Gadgetory

we only just reviewed the msi gtx 1080 TI gaming X and we're immediately back with an even larger counterpart the gigabyte 1080 TI or as extreme a multi expansion slot card with a three fan cooler way over built heatsink and 12 phase vrm using fairchild at 6 8 to 3 C 50 amp power stages for a total possible throughput of 600 apps on the GPU core this makes gigabytes extreme a heavyweight in all departments and brings it to our bench for full review today before getting to that this video is brought to you by the current bundle on the GTX 1060 and VT X 1080 video cards where you can get Ghost Recon wildlands or for honor at checkout this comes alongside new MSRP is for the GTX 10 80 series cards now down to $500 you can learn more at the link in the description below today's review will focus mostly on thermals for the extreme card and it's for a good reason 1 they've got some really interesting stuff going on with the backside here and also as seen in our teardown the cooler is pretty beefed up so be looking at that very heavily noise testing will be included in the article link in the description below if you care about noise levels the gaming X for what it's worth outputs a noise level that is equivalent to all the other twin fros or 10 series cards we've reviewed at least in the 1080 line like the 1080 game in X 10 somebody came in X as well so I've got those noise levels there this will be added that's in the article below other than that what we're looking at for the gigabytes extreme card this one is a 1632 megahertz OC mode clock and that's kind of out of the box type of thing you can use their salt or to switch between OC and gaming modes or a 16:07 megahertz base for the gaming mode if you go to the boost clocks at 1746 OC mode and 1721 gaming mode which puts this technically a little bit ahead of where the gaming X is but there's a power limit difference out of the box so that means that there could be some slight disparity and performance numbers until you equalize them with overclocking or at least boosting the power target and leaving the clock alone because boost 3.0 ultimately has a per a Siq and per power limit type of functionality when you've accounted for thermals which both of these car should be doing speaking of thermals this Auris card as again shown in our teardown has a three fan cooler with the center fan having half of its blades cut off to accommodate the third fan without having to expand the card so that's an interesting point the other items of interest would be these massive heat pipes so we've got five heat pipes on the right side of the card covering the vrm component and then a couple heat pipes on the left side covering the GPU now interestingly the GPU is cooled with a single cold plate that also contacts the vram which generally means your GPU diode reading will be higher than if you were cooling the vram separately by a different base plate because you're sinking all that heat into one area and cooling it with the same solution the back side has a copper insert that we'll be talking about separately and that is sort of embedded within the backplate but it can be removed other than that you've got our GV LEDs everywhere on the card as one might expect at this point and just to get this out of the way it is a $750 card as is the MSI gaming X so starting with thermal this is the most interesting part of this benchmark because ultimately that's what differentiates these things for the most part thermal testing we have the full methodology defined in the article below as always that includes the thermocouples that we place on the FET and on the backside of the PCB these thermocouples as stated before are about a hundredth of an inch thick and they are designed in a way that means that the adhesive pad doesn't really interfere with the thermal transfer ability of the the native cooling solution on the card so that would be from the FET to whatever is above it normally a thermal pad contacting some kind of plate or thin stack these don't really interfere with it we're looking at deltas anyway and they're used to measure the temperatures of the components whereas the GPU is measured with normal diode starting the temperature testing with something easier and then getting into the a/b test with the copper plate on and off on the back side the first test plots the temperature of critical board components against the MSI GTX 1080p eye gaming X that we just reviewed for those who missed that review the MSI card had a very similar FET and PC backside temperature to the founders edition card so this can kind of serve as something of a stand-in for those numbers since they're so close it'll also simplify the charge so if you want to see the EPI card numbers you can check the EPI review and the gaming extra view the FE card naturally it has a much higher GPU temperature of about 84 C when under Otto conditions on both of these cards help account for that with these two cards the gigabyte extreme GP diode temperature runs about 70 to 72 C with the MSI gaming X diode temperature at around 68 see that temperatures place the MSI card at around 70 C four FET seven and sixty seventy four FET to the gigabyte card runs a hotter vrm with its fat's around 82 see that said these VRMs are both well within any reasonable operating range you're not going to get D rating