🔥 Gate Square Event: #PostToWinNIGHT 🔥
Post anything related to NIGHT to join!
Market outlook, project thoughts, research takeaways, user experience — all count.
📅 Event Duration: Dec 10 08:00 - Dec 21 16:00 UTC
📌 How to Participate
1️⃣ Post on Gate Square (text, analysis, opinions, or image posts are all valid)
2️⃣ Add the hashtag #PostToWinNIGHT or #发帖赢代币NIGHT
🏆 Rewards (Total: 1,000 NIGHT)
🥇 Top 1: 200 NIGHT
🥈 Top 4: 100 NIGHT each
🥉 Top 10: 40 NIGHT each
📄 Notes
Content must be original (no plagiarism or repetitive spam)
Winners must complete Gate Square identity verification
Gat
AMD and OpenAI jointly release a powerful AI chip, improving inference by 35 times
Gate News bot message, early this morning, AMD held the 2025 Global AI Development Conference, mainly announcing and introducing the latest AI chips and cloud infrastructure hardware.
Sam Altman, co-founder and CEO of OpenAI, attended the conference as a special guest and jointly launched the Instinct MI400 and Instinct MI350 series of powerful AI chips with AMD. Especially during the development process, OpenAI has been providing technical feedback to AMD to help optimize its GPUs.
At the press conference, when Sam Altman heard that a single MI400 is equipped with 432G of HBM4 memory, he was also stunned and exclaimed it was impossible. Compared to the previous generation product, the MI350 is equipped with 288GB of HBM3E memory and has a memory bandwidth of up to 8TB/s, resulting in a 4-fold increase in AI computing power and a 35-fold increase in inference performance.
AMD stated that due to the lower power consumption of the chips compared to competitors, the MI355X provides 40% more tokens per dollar than NVIDIA chips.
Source: Jinshi