Gadgetory


All Cool Mind-blowing Gadgets You Love in One Place

RTX 2080 Founders Edition Review, Overclocking, & Benchmarks

2018-09-19
this is the world ray-traced with our TX on and this is our TX off let's just turn that back on there we go so the problem with some van videos our TX advertisements the cards are out today by the way this is the review is that they kind of show a picture of here's what our TX looks like the SDK ray-tracing global illumination all these things and then the comparisons are the developers just didn't enable anything even traditional graphics effects or at least didn't put as much time into it as they will by the time of game launches so it's not really the best comparison and the marketing leaves a lot to be desired so then we of course turned to games benchmarking looking at games that come out with presumably our TX features and yet here we are a day before launch because that's when the reviews go up and there are none 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 first of all quick announcement we're gonna do a live stream on launch day or actually on review launch day so that'll be the day this video goes alive if you're catching it the day it went up their marketing for this launch its man what did did you guys like forget how to make charts because can we if we get a chart on the screen from Nvidia let's just look at the lines here there's there's a line it says 60fps and that's it that's the own known 262 done known again so yeah that the charts we had a problem with they were completely useless they did not tell us anything about the product they were misleading we don't know if the scale started at 0 or started at 30 and if you're talking about 4k 60 and acting like 4k 60 some magical thing that's never been able to be done before except by your own product last year we kind of have to question the intentions behind them as charts final fantasy 15 s benchmark got updated actually the benchmark specifically how we lambasted for being trash because it was misleading and had all kinds of even after nvidia fixed the while nvidia and square enix mostly square enix let's be honest fixed the hair works issues what was being drawn when there weren't actually things with hair in the screen there was still object : LOD scaling issues and just just general screwy within the ffxv benchmark such that we wrote it off as completely invalid for testing it's not really a game we can trust right now for an Nvidia benchmark so then were left with things like the asteroids demo which is very cool but it's not a game we're also left with the Star Wars demo which is cool but we have no camera control and it's not a game technologically don't get us wrong here the things on video is trying to do are impressive the future of graphics will involve some form of deep learning almost certainly for gaming whether or not that's here today it's a different question but Nvidia is going the right direction and the stuff they're doing is impressive so marketing is making the product look worse than it should because it's kind of just been a really messy disorganized sometimes misleading launch and that's making the product look worse than it needs to but we can't review something on promises it's not how so how a product purchase works either don't buy stuff on a promise but we'll continue talking out stuff certainly as it comes today we're reviewing the twenty eighty founders edition we have a separate video coming up today on the twenty atti founders Edition and that'll have more information on that card so what we're gonna do is benchmark games that exist on the market which he is rasterization only presently because our TX the namesake of these cards is not a feature that you can yet use however the cards are still video cards and they might in fact be significantly better than the predecessors even in traditional games it's getting too taxing methodology and test results so there'll be an article link below check that out for all this stuff plus a bit more in detail for testing overall we picked each game based on the uses on the API uses so they were selected for very specific reasons we have a pretty good overall selection of games to cover most the market and then we also have the settings and the chart headers to check those for graphic settings for overclocking there's a full overclocking section in here but for the game overclock tests we let the fan run at Auto so we relied upon the founders Edition lookup table to do its own fan speed management as opposed to just maxing out going 100 percent and hitting 59 decibels because that didn't really seem realistic so Auto for the fan speed despite being manually overclocked otherwise and that was very intentional one more methodology note here - we're testing with partner model 1080 TI's 10 80s things like that because we think that most of you will have those and so it's the best comparison if we use the reference 1080i and 1080p forints gaps would be wider because those coolers are garbage and the clocks are lower as a result so it would certainly make the r-tx cards look better to use reference versus reference because of the upgrades to reference but we think it's more fair to use board partner especially with dual fan versus reference for the new cards and that's what we're doing far cry 5 uses the dunia engine developed by Ubisoft Montreal far cry 5 at 4k has the r-tx 28 EF e card at 59 FPS average planting it functionally tied with the board partner 1080i that we tested at 57 FPS average our error margins on the graph to find room for variance but even considering that the 1080i