DIRS.INFO The most relevant business & tech news

☆ New benchmark tests speed of running AI models
▼ + stars: | 2023-09-11 | by ( Max A. Cherney | ) www.reuters.com time to read: +2 min

[1/2] An AI (Artificial Intelligence) sign is seen at the World Artificial Intelligence Conference (WAIC) in Shanghai, China July 6, 2023. REUTERS/Aly Song Acquire Licensing RightsSept 11 (Reuters) - An artificial intelligence benchmark group called MLCommons unveiled the results on Monday of new tests that determine how quickly top-of-the-line hardware can run AI models. The benchmark simulates the "inference" portion of AI data crunching, which powers the software behind generative AI tools. Nvidia's top submission for the inference benchmark build around eight of its flagship H100 chips. Nvidia has dominated the market for training AI models, but hasn't captured the inference market yet.

Persons: Aly, hasn't, Dave Salvator, Eitan Medina, Max A, Leslie Adler Organizations: Artificial Intelligence, REUTERS, Nvidia Corp, Intel Corp, CNN, Nvidia, Habana, Intel, Google, Thomson Locations: Shanghai, China, San Francisco

The MLPerf benchmark is based on GPT-3, an AI model used to train ChatGPT, the viral chatbot developed by OpenAI and backed by Microsoft (MSFT.O). Habana Labs, an AI chip company acquired by Intel, ran the benchmark in 311.945 minutes with a much smaller system equipped with 384 Gaudi2 chips. "You will get a 1.5X to 2X speed up on the Habana results. So that's when we see Habana Gaudi2 being really competitive and lower priced than H100," Plawner told Reuters. Plawner declined to say how much a Gaudi2 chip costs, but said the industry needs a second supplier of chips for AI training, and the MLPerf results show Intel can fill that need.

Persons: David Kanter, Kanter, Intel's Jordan Plawner, Habana, Plawner, Jane Lanhee Lee, Christopher Cushing Organizations: Nvidia, Microsoft, Reuters, Habana Labs, Intel, Products, Habana, Thomson Locations: San Francisco, Habana

OAKLAND, California, June 13(Reuters) - Silicon Valley-based AI chip startup SiMa.ai on Tuesday said it raised an additional $13 million from investors including a key fund in Taiwan called VentureTech Alliance, which has a strong strategic partnership with Taiwan Semiconductor Manufacturing Co (2330.TW). This is at least the third investment in U.S. chip startups by VentureTech Alliance in the past month. British AI chip unicorn Graphcore's struggles have been widely reported. Rangasayee also pointed to one recent benchmark testing result by SiMa.ai that beat AI chip giant Nvidia Corp (NVDA.O) in performance and power of chips used on devices like cameras, drones and robots. The testing data is published by MLCommons, an engineering consortium that maintains testing benchmarks widely used in the AI chip industry.

Persons: VentureTech, Ethernovia, SiMa.ai, Navin Chaddha, Mayfield, Chaddha, they're, Moshe Gavrielov, Krishna Rangasayee, MLCommons, it's, David, Goliath, Jane Lanhee Lee, Lisa Shumaker Organizations: VentureTech, Taiwan Semiconductor Manufacturing, VentureTech Alliance, Ayar Labs, Nvidia Corp, Nvidia, Thomson Locations: OAKLAND, California, Taiwan, British

☆ Qualcomm, Nvidia spar for top spot in AI chip efficiency tests
▼ + stars: | 2023-04-05 | by ( Stephen Nellis | ) www.reuters.com time to read: +2 min

Nvidia dominates the market for training AI models with huge amounts of data. Qualcomm's chips hit 197.6 server queries per watt versus 108.4 queries per watt for Nvidia. Neuchips, a startup founded by veteran Taiwanese chip academic Youn-Long Lin, took the top spot with 227 queries per watt. Qualcomm also beat Nvidia at object detection with a score of 3.2 queries per watt versus Nvidia's 2.4 queries per watt. Nvidia hit 10.8 samples per watt, while Neuchips ranked second at 8.9 samples per watt and Qualcomm was in third place at 7.5 samples per watt.

Search resuls for: "MLCommons"

4 mentions found