Gadgetory


All Cool Mind-blowing Gadgets You Love in One Place

Exclusive: Interview w/ Principled Technologies on Intel Testing (9900K)

2018-10-10
how about we're never afford to so yeah we go set up upstairs okay sure I'm Steve good to meet you good to meet you hit agree with you oh that's why we're here we're not cutting any of that interview there are times when I asked Andrew is audio good does the video look okay we're not cutting at all before that this video is brought to you by thermal grizzly and their high-end thermal compounds thermal grizzly makes cryo not paste for high thermal performance and conductivity without being electrically conductive so you don't have to worry about shorting components Cremona is particularly good for replacing stock GPU pastes as cryo not is a non curing compound learn more at the link in the description below everyone so I am here at principal technologies and their local to us and I'm joined by Bill Catchings co-founder at principal technologies we're gonna be talking about the hot topic of the the Intel benchmarks heartburn box posted some content recently we posted a follow-up to his and we just got back from the Intel 9900 K launch event so bill I'm not even I don't know if you want to start or if I should start because I know you I think you should store it unless you ask me a specific question that guy I'm sure so I guess let's let's get the the housekeeping out of the way for viewers then clearly Intel commissioned the testing right and think it's like the first day happens to work close to the top of the report and and we don't have any problem with that we're fine with people commissioning testing because a third-party validation is also what we do so it's it's just a different different field if I review instead of a private test house so Commission testing by Intel do you know I mean when when you first started working on this project did you guys already have a test plan laid out or did Intel work with you on developing a test plan um I actually I'll be honest I'm not exactly sure on that I mean I know Intel came to us and said hey we want to test these particular things on games and so that was exactly how we negotiated the list or whatever I'm not positive on that all right I think you can look at the list of games I don't think there any surprises no not particularly yeah there are 19 games and I believe they included see they included ashes of the singularity common benchmark total war Warhammer common benchmark as well csgo GTA 5 so forth so did you just just so I understand what your role was were you working on were you hands-on with testing or C my role is somebody you're showing up on my doorstep you have to talk about this yeah I'm prepared to do that I didn't want to grab some of my people and put them in front of a camera right and that's why I'm a well on some of the questions have to say I'll get back to you with an answer on that I'm not trying to hide anything but some things I just don't know chance all sort of as you're fairly well has happened fairly suddenly yeah yes yes so okay and so on on testing we had a few a winners to get through some of the technical questions I guess please do and as I said I remember these Natasha will make sure that me write them down and I will get you the responses okay in a few hours then one other thing I probably should say ahead of time is you you're not the only person talking about this and you're getting emails and and all sorts of stuff you just happen to live closer right you're here and we felt the turning you at the door would not work out well for anyone so we're trying to respond a real-time to all this but there are way more of you than there are of us and right so I'm sorry if things were a little delayed but we are really trying to understand and we spoke with Intel as well yesterday actually and at the event in person so Intel was was scrambling to get answers to and they didn't provide us with a statement which we published and it was they've left a lot to be desired the statement said have you seen the statement already from them or I have no idea so the statement says we are deeply appreciative of the work of the reviewer community and expect that over the coming weeks of additional testing they'll continue to show that the 9900 K is the world's best gaming processor principal technologies conducted this testing using systems running in spec figured to show CP performance and has published the configuration used the data is consistent with what we've seen in our labs and we look forward to seeing results from additional third parties it's not really a lot there but but they did provide a statement and they've been aware of this since yesterday so I guess the from the reviewer standpoint there are two concerns one was data accuracy I know you guys obviously disagree with a lot of those premises we can talk about why yeah I mean I don't write in in wit large I mean to be specific yep and then the other point was more on I don't know if it was Intel or principal technologies that decided to publish effectively a review yesterday I believe or I don't know who is responsible for the publication of it but well we could be responsible to the publication right obviously who sends a piece that's the date I guess well you were in an event yesterday so I think you could surmise why don't know why the date