This is a fine tool but given the wide swath of hardware out there it's not a sure thing it will push all machines as hard as possible. Areas that I see come up short many times are:
1. Not all cores may engage or go to max freq.
2. Temperature readings may be unreadable or off a bit.
This item is just the nature of the PC beast. There are no standards for the temperature reporting so this is not so much a failing of G4 but of the PC designs.
3. In spite of all that, it's unlikely that a PC that survived torture tests like G4 will suffer a stress or heat failure under real world use. Unless you are using them for Bitcoin mining then the test should be Bitcoin mining and watching the machine for trouble over the day of testing.
4. There is another issue worth noting that G4 will never address. Windows. Specifically Windows 10 automatic driver updates. My advice for those that use a version of Windows that automatically updates drivers (not to be confused with Windows Updates) is to disable this so once your PC is working nicely it won't be sabotaged by a driver update.
Then again... Linux is a better choice for long term service and not having to check on the Windows PC often.
My question is about how well the thermal limitations of a machine is tested during the Geekbench 4 benchmark. In other words, how long does the CPU intensive number crunching part of the benchmark run for?
I need to buy some new hardware for my engineering business and need to know if thermals are sufficiently tested so that I can know to what extent I should rely on the Geekbench results. My machines typically run calcs at full capacity for 24 hours or more.