Gadgetory

everyone we are doing a new type of series today not quite sure what it'll be called yet or how frequent it will be but i'm joined by patrick lathan senior editor and he's been helping with all the rise in passing another testing where he's hard not test average for the last week or two issues we've run into solutions for the issues things like that before getting to that this coverage is brought to you by the ten ATSC which has a new MSRP that is it lowered with the launch of the 1080i series at 1080 and 1060 SD cards come with for honor or Ghost Recon wildlands which you can choose that check out learn more at the link in the description below okay so let's start with closing the GTA 5 story line yeah so we've talked a little bit about our issues with GTA 5 in the past I think one of the largest issues that we had was visible stuttering with I five CPUs specifically to high-end modern 7600 k that's I eyebrows and it was affecting test results by really skewing the point one percent lows and yes same times what's really bad via and we think that I mean we we discussed this in a video earlier but we decided that it was probably an engine issue not an engine not an issue with our testing but something that would maybe King gtf up all of our bench right where I fives at least in the future so we did a little more testing with that there were some helpful comments about viewers that had done their own testing and had seen maybe a frame rate cap around 180 7.5 I always say a rage engine yeah yeah so that's actually what we saw as well so yeah yes so basically the story was i five CPUs looked really bad in GTA 5 when they were fast enough so the faster the CPU was the worse the frame times looked which just doesn't make any sense the 7700 k didn't have this issue so the 7700 k with hyper-threading on never had a frame time above 16 milliseconds I was actually just talking to Wendell about this my one taxon on Twitter so we never saw anything over 60 milliseconds you go to the i5 or you disable hyper-threading on the i7 and as long as their new ish you run into the issue where there would be really hard stuttering to the point that it kind of would ruin the experience but you need the height of faster ones because if it's not fast enough to hit the 187 or whatever frame cap we weren't seeing the stuttering and this is we saw a little bit of this issue with the new horizon CPUs not as much obviously to have a much higher core count than rid of four core i5 but the save wayne smt yeah so we've continued to have the issue and we're going to resolve it by not including the tests that have erroneous results in them yeah or yeah and so like that was an interesting thing we tried back when we did the original GTA 5 5 testing mystery video we talked about trying to get the FX 82-73 produce the results by disabling course but it just it's just not fast enough yeah we couldn't get it to be fastened I think even with the lowest settings so pretty interesting stuff I think basically the conclusion is something of the the storyline if you haven't seen the first video the conclusion is it's an engine constraint where the games hitting high really high FPS like 187 and throttles back performance in stuttering at that point on for thread or similar thread count devices I don't remember I don't recall if top my head if we saw the stuttering with cr5 1600 x we did it wasn't nearly as bad I think we might have seen one or two yes but yeah that's a sixth Red Sea fee or not for which tells us that it may be related to thread count below eight but yes are interesting stuff basically the challenge you face the philosophical question you face is this game as a benchmark works well until you get the four threads so as a benchmark if we're trying to get really high frame rate so that we're CPU constrained then it seems fine until you throw an i5 or an smt off our fiber are seven on there and then it starts stuttering but the question is you know is that a real enough use case you know where in GTA 5 will you be playing to a point where you've got that high of a frame rate and you're stuttering on an i5 CPU versus the solution is to just turn up your graphic settings which is really weird or I think the solution that we see offered online a lot if anybody has been having anything close to this issue the solution that people immediately offer is turn vsync on which would REME really solve that video yeah fixes so yes it was a workaround if you have this stuttering problem on a 4th Red CPU then you can solve it by turning up your resolution or your advanced graphics which we didn't do a whole lot with either one of those as soon as you bring that PS down below like 165 or 180 or whatever it is then all the problems go away that's still a high frame rate so interesting weird thing I did not include GTA 5 in the latest r5 benchmarks and I've been slowly phasing it out for the i5 VP use because it's just it's it's showing enough issues where we just need to replace it with different game me which we've kind of done with total war and that brings it to the next us to the next topic which is seeing the total war warhammer uplift and that was a one-off lift with rising TVs initially which we talked about in the rise and revisit yeah and then retested it on until later yeah so more software issues that kind of in the opposite direction this time there's been a lot of reporting on risin improvements in various games but specifically there's been a lot of attention on battlefield 1 and 2 war war hammer