Gadgetory

if you ever wonder what difference 50 bucks makes in a video card with an video somewhat forceful encouragement for manufacturers to hit the $500 price target on the r-tx 2070 which is really difficult for them to do because profit margins are really low if you wonder the difference between a $500 r-tx 2070 and a $550 r-tx 2070 this pretty much sums it up this is only the cooler from the 27th exc ultra which is 50 to 70 dollars more depending on where you look let's call it 70 and this is the entire r-tx 2070 Black Edition and this card fits within the space of the cooler alone for the XE Ultra 2070 and this is an important thing to look at it's it's important to look at what does this 2070 really clearly forced price that manufacturers are mostly hitting at 500 bucks what do you get for that 2070 versus what happens if you spend a little bit of extra money is it worth it because it's it's a really weird market right now with the way the r-tx launch went over and some of the cards were just they're not delivering the way we would like them to and it's not really good for anybody so we're going to look today at the higher-end XE ultra 2070 versus the bare bones basic 2070 black edition at $500 before that this video is brought to you by us and the limited edition foil graph logo shirt this four color foil shirt is the iconic GN graph logo with average 1% and point one percent bar colors it's printed on a soft high-quality and custom made 100% cotton shirt and is available on store documents Nexis night until stock runs out once it's gone we will not be making more of these shirts we sold out within two weeks of our previous limited edition shirts so click on the link below to preorder now to EVGA is credit here they're 20 so d black is actually a pretty damn good card for the $500 price point and we're gonna call $500 cheap because relative to other our tips cards sadly it is but this one's pretty good there are some really not good ones out and we'll probably look at those at some point this is one of the better ones so we'll give them credit that means the difference between this and the XE ultra won't be quite as profound as perhaps one of the low-end gigabyte models and one of the higher-end models either way though we're going to be just focusing on the 20 70s today so the charts will have everything in them they're only really gonna be talking about the 27 TS that were testing and if you want the comparison who have the 27 t compares to like Vega or GTX 10 series check our previous content including the review or the follow-up Vega 56 over power mod where we blasted it with wattage and current and everything blast it with a lot of current basically we increase the power target by 242 percent from baseline and drove 500 watts into the card do you want to see that content check the other video because we'll compare the cards there but this one's just 20 70s so the real discussion here is going to be overclocking thermals and noise primarily on the overclocking side and one of the reasons that is the primary focus is because of the new identifier on NVIDIA GPUs the terrine GPUs have an a signifier at the end of the middle digit so this is a tu 106 - 400 a - a 1 and a 1 is the revision ignore that it's the 400 blank that matters is it 400 or is it 400 a 400 a is going to be a bit of a higher-end GPU that NVIDIA has pre sorted for the Adhan board partner and that will impact the out-of-the-box frequencies and therefore thermals to some extent so we'll be talking about that today let's start with overclock step and we'll get through thermals games all that stuff but it's really overclocking thermals and noise that matter the most today let's start with some overclock stepping charts the 27th EXC ultra gets a 130 percent power target and has a PCB and vrm that are basically from the 28 EF e whereas the 27 t black uses a 27 TF e p ZB v RM we have an analysis of both of these already on the channel the XC ultras higher power target helps in overclocking with no offset at all our stock performance on the XC ultra landed us at nineteen fifty megahertz peak or more realistically 19 10 megahertz average once heated up the performance cap was instantly power wear remains capped until our next step in the chart after increasing the power target we sustained 1920 megahertz average without any core offsets we eventually ran up to a 125 megahertz core offset still with zero megahertz memory offset here and found stability at about 20 40 megahertz when heated up and left to Auto fan speeds of 15 80 rpm the new limit was VL and power mixed together and less so just strict power this time memory overclocked very well on this platform we found stability at 1080 megahertz offset for over 16 gigabits per second throughput the core clock averaged 20 40 megahertz one left to auto fan speeds or 2055 when allowed below 60 degrees Celsius as a result of the boosting algorithm let's bring the 20 70 black on screen now and put that just below the FC ultra this chart shown on the lower half of the screen shows that our maximum overclock was 1950 megahertz average Winnett maximum fan speeds or about 1935 megahertz one left to auto fan speeds and 63 degrees the maximum memory overclock was about 980 megahertz offset the core is significantly faster on the XE ultra and that's a result of the