Hey everyone, this video's endroll was completed ahead of the NVIDIA 4080 12GB Unlaunch on Friday, and snuck past our team before being publicly posted. Please disregard that specific SKU... but the 4090 and 4080 16GBs mentioned are still worth keeping an eye on! - Sean
In regards to the potential scoring system between keyboards; you could have a spreadsheet style interactive chart where users can remove any categories that do not matter to them. That way everyone can get a customized ranking system based on their own preferences and hard data.
I'm really glad LTT seems to be going in the way of info for all concerning benchmarking and testing of hardware. The general public has definitely been waiting for a (very hopefully) unbiased way of knowing what is the best product for them. Bravo.
I love that there is actually a plan for a website. So many YT channels just focus on videos exclusively. And I understand that's how it's monetized, but often I have already watched a video maybe a year or month ago and just need a quick bit of information. So having a website instead of having to find the video and scrub through it looking for the info is great.
Bruh, the LTT Lab is turning into everything I could've hoped for. As a consumer this will be a massive win. Big ups to Linus and the rest of the team 🙏
This is huge as I have almost completely stopped using gaming keyboards and switched to a thin logitech K780 because it feels nice and is most importantly - quiet. I have come to hate gaming keyboards for being the loudest part of my setup. I love gaming keyboards, but I have run into too many ergonomic and noise issues with them.
If you are testing keyboard latency, make sure to test multiple keys being pressed at the same time. Some keyboards send commands differently when there is one key pressed, and when multiple keys are pressed. There used to be some keyboards that had issues when you had more than 3 or 4 keys pressed.
was gonna comment this. For people who do stuff like Rhythm games that is a huge make or break and might help some frustrated people find out why they are missing notes at high speeds
@@LiveType it's always best to use the KISS approach for version zero. If you start with "the best" first you make the problem much more difficult, and it might not be solvable at that scale of difficulty. if you noticed, they started with pictures and mapped that to an XY grid, then ... they just slowly went down till they started feeling resistance, and measured that curve. They did all of that without a machine learning engineer which is much more expensive and harder to hire than a normal software engineer. Then as they saw a monetary value in a more complex solution, then they hired exactly the right talent for that scale of the challenge. To me, it sounds like LTT is doing things much, much better than reviewers before them. I, however, didn't see a simple multi-key setup so it looks like they are optimizing narrowly vs. broadly. Knowing how a keyboard responds to multiple keys is a known issue, and ... it definitely affects max typing speed and gaming. Both of which are probably important to their audiance.
As a robotics engineer, this setup is definitely a decent proof of concept. I'm really excited at the progress and promise of the work! There are some simple changes you guys could try to make the setup more robust, especially for vision. Fully enclosing, specifically with a light box, would really help with consistent lighting as well as integrating the lights. You can also get a 3D image using a 2D camera by stitching together images as the robot moves or using multiple camera angles. Multiple cameras can allow a 3D image also using structured light through a projector. That will be much cheaper $500 USD per camera, than a multi $1k/$10k 3D sensor. Getting a Basler or similar GigE camera allows open source focus control and confocal lenses if you want a really flat 2D image. 2D line scan profilometers are a decent idea, but be careful that robot movements can cause rippling in the image as they move. So take measurements while it is stopped or consider a true 3D sensor (point cloud) as they are about the same cost range as a laser profilometer. Another lower cost option is using a Microsoft Kinect or HP 3D Camera. They used to be off-the-shelf, but you can find used ones. They have limitations in resolution, but might work for identifying the key top centers. Just some attempt at friendly advice that I gained through many hours of banging my head against problems in the industry. I'm sure you guys have already tried a lot of approaches you couldn't fit in the video.
This. Laser topology seems not only overkill but prone to issues such as metallic reflections from both the robot itself and keyboard backplate (or even shiny keycaps), not to mention high sensitivity to jitter and judder. A rigid lightproof enclosure with two or three distinct camera viewpoints should be more than enough to build a reliable pointcloud with high enough fidelity to work here, or spend the big bucks on a 3d scanner that will not require complex positioning to capture the topology of the keyboard.
Every Lab update we get just blows my mind. The fact that a tech youtube channel ive been watching for years and years has developed to the point of legitimately quality testing big name brands to keep them honest, and the fact they HAVE the influence behind them to make this possible is just insane
Let's not forget the fact that if they have been lying, they will need to be scared straight because that can ruin their keyboard reputation status. This can make or break their status. Although, they need to test at least a minimum of 3 of the same keyboards because a single keyboard could be a keyboard that failed and somehow made its way into the stores.
Going from whole room watercooling in a house where they didn't have a permit to run the business, to a multi location quality control facilities is quite the change
Not gonna lie the production value of videos with all these animations and all the extra tests you guys are able to do is absolutely amazing! Keep up the great work!
Not sure if it was mentioned before but one thing you should not forget is that most people won't push the keys 100% vertically so it would be good to at least try a few key by pressing them in an angle like I noticed some Keyboards won't work really well if the angle is to high (usually doesn't happen a lot when you are sitting right in front of the keyboard but sometimes happens when you help someone else who is in front of the keyboard so you have to type from the side.)
@@reaLe-1 As IT support and a "former" gamer that also has friends sometimes sitting next to him I can tell you it happens a lot ... Lots of people asking for help have their keyboard in front of them and don't even think about making some space for you so you type in a very bad angle some times Also if you are sitting in front of the pc with a friend (yeah many years ago online play wasn't really popular) and your pc table is small at least one if not both of you will sit not right in front of the keyboard
I could be way off on this, but I think Alex had a huge hand in all this happening. Him running the workshop really turned LTT from just receiving and reviewing things to actually building things themselves. From 3D printing to CNC, sandblasting, other metal cutting etc. That made LTT really dig deep into the inner workings of things.
@@CamelSmokes23 Alex was the first engineer they had who worked on physical materials, yes (dunno if they had any software/computer engineers before him) They were pretty clear his machining skills were one of the reasons they hired him, so I’d say the drive was there even before Alex was hired
@@CamelSmokes23 Honestly, I would give huge credit to Alex here. Not to say he is the reason LTT got into engineering stuff, but when it comes to them building and engineering their own projects, Alex was always the one there running the show. The success they've had with projects like that are probably a huge part of the reason they stepped further into stuff like this.
Force curves for more esoteric switches would be fantastic. Everyone already knows what Cherry MX curves look like, but less common "hobbyist" switches sometimes don't have any graphs online. It would be great to compare some of these boutique tactile switches against each other, against well known switches like Cherry MX Brown so we have a reference point
"esoteric switches" makes me think of ThereminGoat's switches collection. The guy has like 1.5k switches, and I can maybe name only 10 (not including colors, ofc). I wonder whether there could ever be some sort of collab between them, considering that Goat doesn't record videos (at least not public, afaik) P.S.: For the last "family photo" he posted, the thread ID in the shortlink is "tbqrwf"
Sometimes reinventing the wheel can be tedious. It might be of value to collaborate with iFixit on the teardown schematics and repair parts needed. You both are trusted companies who can utilize each others strengths!
Thought the same thing. If you're talking about "helping consumers make better decisions and access the information", creating a separate platform for the same information that is established on a household name like iFixit seems counterproductive to me.
I really hope Linus' team reads this and takes it in to serious consideration. As soon as I heard his plans on gathering repairability and teardown data, I immediately thought that we really shouldn't fragment this information. They need to work with iFixit on this.
I'd actually be curious to see how custom keyboards compare to the OEM variety, in terms of consistency and latency. KBD67 vs Razer Huntsman, etc. Another test that I don't think I've seen anyone else attempt is two people build the same custom keyboard, and then see how consistent the two resulting boards are. Can give context about whether we can even replicate builds from TH-camrs using the same specs. Their board may sound amazing, but perhaps it's not even feasible for someone else to recreate the same experience.
Soundwise, it's like almost impossible. Like you'll have to know which lube they used for their switches, what mods they used for the keyboard's case, and what kind of textile is used in their desk mat (and its size). The desk material, density, and size. And a lot more. So it's impossible to replicate someone's else builds. For the latency, I think it is only important if you're an elite rhythm game player, otherwise, a 500hz/250hz polling rate is enough.
as far as latency and all of that, there are a few types of PCB used that could effect that, but most of the time it comes down to polling rate as well as if its usb-c or a.
I'm so excited for Labs. Now that Linus finally gets to start showing us what all that time and money went into it's clear that it was totally worth the wait.
