The stuff Hacker News actually recommends.

GPUs and CPUs · NVIDIA GeForce RTX

NVIDIA GeForce RTX 3070

13owners with an opinion
67%of verdicts positive
6/3positive/negative since 2024
1 yrslongest happy owner
0happy after 3+ years

Praised for performance and value; criticized for reliability.

What owners wrote

Likes itOwned 1 year
been playing on a 3070 quite happily on linux for the last year or so, on a dual boot win11/fedora box... but after i installed fedora i never booted into win11 again! Just forked out for a 4070 ti super so i toy with some AI stuff... when it arrives I'm doing a clean fresh build. Fedora only. no windows at home anymore. While I would loved to have bought an AMD GPU, the power suck on those high end cards is an order of magnitude more than the Nvidia
Neutral
i'm not good with paper work, in fact, i'm horrible with anything that's paperwork related. for the past few days, i ran this model on my rtx 4090 + rtx 3070 and told it to check all the bills, invoices, contracts for me and my small company. i used pi with llama and the pi-llama plugin. oh, boy - i hooked it to my email, told it to download all of the invoices and bills i had for both me and my company and organize them by company/date/ and then merge them with the ones i have locally. it did ocr, wrote scripts, organized everything neatly. i am now the most organized i've ever been in my life. Next: RAG on all the documents and bills i have. if you connect staan-search (there is a pi plugin for that) and ctx7 to this it almost does miracles. the downside is i have to sit next to my noisy threadripper as the magic happens and pay for the electricity, but that's about it, i'll gladly do that. and as i finished this paragraph, it also finished organizing all my personal documents on my san. i don't use the expression "game changer" easily, but it's hard to resist in this case. out of all the models i've used locally qwen3.8:27b blows everything out of the water. my setup # Logical CUDA0 = RTX 4090, logical CUDA1 = RTX 3070 export CUDA_VISIBLE_DEVICES=0,1 cd ~/projects/misc/llama.cpp/ exec ./build/bin/llama-server -hf ggml-org/Qwen3.8-27B-GGUF:Q4_K_M --mmproj /xx/xx/xx/xx/xx/mmproj-Qwen3.8-27B-Q8_0.gguf --host 0.0.0.0 --port 8080 --jinja --parallel 1 --split-mode layer --tenso…
Likes it
As RAM prices are so nuts, I picked up a used gaming laptop early this year with an RTX 3070 (i.e. 8GB) to match the specs for my gaming tower that's run everything I'm interested in just fine. That includes recent Unreal 5.x games (Satisfactory, Fortnite, etc.) with high graphics settings on a 4k TV (60hz). There aren't a ton of games that require more than 8GB of GPU ram.
Doesn’t like it
I was able to fit only small models Qwen3.5-4B in RTX3070 which is not very useful for Python and SQL generation thought. When I wan to test larger open LLM models I often just use cloud resources.
Neutral
Yesterday I installed llama.cpp to test it with local AI Data Analyst that I'm building. I was also testing other open LLM providers: Ollama, Jan, vLLM, LM Studio. I had older NVIDIA card (RTX 3070) and llama.cpp instalation was smooth, contrary to vLLM which required me to reinstall CUDA drivers because by default it installed the latest one. I'm curious if there is a speed difference between the same open LLM model served with different runners.
Doesn’t like it
A few months ago I got so fed up of squinting at glowing rectangles to stay informed/updated. I feel like there's a finite amount of podcasts I can reasonably follow/digest so I built a very basic prototype to summarise RSS feeds/transform the content and transcribe it in to a personalised podcast. I put a bit more work in since May have it running almost end to end, still with plenty of gotchas. It's transcribing on an RTX3070 under my desk so pretty limited, but have enjoyed a few morning briefings just as I want them (weather, chance of rain today, BBC/NPR top stories from yesterday, some specific sports news then HN most popular summarised with summary of comments. Got a chuckle out of the LLM summarisation of HN commen threads which was always "Comments reaction: mixed" (which it no longer embeds!) Found traces of it being done before and I understand you can do similar in Notebook LM, but I was hoping to build something set-and-forget.
Likes it