and especially the total war war hammer we've noticed significant improvements and we have confirmed I believe we've confirmed that was due to a an update to the game primarily yeah the windows update so there were there are some stories online about a Windows Update magically fixing all kinds of the albums with risin that wasn't really the case and our testing what actually happened is the windows update build a 970 coincided with updates to the games that posted improvements yeah so if you up get updated both those things at the same time it looks like it's the windows update but in reality was a game update so total war in battlefield both had updates about with one to one point oh eight with the version name or whatever it was something like they shall not pass I think some Gandalf name and the total war added some things and also with somewhat silently improves performance to the point where and this is just recapping the rise and revisit stuff but very quickly basically if you disabled smt previously in the old version of the game now the new version of the game gets your r 7 r r 5 performance up to the levels of where smt off was Mia originally yeah so now disabling smt still gives you performance up lip but it's a couple percent whereas before it was like massive is like 30 frames and really this is exactly the kind of thing that we were hoping to see specifically from total war korean us and we said they were working on an update working with AMD to optimize for Eisen and it seems like they did we did notice improvements across the board we retested the 2600 k and a couple of their Intel CPUs ebony 600 k in the 7700 kmo 7500 yeah we're already tested yeah we noticed improvements with all those so it seems like maybe total wars making better use of a higher thread count see if you just in general yeah yeah so for our r 5 review the 7770 600 k excuse and the saudi 500 we're all retested and then the sub mature that didn't have the asterisk hadn't been retested that means there's actually percentage overhead there we could improve or headroom i should say so we updated for the review if there are reviews where they have not retested intel but have test today and the after the patch then there's some headroom there for intel to improve as well and you can actually look at our old numbers or is the new as for the those CPUs and you'll see the bedroom so that was interesting battlefield Juan had a very slight uplifts not nearly as exciting but there was a flip and then a couple notes I guess maybe we should talk about some of the mem testing stuff so we did in the same rise and revisit and in the archive reviews for our five we retested everything so the 1800 x the 79 or 1700 x to some extent were retested and that included moving to thirty two hundred megahertz memory thanks to efi updates that permit the memory speed increase more easily yeah so 3200 wasn't hard right you didn't have to do any special not hard at all we had initially been having a lot of issues with higher frequency memory I think everybody was having issues with a higher frequency memory on but with the efi updates and whatever other updates that have just like Jesse as I have to hate so I selectively choosing the memory to idea using samsung be die only yeah no hi next stuff yeah that so we we had been using the only approved Benjen memory yeah beforehand and we'd only gotten up to the 29 and 33 but we got up to 3,200 no problem that is the last that's the highest frequency you can set it to without adjusting the be relocating yeah yeah on our a cross-border they were using so AMD sent us a memory kid that worked out of the box I think the other thirty two hundred megahertz kids so we had all worked straight out of the box so we wanted to see if we could push the memory a little bit higher we did do that and at least one of the tests that we published we got it to 34 in 66 yeah 3456 I think that was at CL 16 I think so and we did 3200 at a in some of our original testing we were doing a CL 14 for some stuff and their hands guy offers for the 3200 kid later on yeah so yeah we had 3466 with a g.skill trident ii kit and the I think that may have been I'd have to check our own content horizon revisit talk about whatever the classic lanes he was and then other than that we had pushed to 3,600 finally I think me with the guile kit maybe not what we go okay they might have energy skill we've done a lot of testing yeah there was a realize a test in there the 3600 results haven't had published yet because they were like dick quasi stable yeah they aren't quite stable there was definitely a performance uplift from 3200 to 3600 not a huge one over 3466 but we were kind to be able to hit 33 on it it's just the stability gets really shaky there so there's a couple interesting things one it is definitely worth trying to get the higher speed memory if you can that's a two-factor thing they're one price cost quite a bit ddr4 prices are not great right now oh yeah mn 2 effort so 3200 as we've said effort really is not bad now but if you want to go beyond that then there can be quite a bit of effort involved in memory training which is basically stepping through every single frequency leading up for what do you want you can't go from 2400 to 3600 and it just works so like you went through that process the most yeah I'm not a professional overclocker I don't do memory overclocking very much or at all before they pretty much but I spent a lot of time adjusting frequency timings voltages trying to stay pretty