Man, if you guys manage to apply this level of thoroughness to all the other product testing you're planning, the LTT Lab could become one of the greatest consumer education tools the consumer electronics industry has ever seen. The potential of this undertaking to completely shake up that space is incredible. If I was Intel, AMD, Nvidia, or any other player of any importance in the industry, I'd be watching your progress like a hawk. This can only be a good thing for the end user and I am super pumped!
As someone who gets to work with FANUC robots everyday, I absolutely love that you guys are into testing keyboards with a robot. Still learning my 3DL vision, but the lasers help out a lot with depth. Good luck!
The amount of value LTT is giving to the community with CW's great designs like the backpack and screwdriver and now Labs' MarkBench and now this and who knows what else more is astonishing.
I've always used RTings for this kind of data, and it was unfortunate that no other site had such systematic testing meaning that if RTings didn't have info on the specific device I am interested in, there was nothing else comparable. Hopefully LTT will start to help fill that gap soon!
@@seanmccants1500 The latency tests were the most egregious. They changed methodology a few months back, but previously they had an audio-based setup that produced results so bad that they weren't even internally consistent. I'm talking the same peripherals achieving lower latencies wireless vs when plugged in, massive deltas between models with identical PCBs, Bluetooth devices that scored competitively, etc. Even now there's a bunch of issues in the technical aspect of their reviews. I won't go into all that here since it appears they're trying to improve, but I'd just take everything from them with a massive grain of salt.
@@chroma_aus uhm, there are devices that have lower latency when used wireless vs cable. using a Dualshock 4 on a PS4 through a cable connection has higher latency than through Bluetooth for example. the same is true on the Switch and its pro controller. so I wouldn't instantly say that data like that is automatically inaccurate just because intuitively you think a wired connection has to be faster, there can be complications in the software or hardware that can lead to that not being the case
Hey I also work as an ML Engineer in Computer Vision for product line quality control and I would say your laser scanner is totally overkill. I mean if you have the money, go use it but just for identification of the keys you can use grey scale images with some smoothing and I would guess just edge detection to find all the keys pretty reliable and super fast. Only reflection / polished keys might be a problem. I think if you open source your key identification algorithm you will get help fast an easily
More than overkill, it's like someone doesn't even know what IOL is. As for reflections, image processing on multiple images with different light sources will sort that out, and remove most of the shadows. Most of the code needed is open source, and a decent IOL setup will cost $30?
OTOMH, couldn't they just also invest in another webcam mounted opposite the one they already have, and do some comparative stereoscopic wizardry on the image feeds to determine distances? Seems like a good crossover between cost (versus a laser) and algorithmic error rate (versus an algorithm that only uses one camera and has to rely on ML to discern shapes). Come think of it, they could just use the same one camera and just take multiple images from different XY positions to achieve functionally the same. (I don't know if that's what you guys in this thread were talking about, or if some of what I said is what they were already doing - this is all admittedly beyond my expertise.)
I've been following LTT for I think a little over a decade now, ever since I built my first PC. You guys never fail to make great content, but honestly, the LAB is on a whole different level. This is definitely my favorite initiative I've seen from you guys so far.
Respecting the amount of work they're promising as I definitely look forward to it. Been modding keyboards since last year but will agree that my sound recordings are inconsistent as other content creators also mentioned the inconsistencies they have since it has a lot of factors to really consider but nonetheless with the goal for the website this is aiming for, I am feeling great about this!
I CANNOT WAIT for LTT lab's headphone testing and reviews. So many mainstream reviews focuses solely on features, completely skipping on sound. Having detailed, but accessible information about a headphone's sound signature will be a fantastic push in the right direction for headphone reviews, headphone marketing and manufacturing alike.
Every reply here seems to have missed the OPs point about this information being available on MAINSTREAM channels. Many of the other channels mentioned have fantastic content regarding audio equipment, but none of them can reach the audience that LTT can.
if their peripherals testing is anything like their gpu testing it will be full of errors and bad methodology. hopefully they can improve their own quality control in the time between now and then.
The interesting part is with the 3D laser scanner, you could actually have the robot change the angle of the finger to match ergonomic keyboards. This would help with the testing.
Hopefully laptop keyboards will be included eventually. It would be nice to see some classics like an IBM model M or other never die keyboards from the 80's and 90's too.
Love to see that Antoine comes from the same french school as I am right now. I'm feeling a little better about my future thanks to him :) That stuff sounds amazing, I can't WAIT to see it in production!
The mere fact that this is being served on the main channel shows how much LTT has shifted from a couple guys doing gaming reviews to a MAJOR enterprise - almost on par with Consumer Reports. Kudos for your hard work and drive to filter out the BS and give us real data to work with.
@@eugenehong9668 Yeah, stuff like repairability, how hard is it to get inside etc isn't even a concern in the enthusiast keyboard space, considering most of them come as a kit or are intended to be modded/customised. Only the consumer keyboards have these concerns.
Given you were holding a Keychron, I'm hoping that eventually you guys test many different switches + keycaps as well. This will be tailored towards the Mechanical Keyboard enthusiasts, but would be amazing for data verification.
I can't wait to see more topics about latency in general. Like input-to-display, which is accumulative. Affected by input, game engine, engine settings, drives, drive settings, monitor, and monitor settings.
I like the idea of having a database of information to decide your keyboard, but I think as a product, I would actually prefer that there be some list that is generated of recommended keyboards based on filters set by the user. Not to mention, some kind of "Amazon's Choice" branding on products that LTT deems as a good product overall would also be helpful in making safe purchases.
I think Linus said that the labs is mostly going to be a loss, and is mostly subsidized by earnings form regular ltt content on wan show. He also went into how most of the stuff from the labs is going to be write ups and articles on a website, and how the majority of written media is dying out because it isn't profitable anymore.
I love how he is keeping us informed on not just current projects but where they are protected to be heading to. It's super interesting and i wanna see more
glad to see the redragon keyboards in the mix those are my favorite for no real reason other that I liked their mouse so I got their keyboard and liked it too, so I'm excited to see how it holds up to the competition My K551 is 6 goin on 7 years and still choochin, although the lighting is dimmer (this one is the non RGB one just red) and my RGB K552 is 5 goin on 6 years, ad the RGB is still working fine, and still good brightness I have one on each desktop and they have gotten constant daily use that entire time
The thing about latency with USB keyboards is that USB has to be polled, it doesn't use interrupts like the old PS/2, so at this point you're testing the polling rate, which can, and will be limited by the host more likely than by the keyboard, if you hit a key between polls, it'll have to wait for the next refresh. And not to mention, if you're polling way too much, it will tank your CPU.
High-polling-rate peripherals are a thing and implemented through the driver. If a manufacturer didn't bother to tune the firmware and driver, that's a legit data point.
@@Steamrick the normal HID class will have 1000Hz a maximum polling rate. Which still gives you the 1ms window. Theoretically you can go faster with your own driver, but not sure if anyone does or if it is worth it. And 1ms is an eternity so I would be really surprised if keyboards really differ given the polling latency. Of course statistically you still can extract that latency but probably not worth it...
@@lal12 There's various keyboards and mice with polling rates up to 8000Hz. I don't really see the point since it's already pretty far up the path of diminishing returns.
If and when possible you should try the old butterfly keyboard from apple, it would be interesting to see how it compares to the spec list Or even other laptop keyboards because these are permanent and not replaceable Also, have you thought about testing how many clicks it takes to break the key? Manufacturers claim millions but i dont think in reality this is possible.
Why shall millions of clicks not be possible? Have you ever though about how often you really are pressing the keys? And how long a keyboard lasts? Then there is the matter of dust and other dirt getting into there - what if this was not the case and the switches would wear less?
@@dezpotizmOFheaven thats true but a full keyboard has more than 80 keys and if 90% of them survive, what do you do about the other? Its not like you can replace the keys easily in most keyboards. After reading my comment again, i think the failure rate before reaching the spec would be a better think to see rather than just some keys.
Congratulations on tackling the very difficult task of testing a crucial human-to-machine interface. As a fast typer (500 chars/minute) I have stuck with cheap wired, low-noise, short-throw keyboards. There seem to be very different preferences between gamers, people who touch type and those who don't.
I used to do industrial computer vision. I'm curious if you guys tried low angle lighting instead. It could help make sharper shadows to have more distinct edges of the keyboard. It's a common technique for creating shadows for aluminum components with serial numbers etched on.
