When HWInfo64 is going to have accurate reading for ryzen 3000

Hello. I don't know if this question has been already answered, but I just want to know when HWInfo64 is going to have accurate reading for Ryzen 3000? Atm, Ryzen Master seems to be the only temp reading that has accurate reading for Ryzen 3000 from a lot of people said.
 

Martin

HWiNFO Author
Staff member
Well, the question is: Who knows what is accurate ? AMD's own tool that can report values the way its own manufacturing company wishes to ?
And why are there no other tools reporting the same values ? Because the culprit how to do this is very very secret. Why ?
AMD promised us (and other developers) to provide the information required to implement similar monitoring capabilities long ago. But we choose not to rely on their own Ryzen Master SDK for several reasons that they understood.
But since then they are quiet, they didn't deliver anything. We're getting empty promises only. And not just those few monitoring questions, there are other things too...
Why ?
 
Well, the question is: Who knows what is accurate ? AMD's own tool that can report values the way its own manufacturing company wishes to ?
And why are there no other tools reporting the same values ? Because the culprit how to do this is very very secret. Why ?
AMD promised us (and other developers) to provide the information required to implement similar monitoring capabilities long ago. But we choose not to rely on their own Ryzen Master SDK for several reasons that they understood.
But since then they are quiet, they didn't deliver anything. We're getting empty promises only. And not just those few monitoring questions, there are other things too...
Why ?
I don't know but a lot of people recommend me to use Ryzen Master because it is accurate reading temp for Ryzen 3000. They said HWInfo is close but not accurate as Ryzen Master. I don't like Ryzen Master because it doesn't have a record for minimum and maximum temp like HWInfo and other temp reading software. Might I know what is the reason you can't reply on Ryzen Master SDK?

Also, which sensor is for my cpu (ryzen 7 3700x) temp reading? CPU(TCTL/TDIE) or CPU CCD1(TDIE)? Picture taken after I ran Cinebench R20
Screenshot_10.jpg
 

Martin

HWiNFO Author
Staff member
I believe those people recommended Ryzen Master because AMD did so some time ago.
Integrating their SDK would be problematic as it would require shipping additional AMD DLLs and installing another AMD kernel driver. This adds complexity and can cause undesirable effects. There are also other aspects that I cannot disclose.
We don't know for sure what exactly RM reports as temperature, but most probably it's an average value of all core sensors across some (unknown) interval. Averaging is done in order to flatten potential spikes of instantenous high/low values. A similar technique is done in reporting of core voltages.
CPU (Tctl/Tdie) in HWiNFO should be the immediate hottest temperature in the entire CPU package, while the CCD value covers only the CPU cores of a given Core Complex Die (CCD). Based on our measurements, the CCD value should be closer to AMD Ryzen Master reported temperature.
 

johnwang

New Member
AMD changed how Ryzen Master reported temperatures back in September.

"For temperatures:

• OLD BEHAVIOR: Report the highest temp (“instantaneous temperature”) of any sensor in the whole CPU, no matter how isolated or brief. This approach overstated the prevalence and significance of fleeting peaks, which artificially inflated temperature reports by up to 30°C above the true silicon conditions.

• NEW BEHAVIOR: Report a short-duration rolling average of all temp sensors in the chip. This model will give you a holistic view of what the processor is doing across cores, cache, bus interfaces, etc. It’s the closest any software monitoring tool can come to portraying the true thermal conditions of the chip, and it’s the same model our own Precision Boost 2 algorithm uses to make temperature-related boost decisions."

So basically, they're going to differ from every HW tool out there since they're not reporting instantaneous temps.
 
I believe those people recommended Ryzen Master because AMD did so some time ago.
Integrating their SDK would be problematic as it would require shipping additional AMD DLLs and installing another AMD kernel driver. This adds complexity and can cause undesirable effects. There are also other aspects that I cannot disclose.
We don't know for sure what exactly RM reports as temperature, but most probably it's an average value of all core sensors across some (unknown) interval. Averaging is done in order to flatten potential spikes of instantenous high/low values. A similar technique is done in reporting of core voltages.
CPU (Tctl/Tdie) in HWiNFO should be the immediate hottest temperature in the entire CPU package, while the CCD value covers only the CPU cores of a given Core Complex Die (CCD). Based on our measurements, the CCD value should be closer to AMD Ryzen Master reported temperature.
So which one is the actual temp for R7 3700x? Cpu(tctl/tdie) or cpu ccd(tdie)?
 
AMD changed how Ryzen Master reported temperatures back in September.

"For temperatures:

• OLD BEHAVIOR: Report the highest temp (“instantaneous temperature”) of any sensor in the whole CPU, no matter how isolated or brief. This approach overstated the prevalence and significance of fleeting peaks, which artificially inflated temperature reports by up to 30°C above the true silicon conditions.

• NEW BEHAVIOR: Report a short-duration rolling average of all temp sensors in the chip. This model will give you a holistic view of what the processor is doing across cores, cache, bus interfaces, etc. It’s the closest any software monitoring tool can come to portraying the true thermal conditions of the chip, and it’s the same model our own Precision Boost 2 algorithm uses to make temperature-related boost decisions."

