Here with a brand new 4080S XLR8, mental 5070 Ti prices needed logical solutions. (Coming from RX 9070 XT, I have some.. thoughts.)
^(Please don't judge the poor placeholder Noctua fan, I'm still trying to figure out how I'm gonna hook up 1 extra Lian Li fan to the current setup.)
Last October, I switched from 4070 Ti to 9070 XT. I game at 3440*1440, and sometimes I use DSR to hit even higher resolutions because I'm obsessed with getting the best IQ. My reasoning for that switch was that 12GB wasn't cutting it and in general I needed more brute force for some of the newer titles like BF6, ARC to hit my 175Hz refresh rate.
So I bought Sapphire RX 9070 XT Nitro+, first of all, in terms of build quality and aesthetics, it was a fantastic card. It runs with zero noise, great temps all in all one of the best GPU coolers I've ever owned, kudos to Sapphire.
The last AMD/ATI card I used on my main system was the HD 7950 from 2012, so naturally, I was a total stranger to modern AMD GPUs. At first, there weren't many issues with 9070 XT, it is a powerful card in raster and competitive in RT workloads. However, little things do add up if you're coming from years of Nvidia ease of use.
Before I begin AMD thoughts, why 4080 SUPER now?
Simple math, I'm after premium cards with silent coolers so SSF and really cheap stuff is off the table for me (than again they're still not cheap they still costs around 1100 USD), in my country a 5070 Ti with TUF, Aero, Gaming OC type of beefy cooler costs around 1400 USD and I got this card for 940 USD brand new with 17 months of warranty left.
The card;
I wanted a card with dead silent noise, low temps, so I needed a premium cooler and people asking crazy money for the likes of Strix or premium MSI offerings in used market. I read TPU's regular 4080 XLR8 review and at that time it was the quietest 4080 they've tested 4080 S being practically the same card it was an easy decision. However, out of box its fan profile was different than the regular 4080 XLR8, hitting 1450 rpm at the lowest %30 speed which was pretty audible. Thanks to a random Redditor, I discovered that PNY sent him a new BIOS fixing this issue. Now, the card sits at 1050-1100 rpm with inaudible noise levels while at full load. Currently running at 0.9750v - 2730 mhz, +1300mem pretty good first impressions.
The DLSS 4, 4.5 - FSR 4.1 situation;
One of the biggest head scratcher is still confusing support of their latest upscaler technology. Sure, newer games do generally support FSR 4, but do not hope getting a slightly older game support like Forza Motorsport 2023 (FSR 2) (yes I do play that game boo me) Witcher 3 next gen or RDR 2. I believe Vulkan is totally off the table for them so no FSR 4 for the likes of Doom Dark Ages. We got a brand new Diablo 4 expansion and it still isn't supporting FSR 3/4 so you consume 250-300W while playing an ARPG due to native res even with frame cap. So naturally, this leads you to start to spend your precious time tinkering with Optiscaler. And in general FSR 4 is still not %100 on par with Nvidia's latest DLSS offerings in terms of quality, some games look fantastic, others looks closer to DLSS 3.5. They still have a long road ahead of them and I'm not convinced that they have the pace to match Nvidia's offerings.
The "drivers";
Look, it's no secret that one of the first worries people have when switching to AMD is driver stability and it's not as bad as I thought it would be however the day comes and one or two games catches you off guard with "AMD software detected that a driver timeout has occured" and you start to lose it. Driver crashes in the newest Diablo 4 expansion was so severe that it singlehandedly convinced me to switch back to Nvidia. Oblivion Remastered, for example is completely unplayable, with constant driver crashes happening 5-10 mins into the game.
Confusing undervolting and higher power consumption than Nvidia;
AMD has a pretty decent built-in GPU tweaking tool in their software suite, and it offers great customization. However, architectural implementation is confusing. You undervolt your card so it can boost to higher clocks; however UV doesn't reduce power consumption you have to do that by manually limiting the power limit which causes more performance loss than the Nvidia method. Even if you do find the perfect setup for your card, AMD generally consumes more power. I haven't even done a deep dive into tweaking the 4080S yet, but with a basic undervolt, it consumes at least 70-80W less power than the 9070 XT while being the more powerful card. Nvidia's efficiency is unmatched.
Closing thoughts;
AMD has a really powerful GPU in its hands, generally giving a respectable fight with the likes of 5070 Ti, 4080, and 4080S. In one or two edge cases it is even coming closer to 5080 with a much cheaper price point, which is great for them however, if you are a previous NVIDIA power user you really need to forgive and forget a lot of things and the more you spend time with it the more you realize why you are paying that premium NVIDIA TAX.