GPU Moore Threads AI (MTT)

[safar1980]

China-made Moore Threads AI GPUs used for three billion parameter LLM training — MTT S4000 appears competitive against unspecified Nvidia solutions

Moore Threads claims to be making great strides in its AI GPU development, with its latest S4000 AI GPU accelerator being exponentially faster than its predecessor. As reported by cnBeta, a training regimen of a new Kua’e Qianka Intelligent Computing Cluster sporting S4000 GPUs ranked third fastest in AI testing, outperforming several counterparts consisting of Nvidia AI GPU clusters.

The benchmark run was taken from a stability test of the Kua’e Qianka Intelligent Computing Cluster. Training took a total of 13.2 days and supposedly ran perfectly with no faults or interruptions for the duration of the run. The AI model used to benchmark the new computing cluster was the MT-infini-3B large language model.

[FP91G]

China-made Moore Threads AI GPUs used for three billion parameter LLM training — MTT S4000 appears competitive against unspecified Nvidia solutions
Moore Threads claims to be making great strides in its AI GPU development, with its latest S4000 AI GPU accelerator being exponentially faster than its predecessor. As reported by cnBeta, a training regimen of a new Kua'e Qianka Intelligent Computing Cluster sporting S4000 GPUs ranked third fastest in AI testing, outperforming several counterparts consisting of Nvidia AI GPU clusters.

The benchmark run was taken from a stability test of the Kua'e Qianka Intelligent Computing Cluster. Training took a total of 13.2 days and supposedly ran perfectly with no faults or interruptions for the duration of the run. The AI model used to benchmark the new computing cluster was the MT-infini-3B large language model.

https://www.tomshardware.com/pc-components/gpus/china-made-moore-threads-ai-gpus-used-for-three-billion-parameter-llm-training-mtt-s4000-appears-competitive-against-unspecified-nvidia-solutions

[ccplz]

so how do we make them do calculations?

[FP91G]

so how do we make them do calculations?

I don't think that's a major concern for the manufacturer right now. Video cards for mining are practically not bought now, except for powerful video cards for renting out for computing. But to use these graphics cards in their Chinese infrastructure in the country is very realistic.

[FP91G]

New Chinese office GPU can double as a budget 1080p gaming GPU — MTT S50 wields 2,048 MUSA cores, 8GB VRAM, 85W TGP

Chinese GPU manufacturer Moore Threads has unveiled another S-series GPU in its consumer lineup. This time, it has debuted a mid-range model dubbed the S50, featuring 8GB of VRAM and 2,048 MUSA cores.

The S50 sits firmly in the middle of Moore Threads' consumer GPU product stack, flanked by the S60, S70, and S80 at the top and the S30 (4GB, 2GB) and S10 at the bottom. Specs-wise, the S50 is incredibly similar to the S60. The S50 features 2,048 cores, 8GB of memory, and a 256-bit memory bus, which are the exact same specifications as the S60. Both GPU variants also come packaged in a single-slot graphics card form factor.
https://www.tomshardware.com/pc-components/gpus/new-chinese-office-gpu-can-double-as-a-budget-1080p-gaming-gpu

[Crypt0Gore]

Easy now, these are very good for AI related heavy workloads, mining performance may varies from the RTX4000 and the RTX5000s, until a data shows the performance of the AI GPUs it is not a nice move to buy them for mining.

Since all PoW coins have different mining algorithm I am sure that performance wise will be different across the board, as good as the new Intel graphics are for gaming today, especially that Intel ARC A750, that GPU is a beast for gaming for its selling price and you can't use this GPU for mining.

[FP91G]

Easy now, these are very good for AI related heavy workloads, mining performance may varies from the RTX4000 and the RTX5000s, until a data shows the performance of the AI GPUs it is not a nice move to buy them for mining.

Since all PoW coins have different mining algorithm I am sure that performance wise will be different across the board, as good as the new Intel graphics are for gaming today, especially that Intel ARC A750, that GPU is a beast for gaming for its selling price and you can't use this GPU for mining.

Who told you that this video card doesn't work in mining?
https://www.hashrate.no/gpus/a750



JPR has published another report on the graphics card market. At the end of the first quarter, Nvidia holds 88% of the market in terms of shipments and continues to increase its advantage. Nvidia leads in all sectors, including artificial intelligence, data centres, and video games. Demand for its products is so strong that the company has become the second most capitalised company in the world.
https://t.me/HomishTG/7146

[FP91G]

Moore Threads has confirmed that it will unveil a new GPU architecture at its first MUSA Developer Conference (MDC 2025) in Beijing on December 19–20. The event centers on the company’s MUSA platform, with founder and CEO Zhang Jianzhong set to present the long term GPU strategy and product roadmap. The company promises a new generation of GPU architecture and an updated product lineup, but so far has not named specific boards.





For gamers, this announcement follows a rough first generation. Moore Threads already sells several desktop GPUs, led by the MTT S80, which launched in 2022 as China’s first retail gaming card and the first consumer PCIe 5.0 x16 GPU, with 16 GB of GDDR6 memory and 4,096 MUSA cores. Below it sit the cheaper MTT S70 and entry-level MTT S30, while professional users get models such as the MTT S3000, S4000 and S90 accelerators. On paper the S80 targeted GeForce RTX 3060 class performance, but early testing often placed it closer to older GTX 1050 Ti or GTX 1650 level hardware, and in some games it lagged even further behind.
Link