Gadgetory


All Cool Mind-blowing Gadgets You Love in One Place

GTX 1080 Ti Hybrid Results: Removing the Thermal Limit

2017-03-16
we fix the gtx 980ti founders edition card as noted in the review it had some thermal limitations with the clock the clock boosting behavior was not ideal with one of these on it we kind of proof of concept is this plan by throwing the fan speed to 80 percent but that's way to labs bare so we fix it by adding a liquid cooler instead before getting to that this coverage is brought to you by EVGA and their closed-loop liquid cooling kits you can buy the 1080 model and it will still work with this the showers and Pat ability isn't always perfect but the liquid cooler will work just fine on one of these so if you want to do your mod the way we did ours you can grab one of those kits that link to the description below and throw it on by following our first and second videos as guides so now that we've got that covered this project was a bit different from our previous hybrid builds and the way it was different is sort of exciting we added these thermocouples to the sets and back side of the PCB which is something that I've been doing for a little while now and GPU stuff since the EVGA vrm testing we did and later some other PCB and backplate tests that have been published since adding those allows us to see one really interesting metric that we haven't seen before which is does adding one of these negatively impact the temperature of the vrm components the MOSFETs and the backplate or anything like that as a back side of the PCB because we're taking away this this is a pretty large anyway for for a counters addition card pretty large air cooler it's just a heat sink with a vapor chamber in it pulling that off means we lose some of our service area and means that the fan and here it becomes less efficient at what it's trying to do that said the base plate of the card is still pretty damn big it's still a big metal base plate it's still connected by a lot of thermal pads to the components underneath and we still have a blower fan on the vrm that can now spin out at a lower rpm which again is PWM controlled by the GPU so that's another potential concern because now your blower fan spins up something like 22 percent because the GP is always under 50 by way of using a liquid cooler so there's a lot of interesting concerns with regard to vrm temperatures as a result of improving GP temperatures improving GP temperatures can worsen vrm temperatures because the PWM controlled erm blower fan slaves to the GPU temperature the diode that outputs the GPU temperatures so very interesting stuff to test and we have thermal results we have clock frequency stability over time and then we've got some gaming FPS results as well for volt testing methodology as always hit the link in the description below to the article that contains a fairly sizable section on how the thermal tests are conducted one note I really want to add vram testing as well that wasn't done for this one but looking at it for the future is just a matter of upgrading the thermocouple readers at this point basically but we have the arms covered that's all discussed in the testing methodology game test methods are also down there as well as the CPU using the bench ram all that stuff find that in the link below let's get into it with thermals this chart has a lot going on so breaking it down one line at a time first all the top lines in the legend that represent the TI founders Edition card without our fix or cooler or hack jobs whatever you may call it including the green line representing GPU core temperature that green line sits around 84 see in this endurance test which is NVIDIA thermal limit point under stock conditions the card will not allow the frequency to boost to a point that forces temperature beyond 84 C looking at the GN a hybrid variants of the same card same PCB and same GPU the actual GPU nothing's change I was only cooler or clearly at around 45 C that is substantial and we're looking at a 40 C reduction in temperature there is a further benefit spreading the frequency boosting Headroom which we will look at in a moment the big question here is whether removing the vapor chamber and heatsink negatively impacts that cooling my card still has that massive base plate as discussed but we lose the heatsink the short answer is no the big cluster of lines in the 65 to 70 C range represents the fat to and set seven temperatures for both revisions of the TI card there is no significant difference we can't even really see one line for the other in this mess of them because they're all about the same temperature and for the fats we monitored it's fair to say that the temperature is effectively equal to the founders edition card unmodified even if there were an increase it would have to be substantial to cause any concern the RM components can take a lot of heat upwards of one five to 150 see in some cases most cases really ideally though we're lower than that to reduce D rating or inefficiently concerned and we are well below that range with these results after the impact on thermals at vs. at clock rates we can already hypothesize that e clock will be flatter and less paretic in its fluctuations went on to liquid this conclusion can be formed just by looking at the previous frequency chart from the ten atti review we're running an 80% but sort of intolerable fan speed cloud for greater performance in the boosting behavior of course this also introduces an unbearable level of noise and so isn't accessible for long up times in fact the fan never jumps that high from the stock and Vidia fan speed curve to begin with attempts the plateau around fifty to fifty-five percent so even the stock curve it really doesn't want you to go beyond that unless it's absolutely necessary much for noise reasons well could have a new frequency versus time and temperature charge moving off the old one we see the pink line representing the GN hybrid 1080i which is able to sustain a higher clock overall and a flatter line that's a good thing there's a lot less fluctuation here keep in mind that has stated previously this is with a power virus test we're mostly burning the FETs here and so the cores don't boost as high as they would in a real-world video game test despite being burned at hotter than any other application for thermals we see the GN hybrid card operating below 50 c generally in the 45 t range was a PE version of the same GPM PCB at around 65 celsius won at 80% fan speed or 84 when at auto speeds and that would probably be worse if we were allowed to boost frequency to the same frequencies as in the hybrid version clause fluctuation is that no more than 30 megahertz in the worst cases on the hybrid card which is a pretty market improvement over the stock founders vision card going into FPS tests the things and out here that's most interesting is this card the hybrid version we've built of the founders edition 1080 TI is capable without an overclock applied manually of boosting higher than the stock settings that is because that we removed the thermal