that eighty-two Celsius on the gigabyte card we're so far away from that 125 C kind of threshold point that it's fair to say both cards are doing their job well especially considering gigabyte and MSI use different driver ICS and best to begin with so it's not a perfectly linear comparison we're not comparing just the coolers we're comparing the vrm and the coolers since the overall vrm solution is different in addition to the cooling differences neither card is performing poorly here if you're curious about the noise profile for each device at these fan speeds check the link in the article below and in that article we also have the clock stability over time versus temperature for the gigabyte card this next part of the thermal testing is pretty interesting we've done a beat estimate back plates in the past generally they're more for structural support than for cooling but that doesn't stop manufacturers from saying that they're for cooling there is an exception where when we tested the EVGA FTW ftw2 cards which is the icx version we did actually see a benefit from that backplate and that was primarily after EVGA accounted for their lack of thermal pads on the original FTW versions and FC versions and things like that so we did actually see a benefit in that use case because of the very specific design of the card it doesn't mean it applies to all cards a backplate can act as something of a hotbox if there's no good contact between it and the board below it or the board components at least so that's something we test for this card's a little different because it does have the copper insert which is very specifically designed to at least look like a heatsink for the backside of the GPU itself so we wanted to test that it's easy to remove and we can leave the rest of the backplate on when doing the test for these a/b tests that's exactly what we did and we tested a few different ways one we tested it with the just on and off and then our standard very light airflow from the open bench the second test use a direct cooling solution of a fan mounted about an inch or two above the backplate and so it was pointing straight down at it as basically an intake then that was because when you're testing something like a copper heat sink or what's supposed to be one you really want some kind of dissipation solution because if you're just racking up the heat on the heatsink and you have no way to dissipate it the numbers aren't going to look like they probably should in a more real-world use case now that said this thing's a bit far away from the front of a case and you might have memory in the way so we've accounted for both those scenarios the first test is more of a real case where you're not directing airflow specifically at the insert on the back side and then the other test is lighter airflow in terms of fan speeds we configured them at locked to 60 percent on the card for all tests and that's the common fan speed that the card reaches when under load and real gaming scenarios so that's what we chose for the testing the thermocouples and diodes are allowed to reach a steady state prior to proceeding with the benchmarking that's important as well because it takes a little bit of time to get there and then we're logging ambient seconds a second and using that to calculate a delta value against the thermocouples and the GPIO temperature that way we can account for the one to two Celsius ambient fluctuation and uncertainty that is inherent to every test starting the a/b tests with just one case fan positioned about 10 inches away from the GP backside with the usual sticks of RAM between the family card we see the chart that's on the screen now these numbers are with the backplate on the card so that copper inserts there and we've propped the temperature scale in to 0 to 60 C just to help better show deviation and the numbers once we start adding things into the chart all the values are in delta T we're accounting for ambient in these measurements that helps kind of equalize for variance in the test and and as shown here we're seeing that the 1080 TI extreme runs it's bet seven and bet two temperatures nearly identically they've built a very consistent vrm across the entire board both at roughly 53 Celsius delta T when under load and the power virus scenario the PCB backside temperature with the probe affixed behind the fest and inductors which is about the hottest part of the backside measures at around 49 to 50 Celsius delta T load the GPU diode temperature is bouncing around near 43 to 45 Celsius delta T and next we can remove that insert leaving the rest of the backplate on and run the test again what you're looking at now is a whole bunch of lines on top of each other fortunately it happen to be sectioned off in a very easy to follow fashion with the help of some on-screen highlights the top pairing of lines is for the FETs there are actually four lines here but they're all covering each other because they're about the same temperature that's two and seven with the backplate on and off are effectively identical in performance so there's no real difference here we cannot confidently state that there is a significant difference between these numbers with or without the copper plate and it doesn't really it just doesn't impact that temperatures as you might expect we can next look at the PC backside