Shulou(Shulou.com)11/24 Report--

Thanks to CTOnews.com netizen OC_Formula for the clue delivery! CTOnews.com September 22 news, after the release of the new generation of GeForce RTX 4090 GPU, Nvidia CEO Huang Renxun confirmed to reporters that the company is canceling its NVLink bridge design, and canceling the NVLink connector will free up more space to handle more AI operations.

"the reason we removed [NVLink] was that we needed I / O to do something else, so we used this area to stuff as many AI processors as possible," Huang Renxun explained why NVLink was cut.

For gamers, creators, and workstation users who want to engage in a multi-card exchange with the latest RTX 4000 series, Nvidia can implement similar support with PCIe Gen 5 after canceling NVLink.

"because Ada [Lovelace] is based on PCIe Gen 5, we now have the ability to achieve peer-to-peer transmission fast enough, which is a good deal," Huang added.

Compared to PCIe 4, PCIe 5 doubles the bandwidth by directly accessing the CPU channel, providing extraordinary speed advantages for storage, GPU, and memory.

In addition, RTX 4090 and RTX 4080 graphics cards only have HDMI 2.1and DP 1.4a interfaces, but do not have the latest DP 2.0interface. The NVIDIA official explained this in the community's Q & A, saying that the DP 1.4 interface now supports 8K 60Hz output.

In view of the high price when the RTX 4090 went on sale, Huang Renxun also announced that "Moore's Law is dead, and now the contract manufacturing price of 12-inch wafers is much higher than before, and the idea of chip prices falling is a thing of the past."

Prices rise across the supply chain, he says, and huge intergenerational performance improvements make it impossible to meet the condition of "buying the same performance for half the money" or twice the performance for the same money, just like the previous generation of RTX 3000 series.

"Today's 12-inch wafers are much more expensive, not a little bit more expensive, but a lot more expensive." now the technology is getting more and more expensive, so we have to use more methods, such as RTX, DLSS, SCR, Tensor Cores, etc., so that we can continue to overcome the increase in costs. "

"the performance of Nvidia's $899 GPU or $1599 GPU is much better than it was a year ago or two years ago, at the same price as Ada Lovelace."

CTOnews.com learned that Nvidia RTX 4090 flagship graphics card, equipped with 16384 CUDA + 24GB video memory, the proposed price starts at 12999 yuan and goes on sale on October 12.

According to reports, RTX 4090 is the world's fastest game GPU. In terms of performance, Nvidia said that in the full ray tracing game, compared with the previous generation of RTX 3090 Ti with DLSS 2, the performance of RTX 4090 with DLSS 3 can be improved by up to four times. In modern games, the performance of the RTX 4090 has been improved by as much as 2 times while maintaining the same 450W power consumption.

With 76 billion transistors, 16384 CUDA cores and 24 GB high-speed Meguiar GDDR6X memory, the RTX 4090 runs continuously at more than 100 FPS in games with a 4K resolution.

Top graphics card suppliers such as Asustek, Seven Rainbow, Gengsheng, Yingchi, gigabyte, Yingzhong, MSI and Suotai will launch GeForce RTX 4090 standard frequency version and super version in China. Acer, Alien, Asustek, Dell, Hewlett-Packard, Lenovo and MSI will also launch OEM products. In addition, NVIDIA will also launch a limited edition of the FE version of RTX 4090.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.