Aetina AIE-KT78/68 brings Jetson Thor Blackwell compute to deterministic edge AI control

Aetina launches AIE-KT78 and AIE-KT68 edge AI systems on NVIDIA Jetson Thor, with QSFP28, 8× GMSL2, and an isolated EtherCAT port for cobots, humanoids, and AMRs.

已發布 已更新 4 分鐘閱讀 ✓ 支援 AI 分享

TL;DR — 三句話摘要

  • Aetina announced the AIE-KT78 (Jetson T5000) and AIE-KT68 (Jetson T4000), the first Aetina edge AI systems on the NVIDIA Jetson Thor Blackwell platform. (S1)
  • Top SKUs deliver up to 2,070 FP4 TFLOPS (T5000) and 1,200 FP4 TFLOPS (T4000) with 128 GB and 64 GB of LPDDR5X on a 256-bit interface. (S1)
  • A dedicated 1GbE RJ45 EtherCAT master port is isolated from the Linux data plane, targeting deterministic multi-axis robot control. (S1)

取自 key_takeaways （實際欄位，最多 3 筆）。事實與分析分別標示。

Aetina AIE-KT78/68 brings Jetson Thor Blackwell compute to deterministic edge AI controlEditorial cover for a component sourcing brief. Chip motif; headline "Aetina AIE-KT78/68 brings Jetson Thor Blackwell compute to deterministic edge AI control". Reported key numbers: 2. 1 cited source. Static cover (no animation).Aetina AIE-KT78/68 brings Jetson ThorBlackwell compute to deterministic edgeAI controlREPORTED KEY NUMBERS21 sourceXAIPART
Cover image generated from data already shown on this page: the category motif, the headline, and reported key numbers. No decorative illustration; no third-party image.

您來自 AI 回答 — 這是可驗證的原始內容

AI 摘要可能遺漏了其主張的依據。原始預測、受影響零件及每一項引用皆在本頁，且均可追溯至其來源。

XAIPART 預測

本簡報包含可證偽的預測。

跳至完整預測及其證偽規則 ›
受影響零件

此報導無目錄相符項目 — 寧可不顯示，也不推斷相符。

查看判定方式 ›
正在尋找零件？送出詢價 → RFQ

Reported by EDN on . Summarized and analyzed by XAIPART Editorial.

What happened

Aetina has introduced the AIE-KT78 and AIE-KT68 edge AI systems, built on the NVIDIA Jetson Thor platform that uses the Blackwell GPU architecture for deterministic industrial control.S1 The systems target collaborative robots (cobots), humanoid robots, and autonomous machines, and they run multimodal AI models locally rather than relying on the cloud, reducing the latency that comes with off-device inference.S1 Supported local workloads include generative AI, large language models (LLMs), vision-language models (VLMs), and vision-language-action models (VLAs).S1

The AIE-KT78 uses the Jetson T5000 system-on-module with 128 GB of LPDDR5X on a 256-bit interface, delivering up to 2,070 FP4 TFLOPS of AI compute. The AIE-KT68 uses the Jetson T4000 module with 64 GB of LPDDR5X on the same 256-bit interface and up to 1,200 FP4 TFLOPS. Both platforms run Linux with the full NVIDIA AI software stack.S1

For connectivity, each system provides QSFP28 with up to 4×25-Gbps bandwidth, dual 10GbE RJ45 ports, and support for up to eight GMSL2 camera inputs aggregating data from cameras, LiDAR, radar, and other sensors. A dedicated 1GbE RJ45 port is reserved as an independent EtherCAT master, synchronizing motors, joints, sensors, and actuators for deterministic control loops.S1

Key numbers reported by EDN
MetricReported valueSource
AIE-KT78 AI computeUp to 2,070 FP4 TFLOPS[S1]
AIE-KT78 memory128 GB LPDDR5X (256-bit)[S1]
AIE-KT68 AI computeUp to 1,200 FP4 TFLOPS[S1]
AIE-KT68 memory64 GB LPDDR5X (256-bit)[S1]
QSFP28 bandwidthUp to 4×25 Gbps[S1]
GMSL2 camera inputsUp to 8[S1]
10GbE RJ45 ports2[S1]
EtherCAT master port1× dedicated 1GbE RJ45[S1]
OS / software stackLinux + full NVIDIA AI stack[S1]
PriceNot disclosed[S1]

Background timeline

  1. No prior milestones disclosed by the source.

XAIPART take

Technical impact

The AIE-KT78/68 is the first Aetina platform we have seen pair Jetson Thor’s Blackwell-class FP4 throughput with a hard-isolated 1GbE EtherCAT master on the same chassis, which is the topology robot integrators need to keep real-time motion control outside the Linux scheduler’s worst-case jitter.S1 Eight GMSL2 lanes plus a 4×25G QSFP28 fabric leave headroom for surround camera rigs and high-rate LiDAR fusion, and the 256-bit LPDDR5X channels (128 GB on the T5000 SKU, 64 GB on the T4000 SKU) are well matched to VLA and VLM inference at production frame rates.XAIPART analysis

Market impact

By targeting cobots, humanoids, and AMRs on a single SKU family, Aetina is positioning the Jetson Thor T5000/T4000 as a component-level alternate to x86 GPU edge boxes for robotics OEMs that want to consolidate perception, planning, and actuation on one compact platform.S1 For buyers already on Jetson Orin, the Thor upgrade more than doubles FP4 throughput per watt on paper, but the BOM delta versus the predecessor (T5000 vs AGX Orin 64/100 GB) will determine how fast humanoid and AMR programs re-spec their compute trays.XAIPART analysis

