GPU artifacts at lower clocks?
GPU artifacts at lower clocks?
Artifacts typically occur when increasing memory clock too much, though this issue appears within a narrow range in the low region. The GPU model is Palit GamingPro RTX 5070 with 12GB RAM and dual VBIOS (250W TGP). VBIOS versions 98.05.36.00.4C and 98.05.36.00.4D are installed. Memory clock adjustment varies between -2000MT/s to 6000MT/s or -125MHz to +375MHz. Asrock Taichi X99 PCIe 3.0 is connected via W10 Drivers 572.83, 576.52 and 576.80, with Linux Driver 575.75. Errors appear between 1649.125MHz and 1657.125MHz (DDR 13193MHz and 13257MHz, MT/s 26514MT/s and 26386MT/s). Applying a negative memory offset of approximately -97MHz for PState 0 usually triggers Event ID 14 followed by TDR attempts, which often fail and result in multiple ID 153 errors per second until the system freezes or powers off. It’s worth checking if this is specific to my card or common across other 50 series; be cautious, as aggressive power cycling may cause file corruption.
It appears to be a typical example of mistaken recollection, prompting an RMA process.
Single monitor setup, generally 120Hz @ 2k with DP, though during testing I added another graphics card and ran the RTX 5070 headless (no display connected). Clearly, I didn’t notice any artifacts, but the system froze when the RTX 5070 was placed in that narrow memory range. The issue is fully reproducible. Uncertain if it’s a driver problem or hardware fault, though hardware seems more likely. I was planning to discuss other aspects of this card (RTX 5070), but it looks like an administrator edited the thread into a question. Fine.
It appears to be a typical example of mistaken recollection, prompting an RMA process.
It might be worth checking for other possibilities first, to ensure it isn't related to the driver or VBIOS before proceeding.
adjusting to 60hz and checking which tool is being underclocked.
was there an unintended change in power limit or voltage settings? Please confirm with the afterburner.
Are the newest NVIDIA drivers installed?
Which additional hardware components are connected?
Motherboard plus BIOS version
CPU
PSU
RAM
Which VBIOS versions are active? Is this issue present across both versions?
Usually I use my own SW but have also tried nvidia-settings.
No, sorry, I prefer to skip that.
Yes, the W10 model is 576.88, though I’m not sure about Fedora at the moment. Other versions are mentioned in the first post.
CPU: P1.80, with some modifications from years ago while working with Xeon turbos.
CPU model: E5-2696-v3
Power supply: RM850x shift
Memory: 32GB GSKILL DDR4 2133MT/s quad channel
Yes.
Several questions were addressed in the first post.
I reached out to Palit support, and since the issue doesn't happen with the standard settings, it isn't covered under warranty—no return authorization needed.