Gadgetory

we already reviewed the RTI x28 efe it's on the channel we did all the jokes we did the push the button and turn our TX off and on joke they talk about the bad marketing talk about the missteps by Nvidia and talked about the product positioning of the 2080 when it's viable when it's not when does it make sense when absolutely should you just buy a 1080i and said we did all that check the channel for that one today we're talking about well it's both today but we're talk about the twenty atti that's the interesting one here because the twenty eighty if you saw our previous video and gaming it functionally equals a 1080i for perforins so the twenty atti by brute force alone should be better and more interesting and that's what we're gonna get into you for this one before that this video is brought to you by be quiet and the silent base 601 the silent base 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 to streamline 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 so as previously stated in the other video there'll be a link below for the testing info stuff like that methodology additional resources absolutely check our 20/80 review there's a lot in there that's applicable even here the twenty atti so that's important you should check it out and it also discusses the marketing in more depth and about how Nvidia handled this launch in general because that's part of the review too if we're all honest that affects how people perceive the product in significant ways so we'll leave that to that previous video for this one twenty atti is primarily going against the 1080i and against the twenty eighty for purposes of this comparison we have thermal data we have frequency decay versus thermals so we have thermal frequency charts we have some rpm fan rpm versus thermal charts we've got additional information on just scaling 4k 1440 1080 how does it scale across resolutions that all be in here test methodology more below but we're using 8086 K at five gigahertz that's all in the charts and then if you need game graphics settings those are in the charts as well in the head additionally for anything that says overclock we use auto fan speeds for them for gaming but for our overclock stepping chart we pushed to a hundred for the art 100% for rpm so that we can figure out what is the temperature versus frequency decay how much frequency how many megahertz do you lose as you increase the or increase the temperature so that's part of what we're looking at but otherwise Auto fan for the overclock settings in the game charts because that is the most reasonable user scenario and then also one final note raising board partner ten series cards not the reference models the reference models would make the new reference models look a lot better in the charts that's absolutely true and video would probably pretty happy with that realistically we think that most people are probably using aftermarket cards board partner cards and so it's the fairest comparison but also the previous reference cooler is garbage and doesn't cool a while and it drops clocks and this is a significant upgrade over that so makes more sense to go with dual axial versus dual axial in that regard F 1 2018 uses the ego engine and translates to other codemasters games we'll start this one with a more unique chart just like in the 2080 review that was posted separately looking at a resolution scaling chart we can see how the r-tx 20 atti compares two different resolutions in the past some devices like the Titan V and the fury X have Illustrated stronger performance on one end of the spectrum or the other like 4k we're not seeing anything too special about this one it's pretty straightforward really perhaps other games will show more interesting scaling characteristics across the resolutions as we sometimes see depending on Rob's or other items in the pipeline that benefit 4k vs. 1080p at 4k the r-tx 20/80 TI f e manages to chart top unsurprisingly once again with the 99 to 101 FPS average frame time performance is still all over the place but that's an artifact of f1 and not the cards as we see this behavior on all tested devices this marks the RT x 20 ATT is 23% faster than the RT x 2080 and a 1080 TI alike overclocking also gains the 20 80 TI an additional 15% performance which is actually surprisingly good not being power limited would be even better but Nvidia did restrict overclocking pretty heavily with this series of cards at 1440p the r-tx 2080 TI is one 60 FPS average LANs and as 20% ahead of the RT x 2080 and gtx 980ti which are roughly equivalent to one another that's a gigantic price jump for performance gain which is relative to the price pretty small overclocking does help but then again so does overclocking at 1080i and RT x 2080 it's not like this is a vacuum for the 20 80 TI 1080p still shows scaling as this game is primarily GPU bound with these settings the GPU scale all the way up to 211 FPS average on the overclocks 20 80 TI posting scaling