GeForce RTX 5090 gets Chinese D version, no performance reduction

Special high-end Blackwell model for China due to sanctions

A year ago, due to US sanctions aimed at limiting China’s access to powerful AI acceleration, Nvidia began selling the cut-down RTX 4090D instead of the GeForce RTX 4090 in that market because the high-end gaming GPU was already crossing the performance limits imposed. As expected, the new Blackwell generation will face the same problem and Nvidia is preparing a special RTX 5090D Chinese SKU, but reportedly with full performance.

A Chinese leaker with the handle MEGAsizeGPU has now confirmed that Nvidia’s GeForce RTX 5090D is in the pipeline, showing off a logo from Nvidia’s design assetsapparently sourced from Nvidia materials on (formerly) Twitter. It seems the designation will once again feature the “D” letter. There is no news yet about a possible RTX 5080D model, so this SKU(GeForce RTX 5080) may not require adjustments due to the sanctions.

Later on, the first details on the GeForce RTX 5090D appeared on the Chiphell forum. According to a relatively well-established leaker with the nickname Panzerlied, this model is supposed to have the same configuration as the western (or global) GeForce RTX 5090 card, i.e. 21,760 shaders (170 SM blocks) as well as full memory capacity and bandwidth. The source even states that there will be no hardware difference at all between the GeForce RTX 5090 and the “Chinese” RTX 5090D. Nvidia didn’t even deactivate any Tensor Core units for AI acceleration, which looked like one of the possible means of limiting the AI capabilities without compromising game performance much (if the shader units were kept).

GeForce RTX 4090D specs on Nvidia’s Chinese website, where the card replaced the original RTX 4090 (Author: Nvidia (China), via: VideoCardz)

Instead, it seems that Nvidia will just limit performance in some applications (like AI software) through software means, presumably using some solution similar to what was used to slow down gaming graphics cards in “mining” tasks during the crypto-mining bubble. Back then, graphics cards with the “Lite Hash Rate” (LHR) label had their firmware modified to only accept drivers that had the appropriate cryptographic signature, and detection of various crypto-mining software was then implemented in those drivers. If the card detected such compute loads, it reduced performance accordingly.

Will software protection amount to anything?

Something similar, then, will probably be featured in the GeForce RTX 5090D graphics cards. This raises doubts about whether the measure will be effective. The mining protection on LHR cards was broken or worked-around fairly quickly. Hackers even fount more than one way around it: changes and obfuscation to the mining programs can fool detection, and the cryptographic protection against using hacked unrestricted drivers with the GPU only works as long as nobody finds any vulnerability that allows injecting hacked firmware into the card. In practice, such security measures are often overcome sooner or later.

Additionally, there is the risk that properly signed drivers with the restriction not applied will at some point leak directly out of Nvidia, at which point the protection is compromised forever – which is exactly what happened with the LHR cards. In this case, the adversaries will not just be crypto peddling individuals, but an actual state actor, meaning that Chinese institutions and companies can rely on the deployment of sophisticated espionage or agents to gain means of bypassing the protection.

Exactly what the protection will look like is not yet clear though, so the results cannot be fully predicted at this time. The GeForce RTX 5090D graphics cards will apparently be released in January 2025 alongside the standard RTX 5090 model. So it is also likely to be sold at the same price. However, the exact sums is not yet known. It’s quite possible that Nvidia will again raise it above the 1600 USD the company currently charges for the RTX 4090 as the high-end gaming graphics card model.

Sources: VideoCardz (1, 2)

English translation and edit by Jozef Dudáš


  •  
  •  
  •  
Flattr this!

UDNA: Next-gen architecture will unite AMD’s gaming and AI GPUs

The next generation AMD GPUs with RDNA 4 architecture should be coming soon. The company has now confirmed the rumors that high-end models will not be released in this generation and it will cover only part of the performance and price range. But the company also discussed their long-term roadmaps for the Radeon and Instinct GPU architectures. It seems we could be in for a change as significant as the transition from GCN to RDNA. Read more “UDNA: Next-gen architecture will unite AMD’s gaming and AI GPUs” »

  •  
  •  
  •  

Mobile Zen 5 is here: Ryzen AI 300 “Strix Point” SoC detailed

The Ryzen AI 300 mobile CPUs with Zen 5 architecture officially launched on Sunday. There’s a lot of news to go along with it: a third model has been added to form the top of the range, and we have learned various other architectural details of the laptop version of Zen 5 (and Zen 5c). Including information about the implementation of AVX-512, which as leaked before and will have lower performance than the fully 512-bit desktop Ryzen 9000. Read more “Mobile Zen 5 is here: Ryzen AI 300 “Strix Point” SoC detailed” »

  •  
  •  
  •  

Gigabyte SSD brings back SLC NAND, lasts 109,500 write cycles

The boom (or bubble?) around AI has brought many things, and among them interesting news for those missing SSDs based on MLC and SLC NAND Flash which was more pricy but had better performance and crucially, much longer lifespan so you didn’t have to worry about wearing out the SSD. That said, Gigabyte is launching an SSD that is officially designed for AI applications, but not just for them – its main asset is precisely SLC recording. Read more “Gigabyte SSD brings back SLC NAND, lasts 109,500 write cycles” »

  •  
  •  
  •  

Leave a Reply

Your email address will not be published. Required fields are marked *