AMD Threadripper BIOS Reporting Wrong Temps?

[FormatC]

While testing a new waterblock for AMD's Threadripper platform, we came across a bug in AMD's firmware and immediately investigated.

AMD Threadripper BIOS Reporting Wrong Temps? : Read more

 

[cia1413]

Why would you quote a K value? I would bet less than 1% of people know intuitively what that means in relation to C.

 

[Clamyboy74]

Why would you quote a K value? I would bet less than 1% of people know intuitively what that means in relation to C.

Its just math, subtract 273

 

[Patrick_Bateman]

In case anyone's wondering, -1C=272K; 0C=273K; 1C=274K; you get the picture.

 

[JamesSneed]

Why would you quote a K value? I would bet less than 1% of people know intuitively what that means in relation to C.

Personally it bothered me more mixing Kelvin and Celsius temperatures throughout the article. Just pick a direction for god sakes.

 

[TheTechGen]

Off topic but I want that mother board!

 

[derekullo]

Might as well mix the rest in;
Fahrenheit, Rankine, Delisle, Newton, Réaumur, Rømer and Electronvolts.

Newton, at some archaic time, was a measure of temperature lol.
https://en.wikipedia.org/wiki/Temperature#Units

 

[InvalidError]

My guess about where the reading difference comes from? AMD didn't use a separate ground circuit for its temperature sensing element and the "temperature" reading ends up offset by the "voltage droop" on the ground as current goes up. Electrical resistance also goes up with temperature, which could compound the effect.

 

[Gillerer]

Why would you quote a K value? I would bet less than 1% of people know intuitively what that means in relation to C.

Its just math, subtract 273

In delta temperatures, value in K is exactly the same as value in °C: ∆T of +1 K = +1°C.

 

[willwill56]

lol at all of the comments confused by Kelvin.

Specifying change in temperature in Kelvin is perfectly fine. 10 Kelvin colder is, by definition of the Kelvin itself, exactly the same as 10 Celsius colder.

 

[davidchaoth]

"We have already informed AMD about these measurements, and we are awaiting a statement or a new BIOS, ..."

AMD's statement: All our employees' paychecks have been correctly deposited; the auto-deposit function works well as expected : )

 

[wownwow]

"We have already informed AMD about these measurements, and we are awaiting a statement or a new BIOS, ..."

AMD's statement: All our employees' paychecks have been correctly deposited; the auto-deposit function has been working well as expected : )

 

[Paul17041993]

... which is why I use water temperature probes to control the fan speed on the radiators, CPU temp doesn't get used for anything except monitoring.

 

[alextheblue]

... which is why I use water temperature probes to control the fan speed on the radiators, CPU temp doesn't get used for anything except monitoring.

What happens if your temp probe fails or otherwise malfunctions and relays incorrect values? Not entirely different from a coolant temp sensor failing in a vehicle. I think you should run a redundant sensor or a single "sensor unit" with two independent sensors, and have the system max out the fans if the delta between the two sensors is greater than a few degrees. For safety.

On a related note, most users running LC are using a prebuilt closed loop.

 

[FormatC]

I'm using in my rigs NO fan control. With enough cooling surcface, I can run all fans permanently at 800rpm - their max.speed. But nearly unhearable.

My problem with the current situation:
AMDs current calculation of temperatures is stupid. If I play with a few voltages, I get in Idle (and Ryzen Master Toool) -4°C for the package temperature if the water temperature is 20°C. The curve is going up, depending at the power consumprion and then falling down above 240 Watts. What the hell is this? If I play with the Load Line I can get under full load below the Idle temps.

 

[Paul17041993]

... which is why I use water temperature probes to control the fan speed on the radiators, CPU temp doesn't get used for anything except monitoring.

What happens if your temp probe fails or otherwise malfunctions and relays incorrect values? Not entirely different from a coolant temp sensor failing in a vehicle. I think you should run a redundant sensor or a single "sensor unit" with two independent sensors, and have the system max out the fans if the delta between the two sensors is greater than a few degrees. For safety.

On a related note, most users running LC are using a prebuilt closed loop.

I have two sensors and a warning if either of them surpass 55C, and it's more likely of the fan headers or fans themselves to fail than the actual probes. Unless you were to make a super redundant water loop you basically should avoid letting the system run intensive stuff (such as F@H) while you're not nearby to notice alarms etc, though its entirely possible to set up a service to shut down or sleep the system in the case of bad cooling.

 

[blackbit75]

The Volkswagen of CPU's.

 

[derekullo]

lol at all of the comments confused by Kelvin.

Specifying change in temperature in Kelvin is perfectly fine. 10 Kelvin colder is, by definition of the Kelvin itself, exactly the same as 10 Celsius colder.

You can also say 1 degrees Celsius is 11853248076851000000000 electronvolts .

The object of ridicule isn't that his logic was wrong it was that he bounced needlessly between Celsius and Kelvin for no real reason.

 

[FormatC]

It is a global scientific standard to use Kelvin as unit for temperature differences. It is not an idea from me to confuse somebody, but simply usual. It makes a big difference if I write °K or simply K. But I wrote Kelvin and this is right.

School knowledge BTW. I've learned this in the age of 15. Thermodynamics. 😉

 

[zippyzion]

The CPU is so cool it needs to be measured in Kelvin. Hahaha...

I'll see myself out.

 

[damric]

Tctl is not Celsius, Kelvin, or anything really. It is it's own scale, and not based on any physical temperature probes. It's just a calculation based on TDP. Overclocking does tend to screw up the calculation.

 

[JamesSneed]

I'm using in my rigs NO fan control. With enough cooling surcface, I can run all fans permanently at 800rpm - their max.speed. But nearly unhearable.

My problem with the current situation:
AMDs current calculation of temperatures is stupid. If I play with a few voltages, I get in Idle (and Ryzen Master Toool) -4°C for the package temperature if the water temperature is 20°C. The curve is going up, depending at the power consumprion and then falling down above 240 Watts. What the hell is this? If I play with the Load Line I can get under full load below the Idle temps.

Maybe they should be reporting the temps in Kevlin so they don't go negative? Assuming here the negative is causing wacky calculations.

 

[FormatC]

They are reporting Tdie as Package-Temperature. But this is also more as wrong. It is pure calculation, nothing measured. And currently unusable.

 

[wifiburger]

just threadripper bios ? cause my ryzen bios is stupid too ! I mean I choose the Asrock AB350, multiplyer 36 and 39 is completely broken, temps are 20c+ more then the real values, can't change from 1T to 2T , vcore at 1.375 actually is 1.4 etc...

 

[chuck43105]

My ThreadRipper CPU was reading near 0 degrees C when first installed and is now about 15-25 degrees under mid level load. I did get a few individuals indicating that I was lying, what was I smoking, and other comments like this. All programs are indicating about the same, even AMD Ryzen Master, Core Temp, HWMonitor, HWiNFO, and even the BIOS itself are under the norms. I am using BIOS 0503 on the ASUS Zenith Extreme motherboard. Some at ASUS thought it may be because I used the full 3.5 gram tube of Artic Silver 5 thermal paste on the ThreadRipper chip (one big blob on the Z in Ryzen as shown by Gamers Nexus). Also could be the Corsair H115i cooler or a combination of many of these. I love the readings but doubt whether their accurate (but it would be great if they were).