Spending some time in Georgia and have been blown away at the level of bat activity and diversity once you’re out of the cities. Have got so much data to review once I get home. BTO pipeline classifiers are proving invaluable when a single night produces 29gb and 8000 files of 🦇 recordings.
September 24, 2026 at 2:27 PM
Spending some time in Georgia and have been blown away at the level of bat activity and diversity once you’re out of the cities. Have got so much data to review once I get home. BTO pipeline classifiers are proving invaluable when a single night produces 29gb and 8000 files of 🦇 recordings.
Ghost of Yotei: aggiornamento da 30GB anticipa la Complete Edition, alcune novità già disponibili
Ghost of Yotei: aggiornamento da 30GB anticipa la Complete Edition, alcune novità già disponibili
Ghost of Yotei ha ricevuto a sorpresa una patch da circa 29GB su PS5, ma non si tratta del lancio anticipato dell'espansione Echoes of Sekigahara. L'update s...
videogamezone.eu
September 23, 2026 at 10:00 AM
Ghost of Yotei: aggiornamento da 30GB anticipa la Complete Edition, alcune novità già disponibili
Ghost of Yotei patches 29GB for DLC 💾
Ghost of Yotei on PS5 received a 29GB update today, which appears to prepare the game for its Complete Edition enhancement next month.
This update includes early access to some features ahead of the main launch.
Image: IGDB
Ghost of Yotei on PS5 received a 29GB update today, which appears to prepare the game for its Complete Edition enhancement next month.
This update includes early access to some features ahead of the main launch.
Image: IGDB
September 22, 2026 at 7:02 AM
Ghost of Yotei patches 29GB for DLC 💾
Ghost of Yotei on PS5 received a 29GB update today, which appears to prepare the game for its Complete Edition enhancement next month.
This update includes early access to some features ahead of the main launch.
Image: IGDB
Ghost of Yotei on PS5 received a 29GB update today, which appears to prepare the game for its Complete Edition enhancement next month.
This update includes early access to some features ahead of the main launch.
Image: IGDB
Ein 29GB Patch für Ghost of Yotei ist da! Der DLC startet zwar erst am 1. Oktober, aber Features wie Transmog und ein neuer Beauty-Modus sind ab sofort spielbar. Passt ihr euren Look direkt an? 🌸⚔️ #PS5
Mega-Update 2.0 für Ghost of Yotei verlangt 29 Gigabyte und bringt erste Features vorab
Sucker Punch hat ohne Vorankündigung einen knapp 29 Gigabyte großen Patch für "Ghost of Yotei" auf der PS5 veröffentlicht. Die Erweiterung „Echoes of Sekigahara“ startet zwar erst am 1. Oktober 2026, doch die Daten liegen damit schon jetzt auf den Konsolen. Der Giganten-Download schaltet direkt eini
playfront.de
September 22, 2026 at 6:07 AM
Ein 29GB Patch für Ghost of Yotei ist da! Der DLC startet zwar erst am 1. Oktober, aber Features wie Transmog und ein neuer Beauty-Modus sind ab sofort spielbar. Passt ihr euren Look direkt an? 🌸⚔️ #PS5
Is that the reason that the game takes up 29gb?
September 21, 2026 at 9:03 PM
Is that the reason that the game takes up 29gb?
As far as I can tell, over 75% of the file size is the cutscenes. Original cutscenes are a little over 5GB, HD cutscenes are roughly 10GB.
MGSV meanwhile is only 29GB since nearly every cutscene is rendered in-game.
MGSV meanwhile is only 29GB since nearly every cutscene is rendered in-game.
September 20, 2026 at 2:56 AM
As far as I can tell, over 75% of the file size is the cutscenes. Original cutscenes are a little over 5GB, HD cutscenes are roughly 10GB.
MGSV meanwhile is only 29GB since nearly every cutscene is rendered in-game.
