Nvidia ships longawaited Jetson Thor AI chips
Cette publication existe aussi en Français
Nvidia has finally shipped its longawaited Jetson Thor GPU chip for embedded AI designs.
There are two versions with different power envelopes, the T5000 and T4000, aimed at humanoid robots, image processing and multi-modal edge AI.
The T4000 as 1536 Blackwell GPU cores with 64 fifth generation tensor cores running at a clock frequency of up to 1.57GHz. There are 12 ARM Neoverse V3AE automotive cores at 2.5GHz and 64GBytes of low power LPDDR5X memory. This provides up to 1200 TFLOPS of sparce FP4 4bit inference performance with eight lanes of PCI Express 5.0.
This has a power envelope of 40W up to 130W.
The T5000 has 2560 Blackwell cores with 96 tensor cores and 14 V3AE cores. This supports twice the memory, with 128GB of LPDDR5X. This also supports up to 8 lanes of PCIe5.0 with a lower peak power of 75W.
The chip package measures 100 x 87 mm with an integrated thermal transfer plate with a heat pipe and a 699 pin connector.
This is a key step for the DRIVE Thor chip that is central to autonomous driving in trucks and passenger cars .
Advantech is using the T5000 for its MIC-743 edge AI appliance that is ready to ship. This is ready for deployment in high-performance applications such as video language model (VLM) and large language models (LLMs) with up to 2,070 FP4 TFLOPs of AI performance with 128GB of LPDDR5X memory. This allows the MIC-743 to run the multi-modal AI tasks simultaneously.
www.nvidia.com; www.advantech.com
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
