Gadgetory

today we're going to talk about bottle necking yo J did you find our six in white what do you think fine our six in white okay okay it's an arsenic but it's not new in fact it's not even all white I mean what's with the black front really coming from the guy with the black and white car black hood like okay that's different it's the same that's not the same exact same thing not the same fine change it that how with the custom wrap from slick wraps front panel whole case whatever you want even custom designs us that's right we did do that video about the custom wrapped our six choose any of their existing designs or customize your own upload it and they'll make it for you see wow you really can make these custom with wraps today we're gonna talk specifically about bottle necking a topic we haven't really covered on this channel in quite a long time in my opinion and my reason behind that not being as much of an issue as it used to be as the competition that's taking place in the CPU space since AMD launched horizon has caused CPU manufacturers namely Intel to really have to catch up in terms of what AMD is offering we're talking about overclockable CPUs more cores more threads games I started becoming more multi-threading friendly which means that the CPU itself has been less of an issue as it was back in like the early 2010's when FX was still the deal for core 8 thread was the max you could get on main stream but we're going to do today is we're going to go ahead and show you kind of an extreme example of what bottlenecking really looks like and then we're gonna kind of like tune it out basically to kind of show you where that sweet spot really is now you can actually try this at home with your own CPU if you guys want to if you're running an Intel based CPU anyway I don't know if AMD has as useful of a tool but this is actually the intel extreme tuning utility or XTU you can download this it's free and if you have an overclocking CPU it'll let you do anything you want to it pretty much and then if you don't I don't know what settings will allow you to do I think you can play with the ratio and bring it down I just don't think you can overclock but anyway this is an 8700 K that what we've done so far as you can see we have two active cores because we turned it into basically a 70 100 now yes I know the cache values are different yes I know the cache frequency and all that stuff is different this is not intended to be like here's how performance is on a 70 100 on an 80 183 but that's not what we're doing today we're showing you examples what bottlenecking looks like in a real case scenario with a real game and stuff so what we're doing is we're obviously starting on the very opposite end of the spectrum as bottlenecked as we think we can get to show you what it looks like in its definitive form with a twenty atti on our test bench so obviously you should never pair a 7100 to a twenty atti that's guaranteed to bottleneck which is exactly what we're doing here today so this is just kind of our extreme test right here so we're also going to have MSI Afterburner open we are gonna be running a slight overclock on our graphics card because again we're trying to get these two as far away from each other as possible and then we're gonna be actually using shadow of the Tomb Raider because it has really good tools built into their benchmark to show us things like frame timing CPU timing lots of different it even show us what percentage of GPU bound we are so obviously you see CPU one two three four here because again two cores four threads because we have hyper threading enabled we also have our settings that set to highest right now which is kind of unfair for this test a little bit because of the fact that if we really wanted to see a CPU bound test where we're bottle necking we'd want to have less post-processing filter it's going and higher higher frame rates which would push our CPU even harder but for the sake of this test we are just leaving everything maxed we can see right now with the test running look bouncing between 95 and a hundred percent utilization on the CPU and because of that our GPU utilization is extremely low even in 1080p our fps is only sitting in the 50s with a twenty atti you can see we hit a hundred on all of them for a second there but our GPU is spending a lot of time just waiting for the CPU and that's indicative by these high spikes where suddenly the FPS will climb a little bit and then come back down but you can kind of see a median line sort of drawn through the middle we're starting to you know dip right there down to the forties obviously as we dip so does our frames are our GPU utilization all right so as you can see right here it's not fully rendered because watch this got some flickering there and then we're gonna see a big pop or the mountain come in there it is there's the rest of the scene so that's an example of one thing that happens with an under spec CPU versus the games minimum specs which is clearly not a good thing seeing that popping in like that but once again look our GPU till it's all the way down to the 30s our core clock has even suffered because obviously it's not gonna ramp up the core clock if it's not even being utilized so this scene is only being rendered with thirteen hundred and fifty megahertz with the GPU because the CPU is absolutely 100% pegged so the reason why we use this particular benchmark for this video is because