Question FX 8320 power consumption sensor ?

[NigateloJones]

hello guys recently i overclocked my FX 8320 to 4.7 ghz and by doing that i had to disable every svm, power saving and any other options that stops me controlling everything manually.
So after doing this, the cpu wattage consumption is not displayed anymore by hw info, aida hwmonitor and apps like these.
So my final question is what did i do in order to prevent the cpu from displaying its wattage consumption in real time?

 

[CountMike]

hello guys recently i overclocked my fx 8320 to 4.7 ghz and by doing that i had to disable every svm, power saving and any other options that stops me to control everything manually.
So after doing this, the cpu wattage consumption is not displayed anymore by hw info, aida hwmonitor and apps like these.
So my final question is what did i do in order to prevent the cpu from displaying its wattage consumption in real time?

Try with this
https://www.techspot.com/downloads/4645-amd-overdrive.html

 

[NigateloJones]

Try with this
https://www.techspot.com/downloads/4645-amd-overdrive.html

thanks but could u tell me how is this supposed to help me i mean i used the app before but it didn't do much

 

[CountMike]

thanks but could u tell me how is this supposed to help me i mean i used the app before but it didn't do much

Use it just for reading values, it's an old APP made by AMD just for FX series of their CPUs so most accurate,

 

[NigateloJones]

Use it just for reading values, it's an old APP made by AMD just for FX series of their CPUs so most accurate,

okay will do thanks and one more thong should i leave apm and hpc on regarding perfomance and keeping the hogh clock all the time?

 

[drea.drechsler]

okay will do thanks and one more thong should i leave apm and hpc on regarding perfomance and keeping the hogh clock all the time?

What motherboard are you using?

On my Gigabyte GA990FXA motherboard the only change I make is to enable HPC to allow the CPU (FX6300) to stay at it's set overclock (4.6Ghz) when working hard (Cinebench, Prime95) or even just gaming. Normally (HPC disabled) an overclocked FX CPU drops clocks in heavy useage scenarios as needed to stay within it's average TDP rating. That's a good thing if your motherboard VRM can't deliver adequate power for the CPU without overheating so the decision to disable it depends on what your motherboard VRM is capable of in addition to CPU cooling.

I left APM enabled and as well Cool N Quiet. When they are enabled the CPU will drop clocks and voltage when the CPU isn't working hard and that's desireable IME. If for some reason you want to see the high clocks even when idle then disable them. But the constantly high voltage for overclocking stability only serves to degrade the CPU faster, something to consider if keeping these old beasts alive is also important to you.

SVM is CPU hardware virtualization. If it's enabled HWInfo probably won't be able to read the CPU telemetry which it needs to infer it's power consumption. I'd say that's your problem. You should be able to leave it disabled unless it's needed to run virtual machine environments for your work.

I never used AMD's Overdrive utility and I thought it was deprecated, i.e., no longer supported, a long time ago. I'd be surprised if it works in Windows10 or 11, given how finicky they are about low-level hardware access, unless AMD's made some changes to how it works. I do my overclocking from BIOS.

 

[Karadjgne]

For FX cpus, AMD Overdrive or CoreTemp is the Only accurate way of measuring temps, a vital part of Any OC.

Tcase for all the FX is 63°C, but since there are No thermal strips for temp readings on the cores, there's no way to read an actual temp. Overdrive and CoreTemp use a complex algorithm made up of voltages, loads, socket temps and other parameters to come up with a Thermal Margin.

What's important about TM's is not the exact number, it's not an actual temp, it's a reference to an amount of headroom compared to Tcase, represented as 0. So a TM of 40 is tons of thermal headroom, 20's is about normal, teens is not much room, single digits and you are running hot, and 0 or negatives means you are at no thermal headroom at all or are exceeding Tcase (like a 100° Intel running at 105°C etc).

Pretty much Any post you see of ppl claiming a specific temp on a FX cpu are completely wrong, there's no physical way to read an accurate temp via software regarding the cpu as is. Socket temp is Not cpu temp/core temp, which is what many software titles use or you'll see false readings of ppl at 8°C idle, gaming at 18°C etc.

