Gadgetory

the $1,200 Titan XP 2017 model found its way into our hands on loan from a reader before cranking away on the hybrid bond we ran all the numbers and benchmarked the car just like we would for any gaming GPU on our bench this isn't necessarily something meant for gaming though it still carries a GeForce GTX branding and could certainly throw down some numbers in benchmarks today we're testing the Titan XP versus the best and worst 1080 TI cards on the market all tests were conducted under stock conditions at check back tomorrow for part two of our hybrid mod with part three arian shortly after containing the results for the Titan XP is no thermal limit before getting to that this content is brought to you by ifixit.com who helps provide the toolkits necessary to mob the Titan XP into the hybrid card that will soon be unveiling on YouTube I fix it makes the pro attack toolkit which is $60 on their website you can find a link in the description below for more information use code gamers nexus for five dollars off the Titan XP at $1,200 isn't really advertised for gaming but that won't stop people from buying it for gaming because they want the best that's kind of what a lot of people do in the market and that means we're going to test it from a gaming standpoint now this is something that is more suitable for something like neural-net or machine learning type applications that grant the reader who loaned us the card is one of those developers who works in neural net applications and so could take advantage of the extra one gigabyte of memory over something like a 1080p eye for gaming it's an extra $500 potentially for minimal gain but we will find out shortly and just to kind of throw that in there although NVIDIA doesn't market the cards as a gaming device heavily it does still contain GeForce GTX branding on it and it does have the words gaming appear a few times in its product page information on the official website since we don't specialize in neural networks I took the opportunity to ask grant how he would use the Titan XP nuts more targeted application grant stated the following data sizes can vary and the GPU limits are based on data size and the applied algorithm a simple linear regression can be done easily on most GPUs but when it comes to convolutional neural networks the amount of math is huge I have a four gig data set run at CNN on the 1080i but it can do it on the Titan X also for us CUDA cores matter a lot and the good machine learning algorithms even Google's tensorflow need kudos and and video specific drivers to use the GPU s why does not do anything for us multiple GPUs can be used to split up data sets and run them in parallel but it's tricky as hell as neural networks with multiple cards it's better for us to run one algorithm on one card and another algo on the next card even Google created their own version of a GPU for deep learning that can be farmed much better than any Nvidia option so that's what it's meant for but we are going to be testing it with gaming because again people will be buying it for gaming whether or not that's the best thing they should be doing just to clear of a few things there is GeForce GTX branding on this card in the initial renders of the Titan XP that Brandon was not present they were 3d renders so they were not the actual product the fact that it appears on this card doesn't really mean anything other than nvidia probably took the same tooling and applied it here once again if we look under the hood the card here the Titan XP with the little P uses AGP 102 - 450 GPU whereas a Titan X 2016 edition the one we previously called Titan X capital P that one uses GP 102 - 400 so this is a rev higher its - 450 and if that's not enough for you to tell the difference you could also look at the back of the card the Titan XP lacks DVI ports whereas the Titan X 2016 has a DVI out so that those are the two primary differences here since the branding on the outside looks functionally the same actually in every aspect for the full specs on the card you can check the article linked in the description below which contains the review and all the benchmarks that you'll see here along with some extras the Titan XP again $1200 card is about $500 a head of MSRP for the 1080i reference card and other than that there that's I mean that's really all there is there will be no adding board partner models of the Titan XP the one you see here is our modded version it does not come with a liquid cooler it comes with a normal F equal er and it's Nvidia only you can only buy it from Nvidia com getting to the benchmarks will put the OC stepping table on the screen now since we referencing that as we go through some of the overclock benchmark numbers the hybrid mod will come shortly here after and it will probably contain a bit higher clocks because it's liquid-cooled for full methodology and testing procedures check the link below for the article in the description looking at total system power draw at the wall with 3d mark fire strike extreme the Titan XP system draws around the same power as a 1080 ISC 2 cards from EVGA oh and brother stock each system draws around 345 watts overclocking the Titan XP gets it to around 400 watts total system power draw for an increase of about 15 percent of the stock card the only configuration that draws more powers on