Gadgetory

finding something to actually leverage the increased memory bandwidth of Radeon seven is a bit of a challenge few games will genuinely use more memory than what's found on an RT X 2080 let alone these 16 gigabytes on Radeon 7 and most vram capacity utilization reporting is wildly inaccurate as it only reports allocated memory not necessarily the used memory some applications will allocate more than they ever used for instance and that's what windows task manager GBC will show you two best benchmarks the potential advantages of Radeon seven which would primarily be relegated to memory bandwidth we set up targeted feature tests still looking at gaming to look at anti-aliasing and high resolution benchmarks consider this an academic exercise on Radeon 7s capabilities before that this video is brought to you by Thermaltake score p3 case the core p3 is one of the most unique cases on the market it can serve as an open-air standing chassis a test bench in vertical or horizontal orientation or as a wall-mounted showcase PC the core p3 now comes with a 5 millimeter thick tempered glass panel for its slide but keeps the front top and back open for air the core p3 versatility as a display piece test bench or standard desktop is reinforced by its price of roughly 110 dollars on Amazon you can learn more at the link in the description below so I stayed an academic exercise here what that means is that we're going to be regularly exiting real-world use cases to just better look at how does the card scale how does it scale versus an RT X 2080 it's totally irrelevant what the frame rate is ultimately at 8k resolution because you can't play at 8 K with either device so even with really low settings so that's why we call this academic exercise it's just it's it's looking at what can happen in scaling but not necessarily what's usable for you so what we're looking at here is some unplayable scenarios in a lot of cases although some are playable 8k resolutions included just to see if scaling improves over AMD or Nvidia I mean if we see for example a 10% difference where Nvidia might hold the 10% lead in one particular new 4k does that gap close does it become zero percent or three percent if we switch to AMD at say while 8k resolution as it can an example so another thing that we can do here is in theory test anti-aliasing this is something that Andy indicated should be a bit advantaged on the Radeon card and that makes sense logically anyway so if we change anti-aliasing where you take multiple samples per pixel typically 2 X 4 X 8 X 16 X so forth or sometimes called tap to tap it could have a similar performance impact to just increase in the resolution it's really the same idea hits the same part of the pipeline so if am these memory bandwidth really benefits and in games what we would hypothesize would happen would be a diminishing or reversing performance Delta between the 20 80 and Radeon 7 that anti-aliasing increases as we make it more difficult if we increase the resolution things like that so we already demonstrated how the Radeon 7 card can sometimes close the gap at higher resolutions when we posted our review and we talked about how the gap like an f1 2018 would widen at 1080p and we can show some of those charts again now though of course check the review for fold these help but looking through those charts again at 4k the gap is a bit better in some of these games for Radeon favoring Radeon or at least minimizing the difference and then you see again at 1440p it starts to widen but it's still reasonably favorable at 1080p that gap really starts to widen where Nvidia pulls ahead and a couple of the games we tested and that's just because no longer is it the case that the HBM is really being leveraged in a way that's meaningful because at 1080p your just not hitting the memory quite as hard as at 4k so this is where we can start looking at other scenarios cue in mind that the obvious downside as we enter territory where the content is unplayable at the framerate so you might not really be able to do anything with this information but at least it's interesting we ran a few different types of tests with Unigine superposition this is a great test for a quick synthetic benchmarking and give us a look at performance as it scales across resolutions starting with shader set to low textures set to high and resolution changes at 4k 5k and 8k we saw that Nvidia held a lead of about twenty seven point four percent over baseline moving to twenty seven point nine percent lead with an error over baseline at 5k and then a twenty eight point one percent lead at 8k there's really no change here from 4k to 8k at least with regard to scaling although obviously these resolutions become decreasing lis playable as they increase we ran a one-off test next with extreme shaders instead finding that Andy actually gained ground in this scenario in a pretty significant way Andy operated at fifteen sixty six points on average with Nvidia at 1504 points we reran this test four times and averaged it determining that this Delta is outside of standard deviation between those test passes so it is a meaningful difference after learning that bigger differences emerge with extreme shaders we reran the tests from 1080p to 4k at 1080p and videos are TX 2080 performed about twenty five point eight percent better than the Radeon seven guard with a 100% baseline there at 1440p VRTX twenty eighties advantage rapidly decayed the 16.1% 4k results are too close and create kind of a messy chart so sorry about that but the RT x 2080 ends up actually outperformed by radeon seven here for the first time in all of these tests with Radeon seven holding a couple percentage point lead we can see that as resolution scales with this particular game using the extreme shaders which are pretty compute intensive in this benchmark it looks like Andy does change performance in a meaningful way so in terms of real framerate the benchmark was operate and had about eleven to twelve FPS average clearly this isn't play it's not even a game but if it were it's not playable anyway and that advantage is really not realized since no one would realistically use these settings and a game but it is interesting to see how two cards can flip positions in different testing scenarios and it gives us a bit of insight as to how Radeon seven works and what it does well and this sort of confirms some things we already knew which is that memory bandwidth does matter as you increase the memory the load on the memory boss's Murray pipe and also teaches us or reminds us that he does well in compute scenarios firestrike remains one of the best tools for this type of synthetic workload for this we'll be looking at gt1 and gt2 separately gt1 heavily loads the GP with polygons and tessellation performs its shadow and illumination crunching and uses compute shaders for post-processing and particle physics and simulation gt2 is very heavy on compute shader workload and greatly increases pixels processed