Post by: pablot on March 31, 2006, 05:13:00 AM
I think I have located the 1.6 temperature sensor (inside the xcalibur) and would like it if some one with a 1.6 xbox could try it out. I have tested it but the temperature read out is a bit "jumpy", it jumps with in 2C when idle. I just tested unplugging the fan and the temp whent up from 40C to 60C, I then plugged it in again and it whent back down again. Would like to see some more people testing it though.
Replace "void CFanController::GetCPUTempInternal()" (xbmc/xbmc/utils/FanController.cpp) with this one:
CODE
void CFanController::GetCPUTempInternal()
{
unsigned short cpu, cpudec;
//HalReadSMBusValue(PIC_ADDRESS, CPU_TEMP, 0, (LPBYTE)&cpuTemp);
if (!bIs16Box)
{
_outp(0xc004, (0x4c << 1) | 0x01);
_outp(0xc008, 0x01);
_outpw(0xc000, _inpw(0xc000));
_outp(0xc002, (0) ? 0x0b : 0x0a);
while ((_inp(0xc000) & 8));
cpu = _inpw(0xc006);
_outp(0xc004, (0x4c << 1) | 0x01);
_outp(0xc008, 0x10);
_outpw(0xc000, _inpw(0xc000));
_outp(0xc002, (0) ? 0x0b : 0x0a);
while ((_inp(0xc000) & 8));
cpudec = _inpw(0xc006);
cpuTemp = (float)cpu + (float)cpudec / 256.0f;
}
else
{
_outp(0xc004, (0x70 << 1) | 0x01); // address
_outp(0xc008, 0xC1); // command
_outpw(0xc000, _inpw(0xc000)); // clear possible old errors
_outp(0xc002, 0x0d); // start block transfer
while ((_inp(0xc000) & 8)); // wait for response
_inp(0xc004); // read out the data reg (no. bytes in block, will be 4)
cpudec = _inp(0xc009); // first byte
cpu = _inp(0xc009); // second byte
_inp(0xc009);
_inp(0xc009); // read out the two last bytes, dont' think its neccesary
cpuTemp = (float)cpu + (float)cpudec / 256.0f;
cpuTemp *= 0.8f; // adjust the "too high" cpu temp
}
}
I would be happy if someone could try it and report back.
regards
Pablot
Post by: JayDee on March 31, 2006, 08:26:00 AM
Post by: charlydoes on March 31, 2006, 08:52:00 AM
It was added because someone thought the temperature on his 1.6 box was too high, took a wild guess at what it should be and multiplied by 0.8 to get it.
Post by: pablot on March 31, 2006, 09:24:00 AM
QUOTE(JayDee @ Mar 31 2006, 05:26 PM) 
Since my home pc went down earlier today i havent been able to radmin in and pass it on to any devs. Will make sure they get this info asap.
okay, good.
QUOTE(charlydoes @ Mar 31 2006, 05:52 PM)
First thing you should do is get rid of cpuTemp *= 0.8f;
It was added because someone thought the temperature on his 1.6 box was too high, took a wild guess at what it should be and multiplied by 0.8 to get it.
yeah, I know. But that wasn't the important thing. I still need to see this tested, the exact temperature isn't that important. By looking at the measurements I have I'd say that its constantly about 5-8C too high or something
Post by: pablot on April 01, 2006, 12:47:00 AM
Every thing is done over the SMBus/i2c:
1) GPU temp from reg 0x0A form the PIC
2) Read out the temp from the SMBus controller data reg
3) store it as gpuTemp
4) Request the CPU Temp high byte from the temperature sensor (@0x4C)
5) Read fails, but it isn't checked for so it reads out the SMBus controller data reg thinking its the CPU temp but it is the old GPU temp value.
6) Request the CPU Temp low byte from the temperature sensor (@0x4C)
7) same thing as in step 5 happens
8) calculate and store cpuTemp = CPUhigh + CPUlow/256, which in reality is cpuTemp = GPU + GPU/256
example, my temps are now at:
CPU: 49.79 GPU: 49.60
Remove the 0.8 factor that 1.6 temps get adjusted with:
GPU: 62
cpuTemp = GPU + GPU/256 = 62 + 62/256 = 62.2421875
add the 0.8 factor again
cpuTemp = 62.2421875*0.8 = 49.79375 ~ 49.79
and were back.. so, thats why it looks like your temp sensor is working in a 1.6!
Anyway, I hade pike test the sensor too and it works for him, but the temperature is jumpy for him too. So maybe calculate the displayed value by taking an average of 10 values or something? What do you think?
Post by: pablot on April 01, 2006, 10:20:00 AM
http://www.xboxmedia...=ST;f=5;t=19884