The one thing I can see being a problem with this is that keyboards have so much variation in height - lighting and capturing just the keycaps, and the keycaps alone, would be next to impossible. On flat aluminum pieces to illuminate an etched serial number it may work great, but with keyboards I believe it wouldn’t work that well
I'm surprised they didn't just use the printed keycap symbols as a reference, in order to reduce uncertainty in the location of the edges i.e if the symbol is detected to be outside the bounding area of an inferred keycap surface, then that results in a lower activation/probability in the ML algorithm. Detecting a defined set of symbols is way easier and more reliable than detecting the edges of an arbitrary shape.
@@tomfahey2823 It's unfortunately not as black and white as that. Not all keyboards have symbols on them. Not all keyboards use the same typeface for their symbols. Not all keyboards put their symbols in the same place on the keys. It's a surprisingly tricky engineering problem.
Excited to see potential variances across all the keys on a board - especially gasket mounted boards. Would be pretty cool to see bottom out force graphed as a gradient or something like that, spread across a whole keyboard edit- also force profile consistency of hand lubed switches would be pretty cool to see as well. both immediately after lubing and after a period of heavy use
Would love to see the latency tests extended to mice (I assume you already want to do this), but also hubs/docks and KVM’s, like the USB fiber hubs Linus uses in the new house, or the Level One Techs KVM’s 😀
Man I'm really really pumped to see what people like HipyoTech think of all of this, I feel like this is gonna be ground breaking for all the custom mech keyboard youtubers/builders/reviewers as it's shedding actual concrete data on so many areas where people like Hipyo were previously considered experts just through their vast experience with so many keyboards, but their experience is all from their own testing, and their tools up until now has just been their own hands and knowledge from trying so many keebs, but now we will acutally be able to compare the information we've been getting from people like Hipyo to the actual concrete data we'll see from LTT Labs. I for one am very excited about the future, LTT Labs is so fucking huge for the industry man, grounds will be shook. Side not: Linus suddenly changed how he says Asus..? He said "AH-soos" in the sponsor at the end, I swear he usually says "EY-soos" lol
Honestly, I love the idea of the lab, and it's gonna be amazing for a lot of stuff, but I'm not sure if it's fitting here for keyboards. Chances are, people nerding out over keyboards won't be interested in prebuilt keyboards in stock configuration. It's just a scenario far from any sensible use case. It's like testing current gen graphics cards with random CPUs from 2003. Linus is correct in saying that, especially given sound, we lack consistency in reviews and sound tests. Offering a consistent setup is definitely the way to go, but this needs to be paired with a consistent, true to real world, build. Of course that's a task far from easy, but results matter so little this way. This project will keep manufacturers accountable, which is good, but has no relevance for the enthusiast community. The one (easy) avenue I see is testing switches. Friction could possibly be measured, force graphs of course, wobble and binding. In the end, the keyboard community has to mature towards more scientific methods, "uh listen to this / woaa feels smooth" is just pretty much useless information.
@@lennihein Yeahh I see where you're coming from, but I think there is still value in this for enthusiasts, as you say testing switches for example. But also, not everyone wants to go down the custom mech keyboard route, including some enthusiasts. I think there's a pretty big subset of people who are super into keyboards, who follow people like Hipyo, and who love the whole custom board space, but who don't necessarily have the inclination to go buy all the individual components and build one themselves, I think quite a few people are very enthusiastic about mechanical keyboards but would prefer to just find an ideal keeb that they like, and then maybe make some minor changes like swapping keycaps, or maybe they find a board they love the look of but it doesn't have the switches they want, but the board is hotswappable, or in my case for example I bought an Anne Pro with gateron browns, and I cranked up the thock by applying superlube myself. As much as I find the idea of building my own mech keyboard very attractive, it's also expensive, and is quite an involved process, I think for people like me, these Labs keyboard tests will be useful, because I would rather buy a semi-customisable pre-built and then maybe just swap out the switches to something I like the feel of better, and maybe change the caps for aesthetic reasons. For this kind of thing, Labs will be super useful, because once they have reviewed all the top-selling pre-builts, and all the major switches, then I will have actual data to look at to decide which keyboard I want, and what switches I might want to put in it. But yeah basically my point is, I think there's a pretty big gap where _most_ people reside in between "i don't care about keyboards" and "i'm a custom mechanical keyboard nerd". I think probably _most_ people are into keyboards but aren't likely to ever go to the trouble of building one from scratch, sourcing all the components, researching every different aspect etc. Basically people who want the end result but without having to put all that work in to ensure they buy all the exact right components for them, as it would suck to set out to build a custom board only to find out you bought the wrong parts, or didn't know everything about one or more components until after building it etc...
Dear LTT Team and especially Test Engineers: Please keep up the great Engineering and put your great Engineers into the spotlight. Antoine was perfectly fine to be understood - no need for captions.
With such an in-depth research and documentation, it reminds me of the screwdriver project and makes me think that LTT May release a keyboard in the future using all this data.
I hope that you guys also do historical keyboards that enthusiasts go after as well. Like the fabled IBM Model M, and the Apple Extended Keyboard II. Just to see if those keyboards are worth the hype and to see if buying a 30+ year old keyboards for modern applications are worth it.
One thing that may have been interesting to add is how you baseline and do routine calibration checks on a system like this to ensure your testing method isn't drifting.
You are doing a great work in defining a standard for these products, I’ve been skeptical for expensive keyboard for sometimes and still think paying over 150$ for a keyboard(unless it’s a custom build one) is not gonna get you that much ahead.
I think a really important consideration for sound is not just the sound of a full key press, but also the sound of tapping the key without actuating. Most of the sound of my keyboard (old membrane Dreamcast keyboard, my Keychron K2 is in the mail) is the sound the keycaps make when moving, and the membrane switches are nearly silent. In contrast, my laptop's builtin keyboard has extremely quite keycaps, but the membrane switches on it are much louder. Knowing that keyboard X and keyboard Y have the same switch sound, but different keycap sound will be very valuable, especially if the two keyboards are effectively the same in terms of price, performance, feature set, etc, covering the remaining difference of keycap sound will inform potential buyers of which keyboard they might prefer. Personally, I can't stand the noise of keycap rattle, but for some people that extra bit of clackiness might be a benefit.
I would recommend doing the second test concurrently with the first one. That way, you can analyze the total latency delta across all the keys, see if there are any issues there.
I’m glad to see someone doing this type of thorough testing, and I’m really excited for the results. Also, I will totally forgive you if you don’t finish it all by the end of the year. Imo, it’s better to do it right and with a smaller set, than to rush it. But I’m guessing it’s a time is money situation. Keep it up!
@Michael Mazurczyk I dont think it is possible to show differences between custom keyboard and off the shelf keyboard through videos, you need to actually use and feel it to understand
this is actually insane and I love it. such an incredibly great idea. it's expensive but doing this for like the cables (the website I mean, having that info on this website) and other stuff (Psu stuff since you have a psu tester too) would be amazing. I'd love having a trustworthy website to look up products on
if the only thing the lab ends up being is iFixit for keyboards, still the best timeline. I will be super interested to see how these latency tests go, and I'd LOVE to see you guys eventually doing the same thing with video game controllers via usb and bluetooth as this information is SORELY lacking for all kinds of inputs. I'm also interested in nkey rollover testing as it was a big deal, back in the day - but it really doesn't seem to come up anymore.
This is actually interesting! Hyped about the results. Some of my Keys on my Razor Widow have a delay. I cleaned everything but it didnt fi the issue. Kind of afraid to take it more apart. It was expensive haha
This looks insanely promising. I'm also hoping that they will not only cover the big brands but also do their best to find the hidden gems and no name brands that make good quality products
that's why i also think they should have a suggestion box on what kind of devices or brand they should test on labs. just the back end might need to be a bit smarter with the charting...
It amazes me how many different keyboards are out there now! Back in the 1980s there were just two kinds of keyboards: IBM typewriter style that worked well, and the type with non-standard keys that didn't work well at all. Thank you Linus for setting up your labs, this kind of info is needed, there are so many scam companies selling junk it's impossible to know who to trust.
Is an idea for data visualisation: On the website you can have a tool where a user can select his two most important aspects of a keyboard (f. e. latency and preassure). Then it will plot a graph with this two aspects on x and y axis. And in this graph, all your tested keyboards will be presented as dots. Then the user can see which keyboards perform the best in the two aspects he is interested in and can visually compare alternatives
For key location detection, could stick an IR reflective dot to each key, and segment the IR image very easily. No lighting issue, cheap. No fancy depth laser scanning needed.
