Gadgetory

this is our Titan XP mod we took an EVGA hybrid cooler which is just a CLC stuck it on to the Titan XP and tested it to see if the clock speeds were increased just by removing the thermal barrier or if we could smooth out the clocks over time we have all that testing for you here today thanks to it reader grant for loaning us the card and before we get into that this coverage is brought to you by ifixit.com who provided the tools used for our tear down and this mod you can go to ifixit.com slash gamers Nexus and use code and gamers Nexus for five dollars off of an order $10 or more we would recommend the pro tech tool kit which is linked in the description below so today's goal is to look at how this performs against itself with the stock cooler as it was originally sold the unit is $1200 so the NVIDIA Titan XP that's the new 1 2017 edition and we ripped off the cooler because normally with the founders edition mods if you pull that cooler off we find that the performance in terms of clocks versus time and temperature improves with a better cooling solution like this one now you could run the vrm fan at something like 80% and have similar results but this helps us in a few ways that we weren't expecting because we're now measuring the vram temperatures alongside the power component temperatures which we haven't always done for these mods so the vram temperature in particular is new for us we've got results on that today other than this the mod is it's about $100 liquid cooler kit when you're spending 1,200 bucks it's kind of hard to gauge at what point value still matters this is not a card for gamers as we said in our gaming review of the Titan XP you'd be better off with something like a 1080i because when there are overclocked or stock it's pretty much the same performance but that's for the Titan XP review that we already posted what we're doing here today is just looking at how does it perform thermally and does it improve by changing the cooler not necessarily looking at is it worth buying you can check our review for that and before getting into the test data as always testing methodology is it linked in the description below so we've got the full article on this there which will a few extra charts not found in this video and it's also got the methodology including how we're using thermocouples to measure different temperature of temperatures of components on the card so you can check that link for more information on all of that let's start with an easy chart this one is frequency versus time showing the performance of the Titan XP stock versus our hybrid mod and the tremendous frequency difference here not like we've seen in past mod but there's a big swing in thermals as you'd expect we'll talk more about that aspect in a moment with regard to clocks the hybrid mod is able to boost slightly higher in this power virus scenario averaging about 14 43 megahertz core whereas the stock card with the stock cooler and auto settings ran 1392 megahertz on the court that's an extra 50 megahertz or so without any overclocking effort so not bad well the Titan XP is up against its clock limit already so gains are minimal note also that these tests are conducted with a power virus scenario so that means that the clock is enumerated differently from a gaming workload that's okay because we're just trying to stress the card thermally but if you're testing it with games you'd see a higher clock and lower temperatures overall on the power components getting into tests that involve our thermocouple measurements on the vram and MOSFETs this chart shows our original Titan XP results would be stocked cooler before the mod GPIO temperatures were locked at around 84 C which is the card's thermal threshold for clock limiting MOSFETs were operating around 60 to 70 C in this scenario well within reasonable operating range and the vram was running parallel to the GPU temperature at about 80 to 83 c this is important by looking at how vram tends to follow GPU temperatures though at a slight delay as indicated by the 200 to 350 second mark on the chart we learned that there is a need to pay attention to thermal density with these tests rolling into the next chart with our hybrid model using a 23% vrm fan and 100% radiator we now see the GP diode temperatures around 40 C having the GPU diode reading previously MOSFET and power component temperatures are almost identical to the previous test though note that the hybrid does take longer at the rapid fat temperatures than these card ultimately we're not worse off of the FETs but we can maintain equal temperatures of just 22% on the I'm fan rather than north of 50 to 55 percent theorem is interesting and is the newest item we've added to our measurement the around temperature with the hybrid has dropped down to about 75 C on a module we're measuring whereas it was previously around 82 see our current hypothesis is that this reduction is primarily because of the increased efficiency with which heat is removed from the GPU reducing heat between all of the vram modules the vram is primarily cooled by the base plate so removing that vapor chamber didn't hurt the RAM temperatures at all in fact it actually helped since we can now focus all of our air cooling efforts strictly on power and B RAM components the GPU is taken care of within a closed loop as indicated by the EVGA CLC it mounted to the GPU proper if 100% radiator fan speed is too high for noise reasons we also tested at 40% speed by plugging into the motherboard directly this is where we can see the 100% radiator fan speed exhibiting diminishing returns considering our 40% gpio temperatures only 12 to 13 C higher and still around the 50 C mark max noise is measurably lower at this speed as we'll show in a bit with temperatures still under control the ERM temperature is still improved with this fans filled with fat temperature is still tied with the stock configuration making 40% pretty appealing overall here's a GN EQ chart to make things easier to read over the same test period and averaging a few hundred cells of data after accounting for ambient we see the hybrid modify performance of the measured power and VRAM components alongside the expected GP diode improvement interestingly again it almost makes more sense to run a fan to the motherboard directly then just set it something like 40% while leaving the vrm fan at 23% we're cooling everything adequately and operating at a lower noise level than stock with 100% speeds on the EVGA fan not really posting significant benefits over 40% speeds but still performing louder these components are exciting to start measuring seeing as we've always known that hybrid mods benefit GPIO temperatures but we weren't sure how they affected componentry now we know it's still better all this talk about noise let's take a look at the noise chart first of all the idle temperature listed for the hybrid mod is when running the fan at a 40 percent noise profile as plugged into the motherboard directly I have 40 percent on the radiator fan and 22% of the an hour idle noise output is about 30 2.5 DBA higher than idle for other devices including the original Titan XP but that's because we've added another case fan to the mix again this test uses a bench that is entirely passive other than the