this is expectedly the crossfire 580 plus 480 grouping at 460 watts or about 16% more power draw than the ti XP looking at for honor next to the Titan XP card stock configuration draws about 355 watts about 10 watts behind the 1080 is CPU a card with the boosted power budget on that PCB overclocked MPI XP pushes its power draws at 375 watts total system power consumption for an increase of about 6 percent in this particular workload Ghost Recon wildlands post total system power dried around 370 375 watts for the stock se 2 and for the stock Titan XP cards with an overclocked Titan XP we see that jump to 424 watts total system power drop the wall that's an increase of about 13 percent power consumption from the overclock alone finally idle power consumption for the full system is at about 75 watts the ti XP system and about 73 watch for the 1080 T is e2 system given our less tight tolerances for power testing compared to other testing that this is effectively identical let's get into the most interesting aspect of this cards and thermals traditionally the Nvidia founders Edition coolers have been the most limiting aspects of the cards this can typically be compensated with higher fan RPMs at the cost of noise but our hybrid models resolve both issues with one go as do the AIB partner models which will be present for the ti XP the Titan XP doesn't do too poorly here though we're still hitting a thermal wall around 84 C with a fan stopping at around 55% fan speed from the default profile when under load but clock fluctuations and fan fluctuations are shockingly study with this card but we're still losing at least 100 megahertz off the clock as soon as thermal start rising our range is about 120 megahertz with the important point being that the clock remains steady after its drop that 100 megahertz that's good and bad it's ideal for the classroom and stable as we're seeing here for the most part but it is less than ideal for them to drop as a result of thermals that can be fixed by a better cooler that keeps frame times more consistent in the least our hybrid model talked about this more and attempt to keep the higher clock speed maintained keep in mind that this is a power virus workload so the clock does not enumerate in the same way that it would during a gaming workload fox will therefore be lower here in looks but we're just looking for stability not max speed moving now to a thermal analysis chart we're looking at component temperatures using thermocouples mount the MOSFETs and the vram components we can compare against 1080i FP momentarily for a look at a nearby EPI card the Titan xpf at number 7 hold the temperature around 68 to 70 Celsius with its vram temperature around ADC this is pretty common for the vram temperature though we haven't expanded that test into the 1080p IFE just yet GPU temperatures for the Titan XP are in the 84 C range as you'd expect which is roughly equal to the temperature of a 1080i F II card let's get the F II temperatures on the screen now just to expand this chart of it the temperatures are all roughly the same once we're under load which makes sense seeing as the Titan XP is using a PCBs basically a 1080 TI fe PCB Fe PCB for what it's worth was one that we praise fairly highly in our component level vrm analysis by build void but again it's about the same as we see on the Titan XP here fans ramp at exactly the same profile between the TI x PM at 1080 TI if you are curious as do temperatures and clocks so the 1080 TI runs a higher clock overall the TI GPU temperature is wanted to see higher in the beginning of the test but the cards get closer over time for noise testing we're measuring at 20 inches away and doing so with a passively cooled system the only component making noise in the system is the GPU fans be that idle is the same for the Titan XP as for all other founders Edition cards of this generation running at around 22% idle fan speed with a be partners demonstrating the noise floor with passive operation under idle conditions the Titan XP maintains an output of about 31 DBA idle which should be covered up by the case fans and most systems Auto speeds lands the Titan XP at 47.9 DBA when operating in the 55% fan speed range compared to AIB partner 1080 TI cards in the 30s and 40s this demonstrates the value of those AIB cards but again none will exist for the Titan XP at 50% fan speed we're looking at a noise output of around 40 4.9 DBA with MSI and gigabyte around the same noise levels with EVGA a bit higher you'd ideally never run a 100% fan speed on any of these GPUs but we've included the numbers to demonstrate maximum noise possible anyway moving on to FPS benchmarks we don't expect to see any extraordinary gains over the 1080i cards but we'll see if that $500 difference makes any impact on our gaming tests starting with Ghost Recon wildlands at 4k and very high settings the NVIDIA Titan XP stock card performs at around 60 FPS average with lows at 52 and 48 this puts the card tied with the overclocked 1080i gaming X and our own overclocked hybrid 1080i f:e mod leading stock 1080p I cards by a couple percent versus the 1080p IFE reference card we've got about an 8% gain from 55.3 to 60fps average over the AIB partner 1080 is he to card that we just reviewed configured stock we see a percentage gain of about 2.2 percent average with frame