Home | Notifications | New Note | Local | Federated | Search | Logout
Federated Timeline
Reply to @risahana@mstdn.jp
たかし@tak4@mstdn.y-zu.org (2026-08-02 18:06:33)
@risahana こんありー🦋
:hosimiya_mion::star_stroke:@hos1miya@misskey.0sakana.xyz (2026-08-02 18:06:30)
は？
andGALLERYのランダム、元々キャラ毎に袋分かれてるっぽくて草
https://twitter.com/nan_hobby/status/2083756403983036537?t=mOohkGZvK9kcFOK4MwSPDQ&s=19
Reply to @whtapple@ippandon.hopto.org
たかし@tak4@mstdn.y-zu.org (2026-08-02 18:06:16)
@whtapple ハローハローです
:hosimiya_mion::star_stroke:@hos1miya@misskey.0sakana.xyz (2026-08-02 18:03:45)
ヨドバシのせいで最悪な空目した
Jack@watson554@misskey.gg (2026-08-02 18:02:53)
基本、YTで広報って形だろうなあ...うん...
Jack@watson554@misskey.gg boosted:
@watson554@misskey.gg (2026-08-02 18:01:51)
Webサイト作ったぜぇい
---Attachments---
image: https://sss.misskey.gg/sss/gg/28c0eb79-1312-4bb3-a381-35b18579e746.webp
image: https://sss.misskey.gg/sss/gg/b4ebc926-f0a2-4fab-a63b-349a646957e3.webp
Jack@watson554@misskey.gg (2026-08-02 18:01:51)
Webサイト作ったぜぇい
---Attachments---
image: https://sss.misskey.gg/sss/gg/28c0eb79-1312-4bb3-a381-35b18579e746.webp
image: https://sss.misskey.gg/sss/gg/b4ebc926-f0a2-4fab-a63b-349a646957e3.webp
Ganbold@ganbold@sns1.hydranlab.com (2026-08-02 17:56:11)
I need more people testing RX580
I need to actually write my own result lmao
this is one of the RX580 research done by someone
https://amihart.medium.com/inference-with-an-rx580-13e9c1055472
my way of writing is always writing every cases in a single article, so it will be very bulky but not in a way of research purpose writing.
btw I agree that vulkan is much better with AMD rather than ROCM. ROCM is very weird in my opinion, even in a certain case it could beat Vulkan, but when I try to reproduce it, its kinda flaky.
meanwhile NVIDIA is doing the best with their backend and most of the time it produces better result than vulkan
KOBA789@koba789@misskey.io boosted:
@wakusan0812@misskey.io (2026-08-02 15:00:09)
:box_tissue::tissue_tabe_taikai:
:nantekotoiuno::meowcry:
RE: https://misskey.io/notes/apexxdgc2fsj0a22
KOBA789@koba789@misskey.io boosted:
@giraffe_beer@mstdn.maud.io (2026-08-02 17:35:08)
https://x.com/hologon15/status/2083822624032051365 存在しないピーターラビットすき。
:hosimiya_mion::star_stroke:@hos1miya@misskey.0sakana.xyz (2026-08-02 17:43:43)
絵はホッケーなの草
:hosimiya_mion::star_stroke:@hos1miya@misskey.0sakana.xyz boosted:
@Adore_me_HARAI@misskey.io (2026-08-02 17:36:04)
苫小牧にあったこれ
「HOKKI POWER」って知らん言葉を次の瞬間にはＨＰって略し始めてて怖い
---Attachments---
image: https://media.misskeyusercontent.com/io/ad2551dc-7479-481b-8843-6046629a47db.webp
Reply to @ganbold@sns1.hydranlab.com
Ganbold@ganbold@sns1.hydranlab.com (2026-08-02 17:37:15)
I know that everyone is building on blackwell, since like the research from the manufacturer are pushed that way, and then of course the user are shaping it as well, its more battle-tested compared to RX or Arc, but price follows the trend as well. blackwell getting much more expensive, meanwhile RX and Arc maybe get a little cheaper, since who want to hold on a product that won't sold in short time?
