Hi, this is Wayne again with a topic “The Lab is Only Getting BETTER”.
Hi Linus, have you got any labs updates and is there any big projects in the pipeline? Oh man, Labs was super super exciting because we were doing tours over the last couple of days and I got to kind of listen in on some of the tours and I got to find out about things I didn’t know about. For example, did you know that we have a couple of software-defined radios in the RF chamber now that will go up to seven gigahertz and we’ll be ready for millimeter waves sometime next year? That will allow us to take phones and not just like measure how good their reception is, but also find out how much low reception affects their battery life and we’ll be able to see how good their reception is at every different off axis angle. And apparently this is like a big problem in the wireless industry. Devices are getting approved for Carrier use that have that have antenna issues so bad that sitting on one cushion on a couch, you can have fine signal and be browsingyourx.com or whatever else. It is that you’re doing really doesn’t roll off the tongue, does it no, and then you move two cushions over and your reception sucks and your and your video calls cutting out or whatever it is that you’re doing to be fair, it sounds like something you probably Shouldn’T do going on x.com, which is maybe okay, yeah, okay, but let’s, okay, what about you know, uh being on a video call with your relative that you should do okay, Aunt Muriel deserves the the time spent call your mom, even if she’s only calling, because she Needs tech support, [, Laughter ] anyway, so we’re gon na be we’re gon na be able to. You know, find out. You know which of these devices are slipping under the radar through.
You know, back alley deals um. Oh man, I’m really really excited about what they showed me with mark bench uh. We had live demos running for everyone who came through the tours uh any any of the whales and orcas or anyone who booked a tour. How cool is mark bench? Ah, it’s gon na change everything you guys. We had V1 on display, which is just games with built-in benchmarks and collecting Telemetry data and like automating, the running, which is like pretty cool, but you could have done it with auto hockey. If you really wanted to, then we’ve got V2, where now we’ve got uh replays from games that support that, but do not have built-in benchmarks like rocket League.
Two we had those running live, collecting all the Telemetry data we need. The team has the CPU overhead on the client system down to, I believe, under one percent now, which is absolutely flipping incredible like this is not just about developing a way to run automated benchmarks. This is about doing it in a way that is scientific and repeatable and reliable man it’s really exciting, but what I got to see actually not on the tour. Sorry, you guys didn’t get to see this, but what I have seen is the three which is going to be playing.
Having mark bench play the games, putting all of the inputs into a game for ones that that do not have integrated benchmarks. Do not support replays. That are that can be played back in the engine and put a load on the GPU and just don’t support anything at all, but still allow us to get real world in-game in engine performance results from um and then oh there’s another project that they just finished. That I’m going to see the results, for, I believe early this coming week, where they we bought like a metric whack, which is, I think, 12 or 13., is the exact definition yeah 7800 X, 3D processors, because the problem is that in the old days, if you Wanted to parallelize your testing right.
So, if you wanted to say uh, okay, we’ve got a GPU review coming up and we want to test 30 games and we want to test um. You know 30 different gpus, but what is that 900 tests? In the time that we have with the cards? We couldn’t conceivably run them all in a single bench. Now, in the old days, you could have paralyzed that relatively easily. You just could buy more test benches because the processors ran at three.
Oh, it’s exactly three gigahertz and there was not much performance variation between them, but now with turbo boost or Precision boost yeah, yeah, essentially Auto overclocking, depending on you know how much you won or lost the Silicon Lottery. You can have very measurable performance differences from one identical system to the next one, so what the team did was they bought a metric whack which is 13 or 12. um, 7 800 xvd processors, and then they ran them one after the other, after the other, and Basically took our outliers on the high performance side and on the low performance side.
Maybe just took them home, I don’t know what happened to them. Something happened to them and we ended up with five that are in the middle that are within one percent of each other. So what you’re gon na see is you’re going to see over time as we create these validation processes. You’Re going to see us massively parallelize our testing, which is going to allow us to spit out way more data, and I’m not I’m not just talking like a review, because a review is good. A review is very interesting, but a review only comes around when the product launches and so in the long term.
