Gadgetory

evie J's new icx cooler makes moves to directly measure vram and vrm component temperatures which is a direct response to the thermal pad to moles that we exhaustively tested a few months ago at this point the icx cooler theoretically addresses the fact that GP diode temperature is no longer is the most important point of thermals on a video card and that other components run hotter due to the now lower fan speeds as a result of the GPU governed PWM we have some parts custom-made to validate individual component temperatures of IC X and AC x coolers since the EVGA makes big claims without each individual addition to the new IC x cooling block for example the aluminum fins they call the pin fin the backplate and the holes in the heatsink itself we validated these claims individually rather than just testing the cooler as a whole versus the old cooler and have sc2 and ftw2 cards on the table before getting to that this content is brought to you by our patreon supporters who helped fund in-depth and unique testing like this this type of depth and custom tooling is made easier by supporting at patreon.com slash gamers nexus EVGA Zhai CX and a CX cards will coexist with one another they are gapped by about $30 in price who is a $30 price of Delta between IC x and AC x and that extra cost is basically split between the addition of thermistors across the board and then the cooler overhaul alongside some other additions like to emcee use for processing the data from the thermistors of course the thermistors use our negative type they are NTC thermistors and that means that they respond to higher temperatures with lower resistance that's how the temperature measurement is it created in the various components that we're testing so for these IC x cards they've got placements near the BRM components the memory components and it's placed adjacent not directly atop the casing for each of these devices the services are used to allow individual control of these ii v RM fans but i'd be the fan on the right so the BRM temperature can be independent it's no longer bound to GPU pwm signals which is somewhat flawed at this point because GPUs have become a power efficient enough that they don't require as high of a fan rpm to keep cool but the are still burning because it's not feasible to place these thermistors directly below a vga device like a VRAM module EVGA places the thermistors closeby with some sensors landing on the backside of the PCB we actually acquired some PCB diagrams that show the placement of each thermistor shown on the screen now the top side of the ftw2 PCB has three thermistor is located near but not atop the vram modules there are four more thermistors that located topside for the power component coincidentally EVGA is measuring the same power components we used in our huge EVGA vrm thermal benchmark a few months back and that's been MOSFET number two counting from the bottom up MOSFET number seven and MOSFET number ten and they're also measuring the memory vrm temperature we generally agree with EB J's placement and specific positioning of the thermistors because the number to the second from the bottom that set specifically gets a little warm because it is closest to the motherboard so depending on the airflow and the case use that can get pretty highs flanked by two other power stages and phases on either side so it's going to get warmer the seventh one is in the middle of the board it's going to be the warmest there because it's got components all around it and then the tenth one we didn't really use an hour testing but EB J's added it to their IC x cooler this bottom side diagram shows the rest of the thermistor locations one of which is an additional GPU diode backside measurement the other is the one volt PLL minor rail which is used for non compute functions and can mostly be ignored as for the component level analysis this board is largely the same as the previous EVGA FTW 1080 vrm we previously worked with build Zoid to analyze that board and his analysis remains accurate we'll put a link in the cards for that video EVGA has switched one main stain for the vrm instead of using the NCP eight one three eight to fetes that they reason previously they've now switched to fairchild at thirty thirty five sets the previous fest used on the FTW boards were rated for thirty five amps continuous the new ones are rated for 50 amps continuous and continues EVGA is way overkill designed to the vrm on these cards the NCP FET if you recall our PCB analysis of that is a combination of the high side the lows and the driver I see so moving to Fairchild really we're just getting a higher-quality fetch so 30 35 that's used but in terms of your usability for overclocking you probably won't see a huge difference and the vrm it truly is overkill EVGA is still using a five phase that's doubled with the NCP eight one one six to doublers that builds a detailed in that vrm analysis the same voltage controllers also used with the same four hundred chillers pwm into the doubler which then splits into two 200 kilohertz signals for the phases this means that your power quality is the same as a five phase so the 10 faves design is almost entirely a marketing stunt and not really all that relevant to overclocking quality that said going for 10 phase does mean EVGA can spread the heat out over a wider area which is probably the main thing they're going for here as for the memory vrm that's still the same on semi 4c 805 and a high and low side faites if the thermistor and power design interests you further please check the link in the description below for the full article where we've got a lot of depth on the IC x cooler and the ACX cooler and that mostly covers the basics there are a ton of small changes made to the cooler itself like EB J's newfound obsession with thermal pads wonder what triggers that but we'll test each of those individually in the thermal benchmarking for the records those pads are rated at 1.2 watts per meter Kelvin before getting to the testing to clarify a few things the cards themselves the SC 2 and F TW two cards are largely the same as the AC X 3.0 versions the only difference again that one set change for the VR M is different and then the cooler is different the cooler in addition to the thermistors is really the only thing here that's the main focus point because even the clock rate is the