Think I'm talking to myself, but hey, use a fixed size dot so it will also tell you the distance to the key surface, knowing the camera FOV and resolution.
That sounds like a lot more work in the long run than just having a smart laser vision system. Probably cheaper and easier in the short run but once the hardware has been purchased and the vision system programmed they can just stick a keyboard into the robot, hit run and walk away saving time and potentially money since they wouldn't need to purchase an pen or stickers.
@@grn1 yeah, this would require a couple minutes at least per keyboard, 200 keyboards means 6-10 hours of sticking dots on. Whereas computer vision, while it may take longer to program, results in code and systems which can be reused across many projects and could potentially serve for years without maintenance on a dozen testing setups, saving months of labor.
Detailed testing outside the manufacturer's control environment is important..it really drives manufacturers to create quality products paired with accurate specs that are used for marketing. Thanks Linus!
yes, totally agree on that. I'm also wondering, why not using a small programm on the computer for key pressing detection that interacts with the key pressing microcontroller to determine the time. Measuring it with the monitor while running a game seems really dumb. Also there is a potential flaw: I don't think that they will test every key for latency (because that would be too much work with their current setup), so it could happen, that some keys are faster than others. Combining it with KAVE seems the only good idea.
I highly doubt that they'll even bother with all the custom switches and just do the "popular" ones that are found on majority of prebuilt keyboards, ie. Cherry, Kailh, Gateron and Outemu in their red, brown and blue variants. Maybe sprinkle in some Razer and Corsair switches.
I can't wait to see the full tour of the LAB once it's ready and the data that is being collected by the LAB crew. In the future I can see this LMG project as a "go to" place for a lot of consumers that like to know what exactly they're getting for their money. In even more distant future, we may also have "LTT Certified" stickers on the boxes... who knows. Very ambitious project with very consumer focused goal - you're hands down the best!
Having an "LTT Certified" scheme might seem like a good idea at first blush, but the devil is in the detail; done wrong, something like that could give rise to concerns over impartiality (say, if manufacturers had to pay LTT to do the test in some way, could they be accused of bias in the test results?). But if it could be done in a way that ensures impartiality and integrity of the results, it would be awesome.
@@mnxs I see your point but the same could be said for the current sponsorships that LMG has. Also, you can have the same "fear" with any other certification on the market, so I don't consider this as an problem at all.
Will you also provide the 3D-scan of the boards? This would be useful for designing 3d printed accessories and since you will be already scanning each board it shouldn't be to difficult to pack these as point clouds or stl files.
I would love to see y'all test an original Model M with this. Curious to see if a PS/2 connection has any tangible difference in latency compared to a modern USB keyboard, because I honestly can't tell after using my '87 on my gaming rig for the last 5 years 😂Only time it ever let me down was the GTA: San Andreas driving school, where you need to hold down 3 keys at the same time
I think the difference is that PS/2 input (at least used to be) handled differently from USB input on OS level (PS/2 seems to be interrupt based). I managed to find one comparison study on PS/2 vs USB latency, but it was done in DOS in 2002. Other than that I'm pretty sure that nowadays with >1kHz polling USB either outperforms PS/2 or has negligible difference
Gaming wise, not a major difference. Most game engines poll the inputs rather than get "interrupted" at the system level. So even if the PC itself knows about the new updates, the engine doesn't immediately grab and act on it until the next "tick".
You have such ambitious goals with this, and I'm really excited to see this list of hardware come to be. I genuinely think this could have a large impact on the industry, and if not at the very least it's going to be ultra helpful to so many consumers. I'm really glad you're reinvesting so much of the money you make to do this, and it just makes me hope you'll keep succeeding with everything you're doing. Obviously none of this would be possible without an excellent team like yours, so thanks to each and everyone involved, again, I'm very excited to see this.
Hey everyone, this video's endroll was completed ahead of the NVIDIA 4080 12GB Unlaunch on Friday, and snuck past our team before being publicly posted. Please disregard that specific SKU... but the 4090 and 4080 16GBs mentioned are still worth keeping an eye on! - Sean
literally just came looking for this comment LOL
baba booey
have you thought of having someone actually watch the video before uploading it lol?
nothing beats my K120 from logitech. Bought 10years ago for like 20bucks, still A-fine
@@Dudelee1 🤡
In regards to the potential scoring system between keyboards; you could have a spreadsheet style interactive chart where users can remove any categories that do not matter to them. That way everyone can get a customized ranking system based on their own preferences and hard data.
@@Waterghosteus ill help you with that
This
up vote so that linus could see it
This testing lab will 100% be the lasting legacy of LMG. Love seeing the progress Linus and staff!
I'm really glad LTT seems to be going in the way of info for all concerning benchmarking and testing of hardware. The general public has definitely been waiting for a (very hopefully) unbiased way of knowing what is the best product for them. Bravo.
Between LTT and GN, it is so much easier to make informed decisions than it was even just 5 years ago.
The amount of data that LTT labs is going to be able to produce for free for consumers is crazy Linus is a GOAT
RTINGS is sweating right now.
Or, a collab?
This will be glorious
@@spacetoast7783 Seems they have hired at least one former employee from RTINGS. Apparently Antoine worked there until a few months ago.
Pretty sure Linus is a human. Or did goats recently evolve?
I love that there is actually a plan for a website. So many YT channels just focus on videos exclusively. And I understand that's how it's monetized, but often I have already watched a video maybe a year or month ago and just need a quick bit of information. So having a website instead of having to find the video and scrub through it looking for the info is great.
You can't really index video data either. Like keywords & whatnot.
Cool you will have to scrub through the website then😂
@@larion2336 you can but only if the Chanels put proper titels and descriptions on the videos, unfortunately they get much less views that way though
Bruh, the LTT Lab is turning into everything I could've hoped for. As a consumer this will be a massive win. Big ups to Linus and the rest of the team 🙏
It's Rtings meets iFixit for keyboards. I love it. I imagine more hardware will be added in the future.
between this and GamersNexus Steve, they're gonna be holding manufacturers to their word hard core. and we will be the winners for it!
How would it help you? 🤷♂️
@@ghomerhust this is a reason why we always need some oversight and verification never trust the "word giver", but let the data and science say.
@@jumbomuffin1316 Not wasting money on overpriced products that don't actually perform the way they say they do.
As a QA professional and PC enthusiast I just love this new approach to testing stuff. Wonderful idea, great videos to come.
This is huge as I have almost completely stopped using gaming keyboards and switched to a thin logitech K780 because it feels nice and is most importantly - quiet. I have come to hate gaming keyboards for being the loudest part of my setup. I love gaming keyboards, but I have run into too many ergonomic and noise issues with them.
Fellow QA professional 👍
If you are testing keyboard latency, make sure to test multiple keys being pressed at the same time. Some keyboards send commands differently when there is one key pressed, and when multiple keys are pressed. There used to be some keyboards that had issues when you had more than 3 or 4 keys pressed.
was gonna comment this. For people who do stuff like Rhythm games that is a huge make or break and might help some frustrated people find out why they are missing notes at high speeds
@@LiveType it's always best to use the KISS approach for version zero. If you start with "the best" first you make the problem much more difficult, and it might not be solvable at that scale of difficulty. if you noticed, they started with pictures and mapped that to an XY grid, then ... they just slowly went down till they started feeling resistance, and measured that curve. They did all of that without a machine learning engineer which is much more expensive and harder to hire than a normal software engineer. Then as they saw a monetary value in a more complex solution, then they hired exactly the right talent for that scale of the challenge.
To me, it sounds like LTT is doing things much, much better than reviewers before them. I, however, didn't see a simple multi-key setup so it looks like they are optimizing narrowly vs. broadly. Knowing how a keyboard responds to multiple keys is a known issue, and ... it definitely affects max typing speed and gaming. Both of which are probably important to their audiance.
Linus mentions it in the video.
I could be wrong but I think that's what ghosting means, at least, that's always been my understanding of it.