higher power limit an extra 100 megahertz is noticeable in gaming and thus produce a frame rate difference note that there's also a maximum fan speed difference of 3800 rpm on the 2070 black well it's 3000 on the XC ultra but the fatter heatsink helps the XC ultra keep ahead before getting into the thermal section just some quick notes on methodology it's been a while since we've updated every one so for our testing we previously validated our testing methods for thermals for cases and GPUs and CPUs in a thermal chamber so we have tests and validation data from a thermal chamber it all validated works well the way we do it and we do a couple of things here so one of them is a power virus load on the GPU and that's a hundred percent load it blasts the VRMs another one of them is a 3dmark load looping on the same rendering frame to check for frequency over time and then we also do a more gaming related test as for thermals so we take the GPU measurement of course through software we use gpu-z for that and then for the BRM and the memory you won't see this many other places or any other places we take thermocouple measurements so rather than using a thermal imaging device there MOG Rafi is really not a good tool for this you can't see the vrm or the memory when it's running because there's a cooler there and you can't just take a shot of the back of it or the front of it and really tell much of anything especially if there's a reflective surface or an emissive surface that'll throw off your thermography so we use is we use thermocouples that we attach to probe individual components and for those we're using thermocouples that use an adhesive pad that's 1/100 of an inch thick so it's literally laser thin it does not interfere in any meaningful way with thermal transfer the pad is a combination of polyamide and poly methyl phenyl siloxane and the thermocouple is a K type hooked into our logging meters the calibration offsets are applied as necessary with the exact same thermocouples used in the same spots for each test and this is done as far as calibration goes we you can use an ice bucket bath and then hot bath or you can use individual calibration devices to calibrate the thermocouples and figure out the variance variance T couple TT couple is about 2.2 degrees Celsius for K types and it's perfectly inaccurate so if it's negative one it's always negative one and then you just apply an offset to your numbers you're good to go so thermal testing the point is very accurate when you come here for thermal data you are getting some of the best in the industry in terms of media coverage for thermal so let's get into it moving onto 3dmark frequency over time charts here's a plot for fire strike extreme and it's set rendering the same frame on loop for about 30 minutes the XC ultra average is about 1935 megahertz for the entirety of this test whereas the 2070 black experiences clock decay instantly from about 1815 megahertz down to 1780 1785 megahertz the average frequency bounces between 1785 and 1800 on the 2070 black this is a massive difference for the out of the box frequencies and indicates that performance advantage resultant of the better bins GPUs is favoring the 2070 XE ultra and Vidia cells these higher bin farts to board partners for Premium and it appears that the mixture of the new silicon bidding procedure and higher native power targets allow for a large frequency advantage note that we haven't done any overclocking here either no power offs that's nothing like that this is straight auto performance and even with the better fan speeds on the XE ultra the clock Delta is primarily stem from Nvidia spinning of those better great parts at the premium paid regions of the a versus non AGP use and you can see our teardown for more information on that and then also of course the boosted power target what we really need to know is the noise normalized thermal performance as that'll equalize our primary variable and allow us to see if the huge heatsink actually matters so these are noise normalized thermals at 40 DBA we take these at a 20 inch measurement away from the front of the shroud and this allows us to see something very important which is how does each card perform at a fixed nose level so we are equalizing the biggest variable and looking at GPU performance which allows us to see how good each individual cooler is now the big difference here this time first of all you can't really compare the twenty eighty the ten eighty stuff like that versus the twenty 70s you have to compare intra card versions of twenty seventy versus twenty seventy but even that is difficult this time because the twenty seventy is now sort of Bend within itself it's got two different versions of the twenty seventy and the PCB quality of course is a big thing but that is also why we do the noise normalize thermals normally so anyway regardless despite its higher frequency the XC ultra manages to run lower 40 DB a noise normalized thermals than the twenty seventy black with a 33 degrees Celsius over ambient GPU temperature versus 37 on the twenty seventy black this puts it as similar to some of the 1070 TI cards we test it again not strictly comparable as they are different architectures but it gives you a good idea for where it lands on the stack the MOSFET temperature is significantly lower at 31 