same between an f TW which I have over here at ftw2 the clock rate is the same so the only thing that you're going to see that will cause any variance at all on fps between these two devices is going to be GPU boost 3.0 which is Nvidia's boost functionality there's different boost variants one GPU to the next so unless you had a massive sample size you really can't firmly state that one is better than the other and it doesn't matter anyway because the GPUs are going to boost to whatever o'clock rate they want and performance other than that will remain the same you might see one FPS variance so we're not doing FPS benchmarking with these cards there may be some in the article but not in this video we're focusing entirely on the things that matter here which again is just going to be thermals for this benchmark in terms of the ACS 3.0 cards if you are an owner of SC or FTW before getting some tests you should know that you can upgrade for 100 dollars we'll talk about whether or not that's actually worth it in the conclusion I have that you should probably stick around for if you're thinking of upgrading but the plan is basically $100 step up you can send them there your FTW or your SC and then you get the equivalent IC X version back let's get into the test in our complete thermal testing methodology is defined in great detail in the article includes description below this one is a really detailed article so be sure to check it out if you don't normally with regard to basic testing we use it logging software to monitor the NPC's on EVGA ACX card and we use our own calibrated thermocouples mounted to power components individually which we can also use the further validate EVGA zone thermistor placement and accuracy our thermocouples use an adhesive pad that is 1/100 of an inch thick and 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 up to a login meter we also had a custom plate made to allow us to individually test the components of icx cooler as stated previously note that we are using the same SC 2 card for these side-by-side ACH versus icx comparisons that is critical because the GPU for each test again can have some very intim that's also why there's not much fps difference between one to the next other than GP boost as for the ftw2 testing we're comparing that against the straight a CF 2.0 FTW card strictly because of the vrm issues we talked up previously and we're using the same methodology for our vram and so don't pad towards your testing that we used previously let's start with that one this chart shows temperature versus time for our thermocouples that we mounted all over the AC X and FTW cards yellow orange and red represent temperatures for the ACX cards that be the warmer colors not to do that intentionally but warmer colors for those and then maybe Andrew can highlight those on the screen green teal and blue which is buried by the lighter blue represent the icx card white represents the ambient temperature which is a modifier that we applied to equalized room temperature as it relates to performance as it's done by collecting ambient for each test pass note that getting these numbers is only possible using our trusted thermocouple method that we've tested previously since the ATX FTW did not have EVGA these new sensors with our testing we're seeing MOSFET number seven and number two are mostly equal on the ftw2 represented by the blue colors with both maxing out around seventy to seventy two point five Celsius comparatively the FTW ACX cooler posts fat number seven and number two at around 76 to 79 tells us and that's a temperature improvement of between four and nine C on the MOSFET the back side of the PCB also shows tremendous improvement with the green line indicating we've now mostly achieved parity with the front side set temperatures this is largely attributable to the new backplate and reduced vrm temperatures overall because those do bleed through the PCB although it's not the most efficient transfer now we're moving into the SC two AC x versus IC x comparison which is where things get interesting we're using our custom plates to test the performance on the same a card and GPU with a controlled fan speed of 1700 rpm on both fans even the vrm fan we're testing with combustors per mark again to burnin power stages we're also using EB J's new thermistor readings and first we're looking at the difference of just the back plates this first test was done using the IC X cooler its entirety with only the back plate swapped we ran this test specifically because the EVGA suggested that the new back plate had a big impact on thermals as shown here the temperature difference is roughly within one Celsius for power five power for power one and memory one and also GN zat MOSFET number two thermocouple and even the GPU diode we see a much larger difference in power to and memory 3 which is more of a factor of where those thermistors are located than anything let's look now how the entire cooler performs versus AC x3 no adding in those numbers of the same rpm between the coolers we see that GPU temperature is effectively identical with unusual test variants and our thermocouple on set number two shows an improvement of approximately 6000 versus a CX if we highlight power you'll see that this holds it true there as well we're seeing that several degrees improvement on the second set no matter how its measured the difference between these measurements is that UGA is using a thermistor placed adjacent to the FET while we're using a thermocouple mounted atop effect casing neither is perfect as for other temperatures memory temps on the left side are down in consequentially with performance improving drastically as we approach the inductors and capacitors Bank toward the right that the memory number 3 which needs an improvement of about 14 salties in our testing just to throw in a comparative charge versus other vendors this benchmark shows all cards set to their auto settings and adjusted for the delta T value that we measured during testing EVGA is icx FTW to Land's at around 48 celsius delta T load with it almost eight degrees Celsius delta T idle temperature the latter temperature is seemingly high as a result of the idle fans when it's not actually doing anything of course you can have a higher temperature idle so former doesn't look impressive because the EVGA is opted to allow for an excessively