Just get a keyboard with good key rollover
As a robotics engineer, this setup is definitely a decent proof of concept. I'm really excited at the progress and promise of the work! There are some simple changes you guys could try to make the setup more robust, especially for vision. Fully enclosing, specifically with a light box, would really help with consistent lighting as well as integrating the lights. You can also get a 3D image using a 2D camera by stitching together images as the robot moves or using multiple camera angles. Multiple cameras can allow a 3D image also using structured light through a projector. That will be much cheaper $500 USD per camera, than a multi $1k/$10k 3D sensor. Getting a Basler or similar GigE camera allows open source focus control and confocal lenses if you want a really flat 2D image. 2D line scan profilometers are a decent idea, but be careful that robot movements can cause rippling in the image as they move. So take measurements while it is stopped or consider a true 3D sensor (point cloud) as they are about the same cost range as a laser profilometer. Another lower cost option is using a Microsoft Kinect or HP 3D Camera. They used to be off-the-shelf, but you can find used ones. They have limitations in resolution, but might work for identifying the key top centers. Just some attempt at friendly advice that I gained through many hours of banging my head against problems in the industry. I'm sure you guys have already tried a lot of approaches you couldn't fit in the video.
Or use a tof camera to get a point cloud
This. Laser topology seems not only overkill but prone to issues such as metallic reflections from both the robot itself and keyboard backplate (or even shiny keycaps), not to mention high sensitivity to jitter and judder. A rigid lightproof enclosure with two or three distinct camera viewpoints should be more than enough to build a reliable pointcloud with high enough fidelity to work here, or spend the big bucks on a 3d scanner that will not require complex positioning to capture the topology of the keyboard.
i really hope these changes are implemented
why would fully enclosing help with vision? why not just use a coloured light and a matching filter?
@@Meoiswa I am not sure I can agree with you, but you have some valid points
Every Lab update we get just blows my mind. The fact that a tech youtube channel ive been watching for years and years has developed to the point of legitimately quality testing big name brands to keep them honest, and the fact they HAVE the influence behind them to make this possible is just insane
I think that growth is a testament to how business savvy Linus is. It takes way more than being a simple "youtuber" to achieve what he has.
Let's not forget the fact that if they have been lying, they will need to be scared straight because that can ruin their keyboard reputation status. This can make or break their status. Although, they need to test at least a minimum of 3 of the same keyboards because a single keyboard could be a keyboard that failed and somehow made its way into the stores.
Going from whole room watercooling in a house where they didn't have a permit to run the business, to a multi location quality control facilities is quite the change
I love that you guys are always pushing to get better and better information into the hands of the consumer.
You don't have to remind us Linus
😂😂😂 yeah I felt a tad bit attacked in that intro lol
Linus be looking down on us
pozdrav
25 usd keyboard + mouse combo gang rise up
Remind what?
Not gonna lie the production value of videos with all these animations and all the extra tests you guys are able to do is absolutely amazing! Keep up the great work!
You guys are reaching a new level.
Excited for all of your new projects and their results.
Agreed 😊
Not sure if it was mentioned before but one thing you should not forget is that most people won't push the keys 100% vertically so it would be good to at least try a few key by pressing them in an angle like I noticed some Keyboards won't work really well if the angle is to high (usually doesn't happen a lot when you are sitting right in front of the keyboard but sometimes happens when you help someone else who is in front of the keyboard so you have to type from the side.)
That was addressed by Linus and Jake #3 by edge pressing the keys for seeing if keyboarding happens.
@@logandeathrage6945 pressing the button at the side is something different then pressing it with an angle that doesn't goes straight down
@@XYoukaiX who presses a key from the side? Why would they test for something that the keyboard is not meant to do?
@@reaLe-1 As IT support and a "former" gamer that also has friends sometimes sitting next to him I can tell you it happens a lot ...
Lots of people asking for help have their keyboard in front of them and don't even think about making some space for you so you type in a very bad angle some times
Also if you are sitting in front of the pc with a friend (yeah many years ago online play wasn't really popular) and your pc table is small at least one if not both of you will sit not right in front of the keyboard
I love the normal LTT team and their hardware knowledge and wit, but it’s awesome having some videos with real engineers and specialists.
I could be way off on this, but I think Alex had a huge hand in all this happening. Him running the workshop really turned LTT from just receiving and reviewing things to actually building things themselves. From 3D printing to CNC, sandblasting, other metal cutting etc. That made LTT really dig deep into the inner workings of things.
@@CamelSmokes23 Alex was the first engineer they had who worked on physical materials, yes (dunno if they had any software/computer engineers before him)
They were pretty clear his machining skills were one of the reasons they hired him, so I’d say the drive was there even before Alex was hired
@@CamelSmokes23 Honestly, I would give huge credit to Alex here. Not to say he is the reason LTT got into engineering stuff, but when it comes to them building and engineering their own projects, Alex was always the one there running the show. The success they've had with projects like that are probably a huge part of the reason they stepped further into stuff like this.
Force curves for more esoteric switches would be fantastic. Everyone already knows what Cherry MX curves look like, but less common "hobbyist" switches sometimes don't have any graphs online. It would be great to compare some of these boutique tactile switches against each other, against well known switches like Cherry MX Brown so we have a reference point
This would be fantastic
"esoteric switches" makes me think of ThereminGoat's switches collection. The guy has like 1.5k switches, and I can maybe name only 10 (not including colors, ofc). I wonder whether there could ever be some sort of collab between them, considering that Goat doesn't record videos (at least not public, afaik)
P.S.: For the last "family photo" he posted, the thread ID in the shortlink is "tbqrwf"
Also be cool to have some vintage ones as well
They have curves?
@@graciouscompetentdwarfrabbit ThereminGoat's switch reviews are so good! Hopefully ltt labs is able to reach that level of quality and data
Sometimes reinventing the wheel can be tedious. It might be of value to collaborate with iFixit on the teardown schematics and repair parts needed. You both are trusted companies who can utilize each others strengths!
Was just about to mention the same thing. Seems like a match made in heaven.
Thought the same thing. If you're talking about "helping consumers make better decisions and access the information", creating a separate platform for the same information that is established on a household name like iFixit seems counterproductive to me.
I really hope Linus' team reads this and takes it in to serious consideration. As soon as I heard his plans on gathering repairability and teardown data, I immediately thought that we really shouldn't fragment this information. They need to work with iFixit on this.
I'd actually be curious to see how custom keyboards compare to the OEM variety, in terms of consistency and latency. KBD67 vs Razer Huntsman, etc. Another test that I don't think I've seen anyone else attempt is two people build the same custom keyboard, and then see how consistent the two resulting boards are. Can give context about whether we can even replicate builds from TH-camrs using the same specs. Their board may sound amazing, but perhaps it's not even feasible for someone else to recreate the same experience.
Soundwise, it's like almost impossible. Like you'll have to know which lube they used for their switches, what mods they used for the keyboard's case, and what kind of textile is used in their desk mat (and its size). The desk material, density, and size. And a lot more. So it's impossible to replicate someone's else builds. For the latency, I think it is only important if you're an elite rhythm game player, otherwise, a 500hz/250hz polling rate is enough.
as far as latency and all of that, there are a few types of PCB used that could effect that, but most of the time it comes down to polling rate as well as if its usb-c or a.
@@TryTane right? probably keyboard youtubers that do it already
I'm so excited for Labs. Now that Linus finally gets to start showing us what all that time and money went into it's clear that it was totally worth the wait.
Man, if you guys manage to apply this level of thoroughness to all the other product testing you're planning, the LTT Lab could become one of the greatest consumer education tools the consumer electronics industry has ever seen. The potential of this undertaking to completely shake up that space is incredible. If I was Intel, AMD, Nvidia, or any other player of any importance in the industry, I'd be watching your progress like a hawk. This can only be a good thing for the end user and I am super pumped!
This is truly one of the greatest advancements in tech journalism ever. I’m so glad the LTT team has the resources to do things like this
As someone who gets to work with FANUC robots everyday, I absolutely love that you guys are into testing keyboards with a robot. Still learning my 3DL vision, but the lasers help out a lot with depth. Good luck!
The amount of value LTT is giving to the community with CW's great designs like the backpack and screwdriver and now Labs' MarkBench and now this and who knows what else more is astonishing.
Whats CW?
@@antoniohagopian213 It doesn't mean call of duty cold war, CW is a television network that was founded in 2006.
@@vishalmangla4735 Creator Warehouse
@@CarlosCastilloMusic. Was it really not founded till 06? No wonder I was stuck watching stupid PBS until 7 or 8 years old lol
I've always used RTings for this kind of data, and it was unfortunate that no other site had such systematic testing meaning that if RTings didn't have info on the specific device I am interested in, there was nothing else comparable. Hopefully LTT will start to help fill that gap soon!
