I just finished swapping my 4k 5800x3d system to 9800x3d, thinking all was well but then noticed my Gigabyte 3090 Vision OC (owned from new) appeared to be core throttling in Cyberpunk benchmarks, even with my rock solid undervolt .875mv @ 1935mhz.
This card has always had very hot memory junction around 100C so I always downclocked the VRAM speeds by -300mhz to keep it under.
Tonight however I noticed the hotspot temp (wherever that is located!) was 104C in more intensive stuff, yet the memory temps were still sub 100C and the GPU core is usually floating around 67-69C at max stress. I suspect this is a fairly recent problem as old GPU-Z log files used to show hotspot much lower than this.
I really don't want to start dissecting it before I have a replacement lined up in case I break something (preferably at 5090 at this point), are there any other options here or do I just scale everything back and make it a pet project early next year?
Thanks in advance.
**Edit - so increasing the power limit to 105% stopped the core throttling behaviour bizarrely (on the older version of afterburner I didn't have to do this). I also ramped up the case fans further, however I am using the PCIE16 temp sensor in bios to set curve, which isn't giving me the desired results since it rarely changes. I may need to work out a way to ramp up the system fans based on the GPU core temp.
This card has always had very hot memory junction around 100C so I always downclocked the VRAM speeds by -300mhz to keep it under.
Tonight however I noticed the hotspot temp (wherever that is located!) was 104C in more intensive stuff, yet the memory temps were still sub 100C and the GPU core is usually floating around 67-69C at max stress. I suspect this is a fairly recent problem as old GPU-Z log files used to show hotspot much lower than this.
I really don't want to start dissecting it before I have a replacement lined up in case I break something (preferably at 5090 at this point), are there any other options here or do I just scale everything back and make it a pet project early next year?
Thanks in advance.
**Edit - so increasing the power limit to 105% stopped the core throttling behaviour bizarrely (on the older version of afterburner I didn't have to do this). I also ramped up the case fans further, however I am using the PCIE16 temp sensor in bios to set curve, which isn't giving me the desired results since it rarely changes. I may need to work out a way to ramp up the system fans based on the GPU core temp.
Last edited:
