Investigating the AMD Athlon 64 X2 Overheating Problem

[Shivansps]

So people may know about this issue, the thing is, is a very common issue, and as years go on is getting more common, many people does not pay any attention to it because at this point no body cares as even a Sempron 2650 is faster than these howdays.
The problem presents itself suddenly, one day everything is ok, the next day the CPU is at 65°C idle.
The problem seems to be limited to Athlon 64 AM2 based (Windsor/Brisbane), and there are only 3 ways of fixing it,
1) set the vcore to manual to something like 1.25V
2) changing the motherboard
3) changing the cpu

And thats the thing, 2 and 3 should exclude each other, but they do not.

The first time i noticed this problem was a few years back with my main pc at that time, a Biostar NF61S-M2 TE and a X2 4200+ Windsor.
One day it started to overheat like crazy and it was unusable, i think it was after about one and half year of use, i suspected of the motherboard so i brought a Biostar GF8100, after that i had no more overheat issues, later replaced the 4200+ by a Sempron 140.

Today i have:
Sempron LE-1100
Sempron 3800+
Athlon 64 X2 4200+ (Windsor)
Athlon 64 X2 5000+ (Brisbane)
Athlon 64 X2 5600+ (Brisbane)

MSI K9AGM4
Asus M2N4-SLI
Biostar K8M800 Micro AM2

I used the 5000+ and Asus M2N4-SLI for Bitcoin mining back in 2011 and started to present the same overheating issue on 2012.

The 5600+ overheats with the MSI K9AGM4.

Biostar K8M800 Micro AM2 is fine with every CPU.

If i set the vcore to 1.3V manual on the Asus M2N4-SLI it does not overheat anymore, if i place the 5600+ that overheats on the MSI K9AGM4, on either of the other mbs it does not overheat, the MSI K9AGM4 has no vcore control, if i place the 5000+ on MSI K9AGM4 it does not overheat, is also fine with the K8M800, the 4200+, LE-1100 and 3800+ is fine on every MB, and the 4200+ is the one that used to overheat on NF61S-M2 TE that i dont have anymore.

So this does not make any sence to me, clear cmos and bios updates does not fix it, and it seem to be confined to a specific combination of CPU and MB after some time.
I started to think this is some kind of hardware bug with the auto vcore control or (/Sensationalist mode ON) even a time bomb.

 

[nismotigerwvu]

Well if we had a time machine that could summon up the version of me from a decade ago I would be able to help. It seems like it's just the sort of issues you see with components of this age. Failing components of the system including the PSU or mobo (caps, power management...ect) could be leading to running the CPU out of spec. In the case of overvolting you would certainly see overheating and eventually degradation. Sadly I think you'll just be chasing your tail on this and it might just be best to view it as how these parts finally end up failing.

 

[zir_blazer]

Add in Venice DH-E6. Motherboard was an ASUS A8N-VM CSM which already had a nasty tendency of overvolting by a positive fluctuating offset 0.075-0.1V according to CPU-Z, so the Processor VID was 1.4V but received by default 1.475-1.5V.
After that Thread, I delidded the Athlon 64 and applied standard silicon grease between the Heatspreader and the die. Results were pretty much identical to those before delidding. After that, I noticed that there were bad caps on the Motherboard, so I sent it to get it repaired since I was aware that bad caps could cause overvolting. It came back with new caps but still had temperature issues. I even argued at the shop that it got repaired since they said that I "destroyed the Processor" by delidding it, while it actually worked, but usually shut down due overheating unless you inmediately undervolted it with Central Brain Identifier after Windows booted. Additionally, after the new caps, it started to randomly BSOD, while previously I don't recall any BSOD, just shut downs.
Gave up and purchased newer Hardware.

 

[Shivansps]

Well if we had a time machine that could summon up the version of me from a decade ago I would be able to help. It seems like it's just the sort of issues you see with components of this age. Failing components of the system including the PSU or mobo (caps, power management...ect) could be leading to running the CPU out of spec. In the case of overvolting you would certainly see overheating and eventually degradation. Sadly I think you'll just be chasing your tail on this and it might just be best to view it as how these parts finally end up failing.

Its not because they are old, this started around the time Phenom released about 2 years later, ive see many cases back them, and i keep seeing them today and i think it also has something to do with the quite high dead rate of AM2 motherboards and cpu ive see since them, but it could be just higher volume.

 

[SPBHM]

I never heard of this problem, and my X2 Brisbane from 2007 was used for around 2 years for gaming at 2.6Ghz and now it's been used almost every day for web browsing at 2.4Ghz with the stock volt (around 1.26v), the temperatures are quite good.