Hopefully they end up doing a better job... there's a lot of issues with RTing's methodology that's led to wildly inaccurate results in the past.
@@chroma_aus like?
@@seanmccants1500 The latency tests were the most egregious. They changed methodology a few months back, but previously they had an audio-based setup that produced results so bad that they weren't even internally consistent. I'm talking the same peripherals achieving lower latencies wireless vs when plugged in, massive deltas between models with identical PCBs, Bluetooth devices that scored competitively, etc.
Even now there's a bunch of issues in the technical aspect of their reviews. I won't go into all that here since it appears they're trying to improve, but I'd just take everything from them with a massive grain of salt.
@@chroma_aus uhm, there are devices that have lower latency when used wireless vs cable. using a Dualshock 4 on a PS4 through a cable connection has higher latency than through Bluetooth for example. the same is true on the Switch and its pro controller.
so I wouldn't instantly say that data like that is automatically inaccurate just because intuitively you think a wired connection has to be faster, there can be complications in the software or hardware that can lead to that not being the case
@@kevboard idk what you're smoking but provide a source for bluetooth of all things to be faster than a wired connection
So stoked for Labs to start pushing out data for some epic review videos! So excited with the "new" direction!
Hey I also work as an ML Engineer in Computer Vision for product line quality control and I would say your laser scanner is totally overkill. I mean if you have the money, go use it but just for identification of the keys you can use grey scale images with some smoothing and I would guess just edge detection to find all the keys pretty reliable and super fast. Only reflection / polished keys might be a problem. I think if you open source your key identification algorithm you will get help fast an easily
More than overkill, it's like someone doesn't even know what IOL is. As for reflections, image processing on multiple images with different light sources will sort that out, and remove most of the shadows. Most of the code needed is open source, and a decent IOL setup will cost $30?
@@kaduseus593 what's IOL?
Pretty sure he has the money, and wouldn't be surprised if the laser scanner wound up having multiple purposes in the long run.
OTOMH, couldn't they just also invest in another webcam mounted opposite the one they already have, and do some comparative stereoscopic wizardry on the image feeds to determine distances? Seems like a good crossover between cost (versus a laser) and algorithmic error rate (versus an algorithm that only uses one camera and has to rely on ML to discern shapes).
Come think of it, they could just use the same one camera and just take multiple images from different XY positions to achieve functionally the same.
(I don't know if that's what you guys in this thread were talking about, or if some of what I said is what they were already doing - this is all admittedly beyond my expertise.)
Couldn't keyboards where you can't turn the RGB off also pose a problem for a camera-based system?
I've been following LTT for I think a little over a decade now, ever since I built my first PC. You guys never fail to make great content, but honestly, the LAB is on a whole different level. This is definitely my favorite initiative I've seen from you guys so far.
Respecting the amount of work they're promising as I definitely look forward to it. Been modding keyboards since last year but will agree that my sound recordings are inconsistent as other content creators also mentioned the inconsistencies they have since it has a lot of factors to really consider but nonetheless with the goal for the website this is aiming for, I am feeling great about this!
I CANNOT WAIT for LTT lab's headphone testing and reviews. So many mainstream reviews focuses solely on features, completely skipping on sound. Having detailed, but accessible information about a headphone's sound signature will be a fantastic push in the right direction for headphone reviews, headphone marketing and manufacturing alike.
My dude RTings is waiting for you. Crinacle for in ears
Don't forget Brian as well, BadSeed Tech has solid info.
So many reviewers are already giving frequency response graphs...
Every reply here seems to have missed the OPs point about this information being available on MAINSTREAM channels. Many of the other channels mentioned have fantastic content regarding audio equipment, but none of them can reach the audience that LTT can.
You can tell when the talk is going over Linus's head. But the math and science of this process is amazing!
Im so so excited for everything related with the lab. So much work to do. You are advancing tecnology with this videos/articles
Honestly so excited for this and other projects that LTT Labs might have in the works, you guys are guiding the industry in such a positive way!
if their peripherals testing is anything like their gpu testing it will be full of errors and bad methodology. hopefully they can improve their own quality control in the time between now and then.
@@rdiznfriends You should see the new lab tour
I would also love to see laptop keyboards being tested as a lot of people use them as daily drivers.
Damn! The passion and ambition which needs to do this is just unparalleled!! Keep the good work going!
Really glad to see someone quantifying this marketing lingo. Props to the Labs team, very much looking forward to the future of this channel.
The interesting part is with the 3D laser scanner, you could actually have the robot change the angle of the finger to match ergonomic keyboards. This would help with the testing.
@@RetroDadMTG de Janeiro?
Hopefully laptop keyboards will be included eventually. It would be nice to see some classics like an IBM model M or other never die keyboards from the 80's and 90's too.
I'm a bit worried. Are the Model M keyboards really that hard to kill? Because I killed one through regular typing. I'm quite a heavy typer.
Even just a take on the modern model m remakes would be fascinating!
yes, i really want to see laptop keyboards too
Love to see that Antoine comes from the same french school as I am right now. I'm feeling a little better about my future thanks to him :)
That stuff sounds amazing, I can't WAIT to see it in production!
The mere fact that this is being served on the main channel shows how much LTT has shifted from a couple guys doing gaming reviews to a MAJOR enterprise - almost on par with Consumer Reports.
Kudos for your hard work and drive to filter out the BS and give us real data to work with.
What you're doing for the enthusiast community is SO awesome, I cannot wait to see this in it's final form!
if youre talking about the mechanical keyboard community then this is barely scratching the surface
@@Infinity2437_ seriously?
this is more for the normal consumers unfortunately.
Enthusiast would be about building the keyboards, rather than just buy prebuilt one like this
@@eugenehong9668 Yeah, stuff like repairability, how hard is it to get inside etc isn't even a concern in the enthusiast keyboard space, considering most of them come as a kit or are intended to be modded/customised. Only the consumer keyboards have these concerns.
Given you were holding a Keychron, I'm hoping that eventually you guys test many different switches + keycaps as well. This will be tailored towards the Mechanical Keyboard enthusiasts, but would be amazing for data verification.
I can't wait to see more topics about latency in general. Like input-to-display, which is accumulative. Affected by input, game engine, engine settings, drives, drive settings, monitor, and monitor settings.
I like the idea of having a database of information to decide your keyboard, but I think as a product, I would actually prefer that there be some list that is generated of recommended keyboards based on filters set by the user. Not to mention, some kind of "Amazon's Choice" branding on products that LTT deems as a good product overall would also be helpful in making safe purchases.
This is just so amazing. I don't know how you are going to make labs economically viable but I hope they will be.
I believe the lab can become the standart to be tested. So electronics producers can send their stuff and get the labs’ seal of approval.
@@judassson I wish that was the case. Maybe it will be. If it will be it will take a very long time to get there anyway.
I think Linus said that the labs is mostly going to be a loss, and is mostly subsidized by earnings form regular ltt content on wan show. He also went into how most of the stuff from the labs is going to be write ups and articles on a website, and how the majority of written media is dying out because it isn't profitable anymore.
I love how he is keeping us informed on not just current projects but where they are protected to be heading to. It's super interesting and i wanna see more
I missed the 2014-2017 linus tech tips, looked genuine
glad to see the redragon keyboards in the mix
those are my favorite for no real reason other that I liked their mouse so I got their keyboard and liked it too, so I'm excited to see how it holds up to the competition
My K551 is 6 goin on 7 years and still choochin, although the lighting is dimmer (this one is the non RGB one just red)
and my RGB K552 is 5 goin on 6 years, ad the RGB is still working fine, and still good brightness
I have one on each desktop and they have gotten constant daily use that entire time
Felt mildly attacked by the intro but loved this video 😅
The thing about latency with USB keyboards is that USB has to be polled, it doesn't use interrupts like the old PS/2, so at this point you're testing the polling rate, which can, and will be limited by the host more likely than by the keyboard, if you hit a key between polls, it'll have to wait for the next refresh. And not to mention, if you're polling way too much, it will tank your CPU.
High-polling-rate peripherals are a thing and implemented through the driver. If a manufacturer didn't bother to tune the firmware and driver, that's a legit data point.
@@Steamrick the normal HID class will have 1000Hz a maximum polling rate. Which still gives you the 1ms window. Theoretically you can go faster with your own driver, but not sure if anyone does or if it is worth it. And 1ms is an eternity so I would be really surprised if keyboards really differ given the polling latency. Of course statistically you still can extract that latency but probably not worth it...
