Gadgetory

doom just pushed its Vulcan update finally so now the Vulcan API is available through dooms settings in-game that means that it's got OpenGL and Vulcan it does not run dx11 this is an opengl 4.5 game previously did it also run opengl 4.3 with AMD cards today we're looking at the Vulcan benchmark performance against OpenGL so we're specifically looking for scaling on AMD's cards and on Nvidia's cards I've only got a few cards tested here we didn't do the whole bench suite because really were just looking for kind of architectural scaling abilities it was not a full comprehensive bench of every card out there but we're looking at the GTX 1080 ci Pascal does the GTX at 972 see how the previous generation maxwell does and the RX 480 to see how andy does with the new polaris architecture and the internally was boasting close to 30% gain so we'll be validating that here now one thing that Vulcan does like dx12 they both sort of shift the workload from the CPU more to the GPU in instances where it makes sense like draw calls a draw call is when the CPU sends an actual signal to the GPU and says it's time to draw a triangle or primitive or whatever that is a draw call that's a pretty abusive process if you have a lot of geometry on the screen so for games with heavy geometric complexity and the x11 all these draw calls mean that you're really dragging down one component or the other normally the GPU is not performing to its fullest potential because the CP is dragging it down or something similar to that so with Vulcan it's the same idea as DX 12 same sort of asynchronous compute is available and exposed through Vulcan but it is up to the game developers to actually leverage that technology so just like we talked about with Chris Roberts from star citizen if you just wrap the game with this API and point to it it's actually going to have a negative output then if you build properly from the ground up and in software did work to properly implement Vulcan so let's look at the benchmark results the table on the screen now shows the components that we use for testing this game and most other games and the next interesting thing to do would be to pair these cards high-end cards with a lower-end CPU to see just how far you can get with the low on CPU with Vulcan as opposed to open to Yale but today we're just looking at GPUs so at 1080p with ultra settings the GT X 1080 shows an average FPS that increases 19.4% from OpenGL to Vulcan and that's a one 39.7 FPS with opengl 4.5 and a 160 8.8 FPS average with Vulcan the GTX 970 actually has an average FPS decrease of about 1% or mostly the same otherwise and there's a more variable range in Vulcan testing than with OpenGL so we're seeing between 105 and 108 FPS average and the actual average the average of all these numbers all these test passes is 106 dot 9 fps and that's against a 108 fps for OpenGL there are X 480 shows a 30 point 6 5% increase that's even bigger than what we're seeing on the GTX 1080 and that's moving from 85 FPS average with OpenGL to 111 FPS average with Vulcan so a pretty fair increase there looking at the millisecond time so this is the frame time between frames the average time in milliseconds the GTX 1080s average millisecond between the frames is 5.9 milliseconds and it's very consistent so we don't see any stuttering or spurious frame output the GTX 970 is at nine point three six milliseconds again fairly consistent and the RX 480 is at nine milliseconds and also fairly consistent at 1080p ultra four 1440p this story gets a little bit less interesting on the Pascal and Maxwell side the GTX 1080 only shows up 5% increase and now that we've kind of moved into more GPU intensive frame rates I guess this makes some sense and that moves us from 109 FPS average to one 14.5 five FPS average with Vulcan so 5% look at the GTX 970 we have a 0.36 percent decrease but for all intents and purposes it is effectively identical between OpenGL and Vulcan we're seeing 69 up five versus 60 9.25 FPS and depend how you average it they are pretty much the same so our X 480 it has a twenty nine point three percent increase so it's maintained its increase for the most part and has a pretty heavy lead here moving from fifty-seven FPS to 73 to seven FPS average now the reason I'm presenting these numbers this way as opposed to the normal way we kind of do a head-to-head comparison as I go through it for this I'm strictly comparing OpenGL and Vulcan on specific cards or architectures so when we talk about the RX 480 having a huge increase of almost 30% it's still lower performing than the gtx 1080 which should be given the price disparity and it's still very similar to the gtx 970 and that it just it really just means that polaris scales pretty well in this particular title with vulcan it does not mean it will apply to every game with Vulcan as we saw in the Talos principle for example so it's not direct head-to-head between cards look at it as a head-to-head between AP eyes on specific cards and architectures as for the millisecond times for 1440p ultra the GTX 1080 outputs 8.7 milliseconds with a tight range GTX 970 14.4 for milliseconds and there's more variants in this particular range so we do see some more occasional stutters but it's really not that bad and the RX 480 is at 13.5 7 milliseconds and is fairly fluid at 4k Ultra the GTX 1080s are effectively identical on OpenGL and Vulcan there's really no difference here the RX 480 outputs a twenty four point seven percent increase from twenty nine fps to 30 6.15 fps and is still unplayable you can't play it on either of these frame outputs with either of these api's so that don't read too far into that but the important part is that this scaling has maintained across all three big resolutions and we're still seeing big gains even at 4k which is not really where the 480 is a good performer latency on the GTX 1080 is about six tene 0.9 milliseconds between frames on average and the RX 480 is about 27.7 milliseconds between frames on average again totally unplayable as for 1% and 0.1% lows for all of these benchmarks resolutions they're fairly tightly timed with OpenGL and we don't presently measure that for Vulcan because of the limitations we have with our measurement tools so as stated the gains for the our X 480 are pretty large going from OpenGL to Vulcan the GTX 1080 has pretty good gains at 1080p you're not really gonna be playing at 1080p probably unless you're really desperately trying to hit 144 Hertz or something and then the GTX 970 is basically identical between the two with our X 480 really being the only one that outputs a consistent lead at all three big resolutions that we tested the thin here don't read too far into this because this is one game with one API change or implementation of Vulcan so as we've seen in games with DX 12 or the Talos principle you just because the card does a certain performance in one game with one API doesn't mean it will stretch across all games with that API because these API is even though they're powerful it really depends on how the developers implement them and they've got a build for certain support list so ashes of singularity for example supports MDA and can support really odd configurations like a 970 with 390 X and crossfire slash SLI and not every game will support that just cuz as DX 12 so the point here is that these numbers are for doom and doom only but it does look pretty good for the rx 480 in terms of scalability with Polaris from OpenGL to Vulcan 1080 is not bad at 1080p but otherwise pretty boring for Maxwell from the limited test run we performed on it so that's all for this video hit the patreon link push for video to help us out directly subscribe if you like the content and I'll see you all next time