Gadgetory

everyone said risin 7 is a good processor for streaming but there have been few data-driven tests to look at if this is actually true today's benchmark will be a truly multi-threaded multitasking workload or bogging down the 70 to 100 K and r7 1700 with gaming OBS game streaming and stream playback all on the same system at the same time we've been saying that Rison should theoretically be better at game streaming dollar-for-dollar since its launch but that was based largely on educated assessment of the production performance today we're here to look at whether the 1700 is better than the 7700 K at streaming not just better but better in a meaningful paradigm-shifting way before that this is brought to you by EVGA S at CLC 280 liquid cooler for CPUs which we previously benchmarked and found to be a high performer given its relative silence the temperature output learn more about this $140 cooler at the link in the description below there are a lot of things that make stream benchmarking difficult and a lot of ways to optimize the performance for one CPU or another for example you could tune the h.264 profile you go through xsplit OBS and change all manner of settings in there including the bitrate which certainly helps out and helps with uplink that's problem or you can start getting into windows side things like tuning the affinity or process priority for OBS or for xsplit that makes it challenging that's why it takes a while to get through this so everything here in this video just to get this out of the way now is based on our testing we're not going to make sweeping statements about either of these tube CPUs for all games streaming because frankly will vary based on the game based on if you use xsplit OBS based on your h.264 profile and tuning based on how much after you're willing to put into things like process priority optimization we've got a few of those tests here but it's just not reasonable to do all of them so everything here is based on our tests if you get better or worse performance in one way or another then try to use the settings we used and maybe you'll improve your performance or worsen it depending on what you're using so that's part one now the testing here looks at a few different things we are effectively doing a multitasking test finally but it's by using some very heavy workload so we have the game running we're using dirt rally right now for a baseline and then dota 2 as well just because it's a pretty common stream of old game well you can add other gains for the future but those to set a good baseline and then we also are using OBS or open broadcasting the open source program that you can use for free using that we have settings tunes that are all detailed in the article below and the stream testing methodology but the very quick part of it is based on viewer feedback from you all we took some polls not long ago and found that the most common settings for h.264 tuning were fast and very fast and so we chose the one in the middle faster and then there's a couple of others as well like ultra fast but we chose the one right in between what most of you use when you are streaming and then the next question we asked was what's the bitrate that you use and the answer was basically six to ten megabits per second so we tested six and ten megabits per second for main bit rates our uplink depending on the day it was tested limits us to anywhere from 20 to 40 megabits per second up but 6 to 10 is always within that range that's a non-issue and then finally we had some questions about twitch versus YouTube popularity it seems like originally YouTube was more popular twitch edged it out at the end twitch we had some issues with on the account side on the partner side of things so we ended up using YouTube which is fine because ultimately it allowed us to do more tuning anyway just because YouTube with our accounts anyway allows for better bit rates allows for better resolutions things like that so we did do some twitch testing at 720p60 we ended up scrapping it for some better testing with YouTube so that's the main information during this test and we collected power consumption which is a bit unique did that with a current clamp at the EPS 12-volt cables going to the motherboard straight from power supply so that is not through software and then we also measured thermals which we won't really be going into today in the video there will be some notes on it in the article and then we're focusing on a viewer experience and streamer experience viewer experience is what someone like you would see when you're watching a game streamer and the fluidity of the stream coming out to you and then player experience or streamer experience is what they're seen when they're playing the game which is not necessarily what you see they may be there may be some disparity in the framerate there so interesting stuff to look at that goes through most of it though this will wrap up our storyline for a lot of the 1770 700k testing you're previously tested VR and found no significant difference between the two of them for the most part now we're looking to see if the same is true for streaming we'll start with dirt rally then move to dota 2 on the screen now is a side-by-side output comparison of these 7,700 cavers of the r7 1,700 we have not labeled either right now both are in stock configuration this video playback is the resulting stream that went to youtube so this is what the viewer would experience but not necessarily the player this playback is at 1080p 60 natively with a 10 megabit per second operate now that you've had time to develop some thoughts on the outputs let's reveal which CPU corresponds to which playback this gives you an idea for viewer experience but let's put numbers to it we'll look at the players experience after we go through some of