was then right we do work unlike just a magazine where they well you think I'm Dargo ass right now you have some you also have some imperatives you don't have me the last person to the party all right talking about this but but obviously they set the date for what yeah and that that was one of the concerns particularly of her colleague Steve from hard ground box was the the publication of functionally review ahead of review embargo undermines all the reviewers and it's commissioned by Intel which is fair but the concern of a Commission report by Intel that's published ahead of review embargo is that then the consumers have only one piece of information to read and it comes basically from Intel and that might be more on them than on PT to be fair because I'm sure they set the date yes I mean I don't want to throw them under the Bunner but I don't have any control of it right so yes so the reviewer relations might be a bit of a faux pas on Intel's part technical questions so some of the concerns we had word I guess let me for the GPU it was a GTX 10 ATT I used do you know which model which like specific units number one the specific model of that and I may be able to get some disaster for you okay sure it should be in the the disclosure information if we must didn't have enough details there them it just says gigabyte which is a brand but there as you know they're right 20 different skis so okay so that would be good to know but do you know if it was the same model for all 10 yes it was okay so if it was like a gigabyte it just pick one like horas came in whatever it was it's very much the emphasis of this was not to be testing whatever the graphics card was this right meant to be a processor test so we don't want that to be the bottleneck there this was looking at the processor yes GPU can still have an impact there obviously yeah and we've seen that so one of the concerns we had was with Pascal architecture which is Nvidia's you know 10 series architecture the Pascal boost clocks under boost 3.0 are largely silicon quality dependent and silicon quality various chip to chip so if you have ten different cards and we've done this to because we do CP reviews and at some point we said hey we have two of these identical cards when we get to benches running and we found out that in any scenario where you exit a CPU constraint even for a second like in gaming scenarios the boost clocks between them dependent on silicon quality can impact test results so we have seen impacts upwards of 5 percent so that was one of our main points of concern as a few if you bounce load to the GPU for a second to a 100 % GP load momentarily to do a bunch of draw calls on complex geometry or something then the GP with the higher boost will perform better so we had pairs of machines here you know come the numbers wouldn't be just based on on one of those obviously it's a totally fair point where all the pieces of hardware identical there there can be some level of variability I can't I don't know whether that 5% is correct or not but certainly there so but we were tested with multiple units and we were doing it across multiple runs so this was not just one run and see what happens I'm going to write really good try to you say did you have you had a sub pairs of machines you have duplicates of each testbench say it's 16 benches or yet 8 benches um I believe right no no no it's so hard being on camera because suddenly things pop out your head should take your time welcome back to all they don't yeah I don't watch I feel like machines pretty easily but there I believed it was should be 16 with so but so definitely eight but potentially duplicates yeah in which case you would have based on the document you would have six pieces of data for each for each configuration as opposed to three because it says three test passes yeah come on I was reading that I would have to look we should have the numbers in there but I any any particular reason that the the PT decided to use the median instead of average yeah I mean I've been doing benchmarks for maybe longer than you've been alive but the the there are advantages to median and mean that the advantage to median is the you're actually quoting a real result that happened whereas mean you know you could be between that right run didn't actually happen and so I think median does better it throws out a real high end result or throws out a real low end result but it also doesn't penalize you for high end earlier I personally prefer medium but I can argue persuasively for either okay that's that's a fair answer I think our concern for median was the point of doing averages is and multiple passes is to account for potential variants run to run because there's I mean we run the same tests a thousand times and you might get a dozen different results even if it's by 0.5 FPS and generally the median works fine in that case where you get in trouble is if you have two maximums you're jumping between two result right and I'm not sure that the mean really helps there either because it it depends well then then you're relying on the technician to find outliers and decide if they deserve to be removed and or not right so that's a reasonable answer for median I I think I we still prefer average for a lot of these things because taking multiple passes I guess we're doing to meet a median you can still get some validation in case