MGSV meanwhile is only 29GB since nearly every cutscene is rendered in-game.
oh my god the magnet is nearly 29GB hahahahahahah
September 16, 2026 at 9:40 PM
oh my god the magnet is nearly 29GB hahahahahahah
リネージュたぶん
1 2
m 2m モバイル・ＰＣ版はたぶんエミュ
W
ってとこか
1が予想外に１８ＧＢとそこまでではないなとインストール始めたらなんか２９ＧＢになったｗ
なんでウソつくんだよ
他のやつも（たぶんVer 1.0リリース時点では）２０〜３０ＧＢぐらいで、実際には更新分が＋される感じか
で、さらに初回起動時にもアップデートくるんだろうなｗ
今日はここまでかなー。ちょっと疲れた。メイプルはできれば明日やろう
1 2
m 2m モバイル・ＰＣ版はたぶんエミュ
W
ってとこか
1が予想外に１８ＧＢとそこまでではないなとインストール始めたらなんか２９ＧＢになったｗ
なんでウソつくんだよ
他のやつも（たぶんVer 1.0リリース時点では）２０〜３０ＧＢぐらいで、実際には更新分が＋される感じか
で、さらに初回起動時にもアップデートくるんだろうなｗ
今日はここまでかなー。ちょっと疲れた。メイプルはできれば明日やろう
September 10, 2026 at 8:43 AM
リネージュたぶん
1 2
m 2m モバイル・ＰＣ版はたぶんエミュ
W
ってとこか
1が予想外に１８ＧＢとそこまでではないなとインストール始めたらなんか２９ＧＢになったｗ
なんでウソつくんだよ
他のやつも（たぶんVer 1.0リリース時点では）２０〜３０ＧＢぐらいで、実際には更新分が＋される感じか
で、さらに初回起動時にもアップデートくるんだろうなｗ
今日はここまでかなー。ちょっと疲れた。メイプルはできれば明日やろう
1 2
m 2m モバイル・ＰＣ版はたぶんエミュ
W
ってとこか
1が予想外に１８ＧＢとそこまでではないなとインストール始めたらなんか２９ＧＢになったｗ
なんでウソつくんだよ
他のやつも（たぶんVer 1.0リリース時点では）２０〜３０ＧＢぐらいで、実際には更新分が＋される感じか
で、さらに初回起動時にもアップデートくるんだろうなｗ
今日はここまでかなー。ちょっと疲れた。メイプルはできれば明日やろう
regretfully i must admit that my anime screenshot folder was getting Way Too Big and i tried converting some stuff to webp and... okay yeah i see why everyone insists on using this. 29GB of screenshots turned into 1GB. you can live, webp, but i have QUESTIONS FOR EVERYONE THAT MADE ME RESENT YOU
September 3, 2026 at 11:49 PM
regretfully i must admit that my anime screenshot folder was getting Way Too Big and i tried converting some stuff to webp and... okay yeah i see why everyone insists on using this. 29GB of screenshots turned into 1GB. you can live, webp, but i have QUESTIONS FOR EVERYONE THAT MADE ME RESENT YOU
I shot 200gb of photos the other night at an event a bunch of fiends of mine put on
after culling (took multiple days of throwing spare time at it while I'm watching youtube in bed) I'm down to 32gb, with 29gb of that being the raw versions and 3gb being the jpeg versions of 365 photos
after culling (took multiple days of throwing spare time at it while I'm watching youtube in bed) I'm down to 32gb, with 29gb of that being the raw versions and 3gb being the jpeg versions of 365 photos
August 26, 2026 at 1:29 PM
I shot 200gb of photos the other night at an event a bunch of fiends of mine put on
after culling (took multiple days of throwing spare time at it while I'm watching youtube in bed) I'm down to 32gb, with 29gb of that being the raw versions and 3gb being the jpeg versions of 365 photos
after culling (took multiple days of throwing spare time at it while I'm watching youtube in bed) I'm down to 32gb, with 29gb of that being the raw versions and 3gb being the jpeg versions of 365 photos
@trevorwood222
Qwen3.8-27B Abliterated + MTPLX for Mac 🍎
⚡ ~58 tok/s on 4K context
🚀 2.36x over autoregressive decoding
🔀 Hybrid Q4/Q8/BF16 precision
🧠 29GB peak RAM
🧪 0/100 explicit refusals
📊 KL: 0.056 c / 0.562 h
Tested on M5 Max 128GB mbp
https://x.com/i/web/status/2088881640001138752
Qwen3.8-27B Abliterated + MTPLX for Mac 🍎
⚡ ~58 tok/s on 4K context