@@lal12 There's various keyboards and mice with polling rates up to 8000Hz. I don't really see the point since it's already pretty far up the path of diminishing returns.
If and when possible you should try the old butterfly keyboard from apple, it would be interesting to see how it compares to the spec list
Or even other laptop keyboards because these are permanent and not replaceable
Also, have you thought about testing how many clicks it takes to break the key? Manufacturers claim millions but i dont think in reality this is possible.
That’s such a great idea! I hope they see this
Here's another vote for them to see and do this test.
Why shall millions of clicks not be possible?
Have you ever though about how often you really are pressing the keys? And how long a keyboard lasts?
Then there is the matter of dust and other dirt getting into there - what if this was not the case and the switches would wear less?
People should upvote this. Upvote it to the stars and beyond. This needs to be seen by @Linus Tech Tips (won't work but worth a shot)
@@dezpotizmOFheaven thats true but a full keyboard has more than 80 keys and if 90% of them survive, what do you do about the other? Its not like you can replace the keys easily in most keyboards. After reading my comment again, i think the failure rate before reaching the spec would be a better think to see rather than just some keys.
Congratulations on tackling the very difficult task of testing a crucial human-to-machine interface. As a fast typer (500 chars/minute) I have stuck with cheap wired, low-noise, short-throw keyboards. There seem to be very different preferences between gamers, people who touch type and those who don't.
I used to do industrial computer vision. I'm curious if you guys tried low angle lighting instead. It could help make sharper shadows to have more distinct edges of the keyboard. It's a common technique for creating shadows for aluminum components with serial numbers etched on.
The one thing I can see being a problem with this is that keyboards have so much variation in height - lighting and capturing just the keycaps, and the keycaps alone, would be next to impossible. On flat aluminum pieces to illuminate an etched serial number it may work great, but with keyboards I believe it wouldn’t work that well
@@theEduEnthusiast Additionally, not all keyboards are flat, some have built in angles which a 3D scan could pick up best.
I'm surprised they didn't just use the printed keycap symbols as a reference, in order to reduce uncertainty in the location of the edges i.e if the symbol is detected to be outside the bounding area of an inferred keycap surface, then that results in a lower activation/probability in the ML algorithm.
Detecting a defined set of symbols is way easier and more reliable than detecting the edges of an arbitrary shape.
@@tomfahey2823 It's unfortunately not as black and white as that. Not all keyboards have symbols on them. Not all keyboards use the same typeface for their symbols. Not all keyboards put their symbols in the same place on the keys. It's a surprisingly tricky engineering problem.
@@Sam-dn7jk ah yes that's valid. There's a technique I know that can bypass this issue but then we'd be talking more expensive than a 3d range scanner
Excited to see potential variances across all the keys on a board - especially gasket mounted boards. Would be pretty cool to see bottom out force graphed as a gradient or something like that, spread across a whole keyboard
edit- also force profile consistency of hand lubed switches would be pretty cool to see as well. both immediately after lubing and after a period of heavy use
It surprises me how much this excites me to see the reviews for all the equipment you guys test
The labs content is sooo cool! When showing Frequency Response graphs, could you maybe include what smoothing the graph has?
Would love to see the latency tests extended to mice (I assume you already want to do this), but also hubs/docks and KVM’s, like the USB fiber hubs Linus uses in the new house, or the Level One Techs KVM’s 😀
Man I'm really really pumped to see what people like HipyoTech think of all of this, I feel like this is gonna be ground breaking for all the custom mech keyboard youtubers/builders/reviewers as it's shedding actual concrete data on so many areas where people like Hipyo were previously considered experts just through their vast experience with so many keyboards, but their experience is all from their own testing, and their tools up until now has just been their own hands and knowledge from trying so many keebs, but now we will acutally be able to compare the information we've been getting from people like Hipyo to the actual concrete data we'll see from LTT Labs. I for one am very excited about the future, LTT Labs is so fucking huge for the industry man, grounds will be shook.
Side not: Linus suddenly changed how he says Asus..? He said "AH-soos" in the sponsor at the end, I swear he usually says "EY-soos" lol
Probably treading carefully after that ring BS
Howdy hey
Honestly, I love the idea of the lab, and it's gonna be amazing for a lot of stuff, but I'm not sure if it's fitting here for keyboards.
Chances are, people nerding out over keyboards won't be interested in prebuilt keyboards in stock configuration.
It's just a scenario far from any sensible use case. It's like testing current gen graphics cards with random CPUs from 2003.
Linus is correct in saying that, especially given sound, we lack consistency in reviews and sound tests.
Offering a consistent setup is definitely the way to go, but this needs to be paired with a consistent, true to real world, build.
Of course that's a task far from easy, but results matter so little this way.
This project will keep manufacturers accountable, which is good, but has no relevance for the enthusiast community.
The one (easy) avenue I see is testing switches. Friction could possibly be measured, force graphs of course, wobble and binding.
In the end, the keyboard community has to mature towards more scientific methods, "uh listen to this / woaa feels smooth" is just pretty much useless information.
@@lennihein Yeahh I see where you're coming from, but I think there is still value in this for enthusiasts, as you say testing switches for example. But also, not everyone wants to go down the custom mech keyboard route, including some enthusiasts. I think there's a pretty big subset of people who are super into keyboards, who follow people like Hipyo, and who love the whole custom board space, but who don't necessarily have the inclination to go buy all the individual components and build one themselves, I think quite a few people are very enthusiastic about mechanical keyboards but would prefer to just find an ideal keeb that they like, and then maybe make some minor changes like swapping keycaps, or maybe they find a board they love the look of but it doesn't have the switches they want, but the board is hotswappable, or in my case for example I bought an Anne Pro with gateron browns, and I cranked up the thock by applying superlube myself. As much as I find the idea of building my own mech keyboard very attractive, it's also expensive, and is quite an involved process, I think for people like me, these Labs keyboard tests will be useful, because I would rather buy a semi-customisable pre-built and then maybe just swap out the switches to something I like the feel of better, and maybe change the caps for aesthetic reasons. For this kind of thing, Labs will be super useful, because once they have reviewed all the top-selling pre-builts, and all the major switches, then I will have actual data to look at to decide which keyboard I want, and what switches I might want to put in it.
But yeah basically my point is, I think there's a pretty big gap where _most_ people reside in between "i don't care about keyboards" and "i'm a custom mechanical keyboard nerd". I think probably _most_ people are into keyboards but aren't likely to ever go to the trouble of building one from scratch, sourcing all the components, researching every different aspect etc. Basically people who want the end result but without having to put all that work in to ensure they buy all the exact right components for them, as it would suck to set out to build a custom board only to find out you bought the wrong parts, or didn't know everything about one or more components until after building it etc...
Hi Kave! I'm so excited for these results!
Dear LTT Team and especially Test Engineers: Please keep up the great Engineering and put your great Engineers into the spotlight. Antoine was perfectly fine to be understood - no need for captions.
With such an in-depth research and documentation, it reminds me of the screwdriver project and makes me think that LTT May release a keyboard in the future using all this data.
I hope that you guys also do historical keyboards that enthusiasts go after as well. Like the fabled IBM Model M, and the Apple Extended Keyboard II. Just to see if those keyboards are worth the hype and to see if buying a 30+ year old keyboards for modern applications are worth it.
I don’t see how testing the actuation force of an old keyboard tells us if it is “worth the hype”
One thing that may have been interesting to add is how you baseline and do routine calibration checks on a system like this to ensure your testing method isn't drifting.
The more LTT Labs content, the better!
Keep it up!
You are doing a great work in defining a standard for these products, I’ve been skeptical for expensive keyboard for sometimes and still think paying over 150$ for a keyboard(unless it’s a custom build one) is not gonna get you that much ahead.
The first guy was so knowledgeable that Linus mostly shut up and let him explain. That was neat to see.
I love what you guys are doing with the labs! This is BADLY needed.
I think a really important consideration for sound is not just the sound of a full key press, but also the sound of tapping the key without actuating.
Most of the sound of my keyboard (old membrane Dreamcast keyboard, my Keychron K2 is in the mail) is the sound the keycaps make when moving, and the membrane switches are nearly silent.
In contrast, my laptop's builtin keyboard has extremely quite keycaps, but the membrane switches on it are much louder.
