Riku Pasonen🌞 retweetledi
Riku Pasonen🌞
151.5K posts

Riku Pasonen🌞
@Raitziger
Investor with ERP-Invest Oy. Investing since 2006 🇫🇮🇺🇲🎮📸📈🛸. AI tourist with local models.
Mikkeli, Suomi Katılım Kasım 2017
4.4K Takip Edilen4.3K Takipçiler

@JoelDeTeves Is that there with riser or does the mobo have slots split around cpu like that? I am only families wirh regular desktop atx format.
English

It’s alive!
Power limiting the ICAN 6 to 8 pin one out of precaution.
I have a Lenovo official one on the way that’s rated for 150 watts 👍

Joel - coffee/acc@JoelDeTeves
Just copped another reference 3090, let’s go
English
Riku Pasonen🌞 retweetledi

Rare interview of Arnold talking about his bodybuilding career. Part1 youtu.be/c91eSydwrn8?is…

YouTube
English

@jun_song I would subscribe to cloud model instead of mac. I understand smaller models ran on discrete gpu locally.
English

@entrepeneur4lyf @1337hero Liquid cooling is the way to go if you stack cards like that in home environment.
English

@1337hero I built this one for $5k but it has an MSI Godlike z790 in it. 96gb DDR5 6000mhz 14700k with 6gb NVME and 2xRTX4090 Liquid Suprim

English
Riku Pasonen🌞 retweetledi
Riku Pasonen🌞 retweetledi

local LLM people: what are you actually running right now?
everyone talks like they have a DGX Spark under the desk, but I’m curious what the real setups look like
DGX Spark 128GB unified memory?
RTX 6000?
RTX 5090 32GB?
RTX 3090 24GB?
MacBook Pro?
Mac Studio M3 Ultra, if you somehow found one?
or are you running something completely different that people are sleeping on?
English
Riku Pasonen🌞 retweetledi

@ThaRavvBloke @LaeppaVika Eipä noin nopeaa verkkokorttia edes koneeseen saa kiinni ymmärtääkseni. Jotain erikoislaitteita ja ehkä kaksi kuitua vedetty ja yhdistetty. Kapasiteetin puolesta hinta voisi olla luokkaa 150-200€/kk
Suomi

The man is close but not quite there yet.
Know Your Computer 💻@ComputerrBooks
The man aiming to have the world's fastest internet shered a speed test video 😱😱
English

@JormaPoranen @JukkaLepikko Paljon ne mällää mutta vertailuna kukaan ei muista metan virtuaalilaseja tai ympäristöjä. Rahat kaikki about meni hukkaan va vain nimi jäi.
Suomi

@Raitziger @JukkaLepikko Business as usual. Onhan toi capex ihan älytön mitä tekoälyyn satsataan. Väkisinkin joku osa-alue kuplaantuu.
Suomi

@JukkaLepikko Oli aika kun luulin kuplaksi. Nyt rakennan malleja ja koodia päivittäin pelikoneellani. Tässä on iso harrastajien ryhmä nyt ammattikäytön lisäksi.
Suomi

@loktar00 6000 has memory but 5090 is about as fast if model fits in. Massive price difference between these also.
English

Imagine being offered an 8 3090 machine less than the price of a 6000 pro and passing on it.
Is the 6000 pro good? Of course, but to say not to get other cards is just plain silly. An 8x3060ti machine is even incredibly valuable to have for local AI.
Behnam@OrganicGPT
If you wanna run AI models locally, the best option is an RTX 6000 Pro. DON'T get a 5090/4090. And DON'T listen to people who hype the 3090; those cards are beat at this point. Get the RTX with education discount through Nvidia. These used to be $8000, now they're +$9000.
English

@TeksEdge If you have Linux and 5090, lucebox can hit q5 model 120 to 180 t/s on with Dflash and DFlash drafts.
English

🤯 Unsloth released the fastest Qwen3.6-27B MTP GGUF I've tested. Time to upgrade.
Compared to the previous GGUF, Q4/Q6 XL versions are 👀 ~55% faster!
On a single RTX 5090:
✅ 114 tok/s — UD-IQ2_M (MTP)
✅ 93 tok/s — UD-Q4_K_XL (MTP)
✅ 75 tok/s — UD-Q6_K_XL (MTP)
💨Fastest MTP quant is 3.3x faster than the old Q8_0 baseline (35 tps)
262K context + tool calling. All on one 5090.
* compiled from the MTP PR branch ('am17an:mtp-clean', build b9117-ebe4fca4b)

English
Riku Pasonen🌞 retweetledi
Riku Pasonen🌞 retweetledi



