per frame but reduces tessellation workload by more than 50% gt2 should therefore run better on Radeon 7 and gt1 wood relative to the RT x28 efe so this is where it's really important to understand what the benchmark is doing and not just run it blindly because if we understand those two differences we can look specifically at the two workloads and trying I'm trying analyze where would one card theoretically do better than the other starting with gt2 we see that the stock firestrike ultra settings have nvidia below and these performance at eighty six point seven percent of the baseline radeon seven 100% performance this is unique to this benchmark thus far at 2x MSAA that gap shrinks to ninety point seven percent of baseline performance for x MSAA brings us to ninety four point six percent of baseline with 8x MSAA finally allowing the RT x 2080 to surpass AMD and hold at one hundred four point seven percent of baseline performance it makes sense that AMD would generally be more competitive in gt2 where compute workload and memory bandwidth are both stressed as for why nvidia starts to pull ahead once msaa is brought to 8x we're working with 3dmark steam and others to try and better understand the specific behavior that we're seeing here our estimation is that this may have to do with memory compression technology on nvidia or maybe edge detection and anti-aliasing with this specific hardware and software combination the software implementation especially so there may be implementation level advantages here but we're working with the teams to try and better understand them for GT 1 getting that chart on the screen there's more tessellation and polygons are a bit heavier here so we see the two cards start to functionally equal each other under stock fire strike ultra settings with nvidia gaining a notable 16% lead at 2x MSA a 21% lead at 4x MSAA and 27% at 8x MSAA and that will also note that there was a massive difference between Radeon 7 and the 28 efe when running 8k resolution with 8x msaa now as a percentage this difference looks like something like 60% favoring the Radeon card however in reality you're talking about something like four point something FPS versus one point something FPS so it is clearly unplayable on both however very interesting to look at the differences the reason this particular difference emerges is because we were running out of memory on NVIDIA card so in order to exhaust properly truly exhaust that 8 gigabyte frame buffer we had to run an 8k resolution with 8 tap multi sample anti-aliasing that's a lot of pixels and and that's what it took to exhaust the frame buffer at which point AMD was able to produce a significantly higher FPS when looking at percentages but in reality it's a 3 FPS difference and it's still unplayable it's just that that's where you start seeing that deviation of capacity versus brute force in other ways so interesting data but not particularly useful considering how unplayable it is anyway GTA 5 has a lot of anti-aliasing options so we'll exercise both msaa from 2 X - 8 X and then the reflections MSAA at 8 X well start this with just using an average FPS chart at the different settings which will best illustrate how the gap changes between each test at 4k and very high in ultra settings for everything 2x MSAA puts us at 58.8 FPS average for the 2080 and 51.4 for the Radeon 7 card the more important indicator is that the 2080 holds a 14.4% lead over the Radeon 7 card at 4x msaa that lead doesn't really change much and scales down to 14.1% and it's pretty close to error at that point at 8x msaa we start to see some movement and the average FPS of 27.5 on Radeon 7 verses 31 FPS on our TX 2080 shows that the improvement for the r-tx 2080 the phrase down to 13% down from an initial 14.4% by then also adding 8x MSAA reflections in addition to the other AAA we see that the difference is now changing to a 10.8% advantage for the 2080 fe which diminishes its benefit looking at the line graph in total we can see that andy does start recovering some losses toward the extreme end of the scale it's just that again we're in unplayable territory for other reasons but the gap does close so this sort of reinforces that initial hypothesis from the review that performance will be better at high resolutions and heavier anti-aliasing for reference with the amount of test passes we ran standard deviation run to run was about 0.3 FPS average so these results are very accurate far cry 5 is next for this one we test it with SMA Ataa and HD textures under different scenarios unfortunately we can't really expand the anti-aliasing it in any other way beyond these two in the first test at 4k and using HD textures with SMA we measured an Nvidia at 62.8 FBS average and AMD at 60 point 6 FPS average put in Nvidia as about 3.5% ahead with 1% low similarly spaced as were 0.1% lowest to test scaling with different settings we also ran a test with the reduced vram consumption but otherwise identical settings we can switch to that chart now the results disabling HD textures were functionally identical to the results with HD textures and nothing else changed it's no surprise that we're within margin of error as textures don't really impact performance unless vram becomes a limitation on both devices quantitatively the experience is unchanged from the previous chart qualitatively the output is the same on each device so one to the next there's no visible difference if you're ignoring framerate and things like that again Nvidia is about 3.3 percent faster in this test as well switching to taa instead performance falls off a little bit for each device but Nvidia still maintains a 3.3 percent advantage in average FPS HD textures and anti-aliasing changes in far cry 5 aren't enough to reduce or change the performance delta from card to card closing out then we see some interesting scaling and a few of these titles far astray the most useful it gives us the most the clearest picture of what's happening and why because of well we know how fire strike is built it's extremely well documented so that one's interesting to look at it's also potentially a bit useful to know that it required an 8k resolution with 8 tap msaa to exhaust the frame buffer on the r-tx 2080 fe and start to gain an advantage in a simulated gaming scenario it's synthetic obviously with the Radeon card now again unplayable anyway at that point but this might scale elsewhere and maybe it would be useful to you to know this information so that's it for this one as stated it's an academic exercise this is not particularly useful in the real world but it's very interesting and if you want actual performance numbers as it pertains to playing the game we would of course advise you check out our review of the Radeon 7 card for more information so that's it for this one thank you for watching as always subscribe for more go to store documents access net that helps out directly or patreon.com slash gamers Nexus and we'll see you all next time you