your first pass is invalid for some reason but typically we've we've just kind of seen averages works better with multiple passes because you end up with more data that you can work with depending on the length of the past as well I don't know how long you guys tested I think there's kind of variable depending on the game but the length of the pass plus like the first load in a lot of instances as you're pulling in all the textures and assets and meshes and resources that first load can sometimes dip below the average in a significant way and then you know you might see a performance climb depending on how many passes you've done how hot the video card is as I said I can think I can argue persuasively for either charge but I don't if that's all you want to fight with me about I think we're good ok well let's pick another topic then so thermals yeah the the testing was done in a thermal take suppressor and for the case and that's a fairly closed off front panel case the CPU cooler was a knock to Nhu 14s for all intel parts was that used for the thread ripper part as well you know I believe so same one at all so that that cooler is not compatible with thread Ripper there's a tr4 model of it but it's it's pretty different it has a larger cold plate okay the and the 2700 acts used the stock box cooler any particular reason the PT decided to use the Box cooler versus home again I remember what the the rationale was there we do one of the things one of our overriding goals here was try to make the playing field is level spun yeah that's exactly what you deal with all the time how do we what do we change to make it fair and what are we not changed to make it fair and so as much as possible we were trying to keep things exactly identical and but again on this particular one so could you just write down the exactly why we chose andrew has your audio on okay shotgun mics okay for that just think about that okay so coolers sounds like we might get an answer later yeah sure I mean I get most of these I hopefully can send you the answer in the next few hours but okay will depends who who's knocking on my door next right right not anyone from Australia so you're there for another couple days it's what about fan speeds do you know where those how are the fan speeds configured um I believe on all this we went with the defaults unless it was a specific reason okay so I would assume that but I don't know okay my my argument would be towards max fan speeds because if you're thermally constrained and I and these processors behave more like GPUs these days we're boost is thermal and voltage dependent as well so if your temperature dependent and you're allowing auto especially with cool and quiet and the operation it might end especially with a thermal take suppressor which is a restricted front panel case it doesn't have much air flow coming in and then you have a box cooler my concern would be that the temperatures are uncharacteristically fair I mean the thing that what we try to do in all these things is just obviously not an out-of-the-box thing because you can't just buy these out of the box but very much what we want to do is try to do what we think typical person is now a typical gamer is not the same as a typical person at gamers don't put way more time into trying to optimize our system but you know there are game version there are gamers and so what we've tried to do as much as possible is go for that not the absolute top guys but what are the so I guess the question is what is the what's the goal what's the objective of the test what are you what are you trying to accomplish with your test and the goal is level playing field as that was the right thing but is out of the box as you can level playing field and see what we do well out of out of the box for in town this instance would be no cooler because it doesn't come with one so then why doesn't I'll get an aftermarket cooler but Andy doesn't so I believe we said all of them have coolers correct all of them have cooler but out of the box the Intel units specifically in the PT document were and this is correct were noted as not having a stock cooler and that's because Intel stop stripping stock colors their case keys actually no and that's my that's that's my point well that's why I'm saying we go with out of the bottle and we can but you can't do something good just makes no sense I mean so that you see though you see though how it's leaning towards benefiting Intel or I mean is that because you're putting a a significantly objectively better cooler on the intel part and you're using the stock part on AMD where the stock part with Intel would be nothing believe we use the same cooler on all of them so the document the document yeah knock to us the brand and the document is States let see Stockman says CPU cooler brand knock to Nhu 14s exception the risin 7 2,700 X use the included anti-racism cooler the other sea views did not come with CPU coolers okay so I mean I guess we we couldn't run him without a cooler yes correct so we chose that one so okay so you're maintaining the fairer to put that if we did not to put that same cooler on that right and I speaking with you more as a peer for a second protesting my my argument from a stance of scientific methodology would be your goal is to show CPU performance not cooling performance and so we're trying to show gaming performance than