warm GPU in exchange for improved noise levels while leveraging their new independently controlled BRM fan to keep power under control this is how the icx card is meant to be used and is why the GPU temperature is no longer the full story speaking of noise well we'll take a look at those with Auto fan speed and stress testing workloads we generally saw the FTW too icx cooler opting to run at CPU fan 1900 rpm and via ramp an at 2100 rpm with these configured and with all passive cooling on the CPU power supply and everything else in the bench we're seeing a noise output of 46.2 DBA this is the only components in the system making noise when we moved to a fixed 1,700 1,700 fan speed we're seeing a noise output of about 40 1.8 1400 by 1600 which is where we were operating when averaged over low load gaming puts us at 38.2 DBA and 1700 by 1800 is where we operated during most gaming workloads of sort of triple-a titles that's what you should expect for the most part and that configuration had us just over 42 da we also have before and after V BIOS update numbers for the original AC fftw card which uses fans that are physically identical other than the MCU control those are on here as well EVGA is decision to launch IC acts at this point is a little interesting Pascal is not new and there are new cards theoretically around the corner so why would they do it well it's really just because they've got this big AC X thing looming over them you search for any of their cards on Google you'll probably run into thermal testing maybe our vrm testing maybe their forums and it's enough to cause concern for new buyers now just to kind of recap what we said previously with the updated VIII bios which is not necessary if you have the thermal pad and the thermal pads the AC x cards are fine there well with an operating temperature the V RMS can handle a lot more temperature than the cards are putting out especially with those updates so if you're on one of those you're fine but from EVGA is perspective their strategy is to get out from under this by putting out a new card with a new name so it's not going to bring up all this controversy from a few months ago so that's the reasoning for launching icx now the next thing is if you own an AC X card you could trade it in with a hundred dollar bill in the envelope and get an ICS card is it worth it well not really the only main reason to do that is if you want new toys to play with because again if you have an AC X card and you've done the thermal pad mod which is free for most countries anyway it's fine those temperatures are operating just fine the only reason you'd really need to do the trade in would be for resale value because that will probably go up on my CX cards the AC X ones are again painted by that image even though it is largely a non-issue at this point as we've stated in a very detailed long video but resale is really the main reason to upgrade because clock rates the same fps is basically going to be the same temperature is yes are really improved in a lot of places on the board not all of them but everywhere that we would like to see the improvement EVGA has improved temperatures but that's not necessarily the reason to upgrade a card not for 100 bucks anyway if you're a new buyer though and you don't own an AC X this should absolutely be on your raid icx has made excellently informed engineering decisions and from a testing standpoint these cards are going to be much easier to work with from an enthusiast standpoint you have a lot more sliders and toys to play with which is always a cool thing folks who don't yet only ten series cards should be seriously considering icx and that's really mainly for the finer tune control over temperature noise and these fans are exceptionally rare these days generally for a heatsink for a video card which is all these are heat sinks and coolers you don't see much improvement the innovation is in different thin design and this has gone a bit further than that so we hope EVGA licenses its patents that are relevant to this to other vendors so they can begin adding thermistors to critical non GP components as well for another point if you're curious about EVGA thermistor accuracy we validated them with our own thermocouples by being basically adjacent and the thermistors are perfectly precise from one task to the next showing almost no variance under controlled conditions it's a good thing that means you can run the tests and trust the numbers which was the first question we had we also found EVGA temperatures to closely match our own thermocouple temperatures despite be in place technically off the package from the components they're measuring and another point still EVGA has added a 10 amp fuse to the icx cards that will blow before any major power issues occur so the chance of this happening should really be indefinitely small to begin with but you'll now have a fuse protecting the rest of your computer in the event something does go wrong that alone is a major critical move and hopefully other vendors adopt this if they haven't already so with the $30 price difference value is good for folks who don't yet own a video card of this class but again we say skip it if you're on a CX with the thermal pad mod and if you don't have a CX or anything else get a CX and get I see a and a final reminder again FPS performance is the same so look up FTW or SC for more benchmarks on that we've tested both of them Pascale still has a hard stop around 2100 megahertz or thereabout so you're not going to get much more overclocking out of the new cards either despite having a different vrm overall job well done by EVGA with this improvement and iteration the $100 upgrade is a nice way to try and help people who might feel burned by the previous purchase but it's not necessary either for you to actually upgrade so defuse the cooling apparatus the thermistors all good additions hopefully these get carried onward into new cards if you like this coverage is always patreon.com slash gamers Nexus to help us out directly or click the link in the post role video link to the description below for the full review with a lot more depth if you want a different perspective on the EVGA ACX cooler you can check out our friends at Paul Kyle and Jay who are all at the event with us will link them in the description below thanks for watching I'll see you all next time you