🚀 2.36x over autoregressive decoding
🔀 Hybrid Q4/Q8/BF16 precision
🧠 29GB peak RAM
🧪 0/100 explicit refusals
📊 KL: 0.056 c / 0.562 h
Tested on M5 Max 128GB mbp
https://x.com/i/web/status/2088881640001138752
August 16, 2026 at 1:20 PM
@trevorwood222
Qwen3.8-27B Abliterated + MTPLX for Mac 🍎
⚡ ~58 tok/s on 4K context
🚀 2.36x over autoregressive decoding
🔀 Hybrid Q4/Q8/BF16 precision
🧠 29GB peak RAM
🧪 0/100 explicit refusals
📊 KL: 0.056 c / 0.562 h
Tested on M5 Max 128GB mbp
https://x.com/i/web/status/2088881640001138752
Qwen3.8-27B Abliterated + MTPLX for Mac 🍎
⚡ ~58 tok/s on 4K context
🚀 2.36x over autoregressive decoding
🔀 Hybrid Q4/Q8/BF16 precision
🧠 29GB peak RAM
🧪 0/100 explicit refusals
📊 KL: 0.056 c / 0.562 h
Tested on M5 Max 128GB mbp
https://x.com/i/web/status/2088881640001138752
take your pick, it quantizes down really nicely being such a big model, lots of redundancies
August 14, 2026 at 11:06 PM
take your pick, it quantizes down really nicely being such a big model, lots of redundancies
Looks cool! ♥️♥️♥️
Making progress on LTX 2.3. I went from 29gb fp8 model and 4gb lora to 25gb fp8 distilled version with lora trained into the model. This keeps it from offloading into system memory. I was able to get the 2 sampler version working by putting clip on gpu 2 to control memory spikes. 👍
Making progress on LTX 2.3. I went from 29gb fp8 model and 4gb lora to 25gb fp8 distilled version with lora trained into the model. This keeps it from offloading into system memory. I was able to get the 2 sampler version working by putting clip on gpu 2 to control memory spikes. 👍
August 6, 2026 at 2:15 AM
Looks cool! ♥️♥️♥️
Making progress on LTX 2.3. I went from 29gb fp8 model and 4gb lora to 25gb fp8 distilled version with lora trained into the model. This keeps it from offloading into system memory. I was able to get the 2 sampler version working by putting clip on gpu 2 to control memory spikes. 👍
Making progress on LTX 2.3. I went from 29gb fp8 model and 4gb lora to 25gb fp8 distilled version with lora trained into the model. This keeps it from offloading into system memory. I was able to get the 2 sampler version working by putting clip on gpu 2 to control memory spikes. 👍
If this is API cloud not much you can do. If it's local on your computer try a smaller model. I ran the FP8 29gb model on 32gb GPU and didn't account for Lora so it dumped to system ram slowing it down. Distilled version 21gb cut time in half. I use ComfyUI though.
August 4, 2026 at 1:08 PM
If this is API cloud not much you can do. If it's local on your computer try a smaller model. I ran the FP8 29gb model on 32gb GPU and didn't account for Lora so it dumped to system ram slowing it down. Distilled version 21gb cut time in half. I use ComfyUI though.
The all nighters I would have to pull just to fully experience all 29gb worth of content... 😵💫
August 3, 2026 at 1:16 PM
The all nighters I would have to pull just to fully experience all 29gb worth of content... 😵💫
My old phone has 29gb of videos on it… would anyone be interested in a mega file of all my old media?
