The GeForce RTX 5090 D, a variant of NVIDIA’s new flagship “Blackwell” GPU for gamers, creators, and developers that was designed for the Chinese market and tweaked to comply with U.S. export rules, may fail to be recognized upon installing the latest GeForce driver, which NVIDIA rolled out last week to usher in DLSS 4 and its new family of GeForce RTX 50 Series GPUs, according to several complaints that have surfaced over Chinese social media channels following the 5090 D’s launch, including some who say that their 25,000 yuan graphics card have turned into a “brick.”
“After installing the driver, the screen goes black, the graphics card cannot be recognized, and the same thing happens after changing the DP port to HDMI,” reads one machine-translated complaint that can be found on the Chip Hell forums. “This time, the 90 series only released a small amount, and several of them were affected,” the poster goes on to claim.
“The latest news is that our group member’s newly bought 5090D became a brick after the driver was updated,” another said on China’s search giant Baidu before going on to point out some of the premiums that they had to pay in order to obtain the GPU. “The group member paid 6,000 yuan more to buy it. After the driver was updated, the graphics card could not be recognized. It has been sent for personal insurance. There are similar cases abroad. Please be cautious when buying the first batch of 5090.”
“RTX 5090D first bricked, don’t rush to buy 5090 at home and abroad,” warns another poster on China’s Goofish platform. “All 5090 and 5090D have exploded, which may be due to core burn. At present, even if you don’t enter the system, the card is not recognized in the bios.”
“The current known information: the driver will cause the graphics card to crash with a high probability, a small probability of screen distortion, and an extremely small probability of IC burnout,” a machine translation of the post continues. “I got it and plugged it in. It had all kinds of problems and became a brick. It was so miserable. It seems to be a driver problem. I installed it and it became a brick. Renaissance Starship 5090d. (I have asked the victim for authorization. I bought a new card for 25,000 yuan)”
“I bought an RTX 5090 but it is not detected on Windows in the device manager, nor by the NVIDIA drivers, nor in the BIOS,” reads another post on the ASUS subreddit, one that would suggest this issue may extend to the regular GeForce RTX 5090. “The LED on the graphics card is lit and the fans are running.”
“It is fixed by changing PCIE5.0 into 4.0 in bios. You can have a try,” recommended one user, although that didn’t seem to help.
Discussion (13 replies)
Join Discussion →I would not be surprised if (some of) these issues are because of tampered with/modded BIOSs to bypass the restrictions
Also who updates their GPU driver after installing the graphics card?
Uhhh.... everyone I would hope.
? I have upgraded many a GPU and I've never installed the driver after install unless it's a brand new build or the driver was bespoke. Never had a problem.
I never want Windows to install it's own driver.
Are you being serious? Every new to market GPU requires the latest driver.
And that driver is available and installed before I get the card 100% of the time (again unless it's bespoke or a new build)
Do yall not update drivers until after you upgrade your GPU?
Uhhh... yea? I'm shook that you don't.
1. I don't let windows install drivers. I get them from the chipset manufacturer, be that Intel, Nvidia, or AMD. And I tend to upgrade when the next newest one comes out or shortly after a few others on chats and forums do.
2. Drivers often give performance improvements, as of late with specific targeted games being released. As well as new features unlocked in updated driver sets.
By NOT updating drivers you're just part of the problem. Do you NOT patch your Operating systems and various downloaded software as well?
Mostly depends. New build as you mention, probably not; gotta have an OS loaded to install drivers (beyond the included generic ones). Already have that brand of GPU installed? Leave alone. Switching vendors? DDUing in safe mode, so it'll come up with no drivers again.
I think you are either misunderstanding what I posted or I didn't explain it properly.
I always update drivers and my OS. When a new GPU launches I install that new driver. When I get a new GPU the new driver is already installed on my system. I don't wait weeks or months to get my new card and then install that driver.
Ah I did misunderstand what you were trying to say. It sounded to me like you were one and done.
Since the new GPU's came out there was the launch driver... and a new driver just released yesterday. Hence the issue. Updated driver caused issues.
I suspect these same people have Less than NVME 5.x rated drives on the shared M.2 slot with the GPU PCIE 5.x rated slot. Hence a problem. I had similar when I had a NVME 3.x drive and went from a 2080 to a 6800xt. Later to a 7900xtx. Had to move my nvme drive to a different PCIE lane set to get my GPU to work right. I suspect they unlocked or modified how the GPU utilizes the PCIE slot with the updated driver. Especially since some have had success turning the PCIE slot down to PCIE 4.x standard instead of 5.x.
They seem to be stabbing into the dark about the PCIe 5 issue - but it's been an issue between generations before.
I can't really think of anything useful though. Something like pulling other PCIe devices like NVMe off of the CPU seems like a poor answer if that's necessary.
And I remain exceedingly cautious about boards that hack up the main PCIe x16 link for NVMe drives. I really, really wouldn't want to be dealing with issues there especially if I were counting on those NVMe slot(s) being useful.
That's why I'm having a hard time finding a good AMD mobo with enough NVMe options.
Look at ASRock. Their X870E options seem to have a better setup than most; I'm particularly fond of the Nova.