(my memory controller load is really low, running a single stick underclocked for most of the years)

 

[VirtualLarry]

I've seen this issue with OEM systems, often the 5600+ / 5800+ / 6000+ AM2 125W CPU-based systems. Swapping heatsinks and re-pasting doesn't seem to help much.

 

[coercitiv]

Add in Venice DH-E6.

I never had any problems with my Venice, but then again that Athlon 3000+ and a DFI nF4 Infinity board spent most of it's life at stock speeds and undervolted (1.25V if memory serves me right). The system operated 24/7, with shutdowns months apart.

 

[zir_blazer]

I never had any problems with my Venice, but then again that Athlon 3000+ and a DFI nF4 Infinity board spent most of it's life at stock speeds and undervolted (1.25V if memory serves me right). The system operated 24/7, with shutdowns months apart.

So did mine. Since the BIOS of the A8N-VM CSM didn't let me do anything at all, there were two options: Using SetFSB in Windows for overclocking, (Which I think I did only once or twice, since I had to slowly increase it in order to not freeze/BSOD. I think that even I managed to corrupt the BIOS, which fixed itself aftet a few reboots. Decided the risk wasn't worth it), or take advantage of the headroom to undervolt. After I learned that I could do the latter with Central Brain Identifier, the very first thing I did after a reboot was opening it, and lowering the Voltage to 1.1V, which was the minimum. So yes, my 3000+ was 24/7 and undervolted during nearly its entire life.
Problems were already apparent when I builded my Athlon II X4 620 during 2009, since it was at around 70°C during load when undervolted. It lasted like two or three more years on a family member hands (Which I actually teached how to undervolt) when the shutdown issues kicked in everytime he forgot to undervolt.

Also remember that there were several A64 3000+ versions, and two Venices, DH-E3 and DH-E6. The latter was when the RoHS craze started and didn't use lead if I recall correctly.

 

[mizzou]

i own and still use a phenom ii and athlon x2. OC d the heck out of them and never had a heat problem

 

[coercitiv]

Also remember that there were several A64 3000+ versions, and two Venices, DH-E3 and DH-E6. The latter was when the RoHS craze started and didn't use lead if I recall correctly.

Well, here's mine bellow: using CPU World info results this is a DH-E6 as well.

 

[Shivansps]

mmm maybe i was not clear, this issue happen only on AM2 Athlon, not on every one, but is a fairly common issue, i never heard of any issues on athlons of socket 754 and 939, this is the first time, also AM2+ cpus and on seem to be unaffected as well, Phenom, Phenom 2 and Athlon 2 are unaffected.

 

[Deders]

Is it definitely the chip that is overheating. I seem to remember problems with the northbridge overheating leading me to use a desktop fan directed into the open case.

 

[SPBHM]

Is it definitely the chip that is overheating. I seem to remember problems with the northbridge overheating leading me to use a desktop fan directed into the open case.

by northbridge do you mean the integrated memory controller in the CPU die or the external chipset (like the x16 PCIE controller and IGP)?

on my AM2 board the chipset (6100-430, single chip) used to run really hot (over 80C), but since new I had a fan near it and it runs with decent temps (under 60C), but that's because Asrock used a ridiculous heatsink

the CPU is really far from bad, at one time it was going over 60 under full load, but after cleaning it the temp was reduced by almost 10ºC

 

[Deders]

On the board itself. Iirc Brisbane especially had a notoriously innacurate temperature sensor.

 

[Shivansps]

Its not, they are really overheating, also remember is does not happen if i change the cpu/mb combo, or i do set the vcore to manual.

 

[Burpo]

I have 2 Acer mini towers, both STILL running original Athlon 64 X2 4200+ cpu's. One is original mobo & the other was replaced about a yr ago (ASUS M2A-VM AM2 690G). I apply new TIM when I clean them, but they're both used daily, and chugging along on Win7 with no overheating (original HSF). 'Shrug'..

 

[SPBHM]

On the board itself. Iirc Brisbane especially had a notoriously innacurate temperature sensor.

the temperature readings are not the most reliable, but, I always used the highest as a reference and it always looked believable enough in terms of idle vs load temperature (just did 10 min of Aida burn test and it was around 20-25ºC from lowest to highest core from idle 1.15v to load 1.26-1.28v for 10min, 63ºC was the highest temp), also as I said when I cleaned like 4 years of dust out of it it had a big effect on the readings.

right now it's probably been over a 1 year since I last cleaned it and the room temp is pretty high but it all seems fine, it could be even better with lower room temp and some cleaning, same CPU, cooler and board since 2007... if some A64 are having problems, this is not one of them,

 

[Deders]

Proportionally it could look right but you'd need a laser thermometer or something similar to be sure. I always found it strange that the heatsink never seemed to feel as warm as I would have expected at around 60c.

