.

I'm not insisting on adding the CPU temperature features in BOINC.

I just say it is possible for some processors to read their core
temperature without the need of motherboard sensors knowledge.

.

.

Still, I'm rather "old school" and I'd rather use an external gauge instead of one based on software, but that's just me.

So I'm "ancient school" (born 1960) and Deep Purple, Rainbow, Pink Floyd, Queen, Dire Straits ... fan :)


.

.

Well, a couple things to mention is that page you link to is only good for Intel Core architecture CPUs and AMD K8/K10 CPUs, which severely limits it to only the newer processors. This project has such a large range of CPUs that programming something like this into BOINC would only benefit newer CPU owners and would be of little use to everyone else.

Further, it states that the sensor is digital and that a value is stored in a register, but it does not show the instruction for retrieving that value from that register, nor are there any instructions that I'm aware of that would do so. I'd have to study up on my newer instruction sets to be certain.

Its great if they did add this feature, but it does little for 95% of the active crunchers who are still using Pentium IIIs, Pentium 4s, Athlon XPs and Athlon 64s.

I think Athlon 64 is K8 -
Core Temp works on my
AMD Athlon(tm) 64 Processor 3500+ [x86 Family 15 Model 95 Stepping 2]
.

You're right about that. That at least increases the installed user base of BOINC systems, but without enough documentation it would be hard to implement into BOINC.

Still, I'm rather "old school" and I'd rather use an external gauge instead of one based on software, but that's just me.

.

Well, a couple things to mention is that page you link to is only good for Intel Core architecture CPUs and AMD K8/K10 CPUs, which severely limits it to only the newer processors. This project has such a large range of CPUs that programming something like this into BOINC would only benefit newer CPU owners and would be of little use to everyone else.

Further, it states that the sensor is digital and that a value is stored in a register, but it does not show the instruction for retrieving that value from that register, nor are there any instructions that I'm aware of that would do so. I'd have to study up on my newer instruction sets to be certain.

Its great if they did add this feature, but it does little for 95% of the active crunchers who are still using Pentium IIIs, Pentium 4s, Athlon XPs and Athlon 64s.

I think Athlon 64 is K8 -
Core Temp works on my
AMD Athlon(tm) 64 Processor 3500+ [x86 Family 15 Model 95 Stepping 2]
.

.

Bil: One of the Everest developers told me
(on Lavalys Discussion Forum > Lavalys EVEREST > Hardware monitoring > What is CPU Diode temperature?)
that (modern) CPUs with internal CPU Diode have instruction which returns "CPU Diode temperature"
even if the motherboard has no sensor chip!
(like CPUID instruction returns the CPU name, etc.)

Look at the end of this page:

http://www.lavalys.com/forum/index.php?showtopic=494

OzzFan: Can Fiery support his/her claim? Most developers don't work at the hardware level. Is Fiery certain that the temperature is being read directly from the CPU diode or is it still accessed through the motherboard without Fiery's knowledge? Can Fiery show documentation on how this is achieved?

All the x86 instructions I've seen do not support software to be able to access the thermal diode directly; only hardware solutions such as thermal monitoring chips which have the correct circuitry to access the proper information, which can then be accessed by the software via the circuitry's supporting instruction set, which can vary with each revision or newly discovered methodologies.

The only instruction I've seen that CPUs support is simply one that tells the software if the CPU supports thermal monitoring, but not a direct reading of exactly what temperature the CPU is at.

Look also this:

http://www.alcpu.com/CoreTemp/howitworks.html


"The temperature readings are very accurate as the data is collected from a Digital Thermal Sensor (or DTS) which is located in each individual processing core, near the hottest part. This sensor is digital, which means it doesn't rely on an external circuit located on the motherboard to report temperature, its value is stored in a special register in the processor so any software can access and read it. This eliminates any inaccuracy that can be caused by external motherboard circuits and sensors and then different types of programs trying to read those sensors.".

Well, a couple things to mention is that page you link to is only good for Intel Core architecture CPUs and AMD K8/K10 CPUs, which severely limits it to only the newer processors. This project has such a large range of CPUs that programming something like this into BOINC would only benefit newer CPU owners and would be of little use to everyone else.

Further, it states that the sensor is digital and that a value is stored in a register, but it does not show the instruction for retrieving that value from that register, nor are there any instructions that I'm aware of that would do so. I'd have to study up on my newer instruction sets to be certain.