from the 20 80 TI f e is 196 FPS average of 7.7% that's about the ceiling for this game far cry 5 and the dunya engine are up next at 4k the r-tx 2080 TI fe manages to chart top again not surprising at 74 FPS average this places it ahead of the stock RT x 28 EF is 59 FPS average by about 27% which is one of the most noteworthy gains we've seen of the two cards thus far the card outperforms the stock GT X 1080 T is C 2's is the 7 FPS average by about 29 percent which is it good for generational gains from TI to TI unfortunately this 30% improvement also bears with it a price increase of about 70 percent moving from around 700 or so for a decent partner 1080 TI to $1200 for similar quality RT x 20 ATT is or even the founders Edition which is definitely not worth purchasing at that price the partner models would make more sense at this price but that's still only assuming that you already want to buy a 20 ATT I will talk about the greater value proposition at the end of this video overclocked in the 28 e TI gets it to an impressive 83 FPS average for a gain of 12% over the stock 2080 TI f e this is one of the better over clocks we've seen on the FE cards thus far the 2080 and 1080i remain equivalent but the 2080 ti does post actual generational gains it's just very expensive to get them here's a frame time comparison at 4k showing the RT X 2080 2080 TI and GT x 1080 D I remember frame times are all about frame to frame intervals pacing so more consistent delivery is better a lower number is also better as 16 milliseconds represents 60fps the r-tx 28 ET i posts both measurable and observational improvement in frame time consistency and overall frame rate although much of this is through brute force and the value proposition doesn't quite line up with the gains at 1440p far cry 5 positions the RT x 28 ET i at around 127 FPS average with low as reasonably times 288 FPS one percent and 80 FPS 0.1% the RT x 2080 lands at 105 FPS average allowing the 2080 TI a gain of about 21% intergenerational versus the 1080 t is c2 we see a generational gain of 22 percent this makes sense as the 2080 and 1080 TI are functionally equivalent in performance overclocking the 28 ET i puts it at 138 FPS average a reasonable gain over stock by about 9% we could get more with an unlocked power limit as found on partner cards but those are still pretty locked down just better than this one 1080p proves to be an invalid test as it becomes CPU bound and demonstrates the 28 ET i 2080 and 1080 TI all within a couple of frames of one and others but we're going to show the chart anyway just to show what a CP bound scenario looks like my pearly 4 is up next and uses the essaouira engine this title as stated in our RT x 28 EF e review is one of the most important for a well optimized low-level API implementation in a game this game has shown scaling for asynchronous compute in Volta previously and we now turn to it for a Torian analysis at 4k Sniper Elite for posts the RT X 28 e TI fe stock card at 108 FPS average with lows exceptionally tightly timed though this isn't a behavior of the card more so of the game the RT x 28 ET i ends up 25% ahead of the GT x 1080 t is c2 the stock card which operates an average FPS of 87 that's not as much as both have led us to believe sadly but it's a start scaling versus the RT x 2080 is functionally the same as the RT x 2080 basically equals the 1080 TI and performance overclocking gets the 1080 TI and the arts 2080 both to around 90 FPS plus or minus 1 which closes the lead of the r-tx 2080 TI fe stock card down to 19% overclocked in the 2080 TI fe widens it once again scaling to 123 FPS average for a gain of 13% over the stock 2080 TI fe not bad but still somewhat limited shadow of the Tomb Raider runs on a modified crystal engine and is still a brand new title unfortunately despite being an RDX title it doesn't have any RT ax implementations at this time at 4k the r-tx 2080 TI fe performs at 67 FPS average leapin the RT x 2080 and GT x 1080 TI by about 24% each these hold at about 54 FPS average overclocking boosts us up to 79 FPS average on the TI f ii which is an impressive 17% gain over stock and distances the card further from its counterparts at 1440p is scaling wanes from a 24% lead over the 2080 at 4k down to a 20% lead at 1440 the distance grows as resolution increases which is indicative of a ROPS advantage on the 2080 TI the 1080 TI is again roughly equivalent with the 2080 so we'll skip past these comparisons let's see if our proof of concept is correct at 1080p the r-tx 2080 TI is now 18.7% ahead of the r-tx 2080 and gtx 1080i measuring at 159 FPS average versus about 134 FPS average this further solidifies that the neutering architecture scales better at higher sampling rates and resolutions which also means that it'll do better in handling anti-aliasing at higher levels than its Pascal counterparts the idea is the same your sampling the pixels a lot of extra times which is functionally a resolution