6/2/2026 at 5:52:48 PM
by kkm
6/2/2026 at 8:49:08 PM
I train on AMD MI250X and managed to get Gemma 4 31B to work - but it took a lot of work on the software side.by maCDzP
6/2/2026 at 8:51:31 PM
This is very interesting, planning to write about it?by kkm
6/2/2026 at 7:30:25 PM
Also the vllm patch accompanying the blogpost: https://github.com/doublewordai/vllm-amd-blog-doublewordby kkm
6/2/2026 at 7:31:26 PM
We at doubleword are bullish for AMD for low-interactivity inference - it does just take a bigger lift on the software side...by mezark
6/2/2026 at 9:10:07 PM
Are you long AMD?by brcmthrowaway
6/3/2026 at 3:01:22 AM
Interesting that you ask that as AMD hits another ATH.by latchkey
6/3/2026 at 6:18:40 AM
hello,sir, I want bulk order Asrock BC-250, is it still available ?by boxking
6/3/2026 at 6:31:59 AM
lol find the discord!by latchkey
6/3/2026 at 7:19:39 AM
yes,sir, any possibility to find 1000pcs or moreby boxking
6/3/2026 at 7:23:17 AM
They are all over ebay. https://www.ebay.com/sch/i.html?_nkw=bc-250I'm super curious what you would use them for.
by latchkey
6/3/2026 at 8:13:58 AM
now many guys want to buy this, I am reseller AMD BC-250,it is popular nowby boxking
6/3/2026 at 8:28:58 AM
thank you sir, actually it is not easy to find on ebay, because they are small seller, hard to find hundred, also ebay price a bit high for meby boxking
6/3/2026 at 3:58:18 AM
Then you are definitely long on AMD.by brcmthrowaway
6/3/2026 at 6:36:46 AM
More accurately... I'm long on a viable alternative to the current monopoly. We have two OS's for phones (android and ios), there is no reason why we shouldn't have the same for all AI hardware and software. The only one even close, is AMD.by latchkey
6/3/2026 at 6:10:16 AM
[dead]by keynha
6/3/2026 at 4:10:40 AM
Checked out this company about a year ago and they only offered small models. Now I see they have GLM-fp8/Kimi and DeepSeek V4 Pro. Since workloads are predominantly cached input, I'm surprised to see no separate price for cached input vs uncached. I hope the prices will drop significantly; with these prices you'll end up with thousands in monthly costs quickly. Hopefully more hardware companies will be on the market in the coming years. If the Chinese eventually start competing with the current memory makers, maybe that will help.by edg5000
6/3/2026 at 5:04:59 AM
[flagged]by zftnb666
6/2/2026 at 7:23:19 PM
Nice work! Would DeepSeek V4 Pro on 8xMI300X work with these patches?by benlm
6/3/2026 at 12:22:01 AM
Nice work and thanks for being a customer.(CEO Hot Aisle)
by latchkey
6/3/2026 at 4:27:01 AM
I wish you guys could partner with Modular to get Mojo inference working on your hardware, e.g. https://www.modular.com/models/deepseek-v4-proby erichocean
6/3/2026 at 6:34:45 AM
Not sure I understand. If they support MI300x, their self-hosted will run on our hardware.by latchkey
6/3/2026 at 8:19:28 AM
It’s just weird Deepseek released a model that was not compatible with any of the usual engines. Without derez’s new project just to support DSv4, how long until it’s actually viable in llama :(by alfiedotwtf
6/3/2026 at 1:33:24 AM
[flagged]by jing09928