Its great if they did add this feature, but it does little for 95% of the active crunchers who are still using Pentium IIIs, Pentium 4s, Athlon XPs and Athlon 64s.

[Edited for spelling and gramar]

.

Bil: One of the Everest developers told me
(on Lavalys Discussion Forum > Lavalys EVEREST > Hardware monitoring > What is CPU Diode temperature?)
that (modern) CPUs with internal CPU Diode have instruction which returns "CPU Diode temperature"
even if the motherboard has no sensor chip!
(like CPUID instruction returns the CPU name, etc.)

Look at the end of this page:

http://www.lavalys.com/forum/index.php?showtopic=494

OzzFan: Can Fiery support his/her claim? Most developers don't work at the hardware level. Is Fiery certain that the temperature is being read directly from the CPU diode or is it still accessed through the motherboard without Fiery's knowledge? Can Fiery show documentation on how this is achieved?

All the x86 instructions I've seen do not support software to be able to access the thermal diode directly; only hardware solutions such as thermal monitoring chips which have the correct circuitry to access the proper information, which can then be accessed by the software via the circuitry's supporting instruction set, which can vary with each revision or newly discovered methodologies.

The only instruction I've seen that CPUs support is simply one that tells the software if the CPU supports thermal monitoring, but not a direct reading of exactly what temperature the CPU is at.

Look also this:

http://www.alcpu.com/CoreTemp/howitworks.html


"The temperature readings are very accurate as the data is collected from a Digital Thermal Sensor (or DTS) which is located in each individual processing core, near the hottest part. This sensor is digital, which means it doesn't rely on an external circuit located on the motherboard to report temperature, its value is stored in a special register in the processor so any software can access and read it. This eliminates any inaccuracy that can be caused by external motherboard circuits and sensors and then different types of programs trying to read those sensors."

.

What type of Dell Optiplex machine is it? mine is an ultra compact micro pc and even the outer case feels like it is on fire sometimes at least the top of it where the hard drive and cd drive are located. They are old machines and slow but they do crunch numbers, I've not had an overheating issue with mine running at 100%. I have had issues when running other programs on it eating away the little memory they have (max mem is 512MB on my machine), causing all sorts of issues MD5 errors,downloading errors, temp machine lockup etc. but after not using but one or two other tasks it corrected the issue.


I didn't know if anyone tried this add on
http://home.twcny.rr.com/setiatease/cpulimit-1.1c.tar.gz
Temperature-based CPU limit
Like CPU limiter, but based on CPU temperature
9:04 PM UTC, December 19 2006

If it does what it claims, this is what you've been looking for

Of course it seems almost too easy, I don't know a thing about it or how to utilize it, but I personally don't have issue with heat

Its a GX150 small form factor. The GX150 has a fan on the side of the cpu heatsink but it just blows the air around the inside of the case. Later models (GX260 and GX270) have a shroud that fits over the top of the cpu heatsink and directs the air outside of the case.

I am getting a couple of 270's today and will be selling the GX150 off as its getting a bit old and isn't that fast (P3 @ 930Mhz), well not when compared with a P4 :-)

Sounds about like the same machine I 've got sitting here at least I only gave %50 for mine without seeing it I was told it was 1 Ghz PC which it is, I just didn't realize it was a small form factor at the time or I might have passed (always seemed like they would have severe heat issues) but I did need a comp at the time my other one had just died so I can't complain to much, but you are right they are very outdated. Glad I have the AMD 2 GHz eMachines and with lots of help (TY AGELESS) it is running. granted its mobo is outdated already too but it works...kinda

What type of Dell Optiplex machine is it? mine is an ultra compact micro pc and even the outer case feels like it is on fire sometimes at least the top of it where the hard drive and cd drive are located. They are old machines and slow but they do crunch numbers, I've not had an overheating issue with mine running at 100%. I have had issues when running other programs on it eating away the little memory they have (max mem is 512MB on my machine), causing all sorts of issues MD5 errors,downloading errors, temp machine lockup etc. but after not using but one or two other tasks it corrected the issue.


I didn't know if anyone tried this add on
http://home.twcny.rr.com/setiatease/cpulimit-1.1c.tar.gz
Temperature-based CPU limit
Like CPU limiter, but based on CPU temperature
9:04 PM UTC, December 19 2006

If it does what it claims, this is what you've been looking for

Of course it seems almost too easy, I don't know a thing about it or how to utilize it, but I personally don't have issue with heat

Its a GX150 small form factor. The GX150 has a fan on the side of the cpu heatsink but it just blows the air around the inside of the case. Later models (GX260 and GX270) have a shroud that fits over the top of the cpu heatsink and directs the air outside of the case.

I am getting a couple of 270's today and will be selling the GX150 off as its getting a bit old and isn't that fast (P3 @ 930Mhz), well not when compared with a P4 :-)

---

BOINC blog

Hi Ozzfan,

Downloaded it (ver 4.33) but it didn't recognize my old Dell GX150, which does have a thermal sensor as the CPU fan speeds up when the thing starts running hot (it doesn't have a case fan, so it doesn't take long to warm up with S@H). Strangely enough it didn't try and guess the temperature, so maybe they removed the "guessing" in 4.33.

MarkJ
What type of Dell Optiplex machine is it? mine is an ultra compact micro pc and even the outer case feels like it is on fire sometimes at least the top of it where the hard drive and cd drive are located. They are old machines and slow but they do crunch numbers, I've not had an overheating issue with mine running at 100%. I have had issues when running other programs on it eating away the little memory they have (max mem is 512MB on my machine), causing all sorts of issues MD5 errors,downloading errors, temp machine lockup etc. but after not using but one or two other tasks it corrected the issue.


I didn't know if anyone tried this add on
http://home.twcny.rr.com/setiatease/cpulimit-1.1c.tar.gz
Temperature-based CPU limit
Like CPU limiter, but based on CPU temperature
9:04 PM UTC, December 19 2006

If it does what it claims, this is what you've been looking for

Of course it seems almost too easy, I don't know a thing about it or how to utilize it, but I personally don't have issue with heat

It's strange why Intel & AMD don't offer such an instruction -
it has to be relatively easy to implement?

As easy as implementing any other instruction, I'd imagine.

Can't it be in the CPUID family?

That wouldn't be the appropriate place to put the instruction, and it isn't there now. See here for more information about the CPUID opcode. Intel has a great page dedicated to explaining the CPUID opcode too.

---

All the x86 instructions I've seen do not support software to be able to access the thermal diode directly;
.........
The only instruction I've seen that CPUs support is simply one that tells the software if the CPU supports thermal monitoring, but not a direct reading of exactly what temperature the CPU is at.

It's strange why Intel & AMD don't offer such an instruction -
it has to be relatively easy to implement?

Can't it be in the CPUID family?

Hi Ozzfan,

Downloaded it (ver 4.33) but it didn't recognize my old Dell GX150, which does have a thermal sensor as the CPU fan speeds up when the thing starts running hot (it doesn't have a case fan, so it doesn't take long to warm up with S@H). Strangely enough it didn't try and guess the temperature, so maybe they removed the "guessing" in 4.33.

The guessing may have been taken out, but that doesn't necessarily accurate. I haven't found a software thermal program that's nearly as accurate as a hardware based solution.

---

One of the Everest developers told me
(on Lavalys Discussion Forum > Lavalys EVEREST > Hardware monitoring > What is CPU Diode temperature?)
that (modern) CPUs with internal CPU Diode have instruction which returns "CPU Diode temperature"
even if the motherboard has no sensor chip!
(like CPUID instruction returns the CPU name, etc.)

Look at the end of this page:

http://www.lavalys.com/forum/index.php?showtopic=494

Can Fiery support his/her claim? Most developers don't work at the hardware level. Is Fiery certain that the temperature is being read directly from the CPU diode or is it still accessed through the motherboard without Fiery's knowledge? Can Fiery show documentation on how this is achieved?

All the x86 instructions I've seen do not support software to be able to access the thermal diode directly; only hardware solutions such as thermal monitoring chips which have the correct circuitry to access the proper information, which can then be accessed by the software via the circuitry's supporting instruction set, which can vary with each revision or newly discovered methodologies.

The only instruction I've seen that CPUs support is simply one that tells the software if the CPU supports thermal monitoring, but not a direct reading of exactly what temperature the CPU is at.

---

Look for a download of Motherboard Monitor 5. It may recognize it better on that old machine. It's an old program, no longer updated. You may also try CoreTemp.

---

Download SpeedFan 3.33 (Freeware) for real time monitoring. Supports almost every PC I've used (Windoze).

According to the SpeedFan documents, if it cannot find the thermal chip on your motherboard, then it'll try to "guess" all the temps that it can, thus making it not very accurate (I've seen systems where the temp is off by a good margin on SpeedFan).

I haven't found a piece of software that can do as good of a job as hardware monitoring can do, or as accurately.

Hi Ozzfan,

Downloaded it (ver 4.33) but it didn't recognize my old Dell GX150, which does have a thermal sensor as the CPU fan speeds up when the thing starts running hot (it doesn't have a case fan, so it doesn't take long to warm up with S@H). Strangely enough it didn't try and guess the temperature, so maybe they removed the "guessing" in 4.33.

---

BOINC blog

As you pointed out there are numerous motherboard combinations. I am suggesting getting the cpu temp, which as far as I know is cpu specific. Please correct me if I'm wrong as I am no expert on these things. Other temps would be too hard to obtain as you say, but the cpu temp shouldn't be that hard to return or require a motherboard-specific version of code.

Sure, maximum CPU temps can be inferred from the manufacturer documents, but how does one go about getting the current CPU temp on a given system?

The only way to do that is to find out if the motherboard supports thermal monitoring. Most motherboards these days do, but so many manufacturers use different thermal monitoring chips, each with their own specific set of instructions to use that there are literally hundreds of different types to program for. The same motherboard manufacturer alone could use two dozen different thermal monitoring chips, depending on which third party they got a deal from or if one is simply newer/more advanced than another.

Thermal monitoring chips don't necessarily use the same instruction sets across all brands/types/models as they are not standardized in the industry.

Getting the temp would require getting the programming access codes for each and every thermal monitoring chip used on every motherboard to ensure maximum compatibility (otherwise I guarantee you there'll be complaints here! LOL) It's just too complex and too much for a project like BOINC to handle.

One of the Everest developers told me
(on Lavalys Discussion Forum > Lavalys EVEREST > Hardware monitoring > What is CPU Diode temperature?)
that (modern) CPUs with internal CPU Diode have instruction which returns "CPU Diode temperature"
even if the motherboard has no sensor chip!
(like CPUID instruction returns the CPU name, etc.)

Look at the end of this page:

http://www.lavalys.com/forum/index.php?showtopic=494

Fair play ... personal preference is all good :)

Personally, I like to support authors of freeware and send diagnositics etc to assist in maturity.

I think that it'll try to identify thermal sensors AFAIK it will not simply guess at temperatures for non existant sensors. Maybe semantics.

Yes, the documentation states that it will try to detect the chip first before guessing. But still, there is guessing going on.

Some MoBo manufacturers haven't implemented all thermal sensors hence on initial run of SpeedFan sensors indicating excessive Temp should be ignored. True SpeedFan does not cover all MoBo/CPU combinations thus not perfect by any means but I'm running it on 3 PC's now and have no issue. I also have separate thermistors pasted to CPU's with digital readout and SpeedFan certainly seems accurate enough.

So yes you're correct it's not perfect by why discount it so readily?

I only dismiss it because, for me, if it isn't going to be accurate, then why use it? This is the reason why I generally don't use software thermal programs but prefer to use a hardware based solution. Software can glitch or have bugs that can cause inaccurate readings. Hardware solutions don't usually suffer from these issues.

It's just a personal preference.

---

According to the SpeedFan documents, if it cannot find the thermal chip on your motherboard, then it'll try to "guess" all the temps that it can, thus making it not very accurate (I've seen systems where the temp is off by a good margin on SpeedFan).

I think that it'll try to identify thermal sensors AFAIK it will not simply guess at temperatures for non existant sensors. Maybe semantics.

Some MoBo manufacturers haven't implemented all thermal sensors hence on initial run of SpeedFan sensors indicating excessive Temp should be ignored. True SpeedFan does not cover all MoBo/CPU combinations thus not perfect by any means but I'm running it on 3 PC's now and have no issue. I also have separate thermistors pasted to CPU's with digital readout and SpeedFan certainly seems accurate enough.

So yes you're correct it's not perfect by why discount it so readily?

Download SpeedFan 3.33 (Freeware) for real time monitoring. Supports almost every PC I've used (Windoze).

According to the SpeedFan documents, if it cannot find the thermal chip on your motherboard, then it'll try to "guess" all the temps that it can, thus making it not very accurate (I've seen systems where the temp is off by a good margin on SpeedFan).

I haven't found a piece of software that can do as good of a job as hardware monitoring can do, or as accurately.

---