the OBS benchmarks getting to stream performance numbers starting with 1080p 60 at 10 megabits per second going to YouTube using dirt rally first we monitored the on-screen performance during the stream the 77 100k CPU stock dropped 40 4.3 percent of its frames during encoding due to inability to meet the sixteen point six seven millisecond deadline you want that number to be as close to zero as possible dropped frames it means that they are completely skipped they never make it to the viewer and so it's a stuttering experience if you're not performing at sixty or thirty or whatever you've configured it to the r7 1,700 CPU dropped 0.1% of its frame amounting to double digit numbers versus thousands of total frames and on for a second though because we do some process priority optimization in a moment that has interesting results and helps Intel out a bit within plus or minus 2% of the sixteen point six seven millisecond the delivery window for frames or 60fps if you prefer the 7700 K delivers 63 percent of its frames at around the 60 FPS delivery cadence while the r7 1700 manages to deliver 85 point five percent of its frames during the 60fps cadence we want this number to be as high as possible and we also want the next two numbers frames above and below 16.7 milliseconds to be as low as possible and keep in mind that this is only looking at the frames that actually get encoded to be delivered so this does not count the dropped frames for example which were never they just never made it this far even though frames faster than 16.7 milliseconds will be well faster it doesn't mean that the experience will be better we want consistency at 60fps not frames delivered at varying intervals which leads to choppiness appending that eight milliseconds followed by a frame at 20 or 24 is going to look choppy end of story the 7700 K delivers about 16% of its frames faster than sixteen point seven milliseconds would be 1700 delivering six point six percent of its frames faster than this window as well we're at twenty point six versus eight percent for the frame times above sixteen point seven knows I can but ideally were just as close to sixteen point seven as possible as shown on the overclocking figures the 7700 K Clause back some ground and drop frames by going to four point nine gigahertz but not enough to really even out the battle drop into six megabits per second stream and will lighten the load on the CPU so we're doing that next at six megabits per second with 7700 K jobs twenty six point four percent of its frames rather than in the 40s before and it is now markedly improved that the six megabits per second bit rate rather than ten the r7 1700 is now dropping zero percent of its frames with frames in our sixteen point seven millisecond window improves a bit as well bringing the 70 to 100 K to 64 point four percent and AMD is delivering about the same as the 10 megabit per second test still on the lead though its lead in drop frames is significant and the most notable the 7700 K is delivering more frames to the player as we'll show in a moment in most instances but that doesn't mean the experience is good for the viewer as we've just clearly seen we can improve it though and we did so it's priority optimisation for OBS on the 77 hard K but this doesn't fix everything by changing process priority to high on the 7700 K s stock we drop down to effectively 0% dropped frames that is an insane change from what we were just looking at a moment ago this results in all rendered frames being delivered to the stream but fewer frames are rendered total so it's a trade chenna hard k is now masked with the r7 1700 technically for streaming performance insofar as ability to deliver rendered frames to the stream viewer and it seemed that the fix was easy except there's still FPS on the other side of the argument remember just because we deliver 100% of the frame so the streaming service doesn't mean that the frame rate is high or that the frame latency is low all it means is that we've delivered all the frames however many that may be listen provement occurs because we've taken resources from rendering the game and have given them to rendering it with OBS for encoding on the streaming side so we're not adding resources anywhere that's not possible what we are doing is reassigning them this means that there must be lost elsewhere in the system or the experience it's just a matter of how relevant or visible that loss is looking at FPS to find out how relevant the losses are our baseline FPS performance shows how each of the CPUs performs without any active streaming or streaming playback the stock 7700 K manages 114 FPS average 90 FPS 1% lows and 82 fps 0% lows for fluid performance across the board overclocking doesn't get us much more in this case the stock 1700 operates a 108 FPS average for this particular workload with lows at 88 and 76 remember this isn't a number we actually care about today we're not comparing these two numbers is just for perspective when we're comparing performance loss on the client side when streaming during the 10 megabit per second stream the 7700 K drops to 104 FPS average width was at 77 1 % and 30s your 100 cent we've traditionally seen a big hit to the 0.1% fame time metrics resultant of streaming so this follows that trend the performance loss from baseline is 9% Intel to Intel with streaming vs. not and the client side streaming performance is consistent for the player with no seriously jarring the hitches despite what the viewers are seen in the stock configuration and the experience is a 16 percent performance loss from baseline to 10 megabits per second streaming but again is able to sustain a better stream output for the viewer as for the 77 100k with high process priority on OBS we drop down to 64 FPS average with 1% low frame time is computing