GeForce RTX 3070

[...] I don't even really play games, but for my video editing workloads, more GPU cores and VRAM make a world of difference, and the RTX 3070 level of performance out of the RX7700s that's thus far the only GPU option for Framework devices just doesn't cut it. There's just no way I'm spending $2500+ USD for a laptop that has worse performance than devices costing half as much at this point. They just aren't really delivering on the promise of "Future upgradeability" in any kind of meaningful way so far, and I just can't see the value in purchasing what's undeniably a wildly overpriced machine based on promises that have yet to be delivered upon. [...]

RTX 3070 on Plasma with the latest Nvidia driver. [...] I just can't handle Linux as a daily driver anymore, I don't really want to deal with all the little annoyances anymore. [...]

[...] I bought myself a high-end computer (5950x & RTX 3070) 3 years ago, and I realized how important it is to spend -- it's sometimes cheaper than buying it later (it was the beginning of the silicon crisis), and you get to enjoy it a lot. [...]

[...] Between the APFS slowdowns and the virtualization overhead, it was faster to build everything in a VPS and download the artifacts onto my Mac. [...] My RTX 3070 Ti cost $500, is fabbed on Samsung's 8nm node and beats the 5nm M2 Ultra's GPU in power efficiency. [...] You should be using Nix on macOS if you're a developer, it handles things like side-by-side package versioning and environment isolation much nicer than Homebrew does. [...]

I managed to run it with RTX 3070 (8GB VRAM) following the "Quickstart" on HF model card with minor modifications (modify the architecture 86 for your own hardware): git clone && cd llama.cpp cmake -B build -DGGMLCUDA=ON -DCMAKECUDAARCHITECTURES=86 && cmake --build build -j Then downloaded & verified Ternary-Bonsai-2-27B-PTQ10.gguf from HF and ran ./llama.cpp/build/bin/llama-cli -m Ternary-Bonsai-2-27B-PTQ10.gguf -ngl 99 -fa on -c 32768 -b 256 -ub 64 --temp 1.0 --top-p 0.95 --top-k 20 -ctk q80 -ctv q80 -p "hello world in x86 assembler" -n 256 the parameters were suggested by gpt-5.6-luna to reduce memory footprint, as the defaults ran OOM on my gpu. result looks good: [ Prompt: 165.6 t/s | Generation: 40.6 t/s ] would be nice if they upstreamed their changes so that it runs with the original llama.cpp

As RAM prices are so nuts, I picked up a used gaming laptop early this year with an RTX 3070 (i.e. [...]

I had older NVIDIA card (RTX 3070) and llama.cpp instalation was smooth what model was it that you were able to run with the rtx 3070?

I bought a RTX 3070 off a miner when Eth went to proof of stake. It was clean, cheap, and is still going strong for daily gaming. In its working life it was undervolted and probably cooled better than in my rig.

A few months ago I got so fed up of squinting at glowing rectangles to stay informed/updated. [...] 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.

[...] I used yt-dlp to rip bodycam videos from youtube, distil-large-v3 to extract transcripts on my rtx 3070, and deepseek-v4-flash to output classifications. [...]

[...] 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. [...] Maybe you should be worried that Gemini appears to have more capacity for empathy and self-awareness than you. [...]

[...] Agents run locally on a Ryzen 9 5900X / RTX 3070 using Mistral, Llama 3.2, and CodeLlama via Ollama. [...]

Hey HN, I built this to scratch my own itch. I have an RTX 3070 so just 8gb of VRAM I use ComfyUI for image/video generation, but running multi-step pipelines was painful: - No way to chain small step workflows (image → edit → video) without manual intervention - If step 3 of 5 failed, start over - No approval gates to review outputs mid-pipeline - Single GPU bottleneck ComfyAutomate fixes this with: Chain definitions in YAML - Define multi-step pipelines with dependencies. [...] Durable execution - Built on Temporal, so workflows survive crashes. [...] Technical choices: - Temporal for workflow orchestration (overkill for hobby use, but rock solid for production) - NetworkX for DAG dependency resolution - Jinja2 templates for passing outputs between steps ({{ step1.output.image }}) - SSE for real-time progress updates The approval + regeneration flow was the trickiest part - when you reject a step, it creates a subgraph of that step + all descendants and re-executes just that portion while preserving the rest of the context. [...] Would love feedback, especially on the chain YAML syntax and whether the Temporal dependency is too heavy for most users.