increase GTA 5 runs on the rage engine and although it's from 2015 it remains one of the most popular games on Steam at 4k the RTX 2080 TI fe holds 77 fps average a generational gain of 21 percent over the 1080 TI sc2 and at 64 FPS average the 1080 TI outperformed the RT x 2080 in this game so the 28 e TI manages 27% over the 2080 and overclocking is still helpful but the overclocked 2080 ti leads the overclocked 1080 TI by about 27% 1440p posts winding results even though bottlenecking isn't really here yet that's not until 170 FPS or higher the 2080 TI's lead over the 1080 TI diminishes to 15% making a mockery of the disproportionate generational price gap our overclocked stepping chart is next completely stock we measured a peak frequency of 1890 megahertz with no changes applied average frequency dwindled down to 17 50 megahertz which is a result primarily of power limitations as usual the voltage slider does nothing at all and has no impact on overclocking performance or maximum clocks although we did end up sliding it all the way to the right at this point just in case it's not until memory overclocking that voltage ticked up to one point zero volts which is it lower than Pascal generations 1.06 three volts it's 1.09 three volt operating range that said we don't know that there's really full confidence in these tools just yet so the voltage we're not sure how accurate that measurement is from the software but it would make sense that it would be lower than Pascal so that does at least somewhat line up with what we'd expect anyway boosting the power it's 123 percent that's the second step in our process or the second row this allowed us to hold an additional 50 megahertz on the clock to 1800 megahertz average that is even in spite of higher temperatures we eventually found a 20 100 megahertz peak similar to the 21 15 megahertz peak we hit on the r-tx 2080 but immediately power throttled and hit a power perf cap our average was closer to 2,000 megahertz where the GPU remained when kept in the 60s Celsius closer to 70 really the GDD our six memory was stable at 850 megahertz offset but not beyond that note again that this is in times by extreme and at this overclock doesn't necessarily hold stable in every single application that exists it's per application generally that these will work this one however did work in all of our games so these offsets applied in all game tests this is also with 100% band speeds to keep a lower temperature which were not used in the gaming test we stick to Auto fan speeds for gaming tests with overclocks for realism but either way the 100% speeds at about 60 DBA 59.4 DBA of noise do boost the clocks noticeably look more closely at the yellow cells next though for these we observed frequency thermal responses at around 79 degrees where the clock dropped by 50 megahertz and again at around 62 degrees although this is reflected in our previous r-tx 2018 on t i overclocked slepping chart from the other video the RT x 2080 review check that for more frequency thermal data for more frequency and thermal scallion we looked to a 3d mark lubing pass that put the card under 30 minutes of strain this is what they completely stock GPU with no overclock applied at all under these conditions we see the clocks immediately drop from just over 2000 megahertz down to around 1900 megahertz within the first 90 seconds the clock eventually settles at 1800 megahertz a significant downgrade from the initial operating frequency with the GPU core temperature of 75 degrees Celsius this is indicative of a limitation of the cooling and power management one of which is intentional the cooler can maintain higher clocks but it have to increase the fan speed further that's what we illustrated in the previous chart and videos fan profile is set to favor lower noise levels which means that we do sacrifice clocks in the process board partners should pick up the ball here because they are able to cool better with a larger cooler fan speed also ticks up with all this as shown in our next charge GPU temperature sits around 75 degrees Celsius with occasional spikes while the fan rpm ramps from 1500 rpm about 35 DBA to 2,000 rpm or about 42 DBA and video doesn't appear to be using its own ásinger to span technology at all here even though the fans are capable of spinning independently it's like they completely skipped on using the feature despite end-user support you could a synchronously tune these to be fair though in the past we've found that they do generally work a bit better if they're synchronous but we haven't tested that with this card let's bring in two more lines onto this chart make it a bit complicated these are our 4G DDR 6 and BRM Maas temperatures or MOSFET temperatures measured at hotspots on the card via thermocouples that we attached on average the vrm Mouse stays below 70 degrees Celsius when the card is under stock conditions and that's not delta T for our regulars that's just straight temperature because I'm