performs functionally equivalently to the 28 EF e under these settings the RT x 2080 gains on the 1080 FTW non TI by about 30% generationally bumped up from 45 FPS previously Vegas 64 is in the same boat it's about a 30% performance gain for the r ZX card for reference the RT x 2080 leads the 980 by over 2x and leads at the 980 TI by about 62% overclocked in the 2080 puts that 63 FPS average at roughly equivalent again with the overclocked to 1080 TI here's a frame time plot of the RT x 2080 and the gtx 1080i as these cards perform functionally equivalently they also make sense to compare when considering the advantaged price position of the gtx 980ti s note also that a lot of Nvidia's launch marketing for RTX has revolved around the word experience in big quotes well between these two the experience is identical there's hardly any performance difference across the cross-section of data were showing here and just as a reminder frame time plots are to be read as time and milliseconds between frame presents was 16 milliseconds equating a rough 60fps throughput lower frame time is better but consistency is even more important and here they're both roughly equally consistent so even outside of the sort of dated FPS metrics the cards are the same 1440p is next at this resolution the RTR 2080 Fe runs at 105 FPS average which places it as roughly equivalent with the 1080 T is c2 again lows are also within our margin for error marking frame time does roughly equivalent generational improvement over the 1018 on Ti is 27% the r-tx 2080 has a bigger improvement at 4k than 1440 but not by much as for Vega 60 for this one the board partner model is running roughly equivalently to the 1080 ftw at just around 81 fps so more or less the same given margin of error overclocking the 2080 fe2 plus 100 plus 800 landed us around 111 FPS average with low as reasonably sustained and not suffering from inconsistencies so that's good behavior from an overclock and note again that the offset is not universal so if you see plus 100 in this game it's not the same as plus 100 in another game because the clock that it boosts too could be different and in this instance plus 100 was stable in all games whereas plus 160 was stable in some but not all anyway the 1080 Ti that is ends up tied again even when both are overclocked 1080p is interesting only for scalability purposes in Far Cry 5 at this configuration the 28 efe runs at 138 FPS average or only 18.3% ahead of the GTX 1080 that's a significant drop from the previous gain even at 1440 where the settings were shared at 4k we use different settings slightly lower in fact and that illustrates further differences depending on anti-aliasing settings and we'll look into that more later the higher percent gains would also reflect to lower resolution heavily anti aliased scenarios as you're sampling the screen X more times with anti-aliasing less producing the same amount of load on the render pipeline the improvement to ROPS here as much of the scaling difference between 1440p and 1080p at the same settings Sniper Elite 4 has one of the best DirectX 12 implementations available and was what tipped us off to voltage shift to asynchronous compute performance on the Titan V this game stands as a means to illustrate how modern API is perform on cards with better asynchronous operations scheduling at 4k and high settings the r-tx 28 efe averaged 84 FPS average with a low is at 68 FPS for 0.1% the GTX 1080 is C 2 managed 87 FPS when stock demonstrating again near equivalence but it is one that exits error margins ETI pulls ahead by 2.6 percent overclocking to both the 20 80 and the TI cards the cards again face equivalents we're at 91 FPS for the 2080 average and 89 average for the 1080i with 1% 0.1% lows both extraordinarily similar the GTX 1018 on TI average is 66 FPS allowing the 2080 a generational improvement of 28% not as large of an improvement as the Titan V led us to expect with this generation but still an improvement unfortunately that improvement also comes with a massive price hike and could be matched for a cheaper 1080 TI F 1 2018 is next this game uses the ego engine and is built by Codemasters standing as a good representation for other ego engine games let's start with a different image this is a resolution scaling chart for f1 2018 demonstrating the performance at 1080p 1440p and 4k high with the same graphics settings the demonstration helps us to visualize average FPS across resolutions at 4k the r-tx 2080 is roughly equal with a 1080i which is actually not what we expected we were expecting to see it to begin to pull away at the higher resolutions and that's something we saw sometimes at the Titan v but it does not appear to be reflected in this particular game with these two cards at 4k the r-tx 28 efe performs at 81 FPS average with loes de monster below for all cards in this game this is roughly tied in all metrics with the gtx 980ti c2 it would lead the fe 1080 TI by a few percent but that's hardly impressive and already said why we're not testing that overclocking both the 1080i and 2080 results in another rough equivalence at 4k scaling from twenty eighty stock over the 1080 stock 59 FPS we're looking at about a 30% jump marking the largest leap we've seen generationally thus far interestingly this particular title is a demonstration of thermal frequency response on the r-tx 2080 when over clocks were running into a thermal cap limit and dropping frames pretty hard full fan speeds permit an additional couple of frames per second at the expense of 60 DBA of noise so expects better gains on the partner models here's another frame