RTX 3080 Ti Review: Rich People Only Please

oh hello i see you've been sent round to gather some information on the latest gaming graphics cards for the exceedingly wealthy patrons who you work for i've been testing the rtx 3080ti and oh well i'm sorry i didn't want to make any assumptions but you just don't really look like the kind of person who could afford this type of hardware i know pc gamers like to make believe but with a starting price of twelve hundred dollars for the founders edition to say nothing of what it and third-party versions will be going for on the secondary market i think you're getting a little bit too invested in your own personal fantasy world if you think that you'll actually be able to buy one of these i did run benchmarks versus the 3080 and 3090 and the radeon 6900xt though so i'm gonna press on but i'd recommend that all the poor and middle class people just close out this video your earthy musk is kind of off-putting for my more affluent viewers anyway excellent this is my first benchmarking video in a while at least hopefully one that i pull off successfully and as usual there are time stamps so you can jump ahead if you want to get straight to the benchmark numbers i'm going to quickly go over this founder's edition rtx 3080 ti and my testing setup and if this card looks familiar it's because it is as far as i can tell pretty much the exact same design as the rtx 3080 with the short pcb with the v-shaped cutout that allows for air to pass straight through the end cooled by two axial fans and an array of heat pipes i like this design aesthetically and the cooler works great for the 3080 non-ti but given how much bigger the 30 90s three slot cooler is and i finally got a founder's edition rtx 3090 by the way but and the fact that the 3080ti is being pitched as having near 30 90 like performance i'm wondering if the 3080 cooler will be enough for this card and yes there is still the poorly placed special connector dongle for pci express power and there is a geforce rtx logo with a white backlight the gpu under the hood is another variant of the ampere ga102 ga102-225-a1 to be specific built with samsung's eight nanometer lithography and sporting 10 240 cuda cores that's almost as much as the 3090 and a lot more than the 8700 that the standard 3080 has the 3080 ti also has 80 sm units and ships with 12 gigabytes of gddr 6x vram with a data rate of 19 gigabits per second relevant stats for all the cards tested today are here the founders edition 3080 and 3090 as well as the reference radeon rx 6900xt and yes i know the launch prices are laughable on many levels at this point but i listed them at the bottom anyway moving on to my test setup all tests were run on this test system with the gigabyte oris master x570 motherboard a ryzen 9 5900x 12 core cpu running with auto oc and pbo enabled cooled by an nzxt kraken x62 280 millimeter all-in-one liquid cpu cooler memory is a 32 gigabyte g-skill trident z neo ddr4 3800 cast latency 16 kit meaning the ram and infiniti fabric are both running at 1800 megahertz and we're on the latest windows 10 21 h1 version with game mode on and the ryzen high performance power plan enabled again here are stats for the rest of the system if you'd like to take a closer look let's move into the benchmarks and here are the actual clock speeds i was seeing out of the cards while in use note that the base and boost clocks for nvidia cards can often pretty much be ignored as the real world measurements are always showing them running at higher frequencies all of the founders edition cards averaged around 1 890 megahertz initially but during longer tests would drop to 1875 or 1860 as temperatures increased i'm still impressed with the rdna ii clocks team red has managed especially when you consider that there's an xtxh variant of the 6900xt gpu that can run even faster but clock speeds are moot if the performance isn't there so let's proceed with the benchmarks 3dmark times by extreme is a synthetic benchmark from 3dmark a directx 12 test and here the 3080s graphics score was about 10.8 percent slower than the 3080 ti that's a nice bump for the ti variant and it's also hewing pretty close to the 3090 which is only 2.7 percent faster many have speculated that the 3080 ti is just here to beat the 6900 xt so keep an eye on this trend the 6900xt beats the 3080 but the 3080ti beats the 6900xt 3dmark port royal is my other synthetic benchmark from 3dmark it's a ray tracing focused test and maybe you don't care about ray tracing because it can be a big performance hit or it's not available in the games that you play