Knowing that keyboard X and keyboard Y have the same switch sound, but different keycap sound will be very valuable, especially if the two keyboards are effectively the same in terms of price, performance, feature set, etc, covering the remaining difference of keycap sound will inform potential buyers of which keyboard they might prefer. Personally, I can't stand the noise of keycap rattle, but for some people that extra bit of clackiness might be a benefit.
The amount of software and hardware development that the lab requires is pretty daunting. I wish you guys all the best.
I would recommend doing the second test concurrently with the first one. That way, you can analyze the total latency delta across all the keys, see if there are any issues there.
I am so incredibly excited for this whole project!! LTT Labs is going to be incredible
I’m glad to see someone doing this type of thorough testing, and I’m really excited for the results.
Also, I will totally forgive you if you don’t finish it all by the end of the year.
Imo, it’s better to do it right and with a smaller set, than to rush it. But I’m guessing it’s a time is money situation.
Keep it up!
Will you be testing individual switches? This would make the lives of people who are into custom mechanical keyboards alot easier
@Michael Mazurczyk I dont think it is possible to show differences between custom keyboard and off the shelf keyboard through videos, you need to actually use and feel it to understand
this is actually insane and I love it. such an incredibly great idea. it's expensive but doing this for like the cables (the website I mean, having that info on this website) and other stuff (Psu stuff since you have a psu tester too) would be amazing. I'd love having a trustworthy website to look up products on
if the only thing the lab ends up being is iFixit for keyboards, still the best timeline.
I will be super interested to see how these latency tests go, and I'd LOVE to see you guys eventually doing the same thing with video game controllers via usb and bluetooth as this information is SORELY lacking for all kinds of inputs.
I'm also interested in nkey rollover testing as it was a big deal, back in the day - but it really doesn't seem to come up anymore.
between this and the GPU testing software i can truthfully say that i feel a little proud of of linus and the LTT team, keep up the great work guys 👏👏
BTW, recording and publishing the USB and HID descriptors of each keyboard would also be very useful.
Thank you Linus & Team for putting so much effort into publishing all that data!
"Your Keyboard Sucks and I Can Prove It." - original title
How does a keyboard create negative pressure?
I expect there's a select number of users who would actually buy a keyboard, specifically, if it 'sucked'.
pressure*
@@SuicideNeil haha yeah, was going too quickly on that one, thank-you :-)
This is actually interesting!
Hyped about the results.
Some of my Keys on my Razor Widow have a delay. I cleaned everything but it didnt fi the issue.
Kind of afraid to take it more apart. It was expensive haha
Its always a good day when Linus uploads ❤🔥
Poor Antoine got subtitles! My Canadian ears can understand him just fine. lol
My french ears too 👌
LOL it reminds me of that old skit with the news reporter
New Jersey here understood him just fine.
This looks insanely promising. I'm also hoping that they will not only cover the big brands but also do their best to find the hidden gems and no name brands that make good quality products
I use Podoro Keyboards and Mice. Small company, good products.
that's why i also think they should have a suggestion box on what kind of devices or brand they should test on labs. just the back end might need to be a bit smarter with the charting...
Liked for the intro, stayed for the promise of quality keyboard benchmarking. Keep up the good stuff LTT!
great works guys! this is exactly what we need in every industry, unbiased testing for the consumers.
The googly eyes are how you KNOW you can trust KAVE.
Kave is now my favorite LTT member
It amazes me how many different keyboards are out there now! Back in the 1980s there were just two kinds of keyboards: IBM typewriter style that worked well, and the type with non-standard keys that didn't work well at all. Thank you Linus for setting up your labs, this kind of info is needed, there are so many scam companies selling junk it's impossible to know who to trust.
Is an idea for data visualisation:
On the website you can have a tool where a user can select his two most important aspects of a keyboard (f. e. latency and preassure).
Then it will plot a graph with this two aspects on x and y axis.
And in this graph, all your tested keyboards will be presented as dots.
Then the user can see which keyboards perform the best in the two aspects he is interested in and can visually compare alternatives
How about 3 important aspects for a 3d cloud to really see trade offs
For key location detection, could stick an IR reflective dot to each key, and segment the IR image very easily. No lighting issue, cheap. No fancy depth laser scanning needed.
... could use a IR reflective pen to just draw a dot on each key. Even faster.
Think I'm talking to myself, but hey, use a fixed size dot so it will also tell you the distance to the key surface, knowing the camera FOV and resolution.
That sounds like a lot more work in the long run than just having a smart laser vision system. Probably cheaper and easier in the short run but once the hardware has been purchased and the vision system programmed they can just stick a keyboard into the robot, hit run and walk away saving time and potentially money since they wouldn't need to purchase an pen or stickers.
@@grn1 yeah, this would require a couple minutes at least per keyboard, 200 keyboards means 6-10 hours of sticking dots on. Whereas computer vision, while it may take longer to program, results in code and systems which can be reused across many projects and could potentially serve for years without maintenance on a dozen testing setups, saving months of labor.
You guys at LTT labs are literally going to compete with VESA and USB forum with the amount of industry standards created! Love you!
Detailed testing outside the manufacturer's control environment is important..it really drives manufacturers to create quality products paired with accurate specs that are used for marketing. Thanks Linus!
Out of curiosity, why not combine the actuation testing with KAVE with the latency testing to knock two tests out at once?
yes, totally agree on that. I'm also wondering, why not using a small programm on the computer for key pressing detection that interacts with the key pressing microcontroller to determine the time. Measuring it with the monitor while running a game seems really dumb. Also there is a potential flaw: I don't think that they will test every key for latency (because that would be too much work with their current setup), so it could happen, that some keys are faster than others. Combining it with KAVE seems the only good idea.
For the sound library, you should record every key and make a "simulator" where the sound plays as you press the button on your keyboard
I wish that you would test all the switches and graph force needed. Too bad there are so many switches.
I highly doubt that they'll even bother with all the custom switches and just do the "popular" ones that are found on majority of prebuilt keyboards, ie. Cherry, Kailh, Gateron and Outemu in their red, brown and blue variants. Maybe sprinkle in some Razer and Corsair switches.
I can't wait to see the full tour of the LAB once it's ready and the data that is being collected by the LAB crew. In the future I can see this LMG project as a "go to" place for a lot of consumers that like to know what exactly they're getting for their money. In even more distant future, we may also have "LTT Certified" stickers on the boxes... who knows.
Very ambitious project with very consumer focused goal - you're hands down the best!
Having an "LTT Certified" scheme might seem like a good idea at first blush, but the devil is in the detail; done wrong, something like that could give rise to concerns over impartiality (say, if manufacturers had to pay LTT to do the test in some way, could they be accused of bias in the test results?).
But if it could be done in a way that ensures impartiality and integrity of the results, it would be awesome.
@@mnxs I see your point but the same could be said for the current sponsorships that LMG has. Also, you can have the same "fear" with any other certification on the market, so I don't consider this as an problem at all.
Will you also provide the 3D-scan of the boards? This would be useful for designing 3d printed accessories and since you will be already scanning each board it shouldn't be to difficult to pack these as point clouds or stl files.
seconded!
I would love to see y'all test an original Model M with this. Curious to see if a PS/2 connection has any tangible difference in latency compared to a modern USB keyboard, because I honestly can't tell after using my '87 on my gaming rig for the last 5 years 😂Only time it ever let me down was the GTA: San Andreas driving school, where you need to hold down 3 keys at the same time
I think the difference is that PS/2 input (at least used to be) handled differently from USB input on OS level (PS/2 seems to be interrupt based). I managed to find one comparison study on PS/2 vs USB latency, but it was done in DOS in 2002.
Other than that I'm pretty sure that nowadays with >1kHz polling USB either outperforms PS/2 or has negligible difference
Gaming wise, not a major difference. Most game engines poll the inputs rather than get "interrupted" at the system level. So even if the PC itself knows about the new updates, the engine doesn't immediately grab and act on it until the next "tick".
You have such ambitious goals with this, and I'm really excited to see this list of hardware come to be. I genuinely think this could have a large impact on the industry, and if not at the very least it's going to be ultra helpful to so many consumers. I'm really glad you're reinvesting so much of the money you make to do this, and it just makes me hope you'll keep succeeding with everything you're doing. Obviously none of this would be possible without an excellent team like yours, so thanks to each and everyone involved, again, I'm very excited to see this.
i'm planning an upgrade next year, and all this things you are doing will really help, thanks.