limit so we're going to be testing this no overclocked stock versus the stock 1080i with the cooling set up this and we'll be able to see it's just doing the model with no overclocking improves performance it does and then we'll add an overclock for a couple of games on this that are not going too deep on it today that's still another topic we'll look at a few games though starting our FPS performance testing with Ghost Recon wildlands at 4k and very high settings we see the gtx 980ti is outperformed by our stock 1080 TI hybrid mod that's without any overclock supplied this is because again the thermal limiter has been removed so the GPU can boost higher clocks when the FE cooler would allow this result would be more exaggerated at least to some extent as play time extends our FPS difference ends up being 55 FPS average at verses 57 so too when only the cooler is exchanged lows also improved marginally from 45 and 42 to 51 and 47 and overclocked in the GN hybrid with a 150 megahertz offset and a 4 50 megahertz memory offset and our frequency in the range of 1952 20 50 mega Hertz depending on workload and voltage and power limitations that lands performance at about 60 FPS average that's an increase of approximately 9% over the stock at ecard or an increase of about 5.8 percent over the GN hybrid stock card moving on to 1440p we see the GN hybrid model again outperform the stock TI by just a couple FPS and no this is not within margin of error the difference is repeatable and measurable lows improve to 81 and 77 fps from 24 and 66 fps overclocked in the 1080i hybrid puts us up to 98 FPS average though the 0.1% values do reveal that our clock is bordering instability here have some of the lower values in that category backing down on the 150 megahertz core and for 50 megahertz that memory offsets would improve this metric a bit though you do lose some in the average department finally 1080p shows that the 1080i is genuinely capable of hitting 120 FPS average in this game at four lovers of high refresh displays at the setting under the right conditions in this case those conditions are the GN hybrid card we can push to 120 FPS from the stock cards 115 average with lows also improves markedly overclocking our hybrid pushes us to nearly 1 six FPS average that's an improvement of about nine percent of the stock MTV gif ecard or about five percent over the GN hybrid with no OC looking at for honor with extreme settings next 4k positions the 1080 TI f e at 66 FPS average superseded by the GN hybrid at 70 FPS average that's an increase of about 6 percent just by changing the coolers and is indicative of the kind of performance gains you should expect from a IV partner models of the 1080p I overclocked in our hybrid we land at 76 FPS average or about 15.6 percent faster than the FPS card pretty impressive or 9% faster than the GM hybrid stock card lows exhibits the same behavior as previously noted at 14 at 40 P is a hybrid model with an overclocked four pals our 1080 TI up to the 144 Hertz range now achieving 148 FPS average versus the 130 average of a stock F II card or 137 of the iris stock card low scale mostly linearly except for the 0.1% metrics again this changed and that's us a 14% increase in performance over the completely stock FP card looking at averages that would be non overclocked hybrid is about 5% better than being non overclocked to F II card 1080p is sort of comical with this came but we're still testing it for a baseline for the vast majority of people still on 1080p displays at 1080p the non overclocked to GN hybrid posts a performance gain of about 5.2 percent of the stock F II card with no improvement in blows the overclocked variant posts a gain of about 20 FPS or around 10 percent over the stock F II card about 5% over the hybrid stock card let's move on to doom looking at doom with Vulcan and 4k resolution we see the 1080 TI reference card at 90 FPS average outperformed by the 1080p IGN hybrid by about 5.8 percent our lows also increased marginally and this is fairly consistent in terms of percentage gains when looking at previous tests this shows a continued at 5 to 6% increase in the head room when better coolers come into play we've also got 1440p and 1080p numbers were doomed published on the website if you're curious that's linked in the description below let's do one more with Sniper Elite we've got ashes and synthetic tests alongside the other 14 14 10 a t-test in the article link below again for more results at 4k with high settings with dec 12 and 18 can abled at the 1080 TIG and hybrid push is 84 fps was low at 67 and 65 comparatively the 1080 TI fe card runs at 76 FPS average at 64 1% lows and 60 10.1% lows the difference of averages is about 10.5% for one of the more substantial games we've measured so that was pretty fun test the thing that I was actually really interested in was just the vrm testing stuff in terms of the phat thermals I really want to do the RAM testing in the future but we need to upgrade some equipment so that I can have more thermocouple running simultaneously because we're maxed out at 4 right now overall the results look good this does a few things for us one if you don't want to do this that's fine just know that the five to six percent gains we see in most cases stock with a better cooler should be representative of what you'll be seen in a couple of weeks when the AIB partner models roll out those obviously might have some pre overclocks that changes things but if you're strictly looking at the cooler with the same clock configuration other than boost behavior this is representative of what you'll see there in terms of overclocking and doing this we get a lot of headroom now of course you could overclocked at 1080p IFE as well and we still need to do some of that it doesn't really really packed lately with trips but overall this project definitely is worth it for a bit of fun if you like doing this kind of thing if you'd prefer not to then wait around for AIB models if you are actually buying a ton of UTI check our review to see if that's if it's a good buy for you but a Ivy models will do more or less the same thing just without the effort involved if you're ok with doing the effort and bind I don't know if this helps for now it's generally is about a hundred dollars buying a hundred dollar kit is a fun project it's not too hard to do if you watch our videos and the performance gain and thermals is tremendous 40 Celsius in the core clock about the same for the FETs the performance gain and stock frequency is also noteworthy couple percent 5 6 percent most games I've stated sometimes more and then the overclocking Headroom is also improved because we've removed some of the thermal limitation and the sets do just fine for the most part well actually they do just fine period though we still need to measure so that's all for this time go to patreon.com/scishow and ice themselves that directly click the link the description below for the article or the sponsor thank you for watching I'll see you all next time you
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.