temperatures those you might think would be more likely to be affected given that the plate is conducting heat away from the back of the PCB and therefore potentially out of a potential hot box scenario that could be caused by the back plate there's no real difference here either though the two lines we've highlighted now show that the PCB backside temperature is effectively identical with and without the plate and let's now highlight the last two lines at the bottom these are for the GPS diode temperature again it's a delta T reading there is no significant difference here and our confidence interval is not tight enough to declare a potential 0.5 C advantage we've got about a 1 C range of confidence these numbers are effectively identical with the help of more video editing we can also highlight the front of the curve before this overtime chart and that's why by the way overtime charts are really valuable we can highlight the 200 to 350 second mark vertically and show that the plate does not even impact the ramped up temperature of the components measured the temperature is increased at the same pace regardless of its presence and because our test is automated we can see this quite clearly well there's a chance we need better air flow Oh to actually see the value of this since if it's soaking heat but not dissipating we're not really testing it to its fullest potential so that's where this next set comes in we've added a 120 millimeter case pan about an inch or two above it going straight down at max rpm powered by the power supplied directly and that will give us those readings another thing to note is the place of measurements you might see more of a difference on the caps on the backside but we'll talk with that in a moment so with the direct fan here's the chart with the back plate on we now have a fan directly pointed out that plate temperatures overall are slightly better on the back plate diode than in our previous test which shows that the extra airflow is getting down the backside of the card and actually cooling things so far our numbers look mostly the same as before though the FETs number 2 & 7 are nearly the same temperature and run around 51 to 53 Celsius delta T over ambient PCB backside temperature is now around 46 to 48 C delta T and GPIO temperatures about 43 to 45 Celsius delta T removing the copper insert and back plate we can now add the rest of the lines that's 2 & 7 in both the a and B tests are nearly identical once again that 7 with the plate off is slightly advantaged though we're not confident enough to state if that's because additional air is getting to the back plate that could be the case since that 7 is also in the exact spot opposite the board anyway where there's no thermal pad coverage to the back plate since the LED tape is there instead and we have some footage of that from the tear down this stated the difference is hardly more than 1 Celsius so we can't really confidently confirm a meaningful difference but if there is one it would probably be because air is getting it where there are no thermal pads and helping get rid of some of the heat the PCB backside temperatures are also effectively identical to end of the day even with the help of the direct airflow and the GPIO temperatures are not meaningfully different either based on our test and it does not appear that the copper insert on the backplate actually does anything of value it's not aiding or hurting temperatures in any meaningful way that said it's probably conducting some of the heat away from the specific spots it's touching via thermal pad or maybe those four caps on the back side of the card you could put a thermocouple there and measure it the thing is it's just it's relevant because temperatures are going to be fine up in that spot anyway the PCB itself the sort of hopper and fiberglass mixture really doesn't care if it's an extra five C warmer we just measure the backside because of the EVGA into them last year and the caps aren't going to be a problem to begin with so any advantage that is had in the areas we haven't measured is irrelevant that's just that's how this goes the backplate itself is more for structural support it is a large card you probably want to leave it on there the copper insert doesn't really do a whole lot this is a marketing stunt it is for marketing reasons it's not something that's necessarily bad all of these companies do it MSI certainly has all of their own marketing as do most of the other cards that are on the market so we're not going to slam them for it it's just don't think it's going to help you either this is not really something that's relevant now that said there are a lot more ways to test this one would be different fan speeds we tested at 60% speed because when you're gaming in our testing the fan profile on this card by default tends to push you to about 60% fan speed max and other than that it's basically idle so that's what we tested at if you ran maybe a 30% fan speed you might see a slight benefit we didn't test it there's a time limitation here but for our tests no difference for noise testing you can check the article below that'll all be down there but moving on to the overclocking notes the gaming X and the extreme gaming cards that would be a Messiaen gigabyte respectively both have similar limitations with overclocking we couldn't