I used 5.1.3 and 4.6.4 on the custom development board. Under the same test environment at room temperature, the PCIE*2 write SATA disk rate is 240M/s, but at low temperature, the 5.1.3 rate will be reduced to 179M/s or even lower, How should I solve this problem?
- Use a different (industrial) SSD type designed for low temperatures.
- Use a resistive heating element to raise the internal temperature
All the hardware is the same
I have not changed the hardware at present, because the two versions, and at low temperatures, 4.6.4 speed is normal, but 5.1.3 speed is too low.
ok. Then open a root shell and do
lspci
Note the bus/slot number of your SATA device (example 3:4.5)
Do
lcspci -vv -s 3:4.5 (or whatever the slot number is in your case)
Look for: (these lines are from an Intel X540, yours may differ)
LnkCap: Port #9, Speed 5GT/s, Width x8, ASPM L0s L1, Exit Latency L0s unlimited, L1 <32us ClockPM- Surprise- LLActRep- BwNot- ASPMOptComp+
and
LnkSta: Speed 5GT/s, Width x8 TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-
In this example the PCIE device is capable of 5GT/s and x8, and the actual Link state is also 5GT/s and x8. You can check if these numbers are the same for both Jetpack versions and temperatur ranges (room/cold).
I will add that the way you measure speed might be showing an effect other than the PCIe itself. if the GT/s
is the same at both low and high temperatures, then it means the link speed has not changed. What could change is perhaps errors, which “sudo lspci -vvv
” (an extra “-v
” and sudo
) would show near the end if AER is available. Perhaps this isn’t even due to bus errors or bus speed, but something else entirely.
As mentioned by @fchkjwisq, this could also be the device being tested. The lspci
speeds are rather definitive (see if the actual link changes between low and high temperatures).
Perhaps the drive itself is being given a different argument for the driver in 5.1.3, or the CPU itself is throttling back in some way. There are a number of ways this could slow down without it being the PCIe PHY or bus. You might need something which better breaks down the speed issue component before anyone can work on it.
I will suggest that you use “sudo lspci -vvv
” (but name the specific device, that’s a lot of output if running lspci -vvv
on every device) both before the issue shows up on 5.1.3 and during the time it shows up.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.