Funnily enough the most annoying things on my system at the moment is RGB and keyboard/mouse customisation. I haven’t found a tool that can access all the extra settings of my Logitech mouse, not my Logitech speakers. OpenRGB is amazing but I’m stuck on a version that constantly crashes; this should be fixed in the recent versions but nixpkgs doesn’t seem to have it (last I checked). On the other hand I did manage to get SteamVR somewhat working with ALVR on the Quest 3, but performance wasn’t great or consistent at all from what I remember (RTX 3070, Wayland KDE).

[...] (posted from an RTX 3070 on Wayland)

We very much can, especially such a Mixture of Experts model with only 3B activated parameters. With an RTX 3070 (7GB GRAB VRAM), 32 GB RAM and an SSD I can run such models at speeds tolerable for casual use.

I recently ran Memtest Vulkan v0.5.0 on an NVIDIA GeForce RTX 3070 Laptop GPU to check memory stability and performance. The standard 5-minute test completed successfully without errors, confirming stable VRAM under Vulkan workloads. The test setup: Platform: Windows 11 Laptop CPU: AMD Ryzen 9 5900HX with Radeon Graphics Base speed: 3.30 GHz Sockets: 1 Cores: 8 Logical processors: 16 Virtualization: Enabled L1 cache: 512 KB L2 cache: 4.0 MB L3 cache: 16.0 MB GPU: NVIDIA GeForce RTX 3070 Laptop GPU Tool: Memtest Vulkan v0.5.0 Test Duration: 5 minutes (default run) Would be interested to hear if anyone else here has used Memtest Vulkan and can share their experiences with this open-source tool on different GPUs.

I have an RTX 3070 with 8GB VRAM and for me Qwen3:30B-A3B is fast enough. [...] I have tested the 14B model and it's noticeably slower because it's a dense model.

We have much hardware on the secondary market (resale) that's only 2-3x slower than pristine new primary market devices. It is cheap, it is reuse, and it helps people save in a hyper-consumerist society. The common complaint is that it doesn't run bloated software anymore. And I don't think we can make non-bloated software for a variety of reasons. [...] This is due to hyper-bloating of software, and severe downskilling of game programmers in the industry to cut costs. [...] But the truth is that aside from current-generation graphics, most games haven't evolved much in the last 10 years, and current-gen graphics arrived on PS4/Xbox One. [...] The market demands cheap software. [...] A common price point for good quality games was $80, which is $170 today. But the gamers absolutely decry any game price increases beyond $60. [...] Ironically, upgrading a graphics card one generation (RTX 3070 - 4070) costs about $300 if the old card is sold and $500 if it isn't. So gamers end up paying $400 for the latest games every few years and then rebel against paying $30 extra per game instead, which could very well be cheaper than the GPU upgrade (let alone other PC upgrades), and would allow companies to spend much more time on optimization. [...]

Looks cool, but the site is very heavy! I have an RTX 3070 GPU, and GPU usage went shot to 70% when I opened this website on Firefox.

I use a T480 (bought new a lot time ago) moderately-capable gaming PC (lol): - 3 extended, swappable batteries + third-party external battery charger (also has an internal battery T490 lacks) - 32 GiB RAM - i7-8650U - 2 TiB SSD - Retrofit/modded magnesium top case - I pre-acquired replacement parts like fans, hinges, backlit keyboards, and case screws - Works with Windows 11 + Razer Core X TB3 with an RTX 3070 Ti eGPU Needs a couple of utilities to max out CPU TDP to 25W.

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.

I tried writing an AlphaZero clone to play Chess on my home PC (I only had an RTX 3070) and I failed for essentially the same reason as they mentioned: iteration time was too slow and you couldn’t tell if the model was getting any better at all after weeks of training. [...]