degrees over ambient versus 43 degrees over ambience on the lower end card ggd are six temperature is warmer on the XC ultra and this is primarily a result of the proximity of the memory module to the hotter components of the VR app we also did one feature test without the back plate just to see actually does at 68% fan rpm controlled and with an overclock and maximum power offset so we've completely controlled the environment here we did not see any meaningful change with the presence of the backplate for G GD r 6v RM or GPU temperatures it certainly does something though the backplate gets hot and that's what you want in a backplate the fact that it gets hot and is hot to the touch means that it is conducting heat away from something but we couldn't measure the differences on the parts that we probed so either we missed whatever it is that it's pulling heat from or more reasonably you might need some active cooling over the plate to see more of a difference or you might need to be in a more thermally constrained environment like a hot case either way although it didn't impact our results positively or negatively it doesn't trap heat and make things worse and that's good not all video cards can say that so the backplate isn't bad it doesn't do anything for the things that we measured does something but we just couldn't find the point of measurement that would show that the next test looks at acoustic performance between the two for this we're measuring noise levels at a 20 inch distance again with a noise floor of about twenty six point seven decibels the only active component in the entire system is the GPU strictly comparing the quote low-end $500 20 70 and the $550 XC ultra model the first reminder is that fan noise caps at 3000 rpm on the XC ultra and 3800 for the Black Edition this is where the larger cooler and better PCB on the ultra allow it to spend slower the XC ultra and 2070 black mostly increase in staff with one another in terms of acoustics and are within reasonable variance for fan quality differences and being completely different in general the XC ultra typically spends at a lower rpm one left to auto than the black that is as both seem to stick around 63 to 64 degrees Celsius as their target temperature in a bench that is not super thermally constrained there isn't much difference between these cards when comparing them at a given rpm but the XC ultra cools better at a given rpm than the 20 70 black does as we saw in our thermal and clock data from earlier moving on to game benchmark we are starting with far cry 5 here this is on the dunya engine and as always we do at least 4 test passes averaged over each game for all three resolutions which allows sufficient time for the devices to heat up and get to their average clock speeds as we showed in the overclock stepping chart so this is heavily controlled and is it's very strict methodology to ensure that the data remains accurate across all the test passes on all the GPUs far cry 5 @ 4k ranks the EVGA r-tx 2070 XE ultra 49.7 mph when fully stock indicating that it is boosting higher than the plain EVGA 2070 card and it's 46 fps this is one part cooling and one part power limitations as you saw in the thermal and overclocking sections overclocking the 2070 black with a + 200 + 980 offset illustrates just how much more the XE ultra is boosting out of the box as it's offset maxed at plus 125 on the core due to a higher stock clock again roughly 1935 vs. 1800 megahertz when both the EVGA are TX 20 70s our overclocks putting us over 2000 megahertz for the 2070 XE ultra especially we ended up at around 51 FPS average on the 20 so d black + 54 FPS average on the 27th EXC ultra they mostly equalize in gaming at 1440p the EVGA RT x 27 t black manages 84 FPS average with the XE ultra at 90 FPS average when stock the out-of-the-box performance uplift is about 6.7 percent on the XE ultra firstly 27 t black overclocking both cards allows the XE ultra to retain its lead as a result of landing a higher on the volt frequency curve from lower temperatures though the gap closes to a difference of 3 fps and that's between 97 fps and 93 fps maybe 4 then so they are basically indistinguishable in frame rate when overclocked noise and thermals are a different aspect and we already discussed those finally will flash 1080p up for a moment the differences between the stock 27 TS at 118 FPS average and the XC ultras 126 FPS average is about 6% once again and overclocking gets the to reasonably close together low frame times are stronger on the XE ultra than on the stock card I've seen in these 67 fps 0.1% low vs. 81 fps zero on for low on the XE ultra this is due to boosting behavior and clock stability let's move Sniper Elite four positions the twenty-seventh exe ultra 71 FPS average with lows at 61 FPS and 59 FPS the r-tx 20 70 stock card runs at 64 FPS average closer to the overpowered Vega 56 mod that we have the difference between the stock XC ultra and the 71 FPS average that it performs and the basic 20 70 cards is about 11 percent favoring the RDX 27 exe ultra frame times are proportionately time overclocking allows exe ultra to hold a significant lead likely in part due to video memory load balancing in this game we're looking at 77 FPS average which is 71 FPS average for the overclocks providing a reasonable advantage here