that's okay here are the numbers anyway and as expected amd still has some ketchup to do when it comes to ray tracing as the 6900xt is behind the 3080ti by 23.3 percent with the 3080 backed by 11.2 percent and the 3090 ahead by just 3.1 percent moving on to some real world games microsoft flight simulator 2020 is a newer addition to my benchmark suite a very pretty game that lets you fly anywhere in the world and while it is still a directx 11 title the focus on detail and draw distance means relatively low frame rates at 1080 the 3080 and 3080 ti are neck and neck with both actually out performing the 30 90 by about one and a half percent and the 6900 xt is behind by just over 14 percent at 1440p the 3090 retakes the lead and is about 2.5 percent faster than the 3080ti while the 3080 falls back 4.3 percent and the 6900 xt again comes in last down almost 14 percent at 4k the 6900xt regains some ground but is still behind almost 12 percent while the 3080 non-ti suffers its worst defeat yet down 18.7 percent next up is resident evil village a creepy game that i was very glad to be testing without headphones on you could also call it resident evil 8. it's a directx 12 title running at max quality the 6900xt came roaring back with this game putting up a 10 win over the 3080 ti at 1080 by hitting nearly 270 average frames per second the 3080 and 3090 both had poor showings at this resolution though at 1440p the rtx 3090 recovered somewhat and is only 2.8 percent behind the 3080 ti but the 6900xt had its best win out of any of the games i tested today and is 14.6 ahead and at 4k the 3090 moves back to the top of the nvidia cards being 1.5 faster than the 3080 ti but the 6900 xt manages to sweep the resident evil benchmark category going three for three by staying six point four percent in front i just can't quit you gta 5. i'm either still running you to troll people or because your load times got so much better recently or because you're still a hugely popular game that a lot of people play but at 1080 we are completely capped by the game engine's 187.5 fps limit which pretty much every card in this roundup hits even at 1440p we're still only seeing a 0.1 percent difference or so which makes me feel a little silly for running this test but surely i will be rewarded with the 4k numbers there we go at 4k all cards are above 120fps but still well below the cap so the lineup is more familiar the 90 is 2.2 percent ahead and the 3080 and 6900 xt are 10 and 11 behind the 3080 ti respectively shadow of the tomb raider is also a directx 12 title and at 1080 the rx 6900 xt manages to eke out another win over the 3080 ti but only by the slimmest margin 0.16 which could easily be attributed to run to run variants the 3090 also had a poor showing here which has been the case for a few of the 1080p tests interestingly at 1440p the 3090 continues to struggle but the 3080 ti is back on top although again it is by a super slim 0.05 margin the 3080 continues to be about 10 percent slower than the 3080 ti at 4k the 3090 reminds us that it is supposed to be the fastest gpu out of all of these being two percent ahead it just only sees 4k as a challenge which is why it wasn't trying hard at 1080 or 1440. do maternal is my only vulcan title and i'm running the normal version of the game not the ray tracing upgraded one that they just announced during the nvidia keynote at 1080 the 6900 xt takes another win hitting a monstrous 452 average frames per second 8.4 percent faster than the also impressive 417 fps that the 380 ti managed at 1440 the 6900xt wins again but by a much slimmer 0.8 margin and the 3090 has recovered somewhat from its poor 1080p showing at 4k the 3090 retakes the lead again though which is becoming more and more of a trend as we move through the games and the 3080 non-ti is once again 10 slower cyberpunk 2077 is running with the ultra graphics preset and at 1080p the 6900xt takes yet another win hitting 134.7 average frames per second which is 5.6 faster than the 3080 ti the 6900 xt wins again at 1440 this time by a 1.9 margin and while the lower res wins for team red are welcome i'm sure usually these flagship cards are marketed to better handle higher resolutions like 4k and at 4k the 6900 xd drops back by 6.3 percent behind the 3080 ti with the 3090 once again taking the lead position dirt 5 is last it's a racing game and i haven't tested many of those for some reason we're running the ultra high preset with resolution scaling turned off at 1080 the cards all fall in line nicely with the 6900xt beating the rtx 3080 but still landing 2.4 percent slower than the 