Gadgetory

we gave thread Ripper a pretty positive review for production tasks and now we're going to actually put the things real work not just benchmarking and show an actual use case for this Corrin thread count because depending on what you're doing normally in day to day you might not be able to find one but we've got a good one for it and that is resolving our data crisis we've talked about this in the past so right now linked to our 4k upgrade and just doing more videos in general we're doing something like 50 gigabytes a day we're using about 1.5 terabytes of data a month I can't afford to buy that many hard drives so what do we do you either delete everything in it which sucks because I'm paying someone Andrew to film that stuff and I don't want delete work that I paid for and the other option is find another solution so buy storage don't want to do that I can't afford a petabytes server like Linus or you compress it and you take the approach of building some kind of software or script to compress this data down losslessly as much as possible and then keep it in a much smaller form factor so that's what we're doing and we will be using Red River for that task because it is going to be best for this particular workload will show more of that as we get through this testing before that this coverage is brought to you by cable mod already well-known for their work in custom sleeved power supply cables cable mod is now venturing into liquid cooler tubes leaving with their new a i/o sleeping kits compatible with Corsair and NZXT as of now check the link in the description below for more information so bring everyone up to speed the last time we talked about our compression script was probably in December and basically what it does is it uses the handbrake CLI it's a PowerShell script we wrote ourselves and it goes through all of our files compresses them and does a couple of other cool things that help us with data preservation while also making a ton of room so this can save us something like hundreds of gigabytes per day but the only reason we're not currently pulling that much data back out from being occupied is because we need the system to do premiere and rendering and video editing because the same system that compresses everything that does all the editing and daily production so that's where this comes in I can't keep thread Ripper occupied permanently for this task so what we're going to do is build a system let it compress the thirteen hundred or so files to need to compress and then probably after a couple of days of sitting there doing nothing about compressing we'll be able to reclaim it for other testing and and articles and things like that but to give you an idea of the outcome here we're looking at a reduction of around 80 percent in the space requirement so we go from about forty one point five terabytes one point four terabytes to about two hundred eighty gigabytes per month used just think about that for a second that's that's buying a 4 terabyte well three four terabyte hard drives if I want raid 5 every couple months every two to three months versus compressing all of that and then just being able to use the same dryers for years actually so that's what we're doing and we're going to be nineteen fifty extra taxes 16 cores 32 threads and handbrake really likes threads so what I'm going to do is build this in basically an open-air bench it's the p3 I'm going to be putting this on the floor so I do actually need a case which I normally don't like a whole lot but where you're going to use this case because basically an open-air bench so I need to mount this first okay so we're going to start with getting the board set up and I'm going to use the new Noctua coolers that came in just because they're kind of interesting and haven't worked with them yet so we're going to get a chance to actually install one of those we'll test it very shortly this from our test in mind you eat we did actually test it on paste application is going to be about enough for thread Ripper so before anyone freaks out with a hole that's way too much what are you doing we actually did do contents on the amount that's required to really cover this thing so here's what we're working with before I install this this is the tr4 it's an Nhu 14 I believe NHE 14s so it's got a wider cold plate than the previous version it's a wider cold plate then actually then one of these this is the Corsair H 100 I or something B - there's your normal cold plate size so we've got a much bigger cold that it'll let us contact more area it's not going to be a better cooler than a liquid cooler I don't think it's possible but I'm not that's not why I'm installing it I mean sonic if it's gonna be easy and because I don't want to mount a liquid cooler so these screws come all pre-installed on the thing I didn't I didn't pre do any of this for the knock to cooler and it's kind of like X 99 or X 299 in that way it's just got screws that mount straight into the socket you don't need a backplate retention kit because it's all on the board as part of the socket okay I don't remember exactly how handbrake responds to memory but we're just going to use this G scope kit g.skill sent over some extra of this this is the Trident II RGB memory and this particular kit is 3200 megahertz and it's a CL 14 which is actually pretty damn good so 3200 CL 14 32 gigabytes I don't know that we're gonna use that capacity with hand brakes but we'll find out okay there's our memory oh this by the way I don't like these covers so I took it off so if gotten exposed family if you're wondering that's not how it actually looked what are we next we need the fan for the cooler well that was that was easy so I mean that was the whole point of installing an air cooler I'm not sure if it'll be liquid or not I don't think it will if we're not concerned about noise and stuff but we'll save that judgment for the review what I really wanted to do was avoid having to actually do work it's called a radiator in the case we need to stand off up there oh this is the actual real reason by the way that I took the cover over here off this cover protrudes up and it was interfering with one of our test benches getting in the way right that might be a problem here I'm not sure damnit Asus okay so the one problem with this cooler that I'm seen immediately is clearly we're not going to the top slot fortunately it's irrelevant for a number of reasons one of which is we don't need the top slot there's three other ones and this build is ever going to need the top slot and it's also not going to be a long-lasting build the other reason it's relevant because I'm just going to install Arizer cable and call it a