Gadgetory

so there's been word going around of a 4.5 6 gigahertz rising overclock for the 2200 or 2400 G we bought those ap use and I wanted to show you why that's not actually a thing so we can all move along and everyone understands where the numbers are coming from this is a timer bug that's been around since Rison came out it was patched out for most of the rise in CPUs at this point but some of the ApS are exhibiting the bug again so it's time to refresh everyone and I mean on the screen right now it says we're at 43 50 megahertz or something like that for the core speed I've configured this to have a multiplier of 39 I've not changed the bus speed which it should be at 100 but it says 111 and somehow even though I've told to go to 3900 megahertz we're showing 40 300 megahertz so I will demonstrate that this is the sleep bug it will actually report higher scores and everything but it's not really all it's made out to be before that this video is brought to you by Thermaltake and the view 71 enclosure the view 71 is a full tower case that's capable of fitting 3 video cards and most configurations it's also one of the better cooling cases in our recent case testing bench lineup the view 71 has hinged a tempered glass doors on either side that make it easy to open and show off and it comes with at least one rain fan though you can get the RGB version if you prefer learn more at the link in the description below so a this works is really simple basically just before the camera started rolling I put the system to sleep with the windows s3 function brought it back and we see a 300 megahertz increase in score and if I run the benchmark here it will actually score higher than it did previously so we'll just let this finish for a second okay so according to Windows that took about a minute to finish and we got a score of 643 CB marks don't take that score too far this isn't like hey this is what the score is because I haven't even installed all the drivers and everything all that matters is relative scoring against itself right now so talk about a minute in Windows that's not real time that's the root cause of this bug basically for every one minute in Windows minute and seven seconds is passing in real life and this benchmark Cinebench as with many others I think blender is included in that those tools that use a start versus end time for scoring are scoring versus the windows time that's reported and the windows time that's reported is incorrect because if you looked at the base clock for example that was showing 111 megahertz a reality that's 100 megahertz so somewhere in here Windows is reporting the incorrect time it's actually reporting time slower than is transpiring in real life and because the benchmark is completing in the exact same amount of time say about a minute and seven seconds what's happening is it thinks less time is passing because time is passing more slowly in Windows but in fact same amount of times passing however it's giving us a higher score and again I'll just bring this up so at 4300 megahertz reported in cpu-z and now I'm gonna do is just restart it restarting is different in Windows 10 then shutting down shutting down will give us a hybrid shut down so we're gonna restart that'll clear everything and boot back in okay so we've rebooted completely cold restart which is a thin now and if we open up CPU Z there we go thirty nine hundred megahertz now so now that we are at 3900 megahertz again if we ran that test again we would get a lower score than what we just saw a moment ago even though both tests were actually at 3900 megahertz the timing is different now it's actually proper real-time way so this is very convenient timing and we've ended at 435 434 somewhere in there but either way it's taken more Windows reported time than the first one so 566 for the CPU score this time again don't compare that versus anything else not the point compare it only against the previous one 643 so it looks like there was an improvement after putting the system to sleep but it's not improved all that's happening is you're cheating with the timer bug and software that validates time like 3dmark software would report it as an invalid bench run software that does not validate time Cinebench would report it as perfectly valid and you've just improved it by way of a fake overclock that doesn't exist so I think that pretty much covers it just to kind of prove a point let's get the windows clock up here alongside a cellphone stopwatch and and demonstrate the timing and then we should be done okay so three two one go so what we're doing is timing with the phone real time versus Windows Windows doesn't have access to internet or anything it's not updating actively it's just taking a lawn as as the host to dictate so right now this is pre sleepbug you can see we're at 26 seconds where I are well either way matter what I say it's gonna be a bit ahead so yeah they're tight that's all that really matters we're tied here I would go sleep and wake it back up okay so clearly it went to sleep because everything as still as it was and let's do this one more time just barely good so what should happen is the phone should start pulling away as time ticks on and we'll probably just cut to that or something or speed this up so we've hit a minute here and we're at 54 seconds up there so let it keep going 3 2 1 stop so a minute six point six and versus one minute so we're about six point six seconds over in real time versus what Windows is reporting that's all there is to it it's very simple if you think you got a four point five gigahertz overclock I'm so sorry to ruin your day you didn't perhaps on a future rise in architecture but this is not an architecture that is readily capable of doing four point five six gigahertz on an APU it's fully possible that rise and could get there but right now the only reason that numbers is appearing is because of the timing bug and then the score is aligned with it because they're all based on time so it actually looks like it validates like it looks legitimate but it's not and if you're on 3d mark or some other I think hardware Bob's x265 benchmark also validates against time those would be what to use or you just use a cell phone so yeah that's it that's all for this one we have a lot of really cool content coming out for the 2200 G in the 2400 G that's a bit different from most of the reviews that you've seen thus far we have some some fun feature ideas for it because we did end up buying ours and getting them on the same day as embargo left so because of that we'll save the reviews and just do cool stuff one quick note on this it does depend on the motherboard we were able to reproduce this on MSI boards the Asus crosshair six hero did not have this problem so it seems to be patched out there already and other boards your mileage may vary so if you have the issue just check for BIOS updates for anything that's going to be getting that update later it'll probably come after Chinese New Year and subscribe to catch all of that you go to patreon.com/scishow sexist to helps out directly thank you for watching I'll see you all next time