the 17 and 0.1% 210 1 0 the experience is a stuttery mess and dismal all around despite all friends technically being encoded and alerted by OBS this was where that number is misleading there just aren't that many frames to work with so the job isn't that hard and delivering 100% of frames really isn't a big deal when you have ten for your low-end frame rate six megabit per second numbers will be in the article if those interest you they're largely the same those spoiler performance on the streamer side is relatively good on both processors overall let's now look at power consumption during the 10 megabit per second dirt stream we're measuring power consumption by clamping the 12-volt wires going to the EPS 12 volt CPU cables and then multiplying the current clamp rating by twelve point three volts which is the output of our power supply down those cables the 70 to 100 K plots first with a very stable and consistent 66 watt draw from the EPS 12-volt cables and when we overclock to 7700 K which did improve streaming performance output somewhat significantly in cases if not enough to compete anyway the overclocked version results in about 89 watts drawn at the PSU cables and plotting the r7 1,700 power consumption last we see power consumption that averages out to about 64 to 69 watts with peak consumption at sixty eight point nine that's having 700 K and 1700 to our drawing effectively equal amounts of power in these tests though the 1700 response seems more varied next game here now looking at dota 2 and I side-by-side comparison between the 7700 K and 1700 which we've left unnamed once again dota 2 doesn't run into any limitations other than the CPU for this test given its lightweight GPU workload this was conducted while playing back a match with team complexity during the 2017 invitational so each test is identical to the previous and is easily replicable we start this test around the 15 minute mark of the playback let's reveal which CPU is which the unveil is probably somewhat expected after the results from dirt but on to the numbers next start with streaming the output performance at 10 megabits per second we observe about fifty nine point nine percent of dropped frames on the 7700 K with the overclocked version dropping 54 percent of its frames for a slight improvement of the frames that the 7700 K does successfully deliver without dropping 85 percent are dispatched within the two percent window of sixteen point seven milliseconds and the r7 1700 manages to deliver 95% of its encoded frames within 2% of its 16.7 millisecond window remember this metric isn't counting drop frames since those frames are never made any way they're never encoded so the number it looks a bit better than perhaps it is in reality to the viewer dropping the 6 megabits per second improves performance across the board as expected on this next chart for Intel anyway and these performance really couldn't get much better than it was already the 70 100k now drops just 47% of its frame stock 38% overclocked with 94% of the total delivered frames at landing within 2% of the 16.7 millisecond window that's good for consistency but the actual delivery isn't really that great and can't really make use of it and these are 7 1,700 didn't drop any frames in this test and delivered 98% within these 16.7 millisecond window looking good giving the OBS a processed priority of high however and ranking it above dota for the performance priority in Windows we began resolved the issue of dropping frames and are now delivering all frames to the streaming service theoretically this smooth things out but not of FPS tanks let's take a look at that next we're going to have the FPS testing with dota 2 we found our baseline performance to be about 171 FPS average of the 7700 K what was that 89 and 47 0.1% dota has some issues here as it always has these 17 100's baseline performance is 167 average again and this is important the point here isn't to compare the two against each other right now it's establishing a baseline for comparison against the same product while streaming while streaming at 10 megabits per second the 7700 K manages a 118 FPS average for a 31% reduction in baseline average FPS performance or not streaming the r7 1700 manages 73 FPS average for a 56% reduction from its base line the 7700 K delivers a higher frame rate and overall better experience to the client-side streamer in this instance while the r7 1700 delivers a better experience for the viewer different sides of the same coin the difference is that the 1700 manages a still adequate frame rate for both the streamer despite being lower overall while also offering a significantly better frame rate for the viewers this is the balancing act and an hour tests and he appears to be doing that better our test with process priority on OBS yields an average FPS of 92 still better than a.m. these are 717 100 streaming average fps and our lows are at 34 and 21 overall results are worse than these stocks 7700 K tests with normal process priorities but better than these results it's a hell of a fight for Intel to get this far but in this lighter weight game that's more frequency intensive the 7700 K manages to pull it off I gracefully mind you and you have to do some tuning but it's possible moving on to power consumption at 1080p 60 with a 10 megabit per second up rates the 7700 case sustains a flat consistent power consumption of 60 3.96 watts at the EPS trouble rails with a plus or minus 2% error from our device identical to last time really or close to it 1700 is bouncing around near 62 63 watts of power consumption with minimum at 52 and maximum at 65 B 17 100 averages a couple watts of lower power draw overall versus the studies of 100 K there's no major difference between them this is where all those caveats come into play that we mentioned