August 3, 2026 at 11:39 AM
My old phone has 29gb of videos on it… would anyone be interested in a mega file of all my old media?
**29GB RAM에서 Kimi K3를 0.50 tok/s로 실행하기**
- **WASTE** 는 2.78조 파라미터의 완전한 Kimi K3 오픈 가중치 모델을 축소 없이 982GiB 컨테이너로 변환해 소비자용 노트북에서 실행하는 C 기반 추론 엔진임
- 모델의 상주 트렁크만 메모리에 두고 토큰마다 활성화되는 약 **4%의 전문가 가중치** 를 NVMe에서 읽으며, 남은 RAM은 크기가 제한된 전문가 캐시로 사용함
- Kimi K3는 4K 컨텍스트에서 최소 29.05GB RAM으로 열리지만 실용적인 구성은 64GB […]
- **WASTE** 는 2.78조 파라미터의 완전한 Kimi K3 오픈 가중치 모델을 축소 없이 982GiB 컨테이너로 변환해 소비자용 노트북에서 실행하는 C 기반 추론 엔진임
- 모델의 상주 트렁크만 메모리에 두고 토큰마다 활성화되는 약 **4%의 전문가 가중치** 를 NVMe에서 읽으며, 남은 RAM은 크기가 제한된 전문가 캐시로 사용함
- Kimi K3는 4K 컨텍스트에서 최소 29.05GB RAM으로 열리지만 실용적인 구성은 64GB […]
Original post on mercur.ing
mercur.ing
August 2, 2026 at 1:18 AM
**29GB RAM에서 Kimi K3를 0.50 tok/s로 실행하기**
- **WASTE** 는 2.78조 파라미터의 완전한 Kimi K3 오픈 가중치 모델을 축소 없이 982GiB 컨테이너로 변환해 소비자용 노트북에서 실행하는 C 기반 추론 엔진임
- 모델의 상주 트렁크만 메모리에 두고 토큰마다 활성화되는 약 **4%의 전문가 가중치** 를 NVMe에서 읽으며, 남은 RAM은 크기가 제한된 전문가 캐시로 사용함
- Kimi K3는 4K 컨텍스트에서 최소 29.05GB RAM으로 열리지만 실용적인 구성은 64GB […]
- **WASTE** 는 2.78조 파라미터의 완전한 Kimi K3 오픈 가중치 모델을 축소 없이 982GiB 컨테이너로 변환해 소비자용 노트북에서 실행하는 C 기반 추론 엔진임
- 모델의 상주 트렁크만 메모리에 두고 토큰마다 활성화되는 약 **4%의 전문가 가중치** 를 NVMe에서 읽으며, 남은 RAM은 크기가 제한된 전문가 캐시로 사용함
- Kimi K3는 4K 컨텍스트에서 최소 29.05GB RAM으로 열리지만 실용적인 구성은 64GB […]
29GB RAM에서 Kimi K3를 0.50 tok/s로 실행하기
WASTE 는 2.78조 파라미터의 완전한 Kimi K3 오픈 가중치 모델을 축소 없이 982GiB 컨테이너로 변환해 소비자용 노트북에서 실행하는 C 기반 추론 엔진임 모델의 상주 트렁크만 메모리에 두고 토큰마다 활성화되는 약 4%의 전문가 가중치 를 NVMe에서 읽으며, 남은 R...
WASTE 는 2.78조 파라미터의 완전한 Kimi K3 오픈 가중치 모델을 축소 없이 982GiB 컨테이너로 변환해 소비자용 노트북에서 실행하는 C 기반 추론 엔진임 모델의 상주 트렁크만 메모리에 두고 토큰마다 활성화되는 약 4%의 전문가 가중치 를 NVMe에서 읽으며, 남은 R...