Is that really we’ve we’ve gotten to the point where GPU product Cycles have stretched out past less than a year, which is when I first got into this stuff to about a yearly Cadence to a year ish to two years? To I mean we can have a GPU family, that’s sitting around for two three years at this point, so you could be shopping for this thing, two years down the road and what use is that launch review on Ancient drivers? Looking at ancient games? No, we want to go back and we want to retest this stuff. So how cool would it be if we could get enough processors enough benches that are within our one percent margin of error, and we could basically say: okay, look yeah. This GPU is going to go sit on this bench for like a month, and it’s going to go through every game across every driver revision and we’re really going to track like what’s been going on with the development of this card, I mean particularly for something like Intel Arc that would have been absolutely fine absolutely, but if we don’t have these parallelization processes, we’re not going to be able to do that yeah, because we won’t be able to trust the data. You won’t be able to compare it to anything else in the database.
So um yeah, the the team, was explaining that to some of the people who were going through the tour uh. Oh man, have you seen the mouse robot? Yes, it’s pretty cool, okay, it’s kind of janky right now, it’s all 3D printed, but it’s 3D printed in such a way that everything is, I think they called it like plated or something I can’t remember what they called it, but basically it means that every part Is flat and can be cnc’d on a on a two-dimensional, CNC, router or um or Mill, so that means it’s pretty much ready for manufacturing. Once we’ve got all of the all the Kinks ironed out um. But what was really cool is that they’ve got just like these 3D printer Motors that are like moving the mouse back and forth and then taking comparing the real data.
That’S output from the mouse from the expected data So based on the DPI. That’S configured uh. How far it’s uh supposed to have moved versus how far it actually thinks it needs, so we’re going to be able to really nail down precision and accuracy testing for mice, which is all very fine and cool, but then they started talking about that Juiced up one. I forget what the exact kind of uh Motors that they’re planning to use um, I want to say yes, linear mode they’re planning to uh use high performance, linear motors, with the idea being like right now.
It’S like this big right, because it’s just using like an old 3D printing bed or something like that, and the idea is that they’re going to be able to rip these mice at speeds of around five meters, a second with just unbelievable acceleration. So we’re going to be able to really split hairs, and you know it’s one of those things where I feel like a lot of people, look at the labs content and go well. Why does it matter? You buy a power supply, your computer turns on who cares and but I’ve I have never been of that mindset, like I remember getting into like the text equivalent of screaming matches with people on the forums in the old days where some people had this attitude. That was like look if they want to buy uh a Pentium dual core. You know instead of an Athlon 64 X2 and they want to pay more and they want less performance and they’re happy with that. You know why does it matter, I’m like they could have had more for Less? Why would they accept anything but the most for the least it’s wrong? I love how you say this is back in the day you built literally a whole company to accomplish that goal and then did it again, but you were like this time with a lab, better um, more science, and so like you know, why are we subpar products? Why don’t we care? We should care if this one’s a little bit better and we can prove it well, then we should.
We should get that one and then we should put pressure on everybody else to do better. That’S how the industry moves forward. That’S the whole point: you should care, and so realistically, okay, whatever gaming mouse you choose to buy, I’m actually not going to judge because it’s a peripheral, it’s something you touch. You might judge a little bit okay, but it’s something you touch it’s very subjective.
If one of them makes your hand cramp, even though it has the best sensor, don’t use it. Obviously, right but, and – and so you know in the mouse category, I think there are going to be a lot of you know well really they’re, both pretty good but pretty much. It comes down to you know, what’s more comfortable for you or what has the the software that you find most intuitive to use, but if all other things are equal and you find both of them pretty much comfortable, why wouldn’t you take the one? That’S one percent! More accurate, yeah, yeah, yeah or like why, wouldn’t you get rid of your stupid dinky little like laptop mouse that you’re using on your desktop, because there’s like no way yeah, who was that Electro boom yeah yeah he’s here, yeah uh, Maddie Maddie? Are you okay? We can get you a new mouse, bro yeah. In fact, I I think we actually have a sponsor locked in for the the medi upgrade. I don’t know if he knows that yet, but where we are getting you a new mouse, bro yeah we’re good we’re getting him a makeover nice. Yes, it’s gon na be awesome.
His workspace is utterly unacceptable. Yeah yeah, except for the art which is actually amazing and, and he painted it, which is sick, don’t remember his scopes and actually his treadmill yeah his treadmills okay, so he’s got some pretty cool stuff, but there’s some things that are bad. We’Re gon na fix everything. That’S not cool about it. In fact, I think the sponsor is Asus Rog, who also sponsored LTX, so yeah we’re gon na be hooking up with some re hooking up with some really good stuff, really good stuff. Oh crap Linus leaks, a video is in the bingo [ Applause ] .