 

[drea.drechsler]

For FX cpus, AMD Overdrive or CoreTemp is the Only accurate way of measuring temps, a vital part of Any OC.
....

I still wouldn't call it accurate by any means. IMO, the Overdrive temp margin report is only a way to obfuscate the inherently inaccurate temperature telemetry readout from an FX processor in a way that still leaves it useful.

Can't say much for CoreTemp, I've never liked it when I used it. I prefer HWInfo since at least he gives us a good write-up on AMD's (and Intel's) temperature reporting innacuracies. But it also shows all the temperature monitoring sensor feeds, that's important since many motherboards include a socket temp sensor which helps with a relatively accurate Tcase or Tsocket.

 

[Karadjgne]

OD and CoreTemp have the same thermal margin code, it was co-developed by the authors. And it is the only accurate way to measure but that doesn't necessarily mean the results always make sense.

inherently inaccurate temperature telemetry readout from an FX

That's just it, there isn't any temp telemetry from an FX. Any numbers given to those addresses aren't temps, bios just 'thinks' those addresses should be a temp telemetry because those particular addresses should correspond to a thermal strip inside the cpu. But they do not.

With Intel, software reports the temps of the cores, puts it into a °C format, and you have to do the mental gymnastics to figure out what that result means. It's taken a couple of generations to 're-educate' many users who were indoctrinated to the 70°C + is bad theory, and now have to accept that their 13900k at 90°is just fine.

Thermal margin readings aren't any different, they just don't appear as a set °C figure, but a number representing a combined amount of thermal headroom left. Once you wrap your head around 40's is cool, 20's is ok, 1-9 is getting too warm for comfort, it's no different than looking at an intel at 32°, 60°, 91°-99°.

I like HWInfo, I use it personally on my Ryzen, and as you say, offers a lot for an FX user that is usable and relative, but it does not offer thermal margins, which will be very important to the Op running a FX 8320 at 4.7GHz, which could easily be exceeding Tcase already, and why he has to basically strip the instructions and other settings just to get a stable boot.

CoreTemp works, just does exactly what it says it's for and not much else, it's essentially a temp reporter and that's basically it. No frills. It's only bonus over other such software is the thermal margin ability, which has to be enabled on the last tab, otherwise it reads °C. It's nowhere near as comprehensive as HWInfo, but for many ppl that's fine, they like simple and have no desire to confuse themselves with a ton of extraneous info. One benefit is its ability to close to background, yet still put the temp reading in the Start Bar, next to the resident apps, so temps are always visible as long as the bar is visible. HWInfo only shows temps when you bother to open the app and look at it, which isn't always useful.

Software and Hardware both have limitations, the FX is severely limited in thermal accuracy per core, software is limited by that inability, leaving OD and CT as the only viable software to use, and let's face it, it's an AMD FX cpu, so Intel standards as to reading temps simply do not apply.

 

[drea.drechsler]

....
which will be very important to the Op running a FX 8320 at 4.7GHz, which could easily be exceeding Tcase already,
...

But without delving into meaning of what you may consider 'telemetry', the BIOS and OS utilities have to be getting the information from the CPU that it uses to infer something that's considered "temperature". It's not accurate, never was saying it is and especially at low CPU utilization temperature reported can be wildly hilarious. At high utilization it can be relevant, useful for setting fans, but still inaccurate.

All the thermal margin is is mathematical gymnastics based on the same inaccurate data that's being gleaned from whatever the CPU is providing. It's helpful, though, but so is the temp reading I get from HWINfo.

But then...what means exceeding "Tcase"? what's the specification for that? is it something AMD has established? AMD has established a specification Tjmax and publishes it.

But then... OP never asked anything about temperature he's asking about power reporting, so this is really off-topic.

 

[Karadjgne]

Core temp max (Tjmax) is 61.1°C
Surface of the silicon at the IHS (Tcase) is 63°C.
Tcase temps were measured by AMD engineers using modified engineering samples with thermocouplings at the IHS. The Tjmax is not a measured temp, but an 'educated guess' by those same engineers since they couldn't add thermocouplings to the inside of the cores and still have functioning cpus to give readings on. There were no functional FX made with interior readers in the cores or nodes as that would have changed the properties of the cpus.