push them that much higher than their shipping speeds and that's just because boost 3.0 basically pushes the clock to its absolute limit anyway higher than the advertised speed within the power constraints as defined by the board partner so we've got the speed stepping on the screen ultimately we had to back it down a little bit for all the games though it was able to succeed in fire strike for the most part at higher frequencies it is not sustainable for all gaming but will include some of those numbers to the game bench is going forward now for a game testing methodology drivers test bench specs and all of that you can check the article linked in the description below moving on first to Ghost Recon wildlands we start off with our GQ benchmarking at 4k the msi 1080 TI gaming x and gigabyte nati horas extreme both plays around 57 to 58 FPS average our stock gaming X model runs a higher maximum frequency out of the box given the variance in one basic quality and to the 300 watt out of box power limit on the gaming X which helps increase the clock Headroom over the roughly to 75 watts on the Auris extreme you could max out the power sliders on each to get something similar to our gaming X + or OC numbers in this chart as stated in the gaming extra view there's just not a ton to be gained from overclocking these cards in particular and frame time variants potentially increases which isn't a good thing as clock stability becomes threatened even though the averages might increase the GN hybrid mod of the FE card removed a thermal limits giving it more Headroom than you gain on either of these cards just because it's not pushed as high as these are out of box and both of them are adequately cooled at 1440p it's more the same a 1080i extreme and 1080i gaming X both $750 cards run at about the same FPS roughly 94 in the average Dept and roughly eighty one fps 1% lows roughly 77 fps 0.1% lows the cards provide an identical experience at 1440p overclocking it gives some boost to the average FPS but as seen in the gaming XOC slot we do take a slight hit to frame times it's ultimately a trade of how much you want to give in consistency for the higher overall throughput even the hybrid OC has some stability issues decrease in its frame time consistency so this is a normal thing we can actually plot that too by the way it's not necessarily an appreciable difference to the user depending on how unstable the clocks are but illustrates a point that overclocking these high-end devices means encountering potential voltage or power constraints that choke the clock finally at 1080p the Auris extreme in gaming X again perform more or less identically it we're sort of within the standard deviation test test here overclocking can get us a bump an average FPS with the same potential pitfalls as earlier discussed both the gaming X and horas are about 5.8 percent better than the stock EFI card or about 1.9 percent better than the EFI card when the thermal constraint is removed we're now moving on to for honor which is one of the most GPU intensive games we currently test at 4k resolution Laura's extreme card lands itself around where a liquid-cooled mod of the founders edition card performs or about two FPS behind the gaming X once again that's the power limit coming into play and we're sort of running at full tilt with these 4k benchmarks which is what's showing the difference overclocking produces particularly bad frame times this title so we generally leave it out the experience is not much improved over the stock card at least not unless you're willing to pay out in the frame time variants Department for the higher average or just run a lower overclock in the best circumstances we're getting an extra six to nine FPS on average out of the ten atti cards here's the same game with 1440p testing the Auris card performs about the same as our 1080 TI hybrid fe mod and a few FPS under the 1080i gaming X from MSI there's not a significant difference here yet again though they are about the same price pushing the power limit a bit higher would more or less equalize these cards and for 1080p results you can check the article link of the description below running Doom at 4k with Vulcan for the API we see the Auris extreme card about 3% behind the gaming X once again and out of box performance with overclocking actually proven useful in this game clock stability as a whole is better in Doom than in for honor and so the other games we test and so the frame times generally see a performance improvement rather than a performance hit well highlight do see numbers on the chart between the game and X an extreme card they try to differentiate a bit more in this department but out of box reforms again is mostly equal the MSI gaming X takes a slight lead in stock the 1080p if' e card runs about 6% slower than the oars extreme card when using that stock heatsink or about 0.7 percent slower when using an hour fe hybrid mod for 1440p results see the article below and now moving on to Sniper Elite 4 with the x12 and async compute the Auris extreme card runs at around 84 FPS average with lows at 69 and 64 FPS respectively Emma sighs gaming X operates approximately one point nine percent faster out of box with averages at 86 and gns coin one percent low and 0.1 percent low metrics at seventy two point eight and sixty nine point seven FPS