Gadgetory

we bypassed the problem we had earlier with the 13 50 megahertz lock on one of our tighten our TX cards and now we can properly test them in SLI so these are the same device they're both working this time and they come out to about $5,000 total between the two of them we've done this test before with our TX 2080 TI's and these really aren't all that different we have a review of the Titan r-tx on the channel if you want to see the performance differences between one of them and one of the 20 ATT eyes you can check that review of it today we're really just seeing how far does it go when you put them in SLI which is what mblink is and we look at primarily games with some additional focus on power consumption just just because it's fun to see how much power system can draw under a sort of maximum high end GPU configuration available at present before that this video is brought to you by us and the gamers access store you can go to store documents nexus net to pick up one of our ceramic mugs critically acclaimed mod mats or educational video card tear down and PCB anatomy posters that teach the names and placements of all the key PCB components learn more at store des cameras XS net or click the link below just as a reminder these are really devices that you're gonna use for things like production tasks which is not what we're testing we're testing first with gaming and our previous test was with overclocking power acoustics noise and thermals and the point is really I mean that's just it's what we want to do it's a bit more fun and we're gonna put them in a higher end overclocking setup at some point in the near future but we do hope to eventually look at these from a production standpoint it's just we're waiting on some of the software that we would use for production testing to update so we can leverage our TX and the RT cores properly or the tensor cores as it may be so that's something we're waiting on the reason these are better for production than just straight gaming is because ultimately it's the same GPU it's still tu on it too it's just there are four more s times enabled on the Titan then on the 20 80 TI and then secondarily the only real difference beyond that which that was an insignificant one is that it's got a little more than two times more memory so it's 24 Giga bytes instead of 11 and that is a significant change and one which will impact things like for example blender where blender if you're rendering on the GPU you might run out of GPU memory with some specific project files but we're not testing any of that we're testing gaming so now that you know those differences and where these should be applied we can look at where they shouldn't be applied which is in a gaming computer with $5,000 worth of gilded video cards inside of them Sniper Elite 4 comes up first running at 4k high with DirectX 12 and async compute as always we select our game benchmarks based on what they illustrate carefully choosing each one sniper is the best representation of a well optimized dx12 title with asynchronous compute with this configuration sli tighten our TX cards end up at 219 FPS average just ahead of the RT x 28 et is an SLI which run at 210 FPS average lows are well timed on each device but we'll look at frame times momentarily to see better what's happening underneath the hood the 1080 Ti is in SLI end up at 170 FPS average also strong performers in frame times overall we see scaling of about 96 percent over a single Titan RT X cards 112 FPS average the r-tx 2080 Ti is posted roughly 94% scaling over a single RDX 2080 Ti so the Titan is consistent with our previous scaling findings a well-built game will clearly give plenty of room for scaling although it still doesn't make any financial sense to go with to tighten our TX cards over to our TX 28 ETI cards and even that's questionable frame x should be more interesting to look at as always this gives you a look at frame to frame experience and illustrate to any potential spikes or stutters that would otherwise be averaged out we care about lower frame times but more importantly about consistent frame times with minimum deviation from the mean more than an 8 to 12 millisecond deviation will be detectable as a stutter by users and anything within that range is acceptable with the Titan RT X cards in env link we immediately see a dense frame time bar that takes back-and-forth in an oscillating pattern this isn't an ideal bar every other frame is half the speed of the previous delivery which may speak to alternate frame rendering patterns we're seeing frames at 3.2 milliseconds than 6.3 milliseconds in 3.2 and 6 and so on fortunately this pattern isn't highly noticeable to the user as it's really just three milliseconds but it's certainly less smooth than a single Titan r-tx a single Titan r-tx has frame time variation of less than one millisecond per frame which is incredibly good its industry-leading you can't get much smoother delivery than that the single 20 80 TI looks similar with a variation frame to frame of less than one millisecond on average SLI produces a similarly chaotic pattern for the 1080 T eyes although a less chaotic one for the SLI 28 ET eyes than the Titans it's still less consistent than a single card but ends up somewhere between the dual Titans and a single 28 ECI for consistency the chart is getting a bit illegible at this point so we'll cap it here wherever you see the thicker lines that's an SLI configuration anywhere you see the thinner lines is a single device F 1 2018 gives us the opposite perspective instead looking at a well-built DirectX 11 title this represents most of the games on the market presently and in this title the SLI Titan r-tx cards operate at 175 PS average with frame time scaling minimally from a single card the single Titan r-tx does 110 FPS average with lows at 51 and 29 as opposed to 67 and 28 for the SLI configuration