土曜の夜は最高@Adore_me_HARAI@misskey.io (2026-08-02 17:36:04)
苫小牧にあったこれ
「HOKKI POWER」って知らん言葉を次の瞬間にはＨＰって略し始めてて怖い
---Attachments---
image: https://media.misskeyusercontent.com/io/ad2551dc-7479-481b-8843-6046629a47db.webp
Giraffe Beer@giraffe_beer@mstdn.maud.io (2026-08-02 17:35:08)
https://x.com/hologon15/status/2083822624032051365 存在しないピーターラビットすき。
Reply to @ganbold@sns1.hydranlab.com
Ganbold@ganbold@sns1.hydranlab.com (2026-08-02 17:34:59)
for 16GB market, intel is cheapest in the local market. but building 8GBx2 is cheaper than intel 16GBx1
Ganbold@ganbold@sns1.hydranlab.com boosted:
@mametsuko@mk.absturztau.be (2026-08-02 17:31:55)
good morning
---Attachments---
image: https://misskey-taube.s3.eu-central-1.wasabisys.com/files/webpublic-27afa8a1-2e9c-4c1c-b185-160dcc8f0a8e.webp
image: https://misskey-taube.s3.eu-central-1.wasabisys.com/files/webpublic-1e5ee9e3-4e06-4c8c-b790-4b3f38ac3bc4.webp
Reply to @mametsuko@mk.absturztau.be
Ganbold@ganbold@sns1.hydranlab.com (2026-08-02 17:32:36)
morning hafer soja drink
Reply to @ganbold@sns1.hydranlab.com
Ganbold@ganbold@sns1.hydranlab.com (2026-08-02 17:32:06)
prolly also look for motherboard/cpu that support multiple GPU so I can get like 16-24GB or even 36GB with it which sounds stupid but for the GPU only its less than 500 bucks in total
mame muumimamma@mametsuko@mk.absturztau.be (2026-08-02 17:31:55)
good morning
---Attachments---
image: https://misskey-taube.s3.eu-central-1.wasabisys.com/files/webpublic-27afa8a1-2e9c-4c1c-b185-160dcc8f0a8e.webp
image: https://misskey-taube.s3.eu-central-1.wasabisys.com/files/webpublic-1e5ee9e3-4e06-4c8c-b790-4b3f38ac3bc4.webp
Ganbold@ganbold@sns1.hydranlab.com (2026-08-02 17:28:00)
why is RX580/RX590 is so cheap for 8GB VRAM????????
meanwhile 6GB market are stupidly pricey. what is the catch?
I could go out and buy it right now and running 9B or 12B model with a decent speed
arkw@2日目(日)東W28a@arkw@mi.arkw.work (2026-08-02 17:11:38)
I'm at ガスト 岡崎北店 in 岡崎市, 愛知県 swarmapp
Reply to @novemberhotel@mastodon.art
Mastodon•ART 🎨 Curator@Curator@mastodon.art (2026-08-02 17:00:31)
@novemberhotel yeah! Use the Curator Prompts numbered hashtag for people to find the sketches related to this 'edition' of prompts ^.^
Reply to @Clara_Pirx@mastodon.art
@Clara_Pirx@Clara_Pirx@mastodon.art (2026-08-02 16:57:20)
Frage an den Admin oder wie das heisst: ist hier eigentlich das Thema Suizid erlaubt wenn ich ein CN davor setze? Keine Werbung, mich beschäftigt lediglich etwas. Und bevor Panik aufkommt:
https://m.youtube.com/watch?v=F2cEaWQVsUU&list=RDF2cEaWQVsUU&start_radio=1&pp=ygU9aGVybWFuIHZhbiB2ZWVuIGRlciBtYW5uIGRlciBzbyBnZXJuZSBuaWNodCBtZWhyIGxlYmVuIHdvbGx0ZaAHAQ%3D%3D&ra=m
をとすま@wotosuma@social.nekokawa.net (2026-08-02 16:56:13)
てかPleroma-FEっていつ治るんだろ
@Clara_Pirx@Clara_Pirx@mastodon.art (2026-08-02 16:53:08)
#painting #drawing #sketch #art #MastoArt #kleinekunstklasse #abstract #depression
#comic
Moin. Nein, das Problem ist überhaupt nicht behoben, aber ich hatte halt Sehnsucht.
---Attachments---
image: https://cdn.masto.host/mastodonart/media_attachments/files/117/024/828/228/817/914/original/a6d4e9fafb66c4a5.jpeg
Ganbold@ganbold@sns1.hydranlab.com (2026-08-02 16:50:20)
honestly extremely annoyed with the current throughput per second for either prefill and decode. using cloud inference, it won't give me a time to breath. but using local inference, it gives me plenty of room to breath but I feel like my brain in stale state for a good few seconds which lead it to booting once the inference is done.
I am not sure what can I do with 4GB/6GB/8GB VRAM. kinda afraid with the model quality around that size even if it gives me speed. the amount of money to waste is just too scary
をとすま@wotosuma@social.nekokawa.net (2026-08-02 16:45:00)
nekokawa socialに垢作っても使わんかなーって思ったけど、普通に学校のPCでバリバリ動くからめっちゃ使ってる()
Ganbold@ganbold@sns1.hydranlab.com (2026-08-02 16:39:36)
yoo that is goat
I never knew ivy uses it. what is the tracking pixel one tho
> Lightweight and fast; adds just ~3.5K of extra data to your site. Also has JavaScript-free "tracking pixel" option, or you can use it from your application's middleware or import from logfiles.
going to look at it for sure
:hosimiya_mion::star_stroke:@hos1miya@misskey.0sakana.xyz (2026-08-02 16:38:06)
韓国語なんて全然分からんのに:supiki:のせいで答え分かってしまって悔しい
---Attachments---
image: https://misskey.0sakana.xyz/files/webpublic-4f073641-2d3f-4c4c-9b91-3850e4747731
Older Notes