you can see we get a lot of information including a graph so this is our GPU graph for a GPU CPU relative graph here so as you can see the CPU is taken much longer to do things in the GPU is but if this orange line was above the green line this could be an instance of where you have a very CPU with a very slow GPU where the GPU is then with the CPUs waiting on to be able to get the next scene and then next frame ready so we can see right here a GPU bound 0% and that's a that's kind of a problem so we obviously saw too that our CPU was pretty much maxed out during that test because we're going to do now I'm gonna change the CPU now to from theoretically a 70 100 to an 80 100 which means now we're going to get rid of hyper-threading we're going to enable two more course we have 4 physical cores but instead of having it running at 3.9 which is the maxed speed for these chips to remember they have no - they have no turbo clock on those models we're gonna reduce it down to 3.6 because that's what the 8100 is all right so test 2 we have it running at 3600 megahertz or 3.6 gigahertz with 4 active cores no hyper-threading and our cash ratio is set to match so we're gonna run the exact same test all the same settings as you can see here the graphics are all set to highest same sort of thing in the menu though but you'll notice instead of pegging up in the 90s we go down to the 70s for a little bit so we're gonna look at initially when we start this test is what happens on the initial utilization of the graphics card as well as the CPU alright so in the initial tests we are sitting in the 90s obviously still 1/4 hit at 100 that hundreds gonna bounce around but our utilization came up significantly remember it was sitting in the 50s initially and then dropped down to the 30s well now we hit the 60s and our drop down into the 40s in terms of utilization on the GPU still seeing a hundred percent though on the core is bouncing around so obviously we are still seeing the CPU cause a bottleneck to our GPU because we want this number to be at least 85 to 90 percent or higher if we see this go up into the 90s and we know the GPU is now what we're bound by because the GPU is being utilized to its you know its full capacity or full potential but if this number is sending this low that is what a bottleneck is and that's what we're trying to avoid so by doing nothing more than switching from hyper threading to for physical cores emulating an i7 80 or a 380 101 i7 8100 would be that would be interesting CPU but I digress we've gone up from 72 to 88 and as you can see right here GPU bound is still 0% but we're get close but what you're gonna notice is both charts as a whole moved down because of the time it took because these are measured in milliseconds this is the time for it to do the task and as you can see everything just slid down on that chart so that's obviously an improvement but we still are clearly gonna be bottlenecks although in the menu we went from 7d ish 78% to like 93 which is crazy because the menu does indeed render the game in the background so we're gonna do now is I want to see what happens now if we just give it a crap-ton of course feed so we're gonna overclock this to 5.2 gigahertz and we're gonna see what happens to that fps by having the same core count the same cache the same number of threads obviously and the same CPU settings the same or the same GPU settings we still have the same overclock going on our 28 ETI and see how much more we gained by overclocking so here's what we did we overclocked it to 5.2 gigahertz with a 50 ratio on the cache that's not something that exists it's not something that you can even try to do with your own 8100 because remember it's a locked CPU so that's why we use CPUs like the 8700 K and then and 100 K to do our emulation or emulating of other CPUs because we can take it and dial it to where we want to see what theoretical you know performance is like at different frequencies that you couldn't otherwise do by just popping in an 80 100 so although the cache amount is a little bit different I don't believe that is enough difference between the CPUs to invalidate our test like I said the start though this is all theoretical to show you what bottlenecking looks like it's not meant to be an exact representation of the CPU that we're trying to emulate so it dropped down into the 50s and 60s at the start here did start at about 70 went up to 81 and then came down as this test is happening but look we're seeing a lot less of the 100 pegged there's one it's still pretty high obviously and still clearly bottlenecked even running at 5.2 gigahertz on this one test you can tell to the DP look how cool it's running check this out so we went to GPU bound 2% which is right down here where I said it was gonna potentially happen again everything moved down because the max chart before was 30 milliseconds now it's 20 so you can see things are getting faster specifically because of the CPU but our we went up from 88 to 105 on our average frames per second and then you can compare obviously down here on the bottom with the min max that was but then it really did improve obviously but we are still clearly overall bottlenecking our GPU so the issue here is that if you were running something like a twenty atti and put it on a cpu like this then there's a whole lot of performance that you paid for that you're not getting so we're gonna do now is we are going to go ahead and