WEBVTT 00:00:07.759 --> 00:00:16.720 sometimes you don't realize what you're missing until it's staring you in the face but 00:00:12.880 --> 00:00:18.640 when you notice it it's just 00:00:16.720 --> 00:00:21.439 so sweet isn't it well 00:00:19.760 --> 00:00:25.840 that's what we finally have here today with NVIDIA the opportunity to have our 00:00:23.840 --> 00:00:31.920 cake and eat it too because they have finally enabled resizable bar on their 00:00:28.160 --> 00:00:35.200 desktop GeForce gpus get it resizable 00:00:31.920 --> 00:00:35.200 bar extra sweetness 00:00:36.079 --> 00:00:42.960 in theory this is a simple driver update that will unlock extra performance on 00:00:40.559 --> 00:00:47.440 your existing GPU assuming you're lucky enough to own one but what we're gonna 00:00:44.719 --> 00:00:50.320 find out is if it is truly something for nothing 00:00:48.399 --> 00:00:53.200 or if it's just an illusion 00:00:53.600 --> 00:00:59.440 i'll give you uh this segue to our sponsor for nothing though smart deploy 00:00:57.520 --> 00:01:03.520 can help you manage every Windows endpoint in your environment from one 00:01:01.359 --> 00:01:08.479 centralized golden image grab your exclusive free software worth over eight 00:01:05.439 --> 00:01:11.479 hundred dollars at smartdeploy.com 00:01:08.479 --> 00:01:11.479 Linus 00:01:18.159 --> 00:01:24.000 resizable bar has been a big deal ever since AMD announced their implementation 00:01:22.240 --> 00:01:27.759 of the feature called smart access memory if you haven't been keeping up 00:01:25.600 --> 00:01:33.119 with all the buzz it basically allows a CPU to access all of a pci express 00:01:30.640 --> 00:01:38.640 device's memory at once rather than in small 256 megabyte chunks as has been 00:01:36.079 --> 00:01:42.960 the case until now resizable bar has existed in the data center for some time 00:01:40.640 --> 00:01:47.680 now but it couldn't find momentum on the desktop until AMD kicked off this 00:01:45.200 --> 00:01:51.200 scramble to implement it industry-wide which kind of raises the question why 00:01:50.079 --> 00:01:56.399 now well in the early days of pci express it 00:01:53.840 --> 00:02:00.719 wasn't a problem for gamers because game assets like textures were sized 00:01:58.880 --> 00:02:05.520 proportionally to the video frame buffers of that time but with the 00:02:02.880 --> 00:02:10.800 complexity of today's games and with modern graphics cards having as much as 00:02:07.439 --> 00:02:13.120 24 gigabytes of onboard memory you can 00:02:10.800 --> 00:02:18.560 easily imagine a situation where the CPU could waste a lot of cycles getting the 00:02:15.040 --> 00:02:21.040 data it needs if it can only address 256 00:02:18.560 --> 00:02:25.520 megabytes at a time ultimately eliminating this waste is where our 00:02:23.440 --> 00:02:30.319 performance boost comes from the only problem is that in order to support it 00:02:28.160 --> 00:02:36.800 every device in the chain needs to be capable the CPU the motherboard and the 00:02:33.920 --> 00:02:41.120 pci express device which in this case is a GPU now we've already looked at 00:02:39.040 --> 00:02:46.800 NVIDIA's preliminary implementation of resizable bar on mobile GeForce gpus 00:02:43.760 --> 00:02:49.120 with an MSI ge76 raider laptop and at 00:02:46.800 --> 00:02:52.959 the time we concluded that there was more to the sometimes impressive 00:02:51.200 --> 00:02:58.319 performance improvements that AMD was getting than just enabling resizable bar 00:02:56.239 --> 00:03:02.640 and we also concluded that NVIDIA had some work to do before it was ready for 00:03:00.000 --> 00:03:07.519 prime time but even with it fully cooked it's important to know that not every 00:03:04.720 --> 00:03:11.360 game or application will benefit and how the driver handles memory management 00:03:09.519 --> 00:03:17.040 appears to play a major role in what kind of performance uplift or fall you 00:03:14.480 --> 00:03:22.239 can expect