Gadgetory

in the process of benchmarking envy link for multi-gpu it would seem that a beef has spawned on Twitter one between myself Paul and now even J so if you've been following along we are going to be responding to Paul's attempted overclock where and he tied our score and J's actual number one overclock for two GPUs in times by extreme right now so Paul the wannabe economy clock droppin under Fulton prankster will soon be beaten in times by extreme and J will follow and J I got 99 problems but a chip eight one before that this video is brought to you by be quiet and the silent based 601 the silent based 601 is one of the quietest cases we've tested recently laying claim to the crown of silence on our charts and still manages reasonable air flow by shipping with a pair of high quality fans the 601 also uses well-designed panels and unique ease of installation features the streamlined builds includes a well placed dust filtration and is large enough as a box to fit a full ATX bill learn more at the link in the description below alright so for this video the hell what we are actually doing is benchmarking envy link and the goal is to see how did 228 est eyes do versus 210 ATT is versus a single 20 atti it's been a while since we've looked at multi-gpu scaling it's never been that good typically actually 100% of the time the conclusion from us has been don't bother but it's interesting NV link this is a 100 gigabyte per second bridge for the ti or 50 gigabytes per second for the 2080 so that gives us a lot of options and in this testing we will also be revisiting the pcie 3.0 bandwidth limitation question to see if we're finally exceeding that bus so first of all for the street for the stream discuss you probably saw the slider the lower thirds thing earlier but we are going to be doing a rip J or a rip Paul or some kind of combined stream following up our trip LTTE and we'll be doing that on Wednesday the week coming up Wednesday starting at roughly 7:00 p.m. Eastern Time in the US and that will be live on our YouTube channel so the goal for that is going to be testing our hybrid card that we built hopefully it's actually good I don't know but with probably go with envy link and see how it does with two cards you'll see where we place on time spy extreme scores stuff like that be a lot of fun make sure you tune in for that for today though what we were testing has already been laid out we are using OneNote an EVGA card for the card in the first slot and the efi card heavily modified here in the second slot so that's important to note the scores will be slightly higher by nature of higher clocks than a single card but we're talking single digit percentages here like three on average versus a just straight efi card and then we're doing that in NV link so other than that I guess let's get into it ashes of the singularity is an incredibly interesting benchmarking tool for this scenario ashes it uses explicit multi-gpu via the PCIe bus to allow multiple GPUs a very and make a unique feature of dx12 to work with one another this also means that it communicates entirely along the PCIe bus the cards can't lean on the 100 gigabyte per second bandwidth provided to them by NV link or the bandwidth provided by typical SLI setup instead all that data transacts over the significantly more limited bandwidth of PCIe which is limited to 16 gigabytes per second and x16 mode and is already doing work in our Titan v testing and we'll pop that old chart up on the screen we found that the PCIe bandwidth limits were finally being strained again this is with no supporting bridge and it's the only title we know of that really makes use of multi GPU in this scenario for the 28 ET is removed the NV link bridge and just tested them via explicit multi-gpu through the PCIe bus this is the determine at what point we hit the pcie 3.0 limitations with no assistance from outside bridges with PCIe 4.0 looming there's been a lot of talk of end of life for 3.0 or at least the end of its usable life in high end and through systems and ashes we found our maximum forints as 120 7.2 FPS average average across ten runs running in by 16 and by 8 mode which would be common in Z 370 platforms we had a measurable and consistent loss that exited margin of error the loss was about 1.7 percent not a big deal running in by 8x8 we saw a massive performance penalty unexpected actually the cards were now limited to 107 FPS average resulting in a 16% loss and we measured this a couple of times just for parody times by extreme is an extremely useful synthetic tool for this type of benchmark as well and also runs memory pretty hard for the GPUs we ran times by extreme five times each on these cards automated and found a difference of 0.1 9% between by 8 by 8 and by 16 by 16 for graphics test 1 which is geometrically intensive this is way within margin of test variance for 3d mark and produces zero loss of performance between the dual by 8 and the dual by 16 modes part of this is likely because of envy links additional bandwidth reducing reliance on the PCIe bus for fire strike ultra we observed an FPS difference of about 1% it was a 0.9 percent difference in graphics 1 and a 1 point I will percent difference in graphics 2 we ended up running these in additional 5 times for a total of 10 each and found the results repeated fire strike has variance run to run so we cannot with full confidence state that a difference exists but if one does exist here it amounts to about a 1% advantage in the duel by 16 mode versus duel by 8 so other than ashes where we run into a PCIe bus limitation we know that the envy link bridge and even SLI is enough for a by 16 and by 8 configuration to work well without any meaningful bottlenecking and that's why we're moving back to the 8080 6k bench to