Gadgetory

hey Ron we're back with a quick test so this was just a curiosity on my part but it's a question we ask pretty much every generation of either PCIe or video card architecture and that is what is the impact of PCIe lanes on video card performance so in this case we're specifically looking at this GTA X 1080 just because the fastest performing single card we have right now and we put this into an X 8 slot PCIe gen3 and an X 16 slot PCIe gen3 the way to do this this test is actually very easy to setup so first of all the board we used is one of these X 99 classified boards and it's actually all these are hardwired for 16 if you look into the actual slot you'll see that they are wired for 16 lanes but they don't all actually produce 16 lanes worth of performance on these cards and so EVGA has a manual for this board that does say specifically which slots to use for your video card configuration I use the ones you're not supposed to because we were trying to see how does x8 perform so that's how we set it up you can go into BIOS as well and validate these settings so in BIOS and the top slot on this particular board it will push electrically 16 lanes and it tells you that in BIOS if we move down to the second slot which you really don't want to use for any reason other than something like an ancillary device like PCIe SSD if you move to that slot you'll see that it does push 8 lanes not 16 and that's easy to validate through BIOS and through software in Windows and things like that so that's the setup the reason you would run by 8 lanes I'm not too sure maybe you're putting one of these into a z97 or z170 board or something older ish that is more lane limited than x99 but that would really be the only use case it's still worth testing though always curious about the impact of it and it does give us some Future Sight 2 PCIe Gen 4 and things like that so let's just jump into the results this one's pretty easy and straightforward first of all Metro last light at 1440 key with very high and high settings and tessellation the PCIe gen3 by 16 lane assignment pushed 96 average 74 1% lows and 70 0.3 0.1 percent lows compared against the by 8 assignments 95 73.3 and 67 so the difference between the averages is 1.0 5% negligible but measurable and the difference between the 1% lows is 0.95% so basically there is a 1% difference in this game between an eight-lane assignment PCIe gen3 and a 16 lane assignment for this particular game before moving on further that would mean really you have nothing to worry about if you're concerned over whether to put your high end GPU into a 16 or 8 Lane slot if you're choking on lanes somewhere else like SSDs which really should be pulling from the chip side anyway but you get the idea if that is a concern don't worry about it because it's 1% difference but let's move on the next game the next game is Mordor shadow of Mordor and at 14:40 we're seeing 108 average versus 107 average with 8 lanes half the lanes you lose 1 FPS and then a 60 point 7 for both 1 percent lows and a very slightly reduced 0.1 percent load metrics so the average difference average FPS difference between 16 and 8 lanes is 0.93 percent again about 1 percent total between the two and that's consistent with Metro last light at 4k for shadow of Mordor we see about the same thing the average FPS gap is 0.7 between sex 16 is 64 X a to 62 point 3 and that's a 1.1 percent difference so negligible measurable but not something you'd ever perceive as user black ops 3 1440p we see the same average FPS we see a two point four six percent difference in 1% lows and we see the same more or less 0.1% lows and moving on to 4k it's a gap of x16 at 71 point 3 X eights at 70 point seven and that is a 0.85 percent difference between the averages again around 1 percent and the 1% Lowe's shows a 1.0 5 percent gap with the 57 point 3 vs. 56.7 moving on to GTA 5 at 4k with very high and ultra settings effectively maxed out with going without going to advanced graphics the by 16 setup it's fifty eight point three by eight is 58 FPS average and one percent lows are forty four point seven versus forty five point three and that gives us a difference of zero point five two percent between the averages and one point three three percent between the one percent mometrix ashes of singularities on screen now but you'll see that it's basically the same thing this was just a short with dx12 but again performance is effectively equivalent so even with a high end GPU the the best performing single GPU we currently have we're not seeing a difference between sixteen lanes and eight lanes maybe if we add something like that r9 390 devil dual GPU single card setup maybe then we would see a difference I don't have one of those on hand but we'll try and figure something like that out as far as a single GPU setup meaning one card one GPU there is effectively no difference 1% difference between 16 lanes and eight lanes and that's really all there is to it if you're doing SLI or something obviously that's a bit different because the way scaling works the way the bridge works we'll look at that separately but for now don't worry about the difference between X 8 and X 16 when you're pairing up your GPU with motherboard especially if you're using a lower end card than one of these which is supposed to be 720 dollars but if you buy it today it's 1100 and is a terrible deal so unless you're using that what is effectively an $1,100 video card right now despite the MSRP well actually even then you have nothing to worry about so that's all for this test it's the same it's the way it's been for the past few generations I would imagine the interface which is PCIe will continue to push the bandwidth beyond what these GPUs can actually really utilize in a sense that chokes the performance in games noticeably it's been that way for ten years or more now and it will probably continue to be that way for some time so as always patreon link post roll video if you liked this random exploration of a curiosity we'll have more content for you as always coming up shortly this week thanks for watching I'll see you all next time