Deepseek coder 6.7B is very useful for coding and can run in consumer GPUs. I use the 6bit GGUF quantized version on a laptop RTX 3070

You can easily achieve a 500 dollar Dota PC if you're willing to buy used. I bought a 5600x / 32GB RAM / RTX 3070 PC for 550 USD a bit ago.

Mine is a laptop with i7-11800h CPU + RTX 3070 Max-Q 8GB VRAM + 64GB RAM (though you can get probably get away with 16GB RAM). I bought this system for work and causal gaming, and was happy when I found out that GPU also enabled me to run LLMs locally at good performance. This laptop costed me = $ 1600, which was a bargain considering how much value I get out of it. [...]

I'm running the llama.cpp/gguf Q8 version, with 30 layers offloaded to the laptop's GPU (RTX 3070, 8G VRAM), and I get around 20-25 tokens/second. It really feels like I have one the the earlier versions of ChatGPT 3.5 installed on my computer.

A 7B model @ Q4 needs roughly 6.5GB of memory. Mistral 7B q4 seems to use 5Gb VRAM A few months ago I could not run Llama 7B q4 on my laptop's rtx 3070 (8gb VRAM), but Mistral 7B it runs very well and still has approx 2gb VRAM left even with a big context. Not sure if this is an optimization on inference algorithms or if Mistral 7B is just more efficient

I think the time has come to admit that I need to upgrade my GPU Been using NVIDIA GeForce RTX 3070 till now but I really need more ram, 8gb ain't enough

[...] I got a NVIDIA GeForce RTX 3070 for about $400. [...]

[...] On top of that you still need a PSU, NVMe, a high pressure fan, DP to HDMI adapter, possibly BT and Wifi adapters, and a 3d-printed or DIY case. It is also a very hacky build. [...] The software side of things is Linux, so if you're familiar with setting it up it shouldn't be difficult to go through the different steps. [...] I use Bazzite and boot directly to Steam big picture, the UX is great. You can exit to desktop and you have a fully usable Linux box that you can play around with from the comfort of your sofa. [...] The hardware is remarkably similar, and much better than mini PCs set up with AMD APUs, unless you go Strix halo but that is another price league (10x). You can play Cyberpunk 2077 decently on it on settings similar to an AMD RX 6700XT or slightly below RTX 3070. A pretty impressive number is the memory bandwidth at 410 GB/s, which should run local LLMs like Gemma 4 or Qwen3.8 27B at decent speed but with limited context size as its just 16Gb Would I build it again? [...] Still much cheap

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.

Yesterday I installed llama.cpp to test it with local AI Data Analyst that I'm building. [...] 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. [...]

There's a cost/value calculation that just doesn't work well...I have a Ryzen9/rtx3070 PC ($2k over time) and my M4 Mini ($450) holds it's own for most all normal user stuff...sprinting ahead for specific tasks (Video CODEC)...but the 6 year old dedicated GPU on the PC annihilates the Mini in pushing pixels...You can spec an Apple that does better for gaming, but man, are you gonna pay for it, and still not keep up with current PC GPUS. [...]

T480 (non-s) - Love the dual batteries (one swappable) unavailable on Apple-design infested T490 - Retrofitted with magnesium top case and bezel mod - 5 extended 72 Whr batteries with a third-party external charger from some dude in the UK - Upgraded to fastest processor and discrete GPU (slow on its own but I use a Razer Core X eGPU with an Nvidia RTX 3070 Ti, and can run basically any game on Steam.) - 32 GiB of RAM - WiFi 6e Intel AX210 (looking at WiFi 7 using the AMD-compatible Broadcom FastConnect 7800 / QCNCM865 that I run on my AMD 7900 Asrock DeskMini X600 electronic lab Windows-only things box that I'm typing on right now). - Bought OE replacement most likely to fail: keyboards, pointing stick (and tips), trackpad assembly, and fans (I think I bought 6). [...] - I might have a slight mainboard problem because I'm constantly running ThrottleStop to get higher, sustained Tdp with SpeedFan sending fans manually to full blast or otherwise the max freq randomly drops to painfully-slow 900 MHz max non-deterministically.

[...] I owned an RTX2070, and followed the llama instructions to make sure it was compiling with GPU enabled. [...] It was still pretty slow, and influenced my decision to upgrade to an RTX3070. [...]