times effectively equal that means you're spending over $200 per percentage point gained in performance overclocking doesn't change much the ti XP takes the lead at 65 FPS average followed next by the 1080 is C 2 a 2 FPS average the percentage difference here is 5 percent so we're only paying $100 per percentage point once overclocking is permitted at 1440p the Titan XP performs identically in averages to the 1080p is C 2 card and close to the gigabyte extreme RS card both are priced at 720 and 750 dollars respectively currently anyway the Titan XP it falls slightly behind the 1080i cards and frame times are not in any meaningful way however this is a trend we see in a few other places later overclocking boost the ti XP to 104 FPS average that we run at the thermal and power limits that threaten clock stability and therefore or frame time performance we're at 104 FPS average versus 99 average on the closest Natt IOC card that's about a 5% lead again consistent with the 4k results we see the same performance at 1080p with the Titan XP overclocked faltering and frame time consistency in exchange for a 5% lead over the EVGA sc2 and averages Mass Effect Andromeda at 4k positions the Titan XP at 68 FPS average just behind the 1080i gaming X and ahead of the 1080 T is c2 from EVGA these three cards are effectively equal in performance given standard deviation run to run and are completely equal with regard to user experience at 1440p the Titan XP ties again with the MSI 1080i gaming X and slightly leaves the EVGA 1080i sc2 cards all around 131 to 132 FPS average running Doom at 4k with Vulcan and ultra settings with Titan XP star card runs around 94 FPS average ahead of the 1080i reference card by about 5 percent Titan XP is effectively tied with the pen ATI reference card once the thermal constraint is removed as seen in our hybrid mod and falls behind at the higher clock rates of all the other cards in the bench a real argument Titan XP boosts it to 108 FPS average on this clock sensitive game particularly benefiting from a big memory overclock begin about 3% of the 1080 TI extreme horas and that's rounding up if we're being generous leading to about a $160 cost per percentage point gained running for honor at 4k the Titan XP stock performs around 72 FPS average title to gaming X and averages but behind it marginally and lows with 52 fps 0.12 61 FPS yo kwon percent lows 1080 is c2 runs a 71 FPS average tailing the Titan XP by 2.2 percent once again overclocking in this game it proves problematic for nearly all devices we've tested with stability becoming an issue without backing off the OC in substantial ways still the TI XP manages to cling to the top spot just barely once overclocked though does falter again and lows slightly for at 4k resolution with the x12 annexing compute allowed our crossfire rx 518 480 combo to outperform a reference 1080i thanks to friends a multi GPU optimization in this game titin XP runs at 87 FPS average in the title just ahead of the game and acts at 6 FPS average and just behind the overclock 1080p is d2 card we've got a weed of about fourteen point nine percent over the reference 1080i or about ten point four percent ahead of the crossfire cards was particularly well optimized title overclocking gets us a reasonable lead of eleven point five percent over the sc2 overclocked card but in our costs per percentage point increase down to forty dollars since that's apparently a new metric that we need to use with this card clearly this is not a good buy for gaming do not buy this card for gaming you would be far better off if you need something high-end with a 10 80 TI because you could buy either one of these two variants of the 1080i this one is 720 today so that may go up we've been told and this one is 750 either way they're both going to for sure be sums with your below and that is a savings of something like $450 over a $1200 TI XP and the performance is roughly the same it's like two percent in some cases different and once you've overclocked at these outperform until you overclock that in which case it's maybe a five percent difference for that much money $100 per percentage point at 5 percent not such a good deal so we would not really recommend this for gaming however that doesn't mean it's a bad video card big difference right it's an OK card it's just not OK for the price so what this thing is meant for it seems we're not experts in neural networks is stuff like what grant does the guy who loaned us the device if you are doing something like neural net look for a review where they specialize in that kind of thing they can tell you if it's good for that application apparently the extra 1 gigabyte really helps in certain types of algorithms and workloads for gaming the extra 1 gigabyte of VRAM is totally irrelevant and the extra cores are clearly not giving us any kind of advantage over the really anything else we've got on the bench so that's it for the Titan XP review thank you grant for loaning us this card if you want to help us out directly you can go to patreon.com/scishow spawned our in-depth testing we have part two and three of this coming up very shortly for the hybrid mod results should be pretty interesting on that when you get a store that gamers nexus net to buy shirts like these and subscribe for more thank you I'll see you all next time you