all other variables should be equal I tend to agree in general but at some level you know we could be putting another whole level of crazy listen let's operate these overclocked with now I'm Terry's you know and and so that's the always the trade-off how do we get is out of the box mean again wrong phrase here but as normal as possible without doing something stupid right and you were also talking about how a gamer might tune their system more than the average user but then there are 19 games tested in this benchmark and I don't see any mainstream workloads tested so I gave that management working PC markers I'm Paul of course but this was very much game for you I mean right this was not meant to be an overall thing very much about game right so then I guess I would point back to the the note that gamers would be more likely to tune their system for something like thermals like you were saying and counter that this is a gaming benchmark for case Q CPUs so gamers are probably going to be the audience yeah so if you're building if you're building the test for them then I would I would build it with that aspect of you know equalized coolers that are fairly high end in mind or fairly low I'm yeah in your estimation is the AMD core bad one yes I would say I would say I wouldn't say it's a bad cooler standalone just as a blanket statement I would say that in this particular circumstance it is an uncharacteristically worst cooler than the knock to and that's that's definitely provable fact I can go home and do it in my lab and then also that the concern is it's a downdraft cooler and it's in a case the thermaltake suppressor which has a blocked-off front panel it's got ventilation on the sides and so you're pushing I don't know how many fans are in that case one or two 120 is probably pushing air in from the front through side ventilation where you're losing 30% pressure on every 90 degree angle and then it's going into a downdraft cooler so you may get another turn down so you have a thermally constrained case and a thermally constrained cooler versus one which is significantly more optimized for that air path on Intel okay obviously yes okay so let's move on then I guess beyond the thermal notes we talked about there are moles we talked about BIOS a bit server BIOS settings do you want to walk me through should I I mean are you are you close enough to the testing to know what you all used or no I mean again I think what we did is we hmm did what we would consider rational updates I mean I okay anything you get if you go is let you got from a month ago you're gonna do some level of updating but I don't think right again it should say in our disclosure and if not we screwed up there exactly what it does it does yep he did well to specify the BIOS version better than most reviewers yes no I mean definitely the details are there and I know personally from getting called out in a similar fashion that giving a large amount of details is something that of course you want to take pride in and do because right because of science well it also roads here right but but in all services and it's part of the reason why I'm willing to talk with you here is we do want to do the right thing do you want to do good science and right do you want to be as transparent as we can that doesn't mean I mean we are human it doesn't mean we got everything right back I actually believe it's impossible to do something of a certain size and get everything absolutely I agree with or at least we get to argue over what right is I I agree with you on your so we really do want to do the right thing as right as is reasonable for the memory I saw Intel was configured to 26 66 megahertz and either 29 33 and a KITT spec for 3000 Hertz was used any reason for for manually choosing those frequencies I believe those are the specified ones but I don't know where exactly where we got those no that's the yeah that's the jet expects specification for the chipset compatibility and for the CPU compatibility actually not chipset so yes that I mean I guess I guess depends maybe partly on audience because I don't think I've ever seen anyone other than a novice builder intentionally down clock there memory so like it's it's the spec of the processor insofar as jeddak and Intel and AMD has specified that this minimum of 20 666 or 29 33 is compatible but if you're running ax I'm Peter do CP on AMD then you can still achieve the frequency the kit aspects for unless it's unreasonable like 4000 megahertz or something so so I guess I'm a little confused as to why the frequency would be manually down stepped but the profile would be enabled and so you get some of the timings but not the frequency and is there I get what's what's your response yeah I guess and again I would have to get exact details there again the way we try to approach things is we got to see what's written by the manufacturers whether they recommend and then beyond that if there's anything we're aware of this like oh well maybe work them in that but obviously don't do this or that but again even gamers they some gamers won't play games they don't want to spend their whole life overclocking their system and doing all this there are others who do they actually don't ever get around to playing games and I do I run a website called the hammers now as you say once upon a time I have to play games and then I got slow but