It's kinda not off topic, sorta, because power reporting, consumption, use goes hand in hand with temps. OP can easily get power numbers from HWInfo, but those numbers will be useless if he can't then run the cpu because of those temps. He'll have to get better cooling or down clock to workable temps, which then changes the power numbers.

Dear Stephen,

Your service request : SR #{ticketno:[8200488157]} has been reviewed and updated.

Response and Service Request History:

I understand that you are looking for clarity about the 70c maximum operating temperature of the FX-6100 and which specific temperature it references at that point. If this is incorrect, please let me know as the information provided may change.

The maximum operating temperatures for any AMD CPU is measured as TCase, as we find that the actual temperature measured by a thermistor inside the processor is a much more accurate measurement of the processor's temperature. While TJunction is helpful in determining the temperature of the actual pins that are transmitting the data, the fact that every pin doesn't fire at exactly the same time, the results can be somewhat skewed. While we don't have the tech docs up for the FX series yet in terms of max temperatures, you can see that we post TCase maximum temperatures on our Athlon/Sempron/Phenom processor Tech Docs at http://support.amd.com/us/Processor_TechDocs/43375.pdf , instead of using TJunction/TJ/TCore/etc.

As for Tctl, there is a technical definition at http://support.amd.com/us/Processor_TechDocs/41256.pdf on page 73 that (tries to) explain it. As a very crude Cole's Notes version, it doesn't measure temperature as much as it is a sliding scale that refers to the processor's current temperature as it relates to the temperature at which the cooling fan has to get to 100% to hit the maximum case temperature (TCaseMax). Confusing I know, but it allows the system to see how close it is to hitting 100% and subsequently it has to start slowing things down to get the temperature down. It is usually close to Tcase Max, but is more a point of reference for how close it is getting to Tcase Max, or if it is past that point and by how much. Things run normally at Tctl < TCaseMax - 0.125, and when it hits TCaseMax or higher, then things start getting shut down/slowed down/etc to get the temperature down. I hope that makes some semblance of sense, as it took a bit to wrap my head around it.

We appreciate your loyalty to AMD as we continue to strive for the best products in terms of graphics and processors for our customers. I am glad to hear that you are enjoying your FX 6100, and hope that the joy will continue for years to come. If you have any other questions or comments, please don't hesitate to reply directly to this email and I will try to provide any additional information that you may require.

In order to update this service request, please respond, leaving the service request reference intact.

Best regards,

AMD Global Customer Care

 

[drea.drechsler]

Core temp max (Tjmax) is 61.1°C
Surface of the silicon at the IHS (Tcase) is 63°C.
Tcase temps were measured by AMD engineers using modified engineering samples with thermocouplings at the IHS. The Tjmax is not a measured temp, but an 'educated guess' by those same engineers since they couldn't add thermocouplings to the inside of the cores and still have functioning cpus to give readings on. There were no functional FX made with interior readers in the cores or nodes as that would have changed the properties of the cpus.

It's kinda not off topic, sorta, because power reporting, consumption, use goes hand in hand with temps. OP can easily get power numbers from HWInfo, but those numbers will be useless if he can't then run the cpu because of those temps. He'll have to get better cooling or down clock to workable temps, which then changes the power numbers.

Unfortunately, the tech doc links are no long working. But it appears Tcase is a parameter carried over from Athlon/Sempron/Phenom processor spec's that the responder used in the absence of tech docs at the time that response was sent. The best anyone has that I can find is that "Tjmax" spec AMD puts up on their FX Spec's site.

So, I'm still left thinking the Tmargin number is going to be mathematically derived from the known inaccurate temperature readings derived from whatever the processor sends out. It seems to be informative because it's telling you how far away until it starts hitting protections... max fans, throttling and ultimately shutdown. But it's obviously a differential that has to be derived from something like (Tjmax - Tcurrent) at the point in time it's reporting it so Tmargin must share the inherent inaccuracy of Tcurrent.