conservative with the voltages I think we're like 1.45 at the most yeah with the memory yeah and uh 3466 was a highest frequency that we ended up publishing so yeah I wouldn't speak with confidence about any frequency higher than that yeah with our setup yeah it is possible but it's just an effort thing so we were able to you 3,600 it we'll have to see if we can get a more stable with updates allow it to more stable the referencing build Zoid a whole lot throughout that whole process so thank you Bill died for these professional and overclockers inside we worked with asu's as well to figure out kind of what tuning we could do on the crosshair 6 so you have memory testing a huge benefit for remember you ever call you in for risin if you can get it to work and then you can afford the memory so there's a bit of a hidden costs there we're recommending the r5 fairly heavily in the review of it over comparable high five chips but it is somewhat contends that there's there's a bit of a contingency on memory speed so just make sure 2933 and beyond you're not in bad shape 3200 ideally just you know there may be a bit of a sacrificing capacity in order to get the better speeds which if you're not using the capacity beyond 8 gigabytes which probably won't for gaming it's a worthwhile sacrifice right now and we still kind of have to test the the for stick stuff and rise and we will get on tier sticks so far but yes that's the memory stuff very quickly I guess we can mention we have plans to update the CPU benchmarking methods yeah so um as mentioned we've been having some issues with GTA there were issues with battlefield that I think we've discussed yeah metro last light is aging it it's great to have old games I'm always going to have an old game and our benchmarks because people ask like why don't you use new games all we do but we just keep one or two old hunt them there so the one or two old ones are in there because they don't get updated which is really nice that's a great way to keep 16 or 20 parts on the list so that we have at least some level of relative performance over the past so we will keep something on there that's older total war is probably stained as we're still updating yeah they're just like pork two of a trilogy coming out soon oh yeah which makes it even more relevant GTA 5 on it is probably going to get cut but we'll make some changes I've got some ideas for synthetic testing improvements as well there's I don't know I don't wanna make any promises you have there a couple of sweets out there I look really good for testing in areas we don't currently test and basically CP tasking will be updating now what that means the the critical part here is whenever we update an entire test suite like we just did for the GPU stuff all the data goes away like uh which is very sad unfortunate so we lose a huge historical set of data it means that potentially data going forward is not comparable to the old stuff we'll do our best to retain a lot of that total war should be retain about any other games we keep should be retained a little bit so we will still have some historical data there it's just in the new games like ashes escalation it's going to be a cold chart you're going to lose a lot of data and that's just the nature of it so yes cpu testings getting an update and we'll look more into that probably after a couple more revisit pieces I'm not going to name the chips yet but we have some visits in the works those will likely be the last things tested with the current cpu bench alongside maybe there are five 1400 or whatever you yeah there are three tips by the time those come out and whatever Intel is working on we're probably gonna have a new set of tests to run now so that will give us some time to breathe from risin I last thing here there's other testing things but more importantly I want importantly we have this coming up so i'll let you intro this I don't know how much you want me to spoil here which name is good okay well this is a cooler boss cpu cooler I don't really want to reveal much more than that but I can't let me share the window I'll just share that this is a window click into it and neither of us can read eighty percent of the text on the box and the instructions manual is not in English I think they're swimming this tables it there's a little bit of English so we could probably leave that the hair keep an eye out for this one it's uh I mean it's I think it was a thirty-dollar cooler so you know it's right up there with the hyper 212 one of the most popular coolers of all time the thermaltake contact 12 the silent one that's just come out this is in the same price bracket and we discovered it sir great to see if we can blow the socks off of the cooler industry in the sub $35 market let's look at like like what like a month to delivers about five or six yeah yeah that should give you enough insight escameca yeah thanks for watching if you are interested in this type of thing basically the idea is at some interval we're going to get together and talk about all of the testing issues challenges and other updates that we've faced and walk through some of the behind-the-scenes stuff basically what this is you're interested in it let us know thank you for watching as always subscribe for more you can go to patreon.com/scishow help fund the insane amount of hours put into all this stuff including the inevitable test bench update which is very very sad yeah shut it here for data stored our cameras Nexus dot net for shirts and we'll see you all next time you you