• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

FX-8320 Socket vs Core Temp

Joined
Oct 7, 2009
Messages
396 (0.07/day)
System Name Just A Gaming Rig.
Processor AMD FX-8320 @4.1GHz, 1.268V
Motherboard ASUS M5A97 Evo R2.0
Cooling Cooler Master Hyper 212 Evo
Memory 8 GB Crucial Ballistix Sport 1600MHz
Video Card(s) Club 3D R9 280X 3GB royalQueen, @1115/1615MHz
Storage OS+Often used stuff: Kingston SSDNow V300 120GB, Mass storage: WD Blue 1TB
Case Zalman Z3 Plus
Audio Device(s) Integrated
Power Supply Corsair CX600M
Software Windows 8.1 64-bit
Happy New Year guys.

So I built a new rig for this Holiday season, and I'm trying of course to OC my FX-8320. I'm running it on the ASUS M5A97 Evo R2.0 and it's cooled with the Cooler Master 212 Evo. Now from what I've figured out by searching online is that the max safe temperature for this CPU is 62 degrees Celsius. The problem is, I seem to find conflicting info on if that 62 max temp. is for "socket temp" or "core temp" (I'm using the latest version of HWMonitor). Hoping you guys can help me out on this one. If the 62 max temp is for "core temp", what would you guys say is a max safe socket temp?
 
The 62C refers to socket temp because AMD CPUs don't have proper thermal probes in them. The other thing is that AMD chips do not like running at high temperatures so you will get a Blue screen before the chips is anywhere close to getting damaged by heat.
 
The 62C refers to socket temp because AMD CPUs don't have proper thermal probes in them. The other thing is that AMD chips do not like running at high temperatures so you will get a Blue screen before the chips is anywhere close to getting damaged by heat.
Ok thanks. I'm getting 63-64C @4.0GHz after an hour+ on Prime95 small FFTs, but that's probably because my case airflow isn't great, and I'm also planning on re-mounting the cooler. I think I OD'd on the thermal paste. Also considering on buying an extra fan for push/pull. That should help the temps.
 
Having the same cpu and motherboard as you, I should say that it does not run as hot as yours especially at those clocks and voltages, so check your paste, repaste if necessary, also check if your cpu fan spins up under load.
 
Having the same cpu and motherboard as you, I should say that it does not run as hot as yours especially at those clocks and voltages, so check your paste, repaste if necessary, also check if your cpu fan spins up under load.
Yeah I know it's probably the paste, even if my case airflow isn't stellar, with under stock voltage and not that much OC, it should not go above 62C. My CPU fan does spin up, in fact it spins full RPM all the time, even though I plugged it in a PWM connector...
 
Last edited:
The 62C refers to socket temp because AMD CPUs don't have proper thermal probes in them. The other thing is that AMD chips do not like running at high temperatures so you will get a Blue screen before the chips is anywhere close to getting damaged by heat.

This isn't true unfortunately.

The core temp is the temperature sensor on the CPU itself, and this one is "limited" to 62 Degrees. The socket temp doesn't really have a hard limit, but this can go quite high if your VRM cooling isn't too good while still having a good CPU cooler.
The reason the core temp might give some weird results is because it only produces correct temperatures from 45C and up. This isn't really a problem however, as any CPU core temperature below that is "uninteresting" from a cooling standpoint anyway.


Also, the 62C core limit isn't really a hard one: If you want better stability with lower voltage, or less poweruse with the same voltage you should aim for a lower maximum core temperature, because FX CPU stability gets progressively worse the higher the temperature gets, needing more voltage to correct it which in turn results in even more heat. Keeping this in mind, you sometimes can even improve your clocks a bit when up against a heat wall by lowering your voltages a small bit, resulting in a lower temperature and improved stability.
You can also go with the opposite if your motherboard, cooling and power costs are up for it: tweak your system for a higher core max than 62. This can potentially result in higher clocks if your chip likes high voltages and isnt really impacted by heat at those voltages.
This has the downside of going against the advised the max core temp which potentially results in a slightly shorter lifespan of your CPU. However, tbh, I dont know of anyone, nor have I read about, people who had their CPU die on them because of running higher than the advised temp(its usually too much voltage that kills a chip). Furthermore, if you look at the 8350 with stock cooler you'll see that it will already regularly pass the 62C mark on the core, and this is how AMD specced it, so it can't be that bad surely?