that's where NVIDIA's latest drivers and the free chocolate come in 00:03:20.319 --> 00:03:27.519 our previous investigation showed that the impact differs depending on CPU 00:03:24.400 --> 00:03:31.440 performance so we grabbed both a ryzen 5 00:03:27.519 --> 00:03:32.400 5600x and a ryzen 9 5900x to represent 00:03:31.440 --> 00:03:38.959 the mid-range and the high end we'll be using a GeForce rtx 3080 and for comparison 00:03:36.720 --> 00:03:42.319 against team red we've got a Radeon rx 6800 xt 00:03:40.720 --> 00:03:47.200 something to note is that because we're focused on their respective improvements 00:03:44.799 --> 00:03:51.519 over stock performance and not on completely re-reviewing these cards all 00:03:49.440 --> 00:03:57.120 of the results for both sides will be in relative percentages at 1080p things are 00:03:54.799 --> 00:04:01.439 already really interesting shadow of the tomb raider doesn't change much but 00:03:59.280 --> 00:04:06.879 where Radeon loses some performance in the five percent lows on ryzen 9 GeForce 00:04:04.319 --> 00:04:12.480 gains hinting that NVIDIA might be doing a better job of reducing die to die 00:04:08.799 --> 00:04:15.439 latency on AMD's own CPU than AMD's 00:04:12.480 --> 00:04:20.000 graphics team is gta 5 on the other hand gives GeForce a hard time with Radeon 00:04:17.919 --> 00:04:24.479 gaining a bit with verizon 5 which further hints that AMD's implementation 00:04:22.079 --> 00:04:29.759 favors single die cpus as for f1 2020 this is one of NVIDIA's 00:04:27.280 --> 00:04:34.400 call outs for performance uplift and it brings the first major victory for 00:04:31.680 --> 00:04:38.960 resizable bar we ended up with slightly higher average frame times across the 00:04:36.479 --> 00:04:44.080 board for team green which is bad but significantly lower one percent minimum 00:04:41.600 --> 00:04:49.759 frame times that reach toward a 10 improvement on the 5900x that's good 00:04:47.360 --> 00:04:54.560 because in almost any game improving performance during challenging scenes is 00:04:52.560 --> 00:04:58.639 way more important than when animations are already smooth this is especially 00:04:57.120 --> 00:05:02.240 true in competitive titles when performance tends to fall in the heat of 00:05:00.639 --> 00:05:06.880 battle because of all the character models and effects that are on screen 00:05:04.639 --> 00:05:11.759 again NVIDIA seems to be taking better advantage of the extra course on the 00:05:08.720 --> 00:05:14.160 ryzen 9. forza horizon 4 is another call 00:05:11.759 --> 00:05:18.240 out and another win for NVIDIA although the pattern flips here with the ryzen 5 00:05:16.560 --> 00:05:22.479 pulling better improvements across the board than the ryzen 9. this suggests 00:05:20.720 --> 00:05:26.320 that memory access patterns play a significant role in whether the die 00:05:24.000 --> 00:05:30.479 arrangement causes issues Radeon on the other hand struggles especially with the 00:05:28.320 --> 00:05:34.320 ryzen 9 where minimum frame rates are as much as six percent lower than with 00:05:32.320 --> 00:05:38.320 resizable bar disabled which brings us back to our chocolate analogy doesn't it 00:05:36.639 --> 00:05:43.039 it's perhaps unsurprising that there's little change on flight sim 2020 given 00:05:40.479 --> 00:05:46.960 how CPU-bound that game is but NVIDIA does manage to eke out a slight 00:05:44.800 --> 00:05:51.440 improvement with ryzen 5 while Radeon manages a bit better frame rates on 00:05:48.800 --> 00:05:55.280 ryzen 9. and assassin's creed valhalla was the call-out title that we found 00:05:52.960 --> 00:05:59.759 improvements for in our previous video and that continues today NVIDIA pulls a 00:05:57.759 --> 00:06:04.000 respectable six to nine percent improvement across the board but then 00:06:01.600 --> 00:06:09.440 again that pales in comparison to team red pulling way ahead of stock with up 00:06:06.319 --> 00:06:11.039 to 18 percent higher