for the XC ultra weather that's worth $50 is sort of debatable for f1 2018 at 4k we found the 2070 stock card to perform at about 60 FPS average well then I see ultra up to 66 FPS average this is reasonable again for gains although overclocking does push the stock 2070 above the XE Ultra stock and overclocking the XE ultra reestablishes its lead once again over the 2070 Black Edition at 1440p the XE ultra again is about 10% higher frame rate at 1:12 he has average versus 100 FPS average with low-end performance about the same between the two overclocked in both cards gets them to 120 FPS average for the XE ultra and 115 FPS average for the 2070 black again leading to the question of when the extra $50 is worth it the answer is primarily in thermals and noise which are directly related and we're already discussed not in raw framerate 1080p produces similar scallion with the XE ultra again ahead reliably and with the overclocking permitting the XE ultra to remain in the lead again still it's not quite tie in with the GTS 1080 TI for performance here but it is approaching the sc2 is 166 FPS average one left stock shadow of the Tomb Raider is next this one post similar results at 4k with the 20 70 black OC time the 2070 XE ultra stock again we're seeing largely single-digit performance improvements in the performance for the XE old Trevor's of the stock or black edition model have 1440p the overclocked XE ultra is a few points ahead of the overclocked 2070 once again with the same thing happening with stock performance and no overclocking finally for power consumption we measured the following plot and ashes of the singularity for K crazy once all benchmark rounds were aligned and synchronized in data our data log inputs ts 4k benchmark at a total system power draw of about 320 Watts peak for the 2070 black and 27 exc ultra stock cards alike with occasional spikes to 357 watts missus total system power draw just to just be really clear here it would appear that both 27 TS draw about the same power when stock despite the XE ultra offering meaningfully more performance and it's testing these are both power constrained by Nvidia when the 27 exc ultra is overclocked though it peaks at 400 watts total system draw and average is 357 watts to 380 watts again this is total system power consumption but the motherboard and platform voltages are fully controlled to keep consistent power needs for all non GPU components so I real this difference we see here is when we overclocked but XC ultra otherwise they're all power constrained so that's everything we have for you for the most part we've got more game tests but at some point you got to cut it off because did actually just doesn't matter anymore the difference there was a pattern forming it's the same in all the other games we did GTA did a couple others we'll post those in the article linked in the description below if you really want the data but no point going over it here because the story remains the same so is it worth the extra money an extra 50 bucks sure probably although the debate of an RT X 2017 worth it at all is separate so relatives strictly to a $500 RT x 2070 a $550 XE ultra if you can find one would be a decent deal the problem is it's typically 565 75 80 somewhere in that range and this isn't really the fault of the manufacturers the price in there at today and Vidya is taking record profits on their actual GPU component sale and that's hurting the margin for board makers unless they can adapt obviously their pricing so the 550 would pay for it except there are still 10 ATT eyes and they're not that tested in price anymore so you're probably better off buying a 1080i if it's in budget if it's not you can look at 10 ATS and those are pretty close to the same price you might get a better cooler five hundred dollars for ten eighty ninety I than you would for five hundred dollars for this thing and this isn't terrible but you can get a better ten eighty cooler like an F TW or something if we're equating EVGA to EVGA so it's really a question of can you get the performance you want from an in stock Pascal card if you can find one you probably still can and if you can still find a Pascal card you get a 1080i for within a hundred bucks of either of these 2070 s just by that instead unless RT X is really a thing that you want to bet the future on perhaps better for someone like a developer or someone who does really any kind of work making things that will use ray tracing or whatever else you might be able to leverage the tensor device Kent tensor cores for so for just strict gaming 1080i at within a hundred bucks of one of these is a good deal if you can stretch the money if you can't a 1080 is a good consideration if neither of those options apply for various reasons then the question between twenty seventy at five hundred versus twenty seven eight five fifty if you get something like this for five fifty we'd say it's worth the price jump five seventy s kind of difficult to justify but prices are fluid and they'll move around so anyway that's what we got for you on this one thank you for watching as always you can go to stored on cara's axis dot net to pick up one of our limited edition graph logo shirts they'll be out soon and you go to patreon.com/scishow next up side directly subscribe for more I'll see you all next time