now huh but but again I had what we've tried to do as much as possible is conform to what manufacturers are specifying and that sort of thing so again okay yeah XMP and the OCP I don't think are considered overclocking by know anyone right do you know if you had any control for tertiary or secondary timings for the memory okay there's some we've seen motherboard variants board the board where tertiary might and kit to kit for memory some of your tertiary timings could change so like TR FC or TR i if i refresh interval and refresh cycle and those can potentially have large impact on performance we've seen as much as 30 percent going from onesie 370 board to another in fairness you use the same zi 370 board I believe for all C 370 parts just I think it was a different specific unit but the same model from what this week again right trying to keep things yeah yeah I know so that's good I don't know the a C's prime I don't know how much if there's any variance it might be pulling from a profile and be accurate every time in which case this would be a non-issue we have also seen that memory will repro file itself on boot cycles so you can end up with different tertiary timings and versus another and that's that's potentially concerned for performance and these a bit more sensitive to timings than Intel not that much more but so that was one of our points of concern as well is to what level of control you know what will up with control of their was there over the timings so that's that's those we have most of the we've got thermals and power supply is completely fine of course monitored not really relevant 512 gigabyte nvme - I'm that - SSC nvme am the - SSD that's fine as well 64 gigabytes of memory any any reason for choosing that for the desktop platforms I don't seem like a normal amount don't you use but I don't I don't actually have it now data that says 94% right it seemed like people the amount of you do we know we have an actual rationale but I do Corsair Vengeance RGB so Corsair Corsair Vengeance RGB 64 gigabytes an 8x8 which is a bit different but is almost seven six hundred twenty dollars the no definitely not but I don't think that at $600 for the RAM it's something that a normal person would use 64 gigabytes on a 2700 X or an 8700 K I mean if magazine room whoever set up the test I could see them potentially thinking I don't want to be constrained by memory that seems like a dumb thing to be constrained by pentesting CPUs so only get 64 gigabytes make sure there's no constraint but also as you increase the capacity it gets more difficult to to keep the same frequencies and timings okay and your house your house your light on us are we visible or we sell oh okay oh that's pretty good cool audios okay sorry cameras got audio from the chocolate oh yeah so we're talking about how a memory capacity can impact frequency and time and performance yeah I mean part of the thing there is definitely we're trying to make a local playing field and so while some of the lower end parts might normally be sold with lower amounts of RAM obviously because you know 16 on this and 64 on that yeah you could do 32 pretty safely on everything and still have four sticks for the attt platforms and I guess like it comes down to I'll be I'll be an internet commenter and pitch you what they're gonna be writing on the comments so that you have an opportunity to respond to because I deal with these people all day all day right and the people watching this are 50% terrible you're welcome uh-huh but the so I see I like you he's the wonder hey I'm the bad guy so I'll do the service of pre-empting some of these so you have a chance to respond I think what I would I would be concerned about is if you're talking about a normal user is probably using an auntie stock cooler and not buying an aftermarket cooler if they have one stock you can't also say normally users behind 64 gigabytes of RAM for a 2700 X because it's a $300 part and that's seven hundred six hundred dollars of memory so I think there's like a disconnect there between what is a normal person buying and configuring for their system I don't have a particular response to that okay again unfortunately we really can't go out there and say what's a typical configuration for some of these systems that are just not that old yet right right okay so yeah I mean well move on from that topic ashes of a singularity we saw the core I $9.99 hundred K 50 7.3 FPS the 2700 X 38.3 fps Delta or at not Delta but a percent improvement of forty nine point six for Intel ashes is a game it's one of the only games it's more of a benchmark really one of the only games or benchmarks where Andy can do well and and I don't disagree with your data that the oriental the suggestions of the 9900 kay is probably the chart topper it's probably the best it has the highest frequency it's at Corinth read parody with Andy now so they're probably better so I don't disagree with that the 50% gap though performs lead for Intel in a game the only game which Andy does decently in would be a number that I would question so I I would be curious to see the data or see it tested again I guess on that one well and the one thing I do want to say is you know obviously you other people have said one day warning what's going on here part of the reason why I'm not having my testers here I we're off running more tests trying to make sure all right well what if we change this did we didn't really miss something we really we're trying to do our best here and uh you know the but I'm sure you deal with the same sort of thing yes do still have to publish some day because ultimate science takes forever yes and so somehow you know what what can I get out there and our goal is to tell the right story you have to tell the truth to be as honest open and trying and actually the thing you'll probably cut this part out the thing we're not going to examine the thing that has bothered me most about is is the level of attacks I mean you know we know III agree we've been doing this for a long time and could we have screwed up yeah sure but it's like hey we've done some horrible thing and we were trying to hurt everybody no we're trying to you know we're trying to get the data out there I love having rational conversations about it I just don't hear a lot of ration conversations yeah yeah well I mean that's that's the internet and I've dealt with it too so you know on a personal level if I'm to relate to you I posted an 1800 X review a year ago when I came out I got death threats for it right so that's just the Internet so I totally get it and I agree with you that attacks and I mean of course the unfortunate implication of doing paid work even though it's I think it's completely fine if you're not reviewer which PT isn't as far as I can tell you're doing that as we're career sure sure but not not presently it's not like you're publishing this data as a media outlet right so so I I don't personally and I hope the audience would generally agree think that this is an instance of a bias against Andy I think it's more likely a disagreement in the correct way to test the parts so I just to be clear I don't think you guys are trying to show the 2,700 X is an awful part because Intel paid you which is what I think probably a lot of those people think well I've seen one or two right yeah of course and that's just I mean we got those all day so I get it I think it's more likely a disagreement in how it should be tested certainly we thought some of it was just incorrectly tested and and I know that your team likely disagrees but that's why we're talking through it so ashes though I would like to see that tested again a 50 percent game just seems like an awful lot versus the 2,700 X and I and I don't think so I'll say this too I think it's possible you'll get the same data a second time and I think it's possible that the data is correct and so far as the machines that it was collected on but it could be things like the thermal concern is timing concerns GPU variance again people are writing it down right we're recording this as well so we can go back and look at these things you know we're totally willing to we want to do the right thing to get the right data out there but even if I ran right out and did everything and change the their handlers and did all this it takes time yeah so we're definitely trying to do what we can right okay so we've gone through some of the game numbers with Ash's most the others it's I will give you a warning I'm getting a little tired so we could just pick the main things to go through here because I sure so BIOS configuration we got that we got frequencies I saw that game mode was enabled for AMD are you aware that that disables a CC x1 this one on the thread rippers we debt we actually looked and saw what are the results like we learned without it and we definitely receiving better results right with the game mode on you know so depending on the game typically but at least on the thread rippers we saw it better across the board okay and I've never like I say across the board maybe there were one or two members we're not we're right now investigating with the 2700 there seems to be more variable kind of above and below and and I guess that's one of the things and you might know the gamer community better than I do but you know to change back and forth to MIT meiner say he's got a reaping yes typically you do not change and so you know that's one of the challenges you know if you only play this game well then you don't ever have to do this but if I'm switching between that game in this game I don't know what people do I mean I now know typically for 2,700 X I would imagine most people don't install rise and master don't even change it out of the box especially if you're going for out-of-the-box performance testing a thread Ripper it's not really meant to play games if you have a 32 16 core parts something like that you'll definitely run into PSO days or black screens or just it won't work in some games of that many quarter they don't know what to do with the course so in those instances you're probably enabling game a and tossing out the courses right and it actually will improve performance like you're suggesting I think it's the the 2700 accent was do you know if that was tweaked at all with rise master or what yeah that's the point so that was moved game mode okay that's one of the things right now and I people have brought this on this that's the wrong thing to do as I say we saw results both directions okay we tried to say level playing field let's enable it for all of them that may have been the wrong thing to do and that's what we're trying to do right now is get some get the results here and maybe we'll have to publish one this is with right without