GPU proper auto where the radiator fan is running at 100% our output is 39.1 DBA the stock cooler for the Titan XP operated at around 48 DB a year which is nearly two times louder by human perception given the logarithmic scale here's the thing if you wanted to keep that 40% radiator fan speed since we've shown it works just fine in our test configuration the difference would be even bigger we'd be looking at a forty seven point nine DBA for auto on the stock cooler as opposed to thirty two point five DBA for the 40 percent radiator fan and 20 percent PRM fan even when under load thirty three DBA to 48 decibels is a huge jump in perceived noise and makes it pretty convincing to just run a hybrid cooler with a 40 percent fan speed we don't normally get major gains with the overclocking in these hybrid mods but increased clock stability ultimately yields high performance metrics in gaming anyway we saw some increases in peak frequency for each overclock step with this mod ultimately settling at eight to 25 megahertz offset here's the overclock step in for the Titan XP stock card which will bump up to make room for the hybrid OC stepping boosting memory speed beyond 450 to 500 megahertz produced diminishing returns as a drag down the core clock likely due to power limit and availability for gaming the core clock is not only going to matter more than the additional memory overclock so we ran with a 450 to 500 megahertz memory clock depending on how it impacted the core let's get into some gaming benchmarks will have all these in the article includes description below but go over a few results here start off Ghost Recon wildlands at 4k the Titan XP style card operated an average frame rate of 60 FPS with loads at 52 and 48 without any overclock our Titan XP hybrid model improves to 61 FPS average 54 fps 1% low and 50 FPS 0.1% low that's not a huge jump at about a 1.7 percent improvement but it's an improvement nonetheless overclocked in the stock car that landed us at 65 FPS average and overclocking the hybrid bumped that to 68 FPS average result is a 4.6 percent improvement in when comparing Oh see - Oh see I have 1440p it's of similar stories a stock Titan XP runs a 95 FPS average 35 fps 1% low and 32 FPS 0.1% lows the hybrid model low performance up to 85 and 82 it notably with averages at 97 FPS average overclocking performance shows the same limited gains we experience a bump in average with a bigger jump upward in frame time consistency here's a frame time chart showing the non overclocked frame times of the stock tight next P and the hybrid dot next B just to illustrate the difference it's not huge but you can definitely see it running Mass Effect and draw it at 4k the Titan XP stock card performs with an average FPS of 68 loads at 56 and 53 while the hybrid version runs a 71 FPS average and lows at 57 53 the stocks of stock improvement is about 4.4 percent for this one for 1440p and 1080p results for this game check the article link to the description below let's look at doom with Vulcan and async enabled the Titan XP stock card operated at a frame rate of 94 FPS average which improved to 108 FPS average when the overclocked the hybrid mod meanwhile lands at 99 FPS average so it's 5.5 percent ahead of the stock Titan XP cooler and overclocking gets us to 113 FPS average for an increase of 4.3 percent over the Titan XP air-cooled card when overclocked or an improvement of 13 percent over the hybrid models no OC one final game for this video that will close out the rest of these game tabs will be in the link in the description below along with 3d mark Sniper Elite 4 with the x12 and async it places the Titan XP stock at 87 FPS average with the hybrid card bumping up to 94 FPS average without any overclock supplied that's just from the thermal Headroom improvement from the hybrid cooler so not bad this improves our performance by about seven point eight percent over the stock cooler overclocking the Titan XP with its stock cooler it gets the card 298 FPS average for a gain of 12 point 2 percent over baseline the Titan XP hybrid overclocked a bit higher and held onto the frame times better netting 101 FPS average on the OC so that more or less wraps it up this is more about the noise and thermal than gaming improvements with gaming as we just went through it's not a huge deal for the most part it the very best case was snipe early where we seen something like many percent improvement just from that change so we're not really seeing anything major otherwise the noise department's a completely different story because that can be dragged down to less than 40 decibels in our passive testing environment with the noise floor about 26 DBA and performed better thermally and actually better than FPS for the most part then well actually definitely better in FPS it's a limited percentage then the stock Titan XP with the stock air cooler it's not a bad mod if you are spending $1,200 anyway an extra hundred dollars doesn't sound too daunting I would like to see this kind of cooling improved on in the future for reference cards is probably not going to happen there's a reason reference cards are designed the way they are and the in Nvidia both do the blower fan designs they are easier for sis to implement they are more versatile so it makes sense for them to do that especially because if your mass producing them it's cheaper to just buy a whole bunch of one cooler but if you're in the mood to mod something it's an easy mod to do this is just not something we recommend for gaming still but if you're buying it for something like machine learning like grant did then maybe it's worth considering considered look into that for your uses depending on how you feel about noise levels of the stock cooler and how you feel about temperature output depending on what else is in the box as for the rest of one quick note on the game for honor is that interesting bugs with for honor throughout all the testing we've done with it it does just fine when the cards are stock but as soon as we overclocked them a lot of them have weird instability issues and it seems to be just that the that particular game is causing more problems with clock stability and overclock stability than the other games we test just something with how it's optimized or how the shaders are used or something like that this card in at ibrid state actually had problems with for honor to the point where I had to down clock it back to make sure it was stuck at the original stock titin XP settings before the hybrid mod because it would actually start stuttering to the point where the game was unplayable even at 4k but if you drop the clocks down so that they're similar or equal to the Titan XP stock cooler version of the card then it hit better it was fine so really weird issue that was for honor only none of the other games had that for the rest of the information is always that link in the description below you can go to patreon.com/scishow immers Nexus to help us out directly if you like this type of in-depth content and would like to help fund it or if you don't want to do that you can pick up a shirt store dock cameras next it's not net for one like this we're adding tri-blend should be in stock this week so thank you for watching as always I'll see you all next time you