 

[Karadjgne]

The TM is derived from loads per core, socket temps, core voltages, package voltages etc. All read same as HWInfo would read them. All those measurable and specific values are put through a complex algorithm, the end result being a number which is weighed on a map, so if the number was 1.4276, on the temp map that would equate to a 20TM, if the number was 1.4278 it might equate to a 17TM etc.

What it means is a 8 thread cpu pushing 4 threads very hard at 4.7GHz may show a TM of 15, but that same cpu pushing 6 threads moderately hard at 4.5GHz can also show the same TM, but changes to core loads or thread counts will vary the TM by different amounts. Small changes in the 4.7 having larger changes to the TM.

TM isn't an absolute number, not the way a °C is. It's a number representing a range. Closer to 0 you are the more power the cpu is using, the hotter it's running. It's like saying the cpu has a temp range of 5-1 with 5 being cool and 1 being hot, 0 being overheat. So if the working cpu is always a 3, very occasionally a 2, you are no worries, regardless of actual power consumption or voltages or specific temp.

 

[drea.drechsler]

...
TM isn't an absolute number, not the way a °C is.
...

That's kind of the point...not being a simple value in degrees makes it easier to swallow it's inherent inaccuracies. That said it does provide guidance for those who might need it.

But personally, I'd treat thermal margin the same as watching HWInfo CPU temp readings in that I don't push a CPU to the limits anyway (except maybe in something like a Prime95 stress test, briefly) but never in routine use by any means. As has been reported, the temp measuring algorithm is accurate enough when the processor is heavily utilized I'm comfortable the 10C margins I see to Tjmax when it's working hard is probably more than enough. The secret with any successful overclock is to back off settings and leave a decent safety margin once you find the limit.

And also...Windows 10 is a far cry from what it used to be, and a world apart from Windows 7 before that, at CPU core utilization. I remember, in Windows 7, while editing some photoshop, web browsing and playing an online game it would slam 2 cores and ever so lightly use the others, if at all. Now, 10's scheduler does a good job of throwing it's multitude of active threads around to more cores even with an FX6300, whatever it's doing. This isn't Zen 2 or Zen 3 so there are no preferred cores, they all get beat pretty much equally. Point is: I don't see an issue with one core getting burned up while the other 5 sit there in deep sleep states.

 

[Karadjgne]

Me either. That's the nice part about preferred cores, there is always one. Silicon structure is unique, even amongst cores, they have minute differences that will create a stronger or better core, and a weaker one. Treating all cores equally regardless of status, that weaker core sees just as much abuse as the stronger cores, so will be the first to become unstable in a lengthy OC period that's tuned to strict performance heights.

Preferred cores putting more abuse on the stronger cores evens out the playing field, as soon as that core becomes less strong than another, the other is now preferred. Meaning the cpu is always playing to its strengths.

 

[drea.drechsler]

....
Preferred cores putting more abuse on the stronger cores evens out the playing field, as soon as that core becomes less strong than another, the other is now preferred. Meaning the cpu is always playing to its strengths.

Except there's no way I'm aware of to change core ordering as that's fused-in by AMD at manufacture for Zen 2 , 3 and probably 4 too. The Windows 10/11 scheduler, with CPPC enabled, picks it's preferred cores based on core ordering with consideration for CPU architecture; mainly shared resources such as L2 and L3 caches I suppose. CPPC disabled it seems to simply prefer the "best" cores based on core ordering with my Ryzen systems. Although, with all the cores at a fixed clock it may not prefer any cores; I've never tested it since I don't all-core OC my Ryzen systems.

And I do understand any overclocked CPU will at some point become unstable. But that's inevitable in the scheme of things since any semiconductor degrades when used. Using it more heavily only degrades it faster - ANY overclock, especially when coupled with extensive heavy processing, will do. When, or if, that happens, re-establish stability with a lower clock and march on. IMO, overclocking should only be attempted if you accept that your CPU is, at some level, a consumable. Right now, with $10-20 dollar FX6300's widely available on Ebay it's a viable proposition since overclocking these things is enjoyable and with a capable enough motherboard not going to be all that expensive even if things go south quickly.