Anyhoo, long story short. 62C is the (un)official max core temp. Going higher than that *may* result in a shorter CPU lifespan, but will probably cause instabilities. The max socket temp is dependent on what your mobo can take, and how low your core temp stays. Most important thing by far for the health of your CPU is the core temp.
 
Last edited:
This isn't true unfortunately.

The core temp is the temperature sensor on the CPU itself, and this one is "limited" to 62 Degrees. The socket temp doesn't really have a hard limit, but this can go quite high if your VRM cooling isn't too good while still having a good CPU cooler.
The reason the core temp might give some weird results is because it only produces correct temperatures from 45C and up. This isn't really a problem however, as any CPU core temperature below that is "uninteresting" from a cooling standpoint anyway.


Also, the 62C core limit isn't really a hard one: If you want better stability with lower voltage, or less poweruse with the same voltage you should aim for a lower maximum core temperature, because FX CPU stability gets progressively worse the higher the temperature gets, needing more voltage to correct it which in turn results in even more heat. Keeping this in mind, you sometimes can even improve your clocks a bit when up against a heat wall by lowering your voltages a small bit, resulting in a lower temperature and improved stability.
You can also go with the opposite if your motherboard, cooling and power costs are up for it: tweak your system for a higher core max than 62. This can potentially result in higher clocks if your chip likes high voltages and isnt really impacted by heat at those voltages.
This has the downside of going against the advised the max core temp which potentially results in a slightly shorter lifespan of your CPU. However, tbh, I dont know of anyone, nor have I read about, people who had their CPU die on them because of running higher than the advised temp(its usually too much voltage that kills a chip). Furthermore, if you look at the 8350 with stock cooler you'll see that it will already regularly pass the 62C mark on the core, and this is how AMD specced it, so it can't be that bad surely?

Anyhoo, long story short. 62C is the (un)official max core temp. Going higher than that *may* result in a shorter CPU lifespan, but will probably cause instabilities. The max socket temp is dependant on what your mobo can take, and how low your core temp stays. Most important thing by far for the health of your CPU is the core temp.
Yeah, I always though this explanation made more sense... But now I have two people saying the opposite again :laugh: My core temp usually doesn't go past 45-46C when stress testing. I'm not quite sure what VRM-components are rated for temp wise, but keeping the socket temp under 70C should be sufficient right?
 
Yeah, I always though this explanation made more sense... But now I have two people saying the opposite again :laugh: My core temp usually doesn't go past 45-46C when stress testing. I'm not quite sure what VRM-components are rated for temp wise, but keeping the socket temp under 70C should be sufficient right?
Definitely:)

Edit: Most VRM's are actually specced for use up to something like 105C. Other components might not like such high temperatures though.
 
Because all of you have the same CPU as a friend of mine who made a PC built a week ago, which voltage level you think is enough if interested in undervoltaging the FX-8320 to be stable enough for video editing programs?
 
Because all of you have the same CPU as a friend of mine who made a PC built a week ago, which voltage level you think is enough if interested in undervoltaging the FX-8320 to be stable enough for video editing programs?
Well I'm running @4.0GHz (17.5 x 229) at 1.2125V, with the stock volts being 1.25V. Though you should keep in mind that it's not tested thoroughly yet, running some blend Prime95 right now. So keeping it at stock, I could imagine as little as 1.2V being stable, maybe a little more or a little less, depending on the silicon lottery, and maybe the motherboard your friend has.

EDIT: Not stable, had an error 50 minutes into blend. Back to the drawing board.
 