performance given 00:06:09.440 --> 00:06:14.800 the way that both teams are going back and forth here it seems like additional 00:06:12.960 --> 00:06:19.759 optimizations can still be made for resizable bar on both sides which is 00:06:17.039 --> 00:06:23.919 really promising finally cs go it's an older title that already runs well on 00:06:21.600 --> 00:06:28.000 hardware of this caliber but we still managed to see a slight improvement in 00:06:25.759 --> 00:06:32.960 minimum frame rates at the cost of some average FPS which is a price that most 00:06:31.120 --> 00:06:36.880 players should be willing to pay for the reasons that i outlined before there is 00:06:35.199 --> 00:06:41.360 no price by the way for the new sticker packs at lttstore.com free with every 00:06:38.960 --> 00:06:45.919 order checkmate NVIDIA moving on to 1440p you might not expect 00:06:44.240 --> 00:06:50.560 things to get better here but you'd be wrong while g4 stays more or less the 00:06:48.319 --> 00:06:55.840 same in shadow of the tomb raider Radeon pulls two to four percent ahead with 00:06:52.800 --> 00:06:58.240 ryzen 5. and where GeForce loses in 00:06:55.840 --> 00:07:03.440 minimum frame rates for no gain on ryzen 9 Radeon manages to gain about as many 00:07:01.199 --> 00:07:08.400 average frames as it loses in minimums which in a non-competitive title like 00:07:05.599 --> 00:07:12.479 this one might be an okay trade-off gta 5 is a bit all over the place with ryzen 00:07:10.720 --> 00:07:17.280 5 more or less on par with stock performance on both teams with a slight 00:07:14.720 --> 00:07:22.240 tip to Radeon and ryzen 9 pulling as much of a win on g-force as Radeon loses 00:07:20.319 --> 00:07:27.440 it's almost like NVIDIA stole the frames right out of lisa sue's pocket and f1 00:07:24.880 --> 00:07:30.960 2020 again shows improvements as high as 13 00:07:28.720 --> 00:07:35.360 with the only maybe regression coming from NVIDIA's average frame rates on 00:07:32.880 --> 00:07:38.960 ryzen 9 but we're talking a mere one percent difference NVIDIA comes out 00:07:37.360 --> 00:07:43.599 ahead in performance improvements for minimum frame rates here while Radeon 00:07:40.960 --> 00:07:48.319 pulls better averages as for forza there is no other way to put it it is a 00:07:45.599 --> 00:07:54.639 staggering win for Radeon their worst improvement here is 13 and yes your eyes 00:07:51.680 --> 00:08:00.000 are working fine that is a nearly 20 gain for Radeon's minimum frame rates 00:07:57.599 --> 00:08:05.039 now to be clear NVIDIA still wins everybody wins when they go over 100 of 00:08:02.080 --> 00:08:09.360 stock performance they just don't win as hard now with a little extra emphasis on 00:08:07.199 --> 00:08:13.599 the GPU thanks to our higher resolution flight sim 2020 shows us near universal 00:08:12.000 --> 00:08:18.160 improvements with Radeon actually pulling a healthy 8 improvement in 00:08:15.520 --> 00:08:22.560 minimum frame rates on ryzen 9. and then assassin's creed is once again a bigger 00:08:20.160 --> 00:08:27.840 win for AMD than it is for NVIDIA with improvements reaching 16 for both cpus 00:08:25.599 --> 00:08:31.840 and NVIDIA managing a very respectable six to nine percent lead over stock 00:08:29.680 --> 00:08:37.279 performance isn't it nice when everybody wins guys too bad it doesn't always work 00:08:33.919 --> 00:08:39.120 out that way cs go at 1440p resulted in 00:08:37.279 --> 00:08:44.720 straight up performance regressions across the board with the sole exception 00:08:42.000 --> 00:08:50.000 being minimum frame rates on ryzen 9 for Radeon GeForce takes the worst of it 00:08:47.120 --> 00:08:55.279 though here with up to a 20 performance loss with resizable bar enabled that's 00:08:52.560 --> 00:08:59.519 as big a loss as the largest gains we've seen today maybe that's why NVIDIA's 00:08:57.360 --> 00:09:03.760 driver has an easy to find indicator for whether the feature is active i guess 00:09:01.519 --> 