look at comparative numbers versus our previous benchmark results so that we can compare the envy link version of the 2080 TI's whereas the single version of the 28 guys versus some other cards first up is Sniper Elite for Sniper Elite 4 produced some of the best scaling results as it often does this game is also the best DirectX 12 implementation we're aware of so it's scaling will not apply to all games universally it is an outlier but a good one that can teach us a lot with our usual benchmark settings the dual envied links cards push past 200 FPS with 4k high and hit an average of 200 10 FPS under non overclocked settings this outperforms the stock RTX 20 80 TI fe by about 94% this is nearly perfect 2x scaling and has been rare to achieve in the past years but it's always exciting when we see it because this is what multi-gpu should be like and almost never is versus the overclocked single 2080 ti we saw a performance gain of 71% with the stock 2080 TI is in NV link not bad and overclocking the two cards although annoying to find stability would regain that lead the GTX 1080 Ti is an SLI dat about 170 FPS the next major consideration is frame time consistency multi-gpu has traditionally shown at terrible frames a frame interval consistency often resulting in things like micro stutter or intolerable tearing for this pairing as you can see in our frame time plot the lowest scale pretty well it's not a near-perfect 2x scaling like the average was but it's pretty close as a reminder these plots are to be read as the lowest is best but more consistent is it more important than just being a low interval 16 milliseconds would be about the 60 FPS line just for perspective it's very impressive performance in this game which is more a testament to Sniper Elite Force development team than anything else they've continued to build some of the best optimized games in the space we also tested with Sniper Elite for our ultra settings just in case there was a sea view bottleneck here's a quick chart with those results but they're not too different in terms of scaling Far Cry 5 and the dunia engine are up next these are also show some SLI or NV link scaling support and at 4k high Far Cry 5 plots the r-tx 20 atti single card at 74 FPS average for stock or 83 overclocked loads that stick around 55 to 60 fps and each value with dual cards we managed 108 FPS average posting a growth of 46% over the single 20 ATT is card 74 FPS average that's not nearly as exciting as the past result but at least it's still some scaling at 50% scaling though you can't help but feel like you're only getting six hundred dollars of value out of your additional $1,200 purchase for the lows that we're looking at a 0.1% of 60 FPS compared to a 0.1% of 55 FPS on the stock 20 atti so no improvement there let's look at a more valuable frame time plot as these 0.1% metrics don't tell the whole story in our frame time chart we can see the limitations of scaling although the envy links cards run a higher average they failed to sustain similar scaling and frame time consistency frame times are spike here and potentially more jarring although raw frame rate alone makes up for much of this lost frame to frame interval consistency back to the main chart we also have the 1080i cards in SLI to consider in this configuration the SLI 1080 T is operate at 91 point 4 FPS average with spurious lows bouncing around between 42 and 66 for the 0.1% metric although average is high for averages the overall performance uplift amounts to about 60% over a single 1080p is c2 and outperforms a single 2080 TI fe card of course there will be games where SLI gets you nothing but instances like this will permit outperforming new hardware at the same price with old Hardware shadow of the Tomb Raider is next this is a new game and will eventually host r-tx features in theory but didn't at the time of filming the game also uses a modified crystal engine it's got a lot of issues with envy link and SLI and Nvidia is aware of them as of now we have experienced blue screens of death upon launch upon minimizing it upon enabling TAA we've experienced crashes upon minimizing and seemingly other random crashes fortunately we were able to eventually figure out how to work around these issues for just long enough to run a benchmark just know that the game is extremely unstable with multi-gpu one of the other issues we discovered was constant blue screens with again taa enabled which is unfortunately what we've been using for our full GPU reviews so for this reason we retested the 1080 Ti with taa off as well just for a baseline we did not retest all devices only those which are marked at 4k shadow of the Tomb Raider shows a few FPS difference between the 1080i sc2 with TA a on and off this shows that there's minimal overall performance impact but it will offset our data a bit the 2080 TI fe single card averaged 67 FPS originally with a load tightly tied around 56 to 58 this means frame times are very consistent with a single card multi-gpu got us 147 F he has average and the dual 1080 T is got 113 FPS average these two numbers are directly related as they were run under fully identical conditions for SLI it becomes even more difficult to justify the 20 ATT is versus the 1080 T is not that we fully endorse SLI is a good overall option anyway F 1 2018 also showed scaling and its results envied links 20 ATT is managed 168 FPS average here with 1% Lew is at around 69 FPS the RDX 2080 ti single card had an average of 99 with its 1% lows at 47 when stock the results is scaling of about 70% pretty damn good it's not as impressive as sniper but still a better gain overall than expected for SLI configurations over the past few years as for the 1080p eyes in SLI we measured them at 88 FPS average and 57 for the lows a