3080 ti at 1440 the 6900 xt falls further behind by six percent but the 3080 ti extends its lead by about 13 over the 3080 non-ti and finally at 4k the 3080 ti is once again behind the 3090 by just 1.3 percent the 6900 xt is 10.7 percent slower and the rtx 3080 is 11.4 slower so there are my gaming comparisons but raw performance is only one part of the story let's take a look at temperatures and power draw here are the peak temperatures recorded for each card across all tests note that the gpu temperatures are getting more specific there's both an average temperature hotspot temperature and a gpu memory junction temperature listed for nvidia cards and hardware info these days i noticed a lot of fan noise from the 3080ti while testing and while the average temperature looked okay at 77.7 degrees c at the hottest the memory junction temperature was a bit more concerning since it hit 108 degrees celsius and nvidia has already had problems with memory modules mounted to the back of the pcb getting dangerously hot with the 30 series this card seems to suffer at least somewhat from that and while i didn't notice any throttling there was that fan noise as mentioned in the intro the founders edition 3000 series cards thus far have done very well with cooling aside from the pcb backside memory of course but here nvidia is attempting to cool a card with 30 90 like performance but with the rtx 3080 cooler i think in order to do this they had to get way more aggressive with the fan curve and under sustained load i was seeing 3 to 3 700 rpms from the two cooling fans which got very loud i would be curious to see how much of a difference the rtx 3090 founders edition cooler would make if you could somehow fit it onto the 3080ti but more practically it's probably going to be worth it to pay more attention to third-party designs for this card where they can add more fans a larger cooler and thin stack and possibly also figure a better way out to keep the vram cool here are my power draw numbers and both at peak and on average the 3080 ti was a bit more power hungry than 30 90. again a good indication that a better cooler could really help this card out the 6900 xt still wins the efficiency battle though at least in terms of overall power draw without performance being accounted for and here are my overall numbers welcome to any of you guys who jumped straight to this point in the video these are evenly weighted by game using the rtx 3080 ti as the 100 baseline please remember that for some of these tests especially at 1080 we were cpu or game engine frame rate limited which is why there's more of a performance difference at 1440 and 4k here's what i think nvidia was trying to do with this card though before the 6900xt was beating the rtx 3080 across all resolutions and now the 3080 ti wins unless you count at 1080 where the 6900 xt was still just slightly ahead with my selected batch of games i bet some people would even be excited about grabbing a 3080 ti instead of a 3080 if there was any hope of actually getting one for a reasonable price so if it wasn't obvious in the intro i am keenly aware that pricing and availability are kind of sticking points for all graphics cards right now so the listed retail prices i'm just not even really taking them seriously at least for now i didn't even make a price versus performance chart like i usually do and i'm pretty sure based on leaks and stuff that the 3080 ti was supposed to be one thousand dollars msrp and nvidia decided to just up it to 1200 bucks fairly late in the game because they could and there's nothing much that you or i can do about that but i still wanted to test this card to see how much faster it is in the 3080 how close it gets to the 3090 and how competitive it is with the rx 6900xt and hopefully now you guys have a better idea of how these cards stack up that is all for this video though closing reminder to check out my store at paul's hardware dot net for merch shirts pike glasses and other thumb screw related items and of course like this video if you enjoyed it subscribe to my channel if you super enjoyed it thanks as always for watching and we'll see you in the next video

Channel: Paul's Hardware

Views: 108,227

Keywords: computer, PC, PC gaming, personal computer, computer hardware, paulshardware, RTX 3080 Ti, 3080 Ti Benchmarks, 3080 Ti review, nvidia, graphics card, gpu, benchmark, test, 3dmark, time spy, port royal, flight simulator, dirt 5, resident evil village, gta v, doom eternal, cyberpunk 2077, shadow of the tomb raider

Id: 1HFOSUAfGUM

Length: 13min 17sec (797 seconds)

Published: Wed Jun 02 2021