First, I'd be remiss to point out we went from LLMs are vaporware to people are becoming slaves to their LLMs awful quick. > [I'm scared] you are growing dependent on stilts that could disappear any moment. First, I do control the RTX3070 I own, and that can actually do a pretty decent job nowadays with some of the 3B parameter models. Second, maybe if people like you showed as much concern for that LGBT people can expect family violence for daring to come out of the closet -- instead of worrying about Dr. Strangelove scenarios --, then I wouldn't have to lean on LLMs so heavily. Third, it's hilarious that your response to a comment pointing out how difficult it was to get help from another human without being degraded, was to degrade me by calling me an LLM junkie. Self-awareness is also an important human trait, and you appear to be lacking in that department. Fourth, given that you show absolutely zero concern for the difficulties faced by LGBT people or ADHDers, my advice to you is to take your fears and shove them up your ass.
Likes it
> First, I do control the RTX3070 I own, and that can actually do a pretty decent job nowadays with some of the 3B parameter models. Quick question but what model are you exactly running with 3B parameter. The only decent model I can find which can compete sort of with Cloud models without costing a bank in GPU/RAM are the recently launched Qwen models (35A3B or 27B) which were released a week ago > First, I do control the RTX3070 I own, and that can actually do a pretty decent job nowadays with some of the 3B parameter models. My larger question to you is that even if it might not disappear in any moment, the fact of the matter still remains as if that its still a dependency. Is this dependency worth it? This is an open question and something I am still thinking. > Third, it's hilarious that your response to a comment pointing out how difficult it was to get help from another human without being degraded, was to degrade me by calling me an LLM junkie. Maybe you should be worried that Gemini appears to have more capacity for empathy and self-awareness than you. Gemini isn't real tho. It's still linear algebra with no regards to what it says or not. It's just trained on all the corpus data that Google can find and fine tuned to mimic it. By attaching real human qualities to Gemini, we dilute the value of those human qualities in those first place. I don't necessarily know how "Humans" have treated you. They have treated me both good and bad but I am always more greatful to tho…
Doesn’t like it
I used both. I had a terrible time with llama, and did not realise it until I used ollama. I owned an RTX2070, and followed the llama instructions to make sure it was compiling with GPU enabled. I then hand-tweaked settings (numgpulayers) to try to make it offload as much as possible to the GPU. I verified that it was using a good chunk of my GPU ram (via nvidia-smi), and confirmed that with-gpu was faster than cpu-only. It was still pretty slow, and influenced my decision to upgrade to an RTX3070. It was faster, but still pretty meh... The first time I used ollama, everything just worked straight out of the box, with one command and zero configuration. It was lightning fast. Honestly if I'd had ollama earlier, I probably wouldn't have felt the need to upgrade GPU.
Neutral
What’s the bottleneck? Once I’ve got the model and data onto the GPU my only cost is launching CUDA kernels right? Not sure if that blog post is relevant, but even if it is it shows a 3060 gets //way// faster throughout than the igpu it is testing. I suppose I can test this myself by plugging my 3070 into the NVME on my desktop.
Likes it
MLM, Medium Language Model :) Maybe even "small". You can easily run and train a lot of BERT models on a single consumer GPU. I have a small NER model I am fine-tuning/training on my local Nvidia RTX 3070.
Neutral
I've got an RTX 3070 in my main box right now. Currently I choose not to run it for mining processes because I want the hardware to last longer and don't want to pay for 10x the idle power costs. 300w is a lot of wall power, a non-negligible amount of money and potentially a sustained 80c+ heater in my room until I turn it off. That's just me. 100k downloads disagree apparently, but I do question how many people would stick with this in the long-run.
Doesn’t like it
Strong disagree and this is from someone that used to use FreeBSD and now dual boots Fedora for my laptop and runs Debian for servers. I mean it took me more than 15 minutes to figure out how to get a fresh nobara 38 install to simply boot last time I tried this. For the sin of having an Nvidia 3070 graphics card I had to look up to where to change some settings in grub to boot to be able install the drivers and boot properly. My fedora running framework laptop can't detect one of the external monitors in my lenovo usb c dock if it is plugged in when it boots or after sleep. No lay person consumer is going to waste their time diagnosing this stuff.