single 1080 T is C 2 ran at 81 fps giving us a dismal scaling of 9% hell-blade is up next just for a dx11 Unreal Engine title this game has some of the best graphics in the game right now making it a good benchmarking option and represents Unreal Engine 4 well it also did not show any scaling in fact technically we observed at negative scaling with this title which is not all too uncommon with SLI we saw a drop of about 16% in performance with additional jarring and tearing during gameplay doing research for this content we learned that there is a custom hell-blade sli profile out there made by a modder but it's not an Nvidia official profile out of the box that appears that NV link does not work with hell-blade but it also looks like some modding could be used to hack it to work GTA 5 is last this is another dx11 title but it uses the rage engine and has been more heavily tuned for graphics hardware over its three-year tenure it also shows some scaling and averages though not necessarily in low-end frame time performance we posted a 132 FPS average the env links cards as opposed to a 77 FPS average with a single efi GPU the difference is an approximate 71 percent improvement in average frame rate but lockstep frame time performance there is no improvement in low-end performance overall and for dual gtx 980ti cards in sli we managed 117 FPS average for a gain of 83% over a single 1080i SC to allowing it to outperform a 2080 ti when overclocked albeit with similar frame time consistency illustrated in the lows we are nearing CPU limitations in this game with hard limits around 170 FPS average for our configuration there is a results friendly link this was the most popularly requested content in our recent livestream where we overclocked a single card that actually ultra it was super chats non-stop asking when will you test NV link well there it is so I don't know I to give you my genuine opinion on this with SLI at our outlet we have routinely found that it's never been worth it in the past because the risk of things like blue screens as we saw in shadow of the Tomb Raider sure it will be fixed but do you want a game to come out and not be able to play it for a month or play it with one card and wonder if it will ever be resolved if you have that kind of patience cool but you might be dealing with that a lot so that's always been one of our concerns with SLI and then other games just with poor scaling like anything below really 60 70 percent doesn't feel too good when you're spending two times the amount of money on cards and you're getting really something more like 50% more performance not 100% more performance so that never feels great but then again if you're trying to hit we had sown in the stream the other night asking what do I need to hit 240 Hertz frequently or regularly or all the time I mean this would pretty much do it and to be completely fair to Nvidia since the last time we looked at SLI it looks like support overall for the modern games has expanded significantly so there are actually a lot of games out now that do support SLI and this is SLI even though it's called envy link as it behaves it is SLI just with a different bridge that has more bandwidth so it's getting it more difficult to come to a conclusion on whether or not you should actually consider multi-gpu we only tested a few games here it is not indicative of everything if you doom for example for the longest time probably to some degree even now really didn't scale at all it wasn't supported I haven't checked in the last couple months if it's changed cool but a lot of the time these games they just don't get support there there's negative scaling on a lot of games it takes a long time to get support in a lot of games so that's always been the concern with multi-gpu it does look better today than in the past will say that much but is it worth it I we would still generally make the recommendation to go a single GPU because you're less likely to run into those issues that are extremely frustrating when you have not just a $700 card but now a $1200 card that will be doing nothing except for blocking air flow because they're all three slots now if SLI or env link isn't supported in a particular game so that's the concern now it does perform better overall the games we test to do show it in a better light generally although a couple of them were 30 40 % scallion we saw hell or hell blade was negative scaling one of the games had a 9% uplift for the 1080 T is an SLI f1 I think that was so it's a mixed bag whether or not you want to do it I mean we gave you some data but really you need a lot more than we have here you need like every game to get a full picture of it and that's not even the problem the problem is what do the new games do how long do they take to get support so our general recommendations still single card but if you do do already have like a 1080i and you grab a second one for cheap enough maybe used or something it might be worth considering as opposed to for example twenty atti because you can spend half the money and get pretty similar performance or even more just only some of the time that's the big question with multi-gpu anyway hope you like the content as always subscribe for more check out our live stream one more time it'll be on Wednesday coming up Wednesday after this goes live coming up at 7:00 p.m. Eastern make sure you subscribe to catch that go to twitter.com slash gamers Nexus to get updates on if anything changes or to follow along in the rap battle that Paul and I have invested our souls in to go to store documents access dotnet to pick one of our products like the shirt I have on not literally that one or the mod mats patreon.com/scishow sexist help us out directly there thank you for watching I'll see you all next time