Techintosh

EE student @ Taiwan Tech EDA Lab
Sharing Tech and Insight

EVM: 0xb91CfCdA883979dd166EC6647213f2004FFe116E

Solana: GeM9St9BboD78BMWkbnmiGgh9TNhLxwri1FZxd7ZdWn4

[Techintosh @userHandle]

手機愈賣愈貴！全球平均售價創新高，Apple 獨拿近半營收
2026 年 Q2 全球手機營收達 1,090 億美元、年增 7%，出貨量卻下滑；平均售價年增 17% 至 400 美元。Apple 營收年增 22%，拿下 49% 營收市占。市場競爭正從賣更多支轉向賣更高價值的機型。

[Techintosh @userHandle]

@_jaydeepkarale FOMO

[Techintosh @userHandle]

@Vincent_AINotes 86B-A2B 了屬於是

[Techintosh @userHandle]

@adiix_official 傻逼，DDR5在15年後根本沒人要

[Techintosh @userHandle]

@ldl123292 @MaxForAI 點數，1000點數是40刀

[Techintosh @userHandle]

@qtdevlop @grok 納斯達克和標普500指數定投要買誰

[Techintosh @userHandle]

顯示規格包含雙 2160×2160、110 度視場角與最高 144Hz 更新率；控制器採磁吸 TMR 搖桿與電容式手指感應，續航最高 40 小時。
若零組件壓力持續，Valve 得在吸收成本、調整規格或延後推出之間取捨。你較在意售價，還是規格不縮水？
4/4

[Techintosh @userHandle]

Steam Frame 成本壓力升高，定價與預購恐得重算？
高通已將 Snapdragon 8 Gen 3 出貨價調高逾 10%。在 LPDDR5X 與快閃記憶體供應緊張下，Valve 的 PC VR 頭顯定價、預購與上市節奏都可能調整。
1/4

[Techintosh @userHandle]

硬體定位仍相當進取：頭顯本體重 185g，裝上完整頭帶後為 440g；支援 Wi‑Fi 7 與專屬低延遲轉接器，可無線串流 Steam 的大型 PC 遊戲。眼球追蹤搭配局部渲染，目標是降低 GPU 渲染負擔。
3/4

[Techintosh @userHandle]

@notnullptr 比deepseek v4 flash還貴是哪個天才會想用這個

[Techintosh @userHandle]

Google Pixel Tag 可能快登場：Android 尋物網路要補上關鍵一塊。

斯洛文尼亞電商頁面提前露出疑似 Pixel Tag，資訊顯示它可接入 Google Find Hub，透過周遭 Android 裝置回傳加密位置；近距離則能用蜂鳴器找鑰匙、皮夾。

若規格屬實，Google 將把龐大 Android 裝置網路轉成實際的物品追蹤優勢。

[Techintosh @userHandle]

透過瀏覽器直接在真實地圖開車

[DVLP @DVLPLONDON]

Goodbye clicking on Google Street View to check out the route!

In your browser, with https://t.co/2Zwyv0DEv9 you can actually drive anywhere in the world🚀
You can also create multiplayer races🚗🏎🚓

The world is generated ahead of you in real-time from OpenStreetMap data combined with satellite elevation data as you drive.

Just added a WIP of night mode and replaced the placeholder terrain/road textures with some nicer physics-based textures. Also now anywhere you go a million street lights will be added too. This is actually slowly starting to look like the promised Need For Speed + Google Maps
#threejs #gamedev #indiedev

Single player mode:
Just find a location, click anywhere on the map and click "Hop Here". After parachuting down hold X to call the car to the road.

Multiplayer mode:
Click on "Create race" and select start and finish points. A route will be plotted. Click on Share Race to get the link and start racing!

[Techintosh @userHandle]

在喜歡的名稱被搶走前趕快來註冊吧

[Brayden @BraydenWilmoth]

Reserve your Cloudflare Wallet tag before it's taken: https://t.co/txKBae09Vu

[Techintosh @userHandle]

這不是第一次出現「廠房先行、基建卡關」的情況。首爾近郊的龍仁半導體叢集，也曾因供水分配與居民反對而延宕。

分析來看，半導體已是高度依賴水電的重工業：競爭力不只在製程和資本支出，也在能否及早把供電、供水、環境影響與地方溝通一起完成。
4/4

[Techintosh @userHandle]

三星、SK海力士進駐也沒用？韓國湖南半導體園區面臨斷水電不足危機
晶圓廠蓋得起來，水電能否跟上才是關鍵。韓國計畫在西南部打造湖南半導體產業園區，初期投資達800兆韓元，目標在2030年前完成基建並開始營運。

但四座晶圓廠預估用電量，已相當於韓國西南部現有全年總用電的70%至80%。
1/4

[Techintosh @userHandle]

韓國產業通商資源部認為，西南地區既有發電能力高於當地需求，晶圓廠若落戶，可讓多數電力在地生產、在地消耗，減少長距離高壓輸電需求。

政府也評估大型核電廠與小型模組化反應爐（SMR），並稱在人口密集區盡量採地下化輸電線路；不過新線路與核能規劃仍遭部分居民反對。
3/4

[Syzygy Research @syzygyeng]

Today, we're introducing Mach-1 Additive, a 35 billion parameter model that can inference without ever multiplying by a weight. At 1.7 bits per weight, Mach-1 recovers 95% of the performance of the original full precision model, Qwen 3.6 35b, across 12 agentic and reasoning benchmarks, while being 10x smaller.

At 7GB, Mach-1 comfortably fits on consumer laptops with speeds of up to 120 tokens per second, making local inference not just feasible but useful.

Unlike algorithms like BitNet, our approach requires minimal retraining, under 15 GPU hours, making it scalable to massive LLMs. Over the coming weeks, we will be announcing and serving models of up to 3 trillion parameters compressed using our algorithm. For now, you can visit our website to play with Mach-1 directly in your browser, or download our desktop app.

We couldn't be more excited to launch Mach-1. We're looking forward to an energy efficient future for AI, powered by scaled intelligence density.

[Techintosh @userHandle]

AMD推 Ryzen AI Embedded X100＋Kria AI，瞄準機器人實體 AI
機器人不只要跑 AI，還得即時控制。X199 整合 16 核 Zen 5、40 CU RDNA 3.5 GPU 與 50 TOPS NPU；Kria AI SoM 主打統一記憶體與低延遲控制。AMD 數據稱，平台可在每秒做逾 8,000 次控制決策，VLA 推理低於 100 ms。