time plot to help illustrate performance we're comparing the 1080i and the 2080 despite being identified by different names these two cards are the most immediately similar in price again the experience and big quotes is about the same even when we look at frame times instead of frame rate 1440p posts the 2080 at 134 FPS average which is about 1.8% ahead of our 1080 TI performance then again standard deviation of the average is approximately 0.8 FPS resulting in functional and observational equivalence for scaling the 2080 is 36.7 percent faster than the 1080 keeping the same scaling at 1080p scaling drops to 35 percent which is still a big jump this is an instance where the ROPS and FPU jaunt is largely beneficial even in more traditional rasterization for reference the 2080 outperforms the 980 by about 120 percent marking over a 2x gain across a four year period the 988 launched in September on the 18th actually in 2014 shadow of the tomb Raider's next and uses an in-house engine for the developer it's a modified crystal engine at 4k we measured the r-tx 2080 FBE at 54 FPS average with a low as well timed at 46 FPS for 0.1% and 49 for the 1% lows inversely this tightness of latency is a good path to a smoother experience overclocking allows the RT X 28 efe a lead FPS of 5 or at 9% for the 1080 Ti we're at rough equivalence by all metrics that's within error margins the GTX 1080 runs at 41 FPS average affording the RT X 2080 stock card a 33% lead 1440p post the RT x 28 yet 95 FPS average affording it a measurable but imperceptible 2% advantage over the gtx 980ti overclocks we see the same behavior as before the r-tx card plants itself at 104 FPS average or a lead of 6.7% over the 1080i that's not a guaranteed gain mind you and it does cost an extra a couple hundred dollars scaling over the 1080 is 73 FPS average allows the r-tx 2080 a lead of 30.5% and strictly to illustrate this point here's a glance at 1080p for this one the gap between the RT x28 EF e is 134 FPS average and the 10 80s or 108 FPS average shrinks to a twenty three point nine percent difference the lead is dropping against resolution as shown elsewhere note that we also observed a significantly higher standard deviation on the RT x 2080 overclocked card than on any of the others this was a result of the thermal limits being hit it's not shown here but we measured a massive standard deviation of seven FPS on the overclocked to 2084 perspective all other cards including these stock 2080 had a standard deviation of around 0.5 FPS GTA 5 gives us another fully populated chart and runs on the rage engine at 4k the RT x 2080 efi manages 61 FPS average marking it as below the 1080 TI we retested this twice actually and went for 10 total test passes for validation on the retesting for each device even with all the retests the 1080 is c2 is consistently ahead of the art CX 2080 by 5.1 percent that is far outside of our margin for error in this game and is a market at noteworthy improvement in the 1080 Ti eliminating the RCX 2080 from consideration for this game the value isn't there compared to the GTX 1080 FTW it's about 29 percent faster but being outperformed by a 1080 TI overshadows this growth at 1440p we see more of the same the 1080i outperforms the RT x 2080 by 4.3 percent the competition is better except the competition is also made by Nvidia Vega 64 runs closer to the gtx 1070 TI down around 75 FPS average and we'll skip 1080p because it's more of the same 1080i outperforms the RT x 28 EF when using a partner model for the 1080i which what most of you probably have overclock stepping is next you already saw overclocking results but that was her a frequency that was stable across all games we also did additional testing where we found frequencies that were stable in just some applications but higher overall here's our overclock stepping chart for the RT X 28 EF e card we learned a lot in this process full stock we peaked at 19 50 megahertz and average 1875 while maintaining a temperature of 72 degrees or below or on this later the frequency pushed to 1890 megahertz without any offsets when only increasing the maximum power throughput or whatever was allowed by the card at 100% as we began to increase the power offset and the frequency offset we found that 20 70 megahertz was our average when under load and with a 100% fans beat so this is all things maxed max fan max power allowance max voltage allowance although even though Tom Petersson from Nvidia says it's not a placebo slider kind of still is anyway all this is important and terrines frequency under boost 4.0 is dependent upon thermals just like pascal the all-time peak frequency here was 21 15 megahertz but we ran into power perf cap limitations that did not permit higher frequency it's odd because if you watch build Zoids PCB analysis on our channel the PCB and the BRM can withstand far far more amperage nvidia has limited Turing's overclocking capabilities and current throughput severely here and in ways which are nonlinear with its overall improvements over the previous generation in other words even though there's 60 watts additional overclocking Headroom in terms of power throughput that is not a linear translation from the previous generation we would expect more than that even though it's more it doesn't percentage-wise it's not all that much more regardless we also managed an 800 megahertz stable memory offset and these stats were only true while operating at 59.4 DBA with 100% fan speeds mind you allowing auto fan speeds to take control we hit thermal limits hard at 20 40 megahertz a few of the cells in this chart are highlighted for