Last edited:
Definitely:)

Edit: Most VRM's are actually specced for use up to something like 105C. Other components might not like such high temperatures though.
I would imagine that only the immediate vicinity of the socket is at the measured socket temp, with stuff like RAM or NB being a bit cooler than that. But yeah, 105C sounds very extreme, one should back off way before that sort of temps are reached.
 
Well I'm running @4.0GHz (17.5 x 229) at 1.2125V, with the stock volts being 1.25V. Though you should keep in mind that it's not tested thoroughly yet, running some blend Prime95 right now. So keeping it at stock, I could imagine as little as 1.2V being stable, maybe a little more or a little less, depending on the silicon lottery, and maybe the motherboard your friend has.

EDIT: Not stable, had an error 50 minutes into blend. Back to the drawing board.

I got the idea. 1.2V is the sweet spot for stock clocks. And my friend has the same mobo. Thank you!
 
I got the idea. 1.2V is the sweet spot for stock clocks. And my friend has the same mobo. Thank you!
But remember to tell him to test the stability very thoroughly before actually using the settings for daily use! My CPU ran small FFT's on Prime95 for hours @4.0GHz with 1.2V, but on blend I haven't gotten it stable yet. Currently at 1.225V. I've read on another forum where a guy got his FX 8320 stable at stock with 1.18V, it's all about the silicon lottery.
 
Of course it will be tested. Thank you for the info.
 
iv had phenom IIs north of 65C OC'd and well over 70 stock you don't need to worrie about killing anything untill 80c the IMC dies at about 85c Iv tested it lol
 
Hey so OCCT Stress test is hitting 70C-73C cpu temp yet AMD overdrive is showing a thermal Margin of 2-7 for each core. Which do I go by ? Which do I trust?

Also I can't find info on AMD what the max temp is for the FX 8320, it shows up blank.

http://products.amd.com/en-us/DesktopCPUDetail.aspx?id=810&f1=AMD FX 8-Core Black Edition&f2=FX-8320&f3=3500&f4=1024&f5=AM3+&f6=&f7=32nm&f8=125 W&f9=5200&f10=False&f11=True&f12=True
if its pushing 70c at stock you probably need better cooling
I said I Have had them over 70c I didn't say it was a good idea
 
if its pushing 70c at stock you probably need better cooling
I said I Have had them over 70c I didn't say it was a good idea


Yeah but AMD OD is showing my thermal margin as being pretty safe AKA not in the danger zone. It will hit the DZ for like a sec then go back to 2+.

But I'm not sure which source to trust.
 
Trust AMD, the 3rd party programs don't know what to do with TCTL.

900x900px-LL-306a7aae_tclscale.png
 
TCLT is basicly a speed independent version of intels TJmax rating its essentially meaningless
the higher the core speed the lower the temp threshhold
 
So again still not sure what to go off of? My idle temps are at 40C on boot it was 32C. That sounding about normal? Playing Watch Dogs it goes to 52C max. OCCT did bring it to 77C though.
 
idle and boot temps are meaningless stress test load != real load if you are worried about buy a better cooler if not let it be simple ...
 
idle and boot temps are meaningless stress test load != real load if you are worried about buy a better cooler if not let it be simple ...

I don't think they are meaningless. I'm trying to see if I have it seated well to begin with which is what I'm worried about. And I'm trying to find out which source of Temp monitoring to trust because I've got three options to look at and can't come down to a definitive conclusion.
 
There are only 2 temperatures for AMD CPU/APU: socket temp and TCTL. Socket temp is a fair way to judge idle temperature, but is way off under heavy load. Only AOD interprets TCTL correctly by calling it thermal margin, and it is very accurate at load. You can test it yourself. If you get close to zero thermal margin, you will throttle. It works exactly as in the diagram. When people throw these 60'C-70'C temperatures around, you know they are measuring wrong because if you got that high on socket temp you would have already tripped the TCTL cutback due to hitting the thermal margin. If you think I, or AMD engineers are lying, test it yourself. I did.
 
Back
Top