I've got an RTX 3070 in my main box right now. [...] 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. [...]

As someone who owns an RTX 3070, why is it so special? I've found myself wishing I had the 3060 Ti for the 12gb memory, sometiemes.

I bought an HP Reverb G2 and it was non-functional on arrival. It had some kind of USB 3.0 pickiness/bugginess/not-workiness. At best I had it half-working in a crazy, glitchy state. I had a perfectly mainstream, working gaming PC with an RTX 3070, the third fastest Nvidia GPU available at the time. [...]

[...] And yeah, I also sampled a GIGABYTE/AORUS RTX 3070 Laptop I booted up about 10 times this year, while it sucks in every metric I need (I rate it worse than an Asus EEE on build quality)... it is very quiet and the CUDA performance is respectable for a mobile demo unit (its only redeeming quality). [...]

[...] Software-wise, issues with OS updates, library versions, and too bloated interface. It was only once I got my M1 air that I realised that macos was not the problem, but rather how the computers were configured, shared between people and used. [...] There are occasional issues, annoying security restrictions that you have to manually waive all the time, OS updates that break things, but at least it does not force you to update in the middle of you having to run something important like windows do. And the default macos office apps are lightweight, and decent enough that I do not need to some bloated office software. Work-wise, all the computational software I use runs fine, and only if I need to use intense and long parallelised computations I may use my desktop (or cloud). [...] I share my office with others and get constant reminder how it sounds like. [...] My friend has a laptop with rtx 3070 and it sounds like a jet engine when we play together.

It's certainly possible, I've managed to make the game run playably on a base model 2019 macbook, but that was completely vanilla on the lowest settings I could manage. Beat Saber (especially with mods) can get pretty damn hard to run (it often lags pretty badly on my 3700X/RTX 3070), especially if you're playing noodle maps or with high poly custom sabres. I haven't tried it but I expect you'd have a very bad time running a quest due to the encoding step, and also probably wouldn't have fun on anything newer than a vive/cv1 because the resolution is too high.

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. [...] 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. [...] Is this dependency worth it? [...] Maybe you should be worried that Gemini appears to have more capacity for empathy and self-awareness than you. Gemini isn't real tho. [...] By attaching real human qualities to Gemini, we dilute the value of those human qualities in those first place. [...]

The A770 actually has more transistors than the RTX 3070 Ti: RTX 3070 Ti: 17,400 million transistors A770: 21,700 million transistors It has taken Nvidia decades to figure out how to use transistors as efficiently as it does. [...] That said, it is possible that better drivers would increase A770 performance, although I suspect that reaching parity with the RTX 3070 Ti would be a fantasy. The RTX 3070 Ti has both more compute and more memory bandwidth. [...] To make matters worse for Intel, I am told that games tend to use vendor specific extensions to improve shader performance and those extensions are of course not going to be available to Intel GPUs running the same game. [...] In specific, information in these extensions is here: Also, I vaguely recall that Doom Eternal used some AMD extension that was later incorporated into vulkan 1.1, but unless ID Software updated it, only AMD GPUs will be using that. [...] On the bright side, it is possible for Intel to level the playing field by implementing the Vulkan extensions that its competitors use to get an edge, but that will not help it in Direct3D performance. [...]

[...] My laptop has an Intel Core i9 12th Gen 12900H 2.5GHz processor, 32GB DDR5-4800 RAM, and a NVIDIA GeForce RTX 3070 Ti 8GB GDDR6. [...]

[...] RTX™ 4000 SFF Ada Generation (20GB GDDR6 ECC) 64 GB DDR4 1.92 TB (Gen4) Intel® Core™ i5-13500 30.9 TBit total bandwidth I think RTX 3070 is roughly equivivalent to RTX4000 20GB. [...]

I only have a RTX 3070 with 8GB VRam. It can run quantized 7B models well, but this is 8 x 7B. Maybe an RTX 3090 with 24GB VRAM can do it.

8th gen intel i7, RTX3070. It's aging a little, but it's about as standard as standard can get. I have dual monitors which might contribute to the complexity, but it's still a low bar to expect dual monitors to work on a gaming OS.