@rtroar

software working on https://xitter.cf/t.co/uUGlD1N185 gpu kernel and inference optimization

 Santa Cruz, CA
Joined October 2008
  • Tweets4.7K
  • Following1.3K
  • Followers1.6K
164Media
Pinned Tweet
rain
@rtroarSep 11
I've been working on my own inference engine (rainference) to see just how hard I can push the AMD Radeon 9070 XT with qwen 3.8 27b At this point I'm now more than double llamacpp at 84.8tok/s decode (this is with the DE sharing the card) field report here: localmaxxing.com/en/models/Q… @LottoLabs plz add rainference to the engineName enum so I can post speedtests :D

Qwen3.8-27B on RX 9070 XT: custom HIP engine, 84.8 tok/s (MTP) / 80.8 tok/s (DFlash2 drafter)

Single 16 GB RDNA4 card, custom IQ4_S256 quant, WY-form GDN prefill + WMMA flash attention; lossless speculative decoding verified byte-identical across three configs.

localmaxxing.com
3
1
2
38
10,609
rain retweeted
Roy
@usr_bin_roygbiv16h
I pulled the IRS data, 110 people in the United States under the age of 25 made $10M or more total in 2023
Sarah Chieng
@MilksandMatcha23h
We're living in a horrifying, fascinating new era of Silicon Valley where: 1. You can build a startup in 9 months, get acquired (but rly acquihired) by OpenAI, Anthropic, or Cognition, and walk away a multi-millionaire 2. You can join the right company at 22 and become effectively post-economic And I don't think we've even begun to grapple with the degree to which this is warping an entire generation's relationship with money, ambition, work, and success. When your peers are making generational wealth before 25, what does a normal career even look like anymore
37
28
8
1,137
70,939
I think the nepo hiring is going to get much worse
Autism Capital 🧩
@AutismCapital23h
Microsoft laying off 30% of the workforce (voluntary buyouts for now but will become layoffs if not exercised) can be read a few different ways, all valid 1) AI replacement 2) They were 30% overstaffed to begin with 3) Adult day care is over 4) No more (or less) H1B nonsense Hopefully the era of the nepo job is over.
1
1
21
769
Ive been seriously cooking on a generative video pipeline for a few weeks now. It's vaguely like comfyUI but the focus is an e2e workflow that handles HDR and log formats like you'd use in a professional context. All all self hosted.
3
1
13
425
I’ve noticed that the same people who say “we will always need swe, and learning LLMs isn’t a skill because it’s easy to use them” are the same people who say “LLMs aren’t good at coding and can only do slop” Much to consider 🤔
2
1
16
671
rain retweeted
log4j
@log4jOct 24
2027 will be an incredible year for you if you’re technically gifted, neurologically forked and have access to exogenous focus enhancers 2027 for everyone else will be, idk, just be there for them (nothing changes for the amish)
33
105
6
1,711
47,800
bw fans stay winning
Roy
@usr_bin_roygbivOct 24
In terms of who I would hire S: Dota Starcraft Eve OSRS Paradox games Factorio A Tier: Diablo/Poe Traditional MMOs Tekken Minecraft mods Wall of mediocrity/case by case F: League Cod Fortnite Vanilla minecraft It's far closer correlated to IQ than a degree is at this point
5
266
the next decade is just going to be escalating this cycle
zack
@zack_overflowOct 24
the best part of “fuck around and find out” was the dopamine of finding out what you were gonna find out but then agents made the fucking around part so much faster and cheaper, that i find now that i am finding A LOT of things out but because of this new abundance of finding things out, i eventually found that the finding out part was not as exciting anymore so i started upping the ante and had the agents REALLY fuck around so i could find some quite spectacular things out
1
7
479
We have our first offer received through applystead
10
5
144
4,570
rain retweeted
dan
@irl_danBOct 23
Ronald Reagan eats the same Big Mac as you Tim Cook uses the same iPhone as you the NSA runs the same abliterated GLM-5.3 as you
44
192
17
4,985
107,777
The only way to prevent garbage in prod now is to make sure that people understand what they are pushing, and there is no way to enforce that other than hiring smart people you trust to do things properly, which you should have been doing anyway rather than feeding egos
8
13
2
199
3,321
8
126
10
1,994
33,732
she's a 7 but hey that's 3 more than a root CA private key
13
482
13
8,489
81,900
Todays internet fun is brought to you by the number 4
solst/ICE of Astarte
@IceSolstOct 20
I don’t understand? How
19
705
11
9,980
132,820
it feels like the opportunity cost of a dead end job has never been higher
3
38
811
just wait until you see what you can do with [unannounced video suite] that I’m working on
Engineer Girlfriend
@enggirlfriendOct 20
fuck it’s happening - this is the best malfoid video i’ve seen yet and it’s clearly written and scripted by someone with a professional background AI video models are finally good enough and it’s in the hands of people who can script and craft at the highest level
1
8
308
You can just do things
Readers added context they thought people might want to know
This is not built by Farhan, instead it is a screenshot of a video from the original creator @trunarla. x.com/trunarla/statu…
96
88
37
2,579
417,422
What they really aren’t ready for.. is the fact it’s already there.
trundler
@_trundlerOct 19
gamedev artists are simply not ready for how good AI is about to get at 3d art. and the players won't care as soon as they can't tell the difference
1
21
489
this is kinda true if you’re otherwise fit and well manicured
Mia
@miaferrariiiOct 17
dressing down in sf is a form of respect towards the other person, its like bowing in japan, the lower you go the more respect you have
3
147
Replying to @ASUS
@ASUS fix my 150,000 USD paperweight
iArb | 裁定取引
@iarbpairsOct 13
never larp
18
15
6
108
12,927
I’d pay more for a car with no screens and physical controls for everything.
Alex Edgerton
@AlexEdgertonOct 16
On the very expensive Porsche Mission S you have to use your phone for navigation and entertainment, like on the cheap $20k Slate truck, or an old car.
3
18
448
Load more