 

[SPBHM]

Proportionally it could look right but you'd need a laser thermometer or something similar to be sure. I always found it strange that the heatsink never seemed to feel as warm as I would have expected at around 60c.

I think depends where this reading is coming from and the cooler itself, on my i5 sandy bridge I can have a 90ºC reading and most of the cooler doesn't even feel any different from when it's a sub 40ºC readings (it's one of those old Intel coolers with copper core and mostly aluminum, I can touch the aluminum part and it never feels warm, but I think the copper part is pretty warm), while the all aluminum cooler on the A64 feels warm with the temps I mentioned

anyway, from memory I would say the temps I'm getting now with the A64 are similar to what I had when it was new, but it's hard to be 100% sure since it was a long time ago.

 

[SparkyJJO]

Only time I ever ran into an AMD thermal issue was when I had my Opteron 165. Bought the specific stepping, got one with a horrible IHS thermal connection. First CPU I ever delidded, but it went fine and dropped temps like a rock.

Other than that, the others I've run/built/whatever have been fine. Well, aside from the infamous Brisbane sensor that read 15 degrees C too low or something like that. Hey, my CPU is idling at 10C! Sweet... wait.

 

[SPBHM]

Only time I ever ran into an AMD thermal issue was when I had my Opteron 165. Bought the specific stepping, got one with a horrible IHS thermal connection. First CPU I ever delidded, but it went fine and dropped temps like a rock.

Other than that, the others I've run/built/whatever have been fine. Well, aside from the infamous Brisbane sensor that read 15 degrees C too low or something like that. Hey, my CPU is idling at 10C! Sweet... wait.

on aida64 my idle temps (CPU load under 10% for a few mins on win10 x64) for the Brisbane (g1) right now were 5ºC over room temp for the coolest core, and 13ºC for the hottest core

 

[Deders]

I think depends where this reading is coming from and the cooler itself, on my i5 sandy bridge I can have a 90ºC reading and most of the cooler doesn't even feel any different from when it's a sub 40ºC readings (it's one of those old Intel coolers with copper core and mostly aluminum, I can touch the aluminum part and it never feels warm, but I think the copper part is pretty warm), while the all aluminum cooler on the A64 feels warm with the temps I mentioned

anyway, from memory I would say the temps I'm getting now with the A64 are similar to what I had when it was new, but it's hard to be 100% sure since it was a long time ago.

I think I was comparing touching the base of the standard AMD heatsink (purely aluminium I think) with the times I accidentally touched the heatpipes on my graphics cards and got burned.

 

[coercitiv]

mmm maybe i was not clear, this issue happen only on AM2 Athlon, not on every one, but is a fairly common issue, i never heard of any issues on athlons of socket 754 and 939, this is the first time, also AM2+ cpus and on seem to be unaffected as well, Phenom, Phenom 2 and Athlon 2 are unaffected.

Well, here's the problem they way I see it: your problem and your solution don't match.

The solution you presented (software voltage settings) could have an influence on CPU temperature under load. The problem you presented - CPU temp of 65C at idle - strongly suggests a heat transfer problem. That is assuming everything is properly cooled.

Considering your experience with PCs (very good experience that is), I would venture out and say you had both: a heat transfer problem combined with some bug that modified CPU voltage settings. The second would exacerbate the first, but cannot be the main cause to begin with.

That being said, this is Armchair Diagnostics™, so feel free to contradict me :biggrin:

 

[Shivansps]

yeah it could also be a C&Q bug, ive never thought of that.
The thing with manual control is that does override C&Q control over it, mmm im gona test this in the weekend if i have time.

But the fact remains that is strange this bug appears only on specific CPU/MB combination, suddently after a while, and if you break that combination it does not happen anymore.

I've disregarded thermal transfer problem long ago.

 

[ShadowVVL]

I have had this issue on a 5600+ windsor and a phII 955 x4 with stock heat sinks.
They both started over heating at 70c under mid load around 2 years after I got them and the heat sinks get hot and replacing the thermal paste with noctua nt-h1 makes little difference.

my athlon is running fine now at 28c-51c max and the heat sink isnt raging hot so I have no idea what causes it or when it will strike again.