not comparing it to anything else and the vram is around 70 four degrees Celsius these are both well within spec and it would appear that the cooler does fine for the vrm and vram when under stock loads despite being an overall inefficient overly complex and overly expensive design the FE card does have the capability to cool the VRM NVRAM so job well done there if you're wondering how overclocking impacts thermals here's a look at our torture run from the thermal bench the r-tx 28 ET i stock card before the massive teardown that we filmed again before that operated at about 73 to 75 degrees celsius during this 30 minute power virus torture load running over clocks with the settings we use for our gaming benchmarks and with the fan left to regulate itself under Auto settings GPU core temperature climbed to 77 degrees and fan rpm not shown here clients from 2,000 rpm the 2100 not bad and decent on the silenced focus as for vrm thermals under this scenario the overclocked card pushes vrm temperatures upwards of 90 degrees Celsius for the hottest MOSFET we can find and that's with a k-type thermocouple while stock temperatures are closer to 72 degrees technically that's within spec most MOSFETs don't trip OTP until around 150 degrees and they don't deer ate hard until around 125 degrees still it's a bigger climb than is necessary in temperature considering that better cooler designs exist for vrm cooling and at this hefty premium of $200 extra you'd expect good BRM cooling on the founders edition cart GDD our six operates within spec under both scenarios measuring eighty degrees when stock and about the same when overclocked the 2080 ti then is far more interesting than the 2080 it's not the most exciting but there's real performance uplift over things you can't buy you get a Titan V for more money sure but versus a 1080 ty there is actual tangible increase in performance there's a visible and observational increase in measurable increase in frame time or frame to frame intervals and those are good things but then if we look at the numbers again just like with the 2080 you're looking at something like at 15% in the worst cases but generally 22 to 30 percent gains from my ten adi to a twenty atti if we're looking at a bored partner card versus the founders edition card and that will change we're gonna be testing these cards from EVGA or testimony from asus to that number will increase but only so much and it's still going to be roughly thirty to thirty five percent with board partners for the most part can't predict the future but that's about what it's gonna land at two thirty thirty five percent and there's best-case scenarios over a ten eighty TI and the price increase is massive because a lot of these cards are gonna be eleven hundred twelve hundred dollars in the US and then MSRP is officially a thousand which is not a terrible increase in terms of the performance increase but no one wants to see that and realistically the cards even though MSRP is lower than founders edition it's been a while since we've seen cards follow that model and I actually dropped below founders Edition and hit MSRP so I I don't know I don't know how many cards will hit that and if they'll even have the factory overclocked chips because they might not Nvidia is reportedly according to the tech powerup selling chips that are locked from factory overclocking you can overclock but not the factory and that will impact performance too and probably on the lower end cards close at MSRP so let's just assume eleven hard twelve hundred dollars per card there's a 1080i that is a an extremely hard sell unless you have a specific need like if you absolutely can't get what you need with a 1080 Ti and an extra twenty thirty percent will get it then I guess you have one option otherwise developers can get use out of fount of the twenty series cards we talked about that already people using blender as it updates might get use out of it Unreal Engine when it updates to 4.21 4.20 is somewhere around there we'll eventually get it's our TX SDK implementation finalized and you can start really working on things there so there's room for people to buy the cards but generally speaking unless you really need it and you know you need it because you can't get what you need from a 1080i and a 1080i value eyes is better but at least with the 2080 ti you're getting something at least there's a tangible gain for the money you're spending whereas with the twenty eighty verse the 1080i is basically the same product in terms of horns you're just gaining features that have no present use for gamers so we'll revisit this more with the partner models but that's it for this one subscribe for more as always patreon.com/scishow helps out directly go to store documents nexus net to pick up mod mat like this one or when they're back in stock one of our beer glass is something like the shirt that I'm wearing but not literally this one and thank you for watching I'll see you all next time