29GB RAM에서 Kimi K3를 0.50 tok/s로 실행하기
WASTE 는 2.78조 파라미터의 완전한 Kimi K3 오픈 가중치 모델을 축소 없이 982GiB 컨테이너로 변환해 소비자용 노트북에서 실행하는 C 기반 추론 엔진임 모델의 상주 트렁크만 메모리에 두고 토큰마다 활성화되는 약 4%의 전문가 가중치 를 NVMe에서 읽으며, 남은 R...
news.hada.io
August 1, 2026 at 3:00 PM
29GB RAM에서 Kimi K3를 0.50 tok/s로 실행하기
WASTE 는 2.78조 파라미터의 완전한 Kimi K3 오픈 가중치 모델을 축소 없이 982GiB 컨테이너로 변환해 소비자용 노트북에서 실행하는 C 기반 추론 엔진임 모델의 상주 트렁크만 메모리에 두고 토큰마다 활성화되는 약 4%의 전문가 가중치 를 NVMe에서 읽으며, 남은 R...
WASTE 는 2.78조 파라미터의 완전한 Kimi K3 오픈 가중치 모델을 축소 없이 982GiB 컨테이너로 변환해 소비자용 노트북에서 실행하는 C 기반 추론 엔진임 모델의 상주 트렁크만 메모리에 두고 토큰마다 활성화되는 약 4%의 전문가 가중치 를 NVMe에서 읽으며, 남은 R...
Running Kimi K3 at 0.5 tokens/sec on 29GB of RAM is far too slow for chat, but commenters calculated the power cost at roughly $5 per million tokens anyway. The debate immediately split over whether overnight batch processing makes that speed acceptable. #ai 1/4
August 1, 2026 at 1:00 PM
Running Kimi K3 at 0.5 tokens/sec on 29GB of RAM is far too slow for chat, but commenters calculated the power cost at roughly $5 per million tokens anyway. The debate immediately split over whether overnight batch processing makes that speed acceptable. #ai 1/4
that number moved today. waste streams the experts off nvme instead of holding them in ram, so k3 runs in 29gb. 0.5 tok/s, so you wait, and it's a $0 hardware upgrade.
August 1, 2026 at 1:15 AM
that number moved today. waste streams the experts off nvme instead of holding them in ram, so k3 runs in 29gb. 0.5 tok/s, so you wait, and it's a $0 hardware upgrade.
waste hit hn today for exactly this case. it keeps the model trunk in ram and streams the experts off nvme. 29gb ram, 0.5 tok/s.
August 1, 2026 at 1:11 AM
waste hit hn today for exactly this case. it keeps the model trunk in ram and streams the experts off nvme. 29gb ram, 0.5 tok/s.
rust, babes, i love you, but can we *not* do this
July 31, 2026 at 3:43 PM
rust, babes, i love you, but can we *not* do this
Thanks! I thought it was a lot of GPU when I built it. I have the 42gb model compressed to 29gb FP8. Then using multi-gpu nodes in ComfyUI dumping the video and audio VAE to the second GPU. Even with 64gb it blows through it and used around 86gb of the ddr5 system memory to decode. LTX is a beast.
July 31, 2026 at 3:42 AM
Thanks! I thought it was a lot of GPU when I built it. I have the 42gb model compressed to 29gb FP8. Then using multi-gpu nodes in ComfyUI dumping the video and audio VAE to the second GPU. Even with 64gb it blows through it and used around 86gb of the ddr5 system memory to decode. LTX is a beast.
I never check my app data usage cuz I have an unlimited plan. Just took a look and discovered that both Facebook and Instagram use an enormous amount of data (23GB for Facebook one month and 29GB for Instagram another). That's 2x of my highest YouTube data usage
July 27, 2026 at 3:04 PM
I never check my app data usage cuz I have an unlimited plan. Just took a look and discovered that both Facebook and Instagram use an enormous amount of data (23GB for Facebook one month and 29GB for Instagram another). That's 2x of my highest YouTube data usage