this difference isn't appreciable to the end-user you will not notice it but it isn't measurable and repeatable we are again seeing out of box power configuration at play is the easily modified if you really needed that X one frame per second overall versus the unmodified F II card the RS extreme runs about 11% faster or about 1/2 percent faster than the liquid-cooled hybrid as we're not hitting the thermal limit on the clocks Mass Effect Andromeda 4k resolution shows the same scaling as all the other games out of box the Auris card places a bit below the gaming ex the overclocking assists the gigabyte card in some ways the difference between the stock devices is not appreciable to the end-user if once again but is again measurable and repeatable we're at 69 FPS average for the 6 1% low and 53 0.1% low on the gaming X or 66 FPS average 55 FPS 1% low and 51 FPS 0% lows on the Auris extreme 1440p follows this trend with the cards now at 133 and 129 FPS respectively the low scale linearly on these this difference makes the gaming X about 3% faster out of box just like in previous tests yours extreme is that 2% faster than a thermally unconstrained F II card using our hybrid mod and if you want to see 1080p results you can check the article below you can also find ashes of the singularity down there which is a DX 12 bench as was Sniper Elite and you can find our 3d mark fire strike and - buy tests in the article below now this card has one of the biggest coolers we've looked at lately in terms of noise levels the or scarred tends to run a bit higher rpm on its three fans on average than the MSI card because the RS car just has a more aggressive fan profile than the MSI card does emicida card also has large fans and that means that they can spin them a bit slower and move similar air the fans on the msi Kart are also quite obviously limited to a count of two so that's a difference noise levels not hugely different because ultimately you can lower the aggressive fan curve on both of these devices and still retain operable temperatures if that's the kind of thing you want to tune for noise is a very subjective thing that we have DBA numbers in the chart in the article the gaming X is roughly 3% faster out of box and the games we've tested then the Auris extreme card that's not an appreciable difference it is a measurable difference and one that is mostly attributable to the stock power budget or at least power target for these cards out of box it tends to vary about 25 watts between them and of course there's a sick quality model to model as well or at least unit to unit overclocking or at least pushing the power limit to the max will more or less equalize these devices that said it's not necessarily worth overclocking the 1080 TI devices from what we've seen so far they're pretty much at the clock limit of where they can be largely thanks to boost 3.0 and how it functions the copper backplate on this thing the insert doesn't really do anything but it also doesn't hurt so it's not it's not an aid it's not a detriment it just exists and as more copper being used probably where it shouldn't be in our testing there may be a scenario where it's helpful we haven't found it and we did some pretty detailed passing on it so overall looks more or less like a marketing thing for the copper insert on the backplate now the backplate itself is functional it's structural support these cards when they're as large as both of these are can actually use it to prevent sag or prevent any kind of damage long-term especially when you're looking at a size who might ship things although they tend to ship reference cards so it is functional just not for cooling really our price is the same as the gaming X these are both $750 cards they're both $50 more than the F II card which has a cooler we don't really like the FP card the PCB is actually built really well this time so that poses an interesting challenge where you could probably buy an Fe PCB at least a reference one with a better cooler on it or stick your own cooler on it and be quite happy with the result if you want cards that do well out of the box and never want to touch them in overclocking or in physical Hardware mods these are fine it just it's another question of you could buy probably a $700 gigabyte model with the three fan cooler or 700 ish 7 $20 EVGA SC model or anything like that Strix one of the cheaper ones anyway anything like those cards we've got a decent coil you could buy one of those have more or less the same experience within a couple percent so it again comes down to how much money are you willing to pay for anywhere from 1 to 4 or 5 percent difference over the $700 cards if the answer is $50 these are both fine cards the Emma tywanza bit faster out of box this one's just kind of better-looking in terms of the RGB LEDs if you care about that but that more allows some of things up so if you remove the thermal constraint the advantage over the EPI cards about 1 to 3% if you remove the constraints these are more or less equal of them the outer box difference noise tester in the article differences in that huge though as always you can subscribe for more patreon.com slash gamers Nexus to help us out directly store gamers Nexus net for one of our shirts we have three units there and just restock thank you for watching I'll see you all next time you