day because we're not doing anything GPU intensive anyway which is why there's a 1080i well that's why I guess that's why we're putting a 1080i thing if I can make this somewhat presentable I'm not going to pull a Paul or Kyle and make it pretty though tell you guys that right now because this is to be functional for a few days and then reclaim all the parts 6ke 4+4 cpu we can't use this power supply I might not even manage these cables I might just kiss I'm not going to do that Patrick built this for us as a display system there's some b-roll so it's managed very well it looks nice and I'm just going to come in here unbilled it all and then not even manage the cable on done super nova 750g - oh one thing I do like about it is they're all Universal connectors so you just move them around based on where you want them all you said I'm not cable managing this I'm going to treat this like a test patch the only difference is it's going to be on the floor standing vertical instead of on our very limited table space sitting horizontal it's me that's the Israeli test right now so we're just going to connect everything as efficiently as possible which does not involve making it pretty okay what's left drive aren't you gonna mount the drive Steve I'll tell you where to put your drive I think we're gonna go with gothe right there now looks pretty good okay that's not going to move got the video card power cables power cables times 3 power cables power memory am I missing anything that's it other than style and class okay so we just need to make sure this works move it into the other room turn it on and let it compress stuff for a little while yes does it stay lucid it's fogging a power cycle yeah power cycle number one white LED over here that means we're good that one down there went from red to white like green okay so the system works now we just got to set up to compress everything so I was the numbers here now that we've had it doing some compressing we used a test clip to get started and I first of all had a 4960 in that system originally in the render machine and that one was completing the transcode in 11 minutes so that was a 4k file to a 4k output same framerate same pretty much everything but some reduction in quality on the placebo side so it looks basically the same but it's a lot smaller it's 56 to 80 percent smaller depend on there is to compress so that was 11 minutes when we upgrade to the Z on a couple days ago which we haven't really talked about yet that's the 12 core 24 thread part from x79 era as well and that is doing it in 7 minutes so it's pretty good reduction from 11 to 7 and that allowed us to do something like regain about 50 gigabytes of data every three hours or so so leaving the script running for three hours if we're at 50 gigabytes to start we come back we've got 100 gigabytes now we have all the original files dealt with and the new ones are just compressed fairly losslessly so that was every three hours about 50 gigabytes and now with thread Ripper the 1950 X we're doing the same test file instead of seven minutes we're doing it in three minutes 19 seconds so it's like about half actually a little bit better than half of where we were with the Xeon 12-core 24 thread part from X 79 era that's also better than the sub 900 X we'll do mind you not nearly as much but it is better and that means that if we assume about half the speed perhaps the time requirement I should say to perform the same task instead of 50 gigabytes every three hours we're gonna be gaining back 100 gigabytes every three hours so running the thin overnight means that when we go for a let it run for let's take a number like eight hours make it easy we're going to be doing a couple hundred gigabytes every night that we can claim back and because we pretty is about 50 gigabytes of data on that machine per day will be net positive so that's good that means we can turn through the 1300 or so videos compress some gain back probably I don't even know probably like a terabyte plus of data I should think I would think it'd be one to two terabytes back for those files we've got a 3 by 4 terabyte drive set up on that machine raid 5 so I think we should reclaim a couple terabyte stuff so it looks good thread refer the 1950 X this is an actual use case where there's something we've wanted to show for a while when Rison came out we spent a lot of time talking to Andy as did a lot of other folks and the problem we had there was there wasn't very good communication between the technical people and the people with whom we have contact that would be like PR or tech marketing and so they probably run into is when we say back when they see a hard X launched what are some use cases for this help us out we can test premier but we're seeing kudos faster so what we do to premiere to make sure that you can actually see a difference in CPUs because blender could is faster what do we do in blenders to difference with CP is how do we make these things relevant was the question we were asking and there weren't a lot of good answers at the time but now that we've had our own time to work on Rison for what three or four months now we've pretty much figured these things out and this was one of the best real use cases where one we can't gpu-accelerated it accelerated in a meaningful way that produces better result and we just kind of trust the cpu more when you're dealing with compression and then getting rid of the source file and - this thing really uses cores and thread more than a lot of other stuff doesn't care about frequency as much doesn't care about memory as much doesn't care by the GPU at all so this is a real-world application of what you can do with something like thread Ripper and that's $1,000 and it's better than the 7900 X for the same task so great use case for it now how many people do this type of work probably probably not that many in our audience but hopefully this opens some eyes as to the question of what how do I do with 16 cores why do I even care about that processor they see this question and it's generally because they're not good for gaming so you look at it I mean they can game but value wise it's not a good value and that's because they're not built for gaming as we've been saying they're built for things like this so this is the real-world use case of where you are most efficient with something like a high thread-count processor and of course you use your cheaper stuff for playing games because the value propositions are so much better but yeah hopefully that opens eyes as always you can subscribe for more patreon.com/scishow gamers next to stuff that directly thank you for watching this experiment I'll see you all next time you