00:09:07.839 AMD felt it was less likely that users with it enabled would be troubleshooting 00:09:05.600 --> 00:09:11.360 performance looking at the big picture here though 00:09:09.440 --> 00:09:16.240 Radeon has the best average gains running roughly six percent faster than 00:09:13.360 --> 00:09:21.519 stock at 1440p and two percent faster at 1080p while NVIDIA's gains hover around 00:09:19.519 --> 00:09:25.680 two percent regardless of the resolution that means one of two things either 00:09:23.600 --> 00:09:30.240 AMD's memory management for Radeon is worse than NVIDIA's for GeForce or 00:09:28.240 --> 00:09:34.000 NVIDIA has some more optimization work to do the work they've done already 00:09:32.160 --> 00:09:38.399 since our previous video has clearly paid off handsomely though so either way 00:09:36.320 --> 00:09:42.240 it goes we'll hopefully keep on seeing improvements with future driver 00:09:39.760 --> 00:09:46.399 revisions on both sides on that subject by the way get subscribed because NVIDIA 00:09:44.560 --> 00:09:50.959 finally rolled the driver that fixes the code 43 error and we're gonna be all 00:09:48.880 --> 00:09:54.959 over that in a future video changing gears a bit by now we know that 00:09:52.959 --> 00:10:00.640 productivity tends not to be affected much by resizable bar at least for AMD 00:09:58.160 --> 00:10:06.080 but for fun we did some runs so that we could validate that for NVIDIA as well 00:10:02.880 --> 00:10:08.959 blender well uh it's 100 00:10:06.080 --> 00:10:13.600 yep 100 of stock performance so how about that pugetvent shows some dips 00:10:11.600 --> 00:10:17.200 especially in after effects although which CPU causes the dip changes 00:10:15.680 --> 00:10:21.760 depending on which team GPU you're batting for and then luxmark shows 00:10:19.519 --> 00:10:27.440 basically flat with our ryzen 9 g-force combo managing to lose some performance 00:10:24.640 --> 00:10:32.079 this clearly isn't the norm but it's worth taking into account that some 00:10:29.279 --> 00:10:36.480 applications may suffer for those sweet sweet gaming games 00:10:35.120 --> 00:10:41.519 so then resizable bar should you turn it on assuming that your 00:10:39.839 --> 00:10:47.440 motherboard manufacturer has rolled out a BIOS with support and that you have a 00:10:43.920 --> 00:10:49.200 compatible GPU the answer is yes 00:10:47.440 --> 00:10:52.959 we found that so far there are more situations where it helps than 00:10:50.800 --> 00:10:57.680 situations where it hurts and like with our chocolate bar where it hurts the 00:10:55.279 --> 00:11:02.240 benefits of the tasty snack seem far more noticeable than what went missing 00:11:00.240 --> 00:11:05.680 now if only they could enable it for previous gen gpus 00:11:04.160 --> 00:11:09.920 just like i enable you guys to hear about our sponsor the drop sennheiser 00:11:07.760 --> 00:11:14.640 hd6xx headphones are one of drop's all-time best sellers with over 90 000 00:11:12.640 --> 00:11:19.920 units sold they haven't changed the driver or sound structure compared to 00:11:16.640 --> 00:11:21.680 the hd650s so it delivers a beautiful 00:11:19.920 --> 00:11:26.240 balanced mid-range and natural sounding bass and its open back design allows 00:11:24.079 --> 00:11:30.240 auditorium-like acoustics for accurate sound reproduction it's got a detachable 00:11:28.320 --> 00:11:34.480 six foot cable instead of a 10 foot cable this change was based on community 00:11:31.920 --> 00:11:38.320 feedback and has a 1 8 inch plug with a quarter inch adapter included in the box 00:11:36.399 --> 00:11:42.240 so don't wait get yours today at the link in the video description so thanks 00:11:40.320 --> 00:11:46.720 for watching guys go check out our previous investigation into resizable 00:11:44.240 --> 00:11:52.160 bar on NVIDIA's mobile platform for a little more context on how far NVIDIA's 00:11:49.680 --> 00:11:55.839 implementation has come in such a short period of time