thermal and frequency response you'll see those tags yellow just like Pascal terrain drops clocks as temperature increases even if it's well under the new 87 degree throttle point up from the previous 83 84 ish we noticed that the temperature below 63 degrees when it gets there anyway would permit higher frequencies averaging around 20 70 megahertz instead of 20 55 megahertz we also noticed this trend apply at 75 degrees Celsius diode temperature pushing frequency down again to 2040 megahertz and then modulating between that and one step up speaking of noise here's a chart of the RTF twenty eighty founder Edition rpm based fan noise will show comparative noise normalized thermal data versus previous generation 1080i coolers momentarily the fans have a maximum fan speed of approximately 4200 rpm only when bursting up but will stop and stay at 3700 rpm when manually set to 100% and left alone for a second the minimum fan speed on the FE card is 41% you cannot go below that it does not have a zero spin mode which is about 1500 rpm at the minimum fan speed we observed an idle noise output of about 30 4.8 DBA in a room with a noise floor of 26 DBA pretty damn quiet we measured 40.3 DBA at 52% fan speed which is the speed we used for noise normalized thermal testing that will show in a moment I want a hundred percent speeds which are required for peak OC performance we measured a noise level of 59 point four DBA that's pretty damn loud that's about where Vega is for the reference card about the Titan V about Titan XP most commonly during gaming workloads the card will operate maximally around 44 to 46 DBA depending on how hot your case is because it will boost depending on that four perspective we also have this chart a different one to illustrate where and videos fan curve stands in comparison to some of its previous products though our GX 28 efe fan curve verse acoustic response is less aggressive than the overly loud Titan XP and the desperately hot Titan v coolers sitting closer to be 1070 F equal or in comparative data moving into thermal testing and will start with the frequency thermal response on a full stock our TX 28 EF e card this is data for a card that hasn't been torn apart and is exactly as you'd receive it from the factory with no changes to anything over a 30 minutes mark test we observed frequency decay at key temperature thresholds with drops primarily around the 60 to 63 degree mark and again in the 70s you can see a slight frequency fall out from 1965 megahertz to 1895 megahertz by the end of the test which is normal behavior for Nvidia's gpus with these 3.0 and has we can see now 4.0 this chart a different one again shows the fan speed at versus GPU core thermals the fan ramp begins late starting at around 68 degrees Celsius for core temperature after that point it ramps to about 1800 rpm and remains there as the GPU is within operating spec for the fan speed versus temperature lookup table and go higher if the GP core and warmer like it did in firm work comparatively and always normalize thermals for the r-tx 2080 fe aren't completely terrible keep in mind that this card draws a lot less power than the TI which we'll talk about momentarily and then again later but still is relatively power-hungry in some and for clarity TI there is in reference to the 1080 TI not the 2080 we don't presently have the other r-tx device data but we'll be adding atom board partners as we test their cards for now we offer an incomparable comparison against previous generation designs strictly for some perspective the art CX 2080 at 40 DBA of noise for all devices manages a reasonable GPU temperature of 49 degrees Delta C / ambience we monitor ambient temperatures second to second and calculate a delta minus so this is closer to around 72 degrees in hard measurement that puts the 2080 as better than the dismal 1080 TI armor and Titan XP blower the Titan XP is the most direct comparison it's a large high TDP device made by Nvidia with an Nvidia reference design and it's beaten thoroughly in thermals by the 2080 Fe despite being a different GPU core itself with different die sizes so that matters again you can't strictly compare these things but for purposes of getting some perspective on Nvidia coolers generationally it does help the EFI runs significantly warmer than most of the 1080 TI partner models but it's also completely different so anyway use the the results as you may the point is though that board partner cards will do this a lot better most of them and we'll look at those soon enough for VR I'm MOSFET temperatures we stuck thermocouples all over the card and measured AVR emboss temperature from the hottest center mosque of about forty six point nine degrees Celsius over ambient making this better than the amp extreme 1080i armor and the Titan V but worse than all the partner Model T is from last generation finally vram temperature is completely incomparable and a like-for-like as we're now looking at G GD r6 rather than g5 x for this we measured a temperature of fifty seven point two degrees delta T over ambient which is within GTR sixes operating range and is reasonable so getting into the conclusion then quick note on power consumption we did just a an ad-hoc current clamp on the PCIe connectors didn't measure the socket so we don't know the the power coming out of the PCIe socket but for the connectors the 28 EF e sticks at about 15 absence and locks their four PCIe input and it does not move so that's partly boost 4.0 it's partly a power or current limitation and you can go pass that with overclocking but otherwise 15 amps multiplied by about 12 point to