Chipmaker AMD is looking into a report from the GMP project about two Ryzen processors that failed during testing. Could too much math be to blame?
The GMP project, an open source precision arithmetic library, this week said it had “fried two Ryzen 9950X CPUs in a few months by running GMP tests.”
GMP tests are computationally intensive, so sustained testing may have done thermal damage to the silicon. Images accompanying the post show visible discoloration, the apparent result of overheating.
Image of the back of the damaged Ryzen CPU – Click to enlarge
One Ryzen 9950X based system failed in February. And in August, the replacement CPU failed too.
“Before this problem is fully understood, we must caution people against using GMP heavily on any Zen 5 processor,” the project post explains.
Image of toasted Ryzen heatsink, courtesy of GMP – Click to enlarge
“We suspect that GMP’s extremely tight loops around MULX make the Zen 5 cores use much more power than specified, making cooling solutions inadequate.”
According to GMP, the issue here is not the widely reported problem with ASRock motherboards, though it is similar.
The two cited systems had Asus motherboards. Asus has not responded to a request for comment.
“We are aware GMP is investigating observations with their program on Ryzen 9950X processors,” an AMD spokesperson told The Register. “We have reached out to GMP to obtain additional details in order to better understand the situation.”
AMD executives just over a week ago told South Korean tech site Quasarzone that reports of AM5 socket burnouts can be attributed to hardware partners failing to adhere to recommended specifications. The AM5 socket supports Ryzen 7000, 8000, and 9000 series processors.
AMD’s spokesperson confirmed that statement but cautioned that it does not necessarily apply to what GMP has reported. “We cannot speculate on the cause of the GMP observations,” AMD’s spokesperson said.
Torbjörn Granlund, principal author of GMP, told The Register, “AMD is (again) replacing my CPU. They are collecting detailed information on my system and the load I ran when it died. The CPU is on its way to them, therefore you can really say that the CPU has gone to see its maker.
“No other CPU has ever died for me (in the 30 years I have built my own systems). Ryzen 9950X has now died twice. I was under the impression that modern CPUs had some protection to avoid overheating, and that they would clock down when they detected high temperatures.”
Granlund acknowledged that his cooling solution could have been better.
“While the ambient temperature was good, the case ventilation wasn’t great,” he said. “The heat sink was slightly underspec’ed. I am rebuilding the system in a much better case, with a bigger heat sink, and I will run a little script which checks the CPU temperature regularly. A third dead 9950X would not be fun.” ®