a game mode and see what's there there okay I don't want to deal with that do you know if Numa or my memory was used for threader per okay there's an impact there from latency on some games as well Ash's particularly since there's home games care more about latency some care about unified access to memory but most of them are latency intensive so that's also a question I think for thread Ripper and I can't recall the out-of-the-box specification for that I'll know this I don't really have a question for you here unless you want to comment but Final Fantasy 15 the standalone benchmark we actually we broke the story on this one a few months ago but it constantly renders geometry in the background so it's it's it doesn't call properly and because it's not : geometry properly you end up with largely a GP bound scenario which i think is what you're seeing in your results as you're seeing differences that are 0.4% point 1 percent max 6.7 percent and a 6.7% dealt with it just in the benchmark of the games the benchmark specifically so if you wanna if you want to validate it yeah I wouldn't suggest I would suggest checking the game as well the problem is it's constantly drawing geometry outside of the view frustum I know what it would mean what yeah there I just right right good right and it also has LED scaling issues that was a it's not a benchmark we particularly recommend GTA 5 so in the testing methodology you guys specified whenever you went to presets like ultra high or whatever GTA v doesn't have a preset so you manually go through it but identity specified what settings were used you know and that's that also changes with a hardware change so it's that's why I was asking sometimes it can change the settings you had let me ask you did you do you want to address you know I'm happy to give you a platform do you want to address anything you've seen anything other than what we've gone through that you really want to respond to no I don't I don't really think so I'm gonna let up a little bit earlier but sharing basically just this is my company I'm right one of the two people that founded it when people attack and say we're horribly uniting I get it you know and and as you say it is the Internet I you know right but it still it's still not acceptable so don't do that and so III think we've already talked about that I think the thing you know I just read earlier I said but you know we really do pride ourselves on I mean we picked the name principle technologies both of those names matter to us hmm I'm a geek I'm a technology guy I mean I Drive an electric eye on multiple electric car I just love which is just the most expensive iPhone accessory around but um you know I want all the leis tech we just love tech here that's why it's technologies but it's the first word is principle and we really really try to live by that and one of the things I've said to many of our folks is you know assume that one day the New York Times is so you get to be the New York Times extra fee is asking you about what you did you got to be comfortable standing upright and answering it doesn't mean you'll agree with them or whatever and I don't know that I have more than that to say I really do want to do the right thing I do you and I will say in your in your defense you have been very rational in your questions you've asked me I'm trying to make the answer to you've brought up a few things we're like okay we need to go consider that and so I do appreciate that well I mean look I've I've been there I feel like the difference is my New York Times equivalent is viewers right and so we have to answer to them if we mess something up and we have in the past redacted some testing like with Gears of War 4 for example when we tested one group of cards the wrong settings so I totally get it obviously try to mitigate that as much as possible right and again it's so much easier to get it all right yes first it's way easier yeah way he is here and and and again I don't know I mean obviously you disagree with some of the things you did I don't know that I think yet we did anything wrong mm but that's what we're scrambling Tara do you have time do you mind if we see the test systems at least um thanks for some probably gladly do that I don't want the camera going to the lab one of the things I just have to say we have other people's stuff going on in here right and I I don't think I have anything right there are some things we test that actually just seeing it you know yeah beforehand would be it would be a bad thing but I'm glad to walk you in there but I'd rather okay okay understood okay well thank you for your time you may hate it we'll see you all next time it's really wasn't expecting that it wasn't expecting the opportunity to sit down with them I really appreciate that they let us do that I was trying to of course establish that we're trying to have a rational conversation get answers and not be adversarial unnecessarily so I get what he's saying about you know is his company it's his life so you gotta respect that of course it's going to be jarring to have any of this happen that doesn't mean that I agree with their testing methodology I still all the issues I raised in the video I still pretty much stick to those issues I still think they're a problem thermals are a problem memories a bit weird GPU variance is a problem even if it's largely CPU