demonstrating limited gains in the frame time department at SLI the Titan RT X cards hold the lead of 3.8% over the dual 28 e TI cards which is overall unimpressive scaling versus the single Titan RT x is 59% that's better than a lot of games when running SLI but nothing close to what we saw with sniper as for CPU limitations we know from our 1080p test that this game can minimally push 213 FPS average so we aren't hitting a limitation there just yet far cry 5 @ 4k places the Titan RT x NV link cards at 114 FPS average scaling about 5.3 percent of ahead of the MV link RT x 28 e TI cards or about 25% over the dual 1080 ti cards as for scaling over a single device the dual Titan R T X's end up 62% ahead of a single Titan RT X with dual 28 ET is scaled about 65% over a single 2080 Ti and dual 1080 T is scaled about 59% over a single 1080i 1440p Dean shows us that we weren't hitting a CPU limit in the previous chart although it's starting to emerge here and we link Titans do 143 FPS average but scaling is now limited to 12% over a single Titan r-tx it's at this point that we're clearly hitting CPU bottlenecks we won't bother showing it but at 1080p the limit was also 143 FPS average so that does confirm a cap for the CPU GTA 5 deserves a bit of a preload before we get into the numbers if you missed it previously we discovered an issue with GTA 5 where i5 CPU is would bounce off of the frame rate cap at 187 point 5 FPS average it did so in a pretty rough way that's resulted in higher performing i5 CPUs getting punished for their high performance because they'd stutter hard with each hit to 180 7.5 fps but if you pushed the frame rate down intentionally making the graphics for example higher quality that would result in a better gameplay experience despite having technically lowers fps in the average this was never resolved by Rockstar and the best solution was to just increase the graphics quality until framerate became worse on average a higher framerate gave a worse experience that's seen again here despite using an i7 8086 K at 5.1 gigahertz or so we never saw this in our previous testing so it's somewhat unique at 4k our sli Titan RT X cards are hitting 151 FPS average which would look fine if we only looked at averages in reality the frame times are inconsistent with one percent low as hitting thirty-nine and 0.1% low has gained thirteen point five so somewhere in there were hitting stutters it's instances like this where our 1% and 0.1% low numbers work to illustrate well the limitations of the average frame rate but we still need a bit more we'll look at frame times in a second SLI 20 atti cards didn't encounter the same issue as frame rate was low enough to not trip GTA 5 is weird game engine constraints that we found in the past here's the example of why average frame rate is insufficient for a lot of things in this scenario the chart looks great for the Titan RT X card for most of the test that is until we hit about the three thousandth frame where we encounter 180 milliseconds this means that you're waiting for nearly a fifth of a second to get a new frame which is very noticeable this happens again toward the end then again in rapid succession where we encounter multiple 240ml a second frame times in a row performance is dismal in this title despite overall high averages oddly the 2080 TI is in SL i don't seem to get hit as hard the solution is to boost graphics settings until FPS is lower on average which will help dodge these issues it seems like this is probably a GTA engine bug and is related to the previous bug that we discovered with the i5 cpus finally here's a look at power consumption during our GTA 5 test passes the first set of test shows 1080p results plotting 480 watts for total system power with the SLI Titans or about 330 watts total system power for the single Titan r-tx card that increases to 560 watts at 1440p as we were heavily CPU limited and at the 1080p test and the single card climbs to 400 watts under the same scenario at 4k the total system power consumption for SLI and the Titan r-tx cards is 720 watts with a single Titan r-tx card it's still capped at about 400 watts that's it for this one it's just a it's just a what-if scenario we don't suspect many people are going to be buying these for gaming and putting them in SLI which is what envy link is because in this use case envy link is it works like that's a lie it does AFR it doesn't pull the resources so it's SLI but people are still gonna buy these just to have the best just like they bought the Titan XP for the best where you paying about two hundred dollars per one extra frame over the 1080 Ti at the time of launch so they're still going to be buying these and probably still going to be buying them for SLI if you happen to be one of those people who is considering that purchase we would strongly advise that you consider dwell 2080 T eyes instead if even that but if you must have two video cards for gaming you're talking like three percent improvement by spending two times the amount of money so 2080 Ti is even in in dual configurations we don't fully recommend would be a better purchase and if even if you have a lot of money you can still do other things with that money in the computer so that's that's all we'd suggest that you consider one upside of these though if you are buying them is that it's a twenty eighty TI fe PCB so if you wanted to put a water cooler on this it's trivial you buy a 20 80 TI water block from ek or alpha cool or thermal taker any of them and trivially remove about 30 screws and head and you're good to go but that's it for this one we don't recommend the purchase if you wanted the numbers though you now have them so thank you for watching as always go to store documents access net to support us directly by buying a shirt like this one or patreon.com slash gamers nexus subscribe for more I'll see you all next time