🏳️‍⚧️ System Hacker.

Joined July 2012
4,770 Photos and videos
codifyone, all in one OS application
2
158
vickieGPT retweeted
早安!6/15 外電綜合整理 - 2327國巨: 美系升目標 美系大行調升國巨26-28年EPS: $18.2/$30.7/$42.6,以明年27xPE評價,同步升目標。 券商表示AI 基礎建設使用的高容量 MLCC,其製造複雜度遠高於主流 MLCC 產品,而且用量也是大幅增加。舉例:VR200 機櫃中 47μF 以上 MLCC 的需求量相較 GB300 增加約 210%。所以47μF 以上 MLCC 占總 MLCC 數量超過 30%,而 GB300 則不到 20%。不僅單純數量增加,而是產業往更高容量發展。 也因為對其他 MLCC 類別產生的排擠效應,券商看到多項 0201 與 0402 MLCC(1μF 以下)的現貨價格已明顯上漲,部分料號甚至出現數倍的價格漲幅。在供需環境持續緊張的情況下,券商認為國巨有完整產品組合,且PEG 僅約 0.8 倍,調升獲利預估的同時也調高目標。 - 亞洲MLCC產業: 另一家美系對該產業的觀察是導入超高容量新規格 MLCC(1005-47μF)的同時會使產能折損(Capacity Penalty/Trade Loss)快速增加,導致的結果就是未來三年對產能有雙位數百分比的負影響,對消費性也有衝擊。狀況就是預估產業由 2025 年供給過剩 10%,會轉為 2028 年供給短缺 6%,程度更勝於上一波2017-2018年的週期。結論:券商喜歡日本/台灣的MLCC供應商大於韓國的。 - 3491昇達科:美系重申正向 美系券商表示受惠低軌衛星業務成長,下半年營收可望逐月增溫。預估2Q營收年增88%、3Q年增115%。預估衛星佔營收今年上升至78%,明年進一步到89%。重申正向。 - 6669緯穎:美系重申正向 5 月營收低於美系大行預估,券商表示因部分案子改爲Consignment所致,營收認列雖下降但對毛利率有正面助挹,降低營運資金壓力。展望後市,券商預估6月營收月增,預估3Q營收還是季增23%。重申正向。 - iPhone: 美系大行預估2Q: 5200 萬支優於過往季節性表現,預估3Q: 5400萬支(~季增4%、年減2%)。折疊機預估2H26推出,預估量700-800萬支。今年還會推出iPhone 18 Pro、18 Pro Max 。至於標準版1H27再推出。 #下次會考
6
2
37
9,465
1
2
546
vickieGPT retweeted
一个事儿,之前 claude 不是封禁了 openclaw 么对吧。 解封了。
33
2
59
42,133
vickieGPT retweeted
A couple of observations. There is no pure play COPOS player. Basically, existing suppliers adding to their technology capabilities. Interestingly, VisEra is listed as "Pilot Line and that's not coincidental. TSMC's subsidiary VisEra has been designated to establish the first CoPoS pilot line in 2026 making it the closest thing to a "pure-play CoPoS" entity in existence right now. VisEra is where TSMC's CoPoS technology may first be deployed as a pilot line, with large-scale mass production at the Chiayi AP7 site targeted for end of 2028.
Jun 14
LMAO, TrendForce just revealed the entire CoPoS supply chain lol.
2
5
47
9,288
vickieGPT retweeted
As soon as I woke up this morning, I read SemiAnalysis’ Huawei Kirin die-shot analysis piece that came out yesterday, and I was pretty surprised. What particularly caught my eye was the memory section. Specifically, Huawei appears to be using a mix of CXMT and Samsung memory in its smartphones. I had suspected as much. As Huawei’s smartphone volumes continue to grow, there would naturally be limits to meeting its internal demand purely by scraping together supply from Chinese distributors. I don’t know the exact ratio, but the fact that CXMT can provide fairly stable supply for demand from a customer like Huawei is quite meaningful. It suggests that yields have stabilized to a decent level, and that capacity is also becoming fairly meaningful. I found it very interesting.
20
52
574
65,898
claude code on iPad! building anything with wasm and ptx!
3
492
Would world first CXL-enabled Scale out Scale up Fabric deployed by Zettai?
2
241
vickieGPT retweeted
NVIDIA Hopper のCrossBar と SM/L2-MCの接続図 chipsandcheese.com/p/amds-cd… NVIDIA 公式ブログからHBM3には、2個のMCが付いている。 developer.nvidia.com/blog/nv…
8
42
2,550
4
691
vickieGPT retweeted
Jun 14
一个冷知识:NV 已经在所有 GPU 里集成了基于 RISC-V 的加密计算单元,单从 CPU IP core 的角度来看已经出货 1B 颗 详见 youtu.be/KhWtkZmOPn4?is=8kC3…
I had already wondered how Apple manages to perform inference at Google while simultaneously protecting their privacy, essentially their unique selling point. The answer: the heaviest requests run on Blackwell B200s inside Google Cloud, with NVIDIA's Confidential Computing encrypting the data while it's processed, so neither Google nor Apple can see it. "NVIDIA Confidential Computing provides a hardware-based security layer for accelerated AI workloads. The technology protects data while it’s being processed by isolating workloads in trusted execution environments and enabling systems to cryptographically verify that the infrastructure has not been tampered with before any sensitive data is sent to the server."
6
5
54
32,049
vickieGPT retweeted
Much of our CI platform runs on fixed hardware. That changes how we approach scale, visibility, and capacity, especially when auto-scaling isn’t an option. Full breakdown here: bit.ly/4cvtYrB
3
11
110
1,305,176
RT @jukan05: Explain why panel-level packaging is the future in one picture:
74
vickieGPT retweeted
2
8
77
2,446
zettai.us/forum/ cuda moat is crushed

8
650
vickieGPT retweeted
macOS 27 当iPad作为扩展屏幕时,不但可以调整iPad的分辨率,还直接支持 iPad 的屏幕触控,直接控制macOS!#wwdc26
12
33
438
99,826
vickieGPT retweeted
Jun 8
1T 模型跑在 8 卡系统达到 1000 tokens/sec 这不就是明牌 B300 NVL8 了吗?
🚀 1,000 TOKENS/S ON A 1T MODEL! 🚀 We are thrilled to release Xiaomi MiMo-V2.5-Pro-UltraSpeed in collaboration with @TileRT_AI , breaking the 1,000 tokens/s output speed on a 1 Trillion parameter model for the FIRST TIME! Not wafer-scale integration like Cerebras. Not pure on-chip SRAM chips like Groq. We achieve 1,000 tps on a 1T MoE model using just a SINGLE, STANDARD 8-GPGPU NODE. Read the full technical deep dive:mimo.xiaomi.com/blog/mimo-ti… Want to experience the future of real-time AI? 👉 Apply for UltraSpeed now: platform.xiaomimimo.com/ultr… ⏳ Limited-Time Access: Application-based · Jun 8 – Jun 23 (PDT) 💬 Chat Experience: Completely FREE for a limited time — try the blazing-fast web chat now. ⚡ UltraSpeed API: Just 3x the price for a ~10x boost in output experience. 🤝 Enterprise & Large-Scale Needs: business-mimo@xiaomi.com
18
8
106
37,649