constrained so I still have those concerns I of course would ask that you don't go crazy and attack them personally but it is absolutely fair to question the validity of the data and we've done that from the get-go and I will continue to do that so you know it's a mix it's I'm just saying this for anyone who's like Stevie went too easy on him I don't think I did I kind of she cornered a few of the answers so I think I think it went pretty well on our side especially I think we got what we needed I think we saw that there needs to be some more information from some of their team apparently by the way it was 16 total systems we found out later so you get there so it was 16 total systems but anyway I think it was a pretty difficult interview for him I respect that they agreed to it I still disagree with a lot of the answers or the methodology anyway and I think they took some information from what I was saying that might be able to look into it further and see if they can find different results sounds like the 2700 x had some variants where they weren't expecting it with for example the the c CX and game mode questions that I had so things that I I pushed in the interview and I'll go back to now that I thought were problematic there was a disconnect between we're trying to do what a normal person would do in this case that was apparently setting up with the stock cool or not and not going crazy installing a an aftermarket cooler which is not that crazy and I mean if you're gonna do it for Intel that's I guess that's a normal thing to do so you should do it for an AMG as well still believe that then I pushed that when I was talking to him and it seems like he he considered the possibility that they might be worth revisiting that like trucks doors ooh okay it's another point of consideration another thing I pushed him on was the memory and I think you all will see this in the interview where I disagreed that yeah if if you're gonna say we're doing what a normal person would do which apparently is open the box install the stock cooler you can't then also say a normal person by 64 gigabytes of memory and I don't think it was malicious when he said he thinks that people generally build the 64 gigabytes of memory or whatever his exact phrasing was he didn't say that exactly but he said he think that's he thinks that's normal or a reasonable amount I don't think he's trying to be evil here at all just to preempt those cons I think it's just that there's a disconnect of what is reality and people building threaded for sure HED t99 ATX t-shirt maybe legitimate use case workstation you have probably 64 gigabytes but a 2,700 X 9900 k no that's not that's not even close to normal you're paying two times plus as much for the RAM as you're for the processor and I don't know I firmly disagree the idea that 64 gigabytes is normal and I think there is a disconnect between the company what they're doing now the question is doesn't matter it doesn't matter if it's 64 gigabytes and whether or not as normal are we just playing semantics and I think it does matter because the amount of memory will impact the ability to achieve those higher clocks and better time in tighter timings significantly in some instances and although it is possible to run 3,200 on rise in these days where 3200 was very difficult on rising 1 especially with a lot of memory it is possible and Intel as well it's you still have to question did they drop the frequencies because there was too much memory and because the the frequency stability or timing stability wasn't there with the proper with the actual out-of-the-box frequencies with XMP or or what or was it actually just intentional and we asked why the frequencies and we didn't have an answer they might get back to us later he actually doesn't have my email list I don't know how but I'll email them or something so yeah I think from our end the interview went well I take issue still with pretty much everything I took issue with before it didn't really answer many questions for me the median versus average thing is the one that I'll give him as a bit of a whatever not a big deal ice I don't like testing that way for reasons I described in the first video but also it doesn't matter that much compared to all the other stuff like the thermals so I know those are most my thoughts on it I think there are plenty of areas where I kind of provoked an answer that we might not have gotten for most other people and I like having those answers even if I don't agree with the answer so rationally it was a good discussion now we just need to wait and see what they do what's their follow-up if there is one and maybe we'll start validating in our own lab at this point so anyway stay tuned for more we might have a follow-up from them if I can give them my email so that's it for this one though like I said start our gamers exit on a table directly you can go to patreon.com/scishow and access if you like this kind of work and would like to support it I think it's a pretty unique angle we're very fortunate they were close and of course even though I still disagree with a lot of the answers we are fortunate to that they did agree to go on camera and we really can't do anything more than that so I think that's about all we got but that's all for now thanks for watching I'll see you all next time
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.