Hackernews posts about RTX 3090

1. 01
   

   newQwen3.8-Flash-Next non-uniform quantization runs on 2 RTX3090shuggingface.co

   ▲ 9pfeifferj21d💬 2savesaved
2. 02
   

   newTuning a Local Coding Agent: Oh My Pi and Qwen3.8-27B on Two RTX 3090sdoug.sh

   ▲ 5dougcalobrisi20d💬 1savesaved
3. 03
   

   newExtending Raschka's GPT-2: an MoE trained from scratch on an RTX 3090www.gilesthomas.com

   ▲ 2ibobev24ddiscusssavesaved
4. 04
   

   newExtending Raschka's GPT-2: an MoE trained from scratch on an RTX 3090www.gilesthomas.com

   ▲ 1gpjt25ddiscusssavesaved
5. 05
   

   newI Built a System One Model on a RTX 3070twitter.com

   ▲ 3lafalce16ddiscusssavesaved
6. 06
   

   newSingle-slot low-profile 75W RTX 3060 with no power connectors disappointswww.tomshardware.com

   ▲ 2sbulaev29ddiscusssavesaved
7. 07
   

   newRun Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/sgithub.com

   ▲ 906snehesht32h💬 407savesaved
8. 08
   

   newMicroCenter requires photo ID and signed no-export pledge to buy RTX 5090www.tomshardware.com

   ▲ 4felineflock3h💬 1savesaved
9. 09
   

   newMicro Center requires RTX 5090 buyers to show ID and sign no-export declarationvideocardz.com

   ▲ 2haunter30h💬 1savesaved
10. 10
    

    newMicro Center requires photo ID and signed no-export pledge to buy gaming GPUwww.tomshardware.com

    ▲ 9aillia26h💬 2savesaved
11. 11
    

    newLLM from scratch, part 28 – training a base model from scratch on an RTX 3090www.gilesthomas.com

    ▲ 540gpjt307d💬 121savesaved
12. 12
    

    newShow HN: Llama 3.1 70B on a single RTX 3090 via NVMe-to-GPU bypassing the CPUgithub.com

    ▲ 395xaskasdf226d💬 101savesaved
13. 13
    

    newRTX 5080 and RTX 3090 Setup: 80 Tok/s on Qwen 3.6 27B Q8imil.net

    ▲ 292iMil114d💬 108savesaved
14. 14
    

    newWe got 207 tok/s with Qwen3.5-27B on an RTX 3090github.com

    ▲ 165GreenGames168d💬 52savesaved
15. 15
    

    newCounter-Strike's Dust II runs purely within a neural network on an RTX 3090www.tomshardware.com

    ▲ 26bpierre722d💬 3savesaved
16. 16
    

    newLaguna XS 2.1 33B on a RTX 3090: 296 tok/s peak, 152 tok/s at 256K contextwww.lucebox.com

    ▲ 20GreenGames86d💬 1savesaved
17. 17
    

    newBenchmarking Qwen 3.6 35B MoE (3B active) on an RTX 3090www.gilesthomas.com

    ▲ 16gpjt72ddiscusssavesaved
18. 18
    

    newDual RTX 5060 Ti 16GB vs. RTX 3090 for Local LLMswww.hardware-corner.net

    ▲ 14pietrushnic486d💬 5savesaved
19. 19
    

    newStory of How Im Running an Unlimited $6/Month AI Provider on 4x RTX 3090s

    ▲ 8yolo-auto113d💬 3savesaved
20. 20
    

    newShow HN: OS Megakernel that match M5 Max Tok/w at 2x the Throughput on RTX 3090github.com

    ▲ 6GreenGames180d💬 1savesaved
21. 21
    

    newQwen3.6-35B-A3B speculative decoding is net-negative on RTX 3090github.com

    ▲ 5thc1006167d💬 2savesaved
22. 22
    

    newQwen 3.6 93B with MTP on 2×RTX 3090 NVLink=187 tokens/SEC,LLM lost bleat-a-thongithub.com

    ▲ 4devilfileprong113ddiscusssavesaved
23. 23
    

    newQwen3.8-27B on a single RTX 3090: crash fix, 131K context, 9 mythsjonidimo.github.io

    ▲ 3jonaddb48d💬 2savesaved
24. 24
    

    newPFlash: 10x prefill speedup over llama.cpp at 128K on a RTX 3090github.com

    ▲ 3GreenGames157d💬 1savesaved
25. 25
    

    newUpgrading Fans with a Custom Shroud on a RTX3090 – Goodbye Fan Noiseboilingsteam.com

    ▲ 3ekianjo319d💬 1savesaved
26. 26
    

    newI swept all 667 layer duplication configs for Qwen3-4B on 1x RTX 3090austinsnerdythings.com

    ▲ 3auspiv173ddiscusssavesaved
27. 27
    

    newUpgrading Fans with a Custom Shroud on a RTX3090 For A Silent Experienceboilingsteam.com

    ▲ 3ekianjo322ddiscusssavesaved
28. 28
    

    newFine-Tuning Nvidia's GR00T 2B Robotic Model on a RTX 3090twitter.com

    ▲ 3nprobably560ddiscusssavesaved
29. 29
    

    newBenchmarking Qwen 3.6 35B MoE (3B active) on an RTX 3090www.gilesthomas.com

    ▲ 1ibobev70ddiscusssavesaved
30. 30
    

    newClub-3090 Recipes for serving QWEN3.6 27B locally on RTX 3090sgithub.com

    ▲ 1lastdong153ddiscusssavesaved
31. 31
    

    newShow HN: Federation of robots collaboratively train an object manipulation modelgithub.com

    ▲ 7jafermarq628ddiscusssavesaved
32. 32
    

    newShow HN: Open Access Qwen3.6-35B-A3B-UD-Q5_K_M with TurboQuant

    ▲ 4freakynit171d💬 2savesaved
33. 33
    

    newShow HN: ZigNet: How I Built an MCP Server for Zig in 1.5 Daysfulgidus.github.io

    ▲ 4Fulgidus338ddiscusssavesaved
34. 34
    

    newShow HN: I built a 2nd-order PyTorch optimizer for LLMs that runs on 16GB GPUs

    ▲ 2dnosoz159d💬 4savesaved
35. 35
    

    newShow HN: Accidentally built churn detection AI (I don't code)github.com

    ▲ 2willster787417d💬 2savesaved
36. 36
    

    newShow HN: kulyk-uk-en and kulyk-en-ukhuggingface.co

    ▲ 2yehors440d💬 1savesaved
37. 37
    

    newShow HN: Don't trust me bro: fixing GPT-OSS (3.49B tokens, 1k GPU hours, 1x3090)github.com

    ▲ 2iamskeole34ddiscusssavesaved
38. 38
    

    new70M vectors searched in 48ms on a single consumer GPU –results you won't believe

    ▲ 1EffCompute203d💬 4savesaved
39. 39
    

    newShow HN: Localvoxtral – Local real-time dictation on macOS with streaming STTgithub.com

    ▲ 1T0mSIlver223d💬 2savesaved
40. 40
    

    newAluminatiAI – per-job GPU cost tracking (Nvidia-smi shows watts, not dollars)

    ▲ 1AluminatiAi210ddiscusssavesaved

· edit