earlier if you wanted to stream on the 7700 K you could there are just a lot of things you have to do to make it happen and it's really going to depend on the games you're playing for one you could change the quality where we did faster h.264 you could drop to something like very fast or ultra fast which will increase your encoding speed at the cost of the output quality to the viewer on the stream there is some level placebo effects once you get down into the slower settings mind you so there's room to do this tuning if you wanted to you could also set process priority for OBS which clearly picked up the streaming performance but did hurt our gaming performance so depending on what game it is you may or may not have room there to do that if you're already struggling to run the game without process priority given through OBS you're going to struggle even worse when it OBS is given priority and ultimately if the game just can't sustain a good framerate to begin with it's not going to look good on the stream either even if it delivers a hundred percent of the frames if you're delivering 20 frames a second it's still going to be 20 frames a second despite how efficient or consistent that delivery is a next item of note is that you can keep in mind the more intensive games will have more difficulty coping with the process the prioritization when given to obs as resources get assigned to OBS rather than the game we're not creating more resources we're just moving them around and you have a finite amount of resources on the CPU side to give to the game and the streaming service and in this case all to the playback service which is the lightest workload of all finally you could lower the bitrate where we did ten and six you can go down to something like 3,500 kilobits per second or 4,000 basically four megabits per second below that would certainly help out but again you're doing so at the cost of quality for the output product just depends on how important it is to you to have a higher quality stream if you're doing this professionally it's probably still worth looking into a dedicated streaming box that'll make your life a lot easier going forward but it is potentially a more expensive solution if not more expensive depending on the market of RAM and all that stuff right now in the very least it is more space consuming so different options for different folks but right now as it stands in our test today they're rallying dota 2 being the only two examples we have currently Intel is getting crushed in dirt rally it's kind of doing okay and dota if you do all the tuning but again you're entering into arguments where yes it can compete but you have to work for it you have to lower your quality potentially and AMD is doing it a bit easier out of the box which is kind of an interesting flip considering and the out of the box when Rison shift and up until really just recently had a whole lot of issues it was a like we said about Vega it was a project car you bought it because you were okay with the idea of playing on settings you're okay with the idea of BIOS updates of Windows updates of tuning things and trying to get things to work of finding stability for memory timings those things we're hopefully okay to you as a project but now looking at streaming Intel is the one that requires the tuning Intel has become the project car so depending on what you're doing there are different options for different people and right now we have to go with and the Rison 7 CPUs dollar-for-dollar vs. Intel so the 7700 K versus the 1700 for streaming that probably extends down to the r5 line as well we haven't tested it yet but one could reasonably assume that an r5 CPU versus an i5 CPU is probably going to crush an i-5 potentially more so than we've seen here but we haven't tested it yet it's just that looking at the numbers we have core counts we have thread counts we have the processor speeds it's pretty clear how it would perform and the i-5 is don't hold on as well as the i7s do because they've got a frequency deficit and a major thread deficit to the i7s and of course the r5 is by extension so if you're streaming and you don't want to use NV encoder and you don't want to use a capture machine look at Rison for the easier to work with option if you are using one of those things doesn't matter so much pick whatever is going to play your game at the frame that you want and then hook it into NV encoder at the risk of the other new issues you have to run into with that side of things or hook it into a capture box and really resolve pretty much everything quality wise but introduce a whole bunch of new headaches for you in terms of money spent space consumed and learning how to use the tools there's no perfect solution here but AMD is doing well to compete in a market where previously Intel had no competition to the point that Intel CPUs kind of looked ok but today it's different so thank you for watching as well you can go to patreon.com/scishow and access to help support this type of testing it does take quite a bit of time to figure out we have a lot more we want to do with streaming i'm curious to hear what you use for your settings when you stream we've already asked this question a few times now it doesn't hurt to tell us below what settings you use what quality you try to target when outputting maybe we'll consider it for inclusion in our next round of test games of course are also interesting to us if there's a particular game you want to test it maybe overwatch let us know below subscribe for more as always again gamers Nexus dot squarespace.com for this brand new shirt design - actually a teardown of the GN logo it's got a bunch of Easter eggs in there like prm's PCIe slot stuff like that cool style thank you for watching I'll see you all next time a lot of framerate z-type latency