TI for a hundred or in some cases well 00:12:37.500 --> 00:12:46.079 over a hundred dollars less so the 3090 TI slaying claims were a 00:12:44.100 --> 00:12:49.200 little optimistic but hey free performance is free performance and I've 00:12:47.760 --> 00:12:54.000 got to say that our experience with these cards both on the bench and for 00:12:51.899 --> 00:12:58.920 the writers at home lends credibility to AMD's claims of much better drivers 00:12:56.639 --> 00:13:04.860 though there are some inconvenient details in all of this NVIDIA has some 00:13:02.399 --> 00:13:10.139 feature benefits over AMD that simply cannot be ignored although AMD did 00:13:07.860 --> 00:13:16.200 recently improve their AMF encoder to include B frames it still isn't quite at 00:13:13.380 --> 00:13:19.860 the same level as NVIDIA's nvank and if the games that you enjoy support 00:13:17.880 --> 00:13:25.440 real-time raytraced lighting and you wish to enable said Ray tracing feature 00:13:22.440 --> 00:13:28.380 NVIDIA has a pretty clear performance 00:13:25.440 --> 00:13:32.940 advantage and that's before you enable their excellent dlss machine learning 00:13:30.720 --> 00:13:39.420 powered image upscaling technology AMD of course has their own version of 00:13:35.220 --> 00:13:42.540 dlss called FSR but frankly it isn't as 00:13:39.420 --> 00:13:44.339 good at least not yet by the time we 00:13:42.540 --> 00:13:48.360 release this video though the story may have changed because AMD will have 00:13:45.959 --> 00:13:54.959 announced their upcoming lineup of rdna 3 gpus hopefully with decent pricing and 00:13:52.200 --> 00:14:00.779 an even more competitive feature set but with how affordable and tunable rdna 2 00:13:58.440 --> 00:14:07.079 is today I think these cards are going to stay relevant for quite some time 00:14:03.540 --> 00:14:08.940 like our sponsor enlisted enlisted as a 00:14:07.079 --> 00:14:15.779 free-to-play World War II multiplayer shooter available on PC PS5 PS4 Xbox 00:14:12.360 --> 00:14:17.339 series X S and Xbox One with crossplay 00:14:15.779 --> 00:14:21.000 it has a focus on historical authenticity and gameplay that always 00:14:19.440 --> 00:14:26.160 keeps you in the middle of the action you'll be placed in large scale combat 00:14:23.760 --> 00:14:30.060 alongside dozens of soldiers vehicles and aircraft directly in famous 00:14:28.200 --> 00:14:34.320 historical campaigns like the invasion of Normandy there's hundreds of firearms 00:14:32.339 --> 00:14:37.680 and specialized weapons from common ones seen throughout the war to rare 00:14:35.880 --> 00:14:41.639 prototypes that you hardly see in other games there's even building construction 00:14:39.300 --> 00:14:45.180 and destructible items in the environments if you use our link you'll 00:14:43.620 --> 00:14:49.019 get a free bonus just for signing up including three days of Premium Time and 00:14:47.279 --> 00:14:55.079 several orders for troops and weapons so head to the link in the description and start playing enlisted free today if you 00:14:53.279 --> 00:14:59.220 guys enjoyed this video maybe check out our last one awesome video where we used 00:14:57.300 --> 00:15:03.440 clock tuner for ryzen to get free performance out of AMD's CPUs