The stuff Hacker News actually recommends.

GPUs and CPUs · NVIDIA GeForce RTX

NVIDIA GeForce RTX 4080

11owners with an opinion
43%of verdicts positive
3/4positive/negative since 2024
0happy after 3+ years

Praised for performance and battery; criticized for value and reliability.

What owners wrote

Doesn’t like it
I'm on the same boat with a 4080, the extra GPU RAM of the 4090 over the 4080 would have been nice, but it still wouldn't be able to run any decent models. I configured my PC in December 2023 with a kit of 2x32GB DDR5-6000 for €249, thinking I would upgrade that RAM later to 128GB by just adding another kit. Now nearly 3 years later that same kit is over €1500 :-( I've also lost all interest in upgrading to a 50-series GPU. With exception of the ludicrously expensive 5090 the entire 50-series lineup is capped at 16GB, I don't see the benefit of upgrading.
Doesn’t like it
That arbitrage opportunity is remarkable. Maybe duty/taxes, as you say? I'm the guy who (Who plays games and runs molecular dynamics simulations and other CUDA stuff) said 3 years ago "$1600 for a graphics card? That is excessive. I'll upgrade in a few years when ready" And bought a 4080 for $1200 from Nvidia instead of the 4090. Oops! Now there is no reasonable upgrade path.
Doesn’t like it
Perhaps not a good example, I tried running local models a few times, to much disappointment (actually made me skeptical of LLMs in general for a while). My last experiment in January was trying to run a Qwen model locally (RTX 4080; 128GB RAM; 9950X3D). I must have been doing it extremely wrong because the models that I tried either hallucinated severely or got stuck in a loop. The funniest one was stuck in a "but wait, ..." loop. I fortunately had started experimenting with Claude, so I opted to pay Anthropic more money for tokens (work already covers the bill, this was for personal use). That whole experience + a noisy GPU, put me off the idea of running/building local agents.
Doesn’t like it
It is wild thinking how a few years ago, I didn't buy a 4090 direct from nvidia because "$1600 (USD) is too much to pay for a graphics card; if I need a better one, i'll upgrade in a few years. (Went with 4080, which is substantially slower and was $1200) Joke's on me! It will be scarcity mindset from here on out; will always buy the top tier thing .
Doesn’t like it
I'd really like to be a drop in that bucket, but as an nvidia owner it's rough. You take a sizable performance hit and general stability is awful. In Linux I can't turn off my monitor on nvidia, or there's a ~80% chance I'll never get it to come back on. If it does, I don't have sound 100% of the time. It's ridiculous. My next GPU will be a Radeon, but they don't have anything to offer that can beat my 4080 from 2022...
Neutral
Precisely. Even inflated if the inflated 16,000 api calls was accurate for how much the cost of mediocre GPU would get you, that’s not an endless store of api calls. I’m also on a 4080 for lighter loads, and even just writing benchmarks, exploring attention mechanisms, token salience, etc, without image gen being my specific purpose I may trash half a thousand generations from output every few days. More if I count the stuff that never made it that far too.
Neutral
That's an issue with the monitor only supporting HDMI 1.4. A monitor that supports HDMI 2.0 would work with 144hz fine. I have the opposite issue where I run my 240hz monitor on HDMI 2.1 because it has more bandwidth than my RTX 4080's DisplayPort 1.4 ports. It seems a common issue enough with that that model is specifically called out sometimes: https://forums.tomshardware.com/threads/how-to-connect-to-a-...
Doesn’t like it
That completely depends on your expectations and uses. I have a gaming rig with a 4080 with 16GB of RAM and it can't even run Mixtral (kind of the minimum bar of a useful generic LLM in my opinion) without being heavily quantized. Yeah it's fast when something fits on it, but I don't see much point in very fast generation of bad output. A refurbished M1 Max with 32GB of RAM will enable you to generate better quality LLM output than even a 4090 with 24GB of VRAM and for ~$300 less, and it's a whole computer instead of a single part that still needs a computer around it. Compared to my 4080, that GPU and the surrounding computer get you half the VRAM capacity for greater cost than the Mac. If you're building a rig with multiple GPUs to support many users or for internal private services and are willing to drop more than $3k then I think the equation swings back in favor of Nvidia, but not until then.
Neutral
I have a 4080 with 16GB of VRAM. I experimented with llama.cpp by offloading layers onto GPU and doing the remaining on CPU. I found that it gives me max tokens/sec if I set it to 8 CPU cores as opposed to the 16 available on the 7950X. I guess beyond that, the bookkeeping between the cores might be taking up more time than it is worth.
Doesn’t like itHasn’t owned it
Battlefield 6 being "highly optimized" is a joke. Runs pretty poorly on a RTX 4080 with 5800X3D @ 1440p. It also legitimately looks worse than the Battlefields that predate it, even up to Battlefield 1, which is over a decade old now. A better example is Arc Raiders.
Likes itHasn’t owned it
100% this is Apple's fault. They never took gaming seriously. I'd hoped that would change with the M series integrated GPUs, but that hasn't been the case lately, with only a tiny number of titles launching with Apple Silicon support. That's on Apple, not anybody else. For what it's worth, I did buy and occasionally use Crossover (thank you!), but only when the game isn't supported on GeForce Now. My M2 Max, even at its best (as in Baldur's Gate 3 running native Apple Silicon code), is still no match for a 4080... not even close. Crossover works well for simple indie games, but AAA games often have more demanding requirements. The extraordinary performance of DLSS (Nvidia's AI upsampling and frame interpolation algorithm) alone makes non-Nvidia GPUs less viable these days. And that's just the performance side. Then, Crossover's UX is a whole other issue. Crossover's Steam takes forever to launch every time (I'm not really sure why this is; never bothered to really look into it). It's never clear to me which proper combination of D3DMetal/DXVK/Esync/Msync to use (I still don't even know what the "sync modes" mean). Some games only work with a certain version of Apple's GPT (for others: Game Porting Toolkit, not AI) manually installed, or requiring the bleeding-edge preview version of Crossover, etc. The idea of user-managed "bottles" and disks is a lot more complicated than GFN's model of individually vendor-managed sandboxed/containerized games, preconfigured to work right th…