The stuff Hacker News actually recommends.

GPUs and CPUs · Nvidia GeForce GTX

Nvidia GeForce GTX 1070

13owners with an opinion
82%of verdicts positive
9/2positive/negative since 2024
8 yrslongest happy owner
2happy after 3+ years

Praised for performance and reliability.

What owners wrote

Likes itOwned 8 yearsStill going
My sample size is fairly small, but I've not encountered driver issues on anything in the last ~15 years. Most stuff is in tree (IME everything except nvidia), and despite what others report, I've never had an issue with nvidia binary drivers. My desktop with gtx1070 has been running the same installation for 8 years now. Why do you think Linux can't do auto-updates? There's probably a built in option for your distribution, but if not, just run your update command on a timer. Phones and ARM land seem to be a disaster of hardware support, but x86 PCs/laptops have all worked great for me.
Likes itOwned 7 years
My main machine is a 3.8GHz 8-core Ryzen, 64Gb RAM, GTX1070 GPU. Bought and self-built in ~2018 and still seems pretty good for development and the odd game. Even the 1TB Toshiba SSD is claiming to be healthy according to the monitoring app. It just zips along and copes with everything, and I've never felt any temptation to upgrade anything. Me back in 2005 would have though this setup was science fiction.
Likes it
Yeah the fab shortage is really keeping lifetimes long, far longer than in the past. I've brought my old GTX 1070 out of retirement for for some molecular dynamics these days because its "free" except for eh electricity and have Ben pleasantly surprised at what it can get done chugging along in the background. I'm considering buying some more, as it's kind of a steal in pricing.
Neutral
I also get an error: Requested allocation size (41967616) is smaller than the image requires (42778624). at ImportMemory (../../third_party/dawn/src/dawn/native/vulkan/external_memory/MemoryServiceImplementationOpaqueFD.cpp:132) using Chromium on an old video card GTX 1070 with 32GB RAM, Fedora 44 KDE Plasma 4.7. Don't know if it matters but monitor scaling is set to 150%
Doesn’t like it
There's nothing stopping you doing this now. You can get used 16GB P100s on AliExpress for ~$100 if you want obsolete GPUs. Allegedly new AMD BC 250s are only slightly more. I've looked at this some but I already have a GTX1070 which is only supported upto CUDA 11.9. That's precludes some interesting modern optimizations out of the box. I've spend a lot of LLM tokens backporting some things, but I'm really not sure the hassle is worth it. New hardware is just better. I think in maybe 5 years when supply and demand are back in equilibrium we are going to have some killer technology for decent prices, and 15yo H100s won't look attractive.
Likes it
Successfully ran Gemma4-26B-A4B on my 8yo first-gen Ryzen with a GeForce GTX 1070. It actually ran acceptably well; I was surprised. I even did some coding with it, but the wheels fell abruptly off when it tried several times to use a constant I told it doesn't exist. I only have 32 GiB of RAM in this old bucket, and these results are not worth the RAM consumption, so I put it aside. Maybe if I finish that build with more memory...
Likes it
Hustle culture nonsense aside, I find using whisper (AI) transcription instead of typing to be super efficient. My workflow involves setting up a whisper server, downloading the Whispering(1) app on my computer, and binding it to a shortcut on my keyboard and mouse. Whenever I want to write something down, I just hit the shortcut, dictate and it transcribes instantly. With a Nvidia GPU (1070 in my case), transcription is nearly instantaneous. Although I haven’t set it up on my MacBook yet, I suspect it will be just as fast with Apple Silicon (1) https://github.com/braden-w/whispering/ You can also use an API like grok, but I'm generally wary of such services. I'm a bit of an introvert, so I found talking out loud to be awkward at first. But now I can't go back to regular typing, given the efficiency gains.
Doesn’t like it
For local large language models, my current setup is Ollama running on my M1 Mac Mini with 8GB of RAM, using whatever SOTA 8B model comes out. I used to have a more powerful workstation I built in 2016 with three GTX 1070s, but the capacitors were falling off, and I could not justify replacing it when Claude and ChatGPT subscriptions are more than enough for me. I plan on building a new dedicated workstation as soon as the first-mover disadvantage comes down. Today's hardware is still too early and too expensive to warrant any significant personal investment, in my opinion.
Likes it
Nobody needs it. It's a luxury product. If you don't want a 600w GPU just get a different one. I only recently upgraded from the 1070 that's served me well for almost a decade. Didn't really notice much of a difference even though I splurged a bit. Buying top end gpus these days seems to me like a total waste of money unless you have very specific needs. For playing any mainstream game at 1440p you'll be completely fine with a mid-tier previous-gen card.
Likes it
Wasn't expecting to stay in the top 100 for this long! Initial C implementation got me to ~70 for a short time, then very sloppy racy pthreads one kept me at the tail end for a bit longer, but my again very sloppy cuda program running on my GTX1070 has me at 35 at the time of writing. Rough calculation shows it to be doing 28.6MHashes/sec, and I suspect I could do a lot better than that if I knew anything about cuda programming. I didn't read enough to know how to pick good values for blocks/threads per block, so I just benchmarked all the combinations of powers of two to arrive at the best result. Really fun challenge! Would be fun to see the total amount of valid submissions, and maybe the number of leading zeroes of a hash for those of us that need a moment to figure it out from hex in their head :]