Mikewarrior2
Diamond Member
Contrary to popular belief, very few VIA based mb solutions actually measure CPU diode temperature. Internal diode inaccuracies being ignored, most via mb's use a under socket(or near slot) thermistor to measure CPU temp. While most of us have no idea that this is a problem(myself included until recently), it yields very inaccurate temperature readings. ON top of this, the readings from the thermistor vary from mb to mb.
First off, i am not saying that VIA mb's are bad. On the contrary, i believe they are great mb's. But the via 596 and 696 southbridges have inherent difficulties reading cpu diode temps. The one true exception is the asus P3V4x and CuV4x. Both of these mb's rely on the cpu/temperature diagnosis chip that Asus includes on almost all of their motherboards. But most of the rest of via based mb solutions use a thermistor to measure cpu temp.
Of the SLot via boards I have noticed, the thermistor sits within a half inch of the mb, and doesn't even get close to either the core of the cpu or the heatsink. They often sit roughly 1-1.5 inches below where the cpu core would be. A very easy way to test this is to use RC5 or Seti or Prime95(or any other cpu stresser). Set up mbm for 1 sec temp readings. start watching the cpu temps. start your cpu stresser program. Within a few seconds, the temperature should change dramatically(in most cases, 10 to 15C). With a thermistor based solution, the temperature change is not only very slow, it never rises to the true temperature of the cpu.
Onto the socket mb's where the thermistor is inside of the socket. This, while it seems like a good solution, is also very inaccurate. With Flip-Chip CPUs(p3s, c2s, durons/t-birds) the hottest part of the cpu is on the part of it that contacts the heatsink. Among the coolest parts are on the backside of the chip. And even if you bend up the thermistor to contact hte back of the cpu, it is still wildly inaccurate. You aren't measuring core temp at all. YOu may also be contacting a resistor or a OEM label on the backside, which further ruins cpu temp readings. In my personal experimentation, comparing my 650E(975/1.75V) on a soyo 7vca(with a thermistor underneath the cpu) and a Cusl2(which uses Asus Temp chip and reads the p3s internal diode), the temperature differences can range from 10 to 14C difference, with the CUSL2 being higher. on my setup, my cpu should put out around 33Watts of heat. Coupled with a Pal6035, which has a thermal resistance of around .4C/W, my temps should be at least 13.2 C above ambient case temperature. That would be under perfect cpu to heatsink contact, perfect everything. On the 7VCA, my cpu ran at load 5C over ambient. On my cusl2, i run roughly 15C over ambient. Based on thermodynamics and some math, it is impossible for my cpu to be running only 5C over ambient.
This is an inherent VIA chipset mb problem. The biggest problem occurs because most of us are led to believe that the thermistor based numbers are accurate CPU core temps(again, i was a believer in the thermistor system until recently). Most people with duron/t-bird motherboards are getting incorrect cpu temp readings from their motherboards. This potentially leads to problems, for example, when you believe your CPU is at 47C when it could actually be in the 50's. The same occurs with p3 mbs based on the via solution.
The only solution to this is to hope for a bios update to via based mbs to fix this. This seems unlikely, since it is likely a hardware problem(and the fact that they are hardwired to read "CPU" temps from the thermistors. Hope this helps everybody out with the heads up that these mbs don't read true cpu core temp.
Mike
First off, i am not saying that VIA mb's are bad. On the contrary, i believe they are great mb's. But the via 596 and 696 southbridges have inherent difficulties reading cpu diode temps. The one true exception is the asus P3V4x and CuV4x. Both of these mb's rely on the cpu/temperature diagnosis chip that Asus includes on almost all of their motherboards. But most of the rest of via based mb solutions use a thermistor to measure cpu temp.
Of the SLot via boards I have noticed, the thermistor sits within a half inch of the mb, and doesn't even get close to either the core of the cpu or the heatsink. They often sit roughly 1-1.5 inches below where the cpu core would be. A very easy way to test this is to use RC5 or Seti or Prime95(or any other cpu stresser). Set up mbm for 1 sec temp readings. start watching the cpu temps. start your cpu stresser program. Within a few seconds, the temperature should change dramatically(in most cases, 10 to 15C). With a thermistor based solution, the temperature change is not only very slow, it never rises to the true temperature of the cpu.
Onto the socket mb's where the thermistor is inside of the socket. This, while it seems like a good solution, is also very inaccurate. With Flip-Chip CPUs(p3s, c2s, durons/t-birds) the hottest part of the cpu is on the part of it that contacts the heatsink. Among the coolest parts are on the backside of the chip. And even if you bend up the thermistor to contact hte back of the cpu, it is still wildly inaccurate. You aren't measuring core temp at all. YOu may also be contacting a resistor or a OEM label on the backside, which further ruins cpu temp readings. In my personal experimentation, comparing my 650E(975/1.75V) on a soyo 7vca(with a thermistor underneath the cpu) and a Cusl2(which uses Asus Temp chip and reads the p3s internal diode), the temperature differences can range from 10 to 14C difference, with the CUSL2 being higher. on my setup, my cpu should put out around 33Watts of heat. Coupled with a Pal6035, which has a thermal resistance of around .4C/W, my temps should be at least 13.2 C above ambient case temperature. That would be under perfect cpu to heatsink contact, perfect everything. On the 7VCA, my cpu ran at load 5C over ambient. On my cusl2, i run roughly 15C over ambient. Based on thermodynamics and some math, it is impossible for my cpu to be running only 5C over ambient.
This is an inherent VIA chipset mb problem. The biggest problem occurs because most of us are led to believe that the thermistor based numbers are accurate CPU core temps(again, i was a believer in the thermistor system until recently). Most people with duron/t-bird motherboards are getting incorrect cpu temp readings from their motherboards. This potentially leads to problems, for example, when you believe your CPU is at 47C when it could actually be in the 50's. The same occurs with p3 mbs based on the via solution.
The only solution to this is to hope for a bios update to via based mbs to fix this. This seems unlikely, since it is likely a hardware problem(and the fact that they are hardwired to read "CPU" temps from the thermistors. Hope this helps everybody out with the heads up that these mbs don't read true cpu core temp.
Mike