unlock the cpu and go normal eighty seven hundred K at five gigahertz and we're gonna see theoretically what the performance looks like what the GPU bound looks like with not introducing any sort of a bottleneck at all and what you're gonna see right now is that our CPU utilization is gonna stay fairly low but our GPU should jump up into the 90s oh it dropped down to 82 guess no we're bottlenecking but you know look at our CPU it's not so it's possible to that this engine is just not using an entire twenty atti in terms of utilization but this is still considered a good yuto because our cpu as you can see is not pegging on any of the course one of the reasons why we're not actually doing like a 1440p or 4k test is because i if we're talking about bottom lucky and those that are running low-end CPUs I think it's extremely rare that you're gonna find someone running a 70 100 or an 80 100 with a 4k panel possibly a 1440 but if you guys want us to do another video where we test very similar type deals and see how resolution affects various CPU speeds then make sure you guys comment down below that you want us to make that video and we'll do it everybody I give this one I like to and our average went up to 121 FPS so we gained a wife a 105 to 121 by unlocking you know obviously all of our cores so I bet you anything if we ran a ninety nine hundred K or another CPU with more threads we'd probably see even higher FPS with this same GPU so here's we're gonna do now we're gonna pop in at 1660 which is a much more mid-range graphics card with the same low-end CPUs and we're gonna see exactly how it compared we're gonna spare you guys going through all the tests we're just gonna show you all the charts at the end all right here we go 99 percent GPU bound with the 1660 on the fully unlocked 8700 k overclocked but you can tell that were GPU bound because if you look right here in our frame rendering times you can see that our GPU is higher than the cp1 everything so the cpu was like hey here's your frame do what you need to do to it and we're ready for the next one come on are you ready for us so the cpu was literally waiting on the GPU which is 100% non bottlenecks in this case so as you saw the twenty atti is obviously a bad pairing for low-end CPUs because there's a lot of performance you'd be paying for in here that you're not necessarily getting at least with this one title I mean there's there's so many different variations that we could have done we could have chosen any CPU where to be rise in or Intel any particular game title and the results would be different because it's impossible to have any test scenario account for all the different you know combinations that there would be out there so this is just an example of what bottlenecking sort of looks like but obviously with the hypothetical CPUs we had setup there was a lot of fluctuation in fps swinging all over the place with the 28 et I give me a terrible 0.1 and 1% lows which is why you see a lot of reviewers use those figures because average is only one piece of the puzzle are and then obviously if point 1 and 1% lows are really low versus your average then you're gonna see huge stutters which we were seeing with this that's why with the 1660 we saw a very very tight line across the board even all the way down to the 7 100 and I was surprised that we were only 63% GPU bound with the 1660 I thought it was gonna be a lot higher which goes to show that the CPU race over the last several years has definitely made a CPU experience for gamers a whole lot better in terms of bottlenecking because you have to try pretty hard that's because of the 5 gigahertz becoming much more common the thread count on CPUs especially on mainstream and the lower end stuff has definitely gotten more inclusive to include you know 4 6 8 cores plus hyper threading or SMT depending on which platform you're running you're definitely benefitting from that race one other thing too worth pointing out though too is if you were running a 60 Hertz panel with vsync or g-sync on like this is a g-sync 60 hertz panel if we had g-sync turned on with an vsync we would have gotten exactly the same performance between both of these cards because the 1660 even with the 70 100 achieved above 60fps on an average so there you go one of the things we're pointing out to though is that these settings were on max and so if we had reduced any of the post-processing quality we would have introduced an even bigger bottleneck but I didn't do that today because I want to just sort of max out the game as a sort of a baseline test and then see how things went from there because I think most people want to shoot for the most I can do that they can and so we just use that as our baseline but it's worth noting if we had gone to like a medium setting we would have seen even a lower GPU utilization and even in higher CP bottleneck but you guys wants to be more tests like this in the future please sound off in the comments below this is a direct response to some of the comments I saw on our our budget pcs actually worth it video and this is that response thanks for watching guys and as always we'll see you in the next one we realized that face swapping Brian's face on to my youngest daughter looks like Steven Burke from gamers and exes twenty hundred ti twenty twenty hundred twenty hundred twenty hundred it videos next big thing right there but how many more rains were friended though rains were friend or