So basically, they're going to differ from every HW tool out there since they're not reporting instantaneous temps.
I don't want to ditch Hwinfo for Ryzen Master because Ryzen Master has no temp logging like other hw monitoring software like what is my max temp after 5 hours playing battlefield 5 if I use Ryzen Master or 30 minutes handbrake encoding a video.
 

Martin

HWiNFO Author
Staff member
That depends on what do you understand under "actual temp". Modern systems change their operating parameters several (hundreds) of times per second.
If you want to measure the actual state software can sample a discrete value only at a certain time point (i.e. depending on the polling frequency set) and report it. This is the Tctl/Tdie or the Old behavior as described by AMD. It can sample only certain values as it's impossible to performs hundreds (or thousands) of measurement per second without affecting the system. So in this mode you get true actual / instant / discrete values at the edge of the polling interval, but not all. It might catch a spike in temperature or clock, or measure a point in between.
Another approach is the averaging method - the hardware calculates an average value across a certain interval, because unlike software it can do these measurements in a more efficient and faster way. This value is then the average/effective value. It might provide better view of what approximately happened during a certain interval, but it won't catch short spikes as the resulting value is an average.
 

johnwang

New Member
I don't want to ditch Hwinfo for Ryzen Master because Ryzen Master has no temp logging like other hw monitoring software like what is my max temp after 5 hours playing battlefield 5 if I use Ryzen Master or 30 minutes handbrake encoding a video.
I don't see why you would need to ditch hwinfo, especially since you're interested in max temps under load. I believe AMD changed ryzen master to report average temperatures due to complaints regarding ryzen 3000 instantaneous idle temperatures being higher than other processors. As I understand it, due to the smaller die size and boost algorithm, the heat load is concentrated in a smaller area and temps can vary as much as 10C when the system isn't doing much other than processing background services. When you're saturating the heat load by running CPU intensive applications, the temperatures reported by the various CPU sensors will wind up converging anyway as they all get toasty.
 
Last edited:
That depends on what do you understand under "actual temp". Modern systems change their operating parameters several (hundreds) of times per second.
If you want to measure the actual state software can sample a discrete value only at a certain time point (i.e. depending on the polling frequency set) and report it. This is the Tctl/Tdie or the Old behavior as described by AMD. It can sample only certain values as it's impossible to performs hundreds (or thousands) of measurement per second without affecting the system. So in this mode you get true actual / instant / discrete values at the edge of the polling interval, but not all. It might catch a spike in temperature or clock, or measure a point in between.
Another approach is the averaging method - the hardware calculates an average value across a certain interval, because unlike software it can do these measurements in a more efficient and faster way. This value is then the average/effective value. It might provide better view of what approximately happened during a certain interval, but it won't catch short spikes as the resulting value is an average.
So which one should I pay attention to when it comes to temp reading? Cpu(tctl/tdie) or cpu ccd(tdie)?
I don't see why you would need to ditch hwinfo, especially since you're interested in max temps under load. I believe AMD changed ryzen master to report average temperatures due to complaints regarding ryzen 3000 instantaneous idle temperatures being higher than other processors. As I understand it, due to the smaller die size and boost algorithm, the heat load is concentrated in a smaller area and temps can vary as much as 10C when the system isn't doing much other than processing background services. When you're saturating the heat load by running CPU intensive applications, the temperatures reported by the various CPU sensors will wind up converging anyway as they all get toasty.
But it has no logging for maximum temp like I want to see what my temp looks like under 5 hours of BF5, you know?
 

johnwang

New Member
Run BF5 for 5 hours and report back what your max temps are in hwinfo. The tdie vs ccd should be just about the same. Incidentally, that temperature should also correspond to ryzen master as well since ryzen master's average will reach a steady state under load
 
Run BF5 for 5 hours and report back what your max temps are in hwinfo. The tdie vs ccd should be just about the same. Incidentally, that temperature should also correspond to ryzen master as well since ryzen master's average will reach a steady state under load
Okay so which one is the actual temp for my 3700x after playing bf5? CCD and TCTL/TDIE no exact the temp. I just want to know which temp sensor I should look for the actual 3700x temp. TCTL/TDIE or CCD1. Capture 1111.JPG
 

Zach

Member
Okay so which one is the actual temp for my 3700x after playing bf5? CCD and TCTL/TDIE no exact the temp. I just want to know which temp sensor I should look for the actual 3700x temp. TCTL/TDIE or CCD1.
Both... are as accurate as possible by what Martin has already stated. The Ryzen CPUs have over 60 temp sensors across the entire package.
The Tctl/Tdie shows constantly the highest temp, I believe by switching instantly to the hottest one.
The CCD Tdie is probably a sensor to one side of the CCD and the reading is fixed to that one. Or may be is the average on all sensors.

If the CPU is on light load the two readings would have some delta between them, because the light load cant heat up the entire CCD equally so the CCD Tdie would read a lower Max and report a lower Avg also. The Tctl/Tdie for the hottest spot would still report higher, because thats whats for.
If the CPU is on heavy load for a certain period of time then the 2 readings would be much more close to each other if not the same for the Max/Avg, because the amount of heat is enough to equally heat the CCD across sides.

Look at mine after several hours of just browsing the internet, watching videos, and idling.
There is a difference for the Low/Max/Avg between them because the CPU has done light work.

HWiNFO64 v6.12-3930 Sensor Status [4 values hidden] 04-Nov-19 00_05_50.png
 
Last edited:
Top