Action

Engineering teams piloting VLAs, AMRs, or vision-guided cobots should request an Aetina AIE-KT78 evaluation kit and validate end-to-end motion loop stability under worst-case sensor load before re-spending on legacy Orin-based designs. Procurement should treat the T5000 and T4000 SKUs as two separate allocations, since their 128 GB and 64 GB memory footprints map to different VLM/VLA batch-size ceilings.XAIPART analysis

XAIPART Forecast

XAIPART expects that, within 9 months of , Aetina will publish lead-time and unit pricing for the Jetson T5000-based AIE-KT78 and that published lead times will fall to 8 weeks or less for orders placed by ; if lead times remain above 12 weeks at that date this forecast is wrong. The metric (lead time in weeks) is industry-checkable against Aetina’s distributor channel and references the AIE-KT78 platform named in this brief. The threshold is XAIPART’s expectation, not a figure reported by the source.XAIPART analysis

Facts are as reported by the cited source. The design-in, sourcing, and forward-looking statements above are XAIPART’s analysis, not the source’s claims. Pricing, lead time, and availability for the AIE-KT78/68 were not disclosed by EDN; the product-page link points to Aetina and was not verified by XAIPART in this brief.

Corrections: none at publication. This brief will be updated with a timestamp if the source corrects its report.

替代品與生命週期

已建檔的替代品——目錄範圍，並非市場排名
零件關係生命週期依據
— 檔案中無替代品 需要資料 — 需要 match_type 在任何前代標籤出現之前

「前代」與「替代」標籤僅來自已記錄的相符類型、生命週期狀態或來源文字。未記錄相符項目時，不顯示任何標籤。我們自家的刊登內容不構成市場參考。

值得關注

可證偽預測 — 門檻、時間窗與目前狀態。已證偽的預測仍保留可見。
主張門檻作者狀態
XAIPART expects that, within 9 months of October 14, 2026 , Aetina will publish lead-time and unit pricing for the Jetson T5000-based AIE-KT78 and that published lead times will fall to 8 weeks or less for orders plac... lead times remain above 12 weeks at that date Pending

事後證明錯誤的預測會保留可見並加以標記。我們不會刪除。

受影響零件

無目錄相符項目。 此消息未對應到 XAIPART 目前上架的任何零件。我們不會推測相符項目，也不會顯示佔位清單。

各維度成本壓力

Source: FRED · Generated · reference output z and composite are algorithm outputs, not a fetch snapshot

Cost pressure by dimension — composite stateDiverging bars for three dimensions around a zero midpoint. Input Cost has no reading; Supply Demand z -1.4711, notable; Fx Energy z 0.1243, neutral. Composite index -10.1, band neutral.0Input costno readingSupply/demandz −1.4711 · notable46 dFX & energyz +0.1243 · neutral5 dcomposite −10.1 · neutral−100 · −40 · −15 · 0 · +15 · +40 · +100
Underlying values — copyable for verification.
DimensionFRED series IDUnitLagAs of
Input costPCOPPUSDM, PALUMUSDM, PNICKUSDMUSD/mt52 d
Supply/demandPCU334413334413, IPG3344Sindex(1982=100) / index(2017=100)46 d
FX & energyDEXCHUS, DCOILBRENTEU, DHHNGSPrate / USD/bbl / USD/MMBtu5 d

No free authoritative interface for: gold · silver · rare earth · polysilicon · lithium carbonate · market share — listed rather than approximated.

本內容的產生方式

流程證據 — 擷取、來源合理性、模型與閘門
此簡報的稽核軌跡
Source fetched EDN · www.edn.com · 2,724 chars · fetch status success
Source sanity (S1) ok — passed before any model call (no s1_ block recorded)
Model / prompt MiniMax-M3 · prompt Blog Agent - News Summary (v2.8-final) · 7,049 tokens · 31.4s · run completed
Gate verdict pass · 6/6 applicable checks passed · spec v2.9 · C10 ratio 0.103928 · C10 rounds 0
Time contract 3 <time datetime> values in this brief — 0 violations (+08:00 or date-only)

上述每一列皆由流程產生，並非人工撰寫。

編輯評分

確定性閘門結果 — 純 PHP，判定過程無模型參與
閘門評分明細 6 項適用檢查， 6通過： 100 百分比. No hold items or mandatory failures. Gate spec: v2.9 (S0 + S1 + C1–C20) · Source affinity (C20): 0.8085. 適用檢查已通過 6 / 6 = 100% · 判定： pass No hold items or mandatory failures Gate spec: v2.9 (S0 + S1 + C1–C20) · Source affinity (C20): 0.8085 是否發布由審核關卡決定。獨立稽核分數僅供參考。

相關報導

發現錯誤？ 告訴我們哪裡有誤 ——更正會與原文一同發布並記錄。我們不會默默修改文章。

留言

成為第一個留言的人。

來源

  1. Systems reduce latency in edge AI automation — https://www.edn.com/systems-reduce-latency-in-edge-ai-automation/ · rss · 存取於

事實標記為 S1/S2，並可追溯至連結來源。標示為 分析 是 XAIPART 自身的推理，與報導事實分開保留。來源連結會被追蹤，而非 nofollow ——署名為刻意為之。