With the Maia 200, the I/O portion of the chip implemented 28 Ethernet ports running at 400 Gb/sec for a total of 2.8 TB/sec ...
BottleCap AI has released ThinkingCap-Qwen3.8-27B, a fine-tune of Qwen3.8-27B that spends 37.2% fewer thinking tokens across ...
On 16 March 2026, Jensen Huang opened Nvidia's GTC conference in San Jose and gave the first detailed look at Feynman, the chip generation Nvidia plans to ship in 2028. Nvidia first named Feynman at ...
智通财经APP获悉，在 纳斯达克综合指数 一举创下历史新高，以及占据纳指与 标普500指数 较高权重的“AI芯片超级霸主” 英伟达 创历史新高点位之后，美股第三季度财报季即将重磅开启。全球投资者们对美国企业营收与盈利超预期，尤其是对于与AI算力密切相关联的科技巨头们的营收与盈利超出分析师一致预期的期待热情再度升温。
Strataみたいに帯域速度の早いグラボのVRAMに、使うエキスパートだけ乗っけて他のエキスパートはDRAMとSSDに置いておく手法が確立された中で DGX Sparkは帯域速度がないし、分業させることも出来ないため伸び代はあまりないのでは？という疑念が ：その疑念は、かなり本質を突いています 結論から言うと、「DGX Sparkは128GBあるから巨大MoEに強い」という評価は、2026年のMo ...
NVIDIA's DGX Station for Windows is due in Q4 2026. We break down the 748GB memory split, the conditions behind the 20 PFLOPS figure, and what a trillion-parameter model needs in NVFP4, plus ...
What we weren’t ready for was another name change from Qualcomm, as the new chip is called Snapdragon 8 Elite Extreme Gen 5.
The 4-node NVIDIA DGX Spark rig pools 512 GB of memory, which is more than enough for the 476 GB footprint of DeepSeek ...
Geekerwan's teardown and analysis of the Kirin 9050 Pro show how Huawei's first LogicFolding chip works, and the results are ...
NVIDIA announced a new 64GB configuration of DGX Spark — from Acer, ASUS, Dell, Gigabyte, HP and MSI — its GB10-powered ...
A rare Ryzen 9 5900X3D sample has appeared with a single 64MB 3D V-Cache stack. Its second CCD retains the standard 32MB Zen ...
An engineering sample of AMD's never-released Ryzen 9 5900X3D has surfaced on the Chinese forum Chiphell, and it looks quite different from the prototype AMD has shown before. The 12-core, 24-thread ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results