on that power supply to get your wattage the 1080i SC to that we tested was closer to 16 to 17 amps so it's actually more power consumption to get equivalent performance to the 2080 from the PCIe connectors what we don't know is did the slot change because if the if there's more power coming out of the PCIe socket then it's kind of a wash but we'll look into that more later perhaps so that's the power behavior the card as a whole look it's out of state in the beginning it's it's easy to get carried away with disliking everything about the r-tx launch because of how disorganized end video was in its public presentation and how misleading a lot of the charts word that I think has pissed off a lot of the community there is also existing concern that is deep-rooted from potential Hardware lock-ins how well does Pascale do with these things how old is AMD do with these things the answer if RT X SDKs ever get used is probably not very well depending on how they're implemented but this is all stuff we know and so there are a lot of reasons to be concerned about the product and about the launch about RT X in general the card even if we a hundred percent ignore all of those other reasons ignore marketing every problem you've ever had with Nvidia put it all aside how are the cards and the answer is the r-tx 20/80 founders edition is not worth buying today it's more expensive than a 1080i partner card it's hotter than a 1080i partner card power is more or less a wash and the gaming performance it's identical so there there is a generational gain 20 80 to 1080 yes it's also a price game look at the tea I will talk about this more in the TI review for the 20 80 TI $1200 for the founders Edition and most partner cards technically yes there's $1000 MSRP we don't know anything about those cards right now most of them are 1,200 bucks and the founders Edition is 1200 as well that's what you're buying to water cool most likely so 1200 dollars versus a 650 $700 1080i in the u.s. that's absolutely available we checked that's something like a 60 70 percent price increase and the performance increase is closer to 20 to 30 and we'll talk about that again more in the ti review so how do you how do you justify that where where do you find the only justification is buying on a hope that you can use all this cool art tech stuff and that also the game developers don't just keep doing what they've always done and make the games look pretty damn good and cheat the effects and do things like cube maps in the window is to make fake reflections stuff that's tricked people for a long time so there's a lot riding on hopes for RTX the performance versus the price is absolutely not there right now the 1080 TI is the the biggest competitor to Nvidia's r-tx product launch design videos previous product which is the 1080i if you're considering a card in the performance class of a 2080 2084 I don't know 800 bucks maybe if you're lucky bit lower but we don't know what the prices are gonna be on the partner cards yet if it's 800 + and you can get a 1080i for 650 get a 1080i for 650 unless you're maybe a developer this is where it gets different game developers not even production users just straight game developers or someone who's maybe working on movies or high-end 3d effects stuff like that those users it makes sense and they need to adopt it anyway for it to make sense for gaming users but for our audience for gamers Nexus it makes more sense generally speaking to buy a 1080i and not buy based on promises and hopes and that's that's gonna be our stance on that one the partner cards might change things we'll look at those and on the development side we will eventually look into that stuff as drivers released as Unreal Engine updates we have the ability to do 3d modeling and ray tracing and all this stuff and software and as blender up days Unreal updates we plan to revisit these these cards and look at them from a production standpoint but today from a gaming standpoint is what we can look at and it's not quite there right now so perhaps in the future perhaps as these technologies actually get adoption I don't think this is really surprising to anybody that performance is good the product is uplifted the products are good objectively in terms of performance the value is terrible for the 2080 TI and is questionable at best for the 2080 the 2080s is more tenable versus a 1080 TI then a 2080 TI is versus anything else 1080i or 2080 because they're the same card so in terms of performance so I think that'll wrap it for us performance gaps will increase as we start switching to things like he BGA's cards and ace uses cards because they're gonna be pre overclocked they'll have more power Headroom so the gains there will be more significant the price might also be more significant we'll visit that soon enough but for today we're comparing what you probably have which is an aftermarket or a board partner TI 1080 versus these and then we'll get into the rest later so anyway that's it kind of kind of weak overall for the availability of what the products namesake is but so be it it'll come out eventually probably more likely then things like I guess VX AO but that's it for this one subscribe for more toys go to store documents nexus net to make about mod mad support us directly these beer glass will be back and stock pretty soon hopefully this week and then remember we have a live stream today the day this goes up at roughly 6 or 7 p.m. Eastern Standard Time in the US and that will be over clocking for the cards so subscribe for that I'll see you all next time
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.