AMD 8350 Maximum "safe" Temp

Status
Not open for further replies.

Lawrence Orsini

Distinguished
Jun 10, 2013
29
0
18,540
Okay guys, this question is a little different than the other 200 'is my temp okay?' questions I've seen posted. Unlike most folks, I am actually looking to make the most heat possible from my CPU/GPUs. So my question is, what is the highest sustainable temperature I can run this AMD FX 8350 indefinitely without melting it down. Please don't suggest anything under 67C at the cores, I've already had it there for more than 24 hours and it never skipped a beat.

AMD has no official max temp listed: Specs


Know that I am running full liquid cooling across all things that go 'ping!' with a 10 gallon thermal reservoir and dual liquid/air radiators plumbed in parallel with 3 Noctua NF-F12 PWM running a Push/Push/Pull config. I have very a stable thermal platform that doesn't get jittery, even at high temps.

Who has run their rig the hottest?
 

robnof

Honorable
Oct 9, 2012
491
0
10,960


I believe it starts to throttle after 70 degrees.
 


assuming your cpu is reporting it's temps right, you're already 7C-12C over the max "safe" temps for the FX cpu. Generally speaking they are no longer stable or for that matter all that safe over 60C... in reality you probably shouldn't get up much higher then 55C...

That said i doubt your cpu is reporting it's temps accurately. since the PhII, AMD has shifted away from using an actual thermometer on it's chips, instead they use some weird algorithm to calculate the closest accurate temperature. The result is many AMD chips are as much as 5-20C off on their reported temps (plus or minus)... generally it isn't much more then 5 to 10, but 20 isn't all that unheard of either.

Not saying you're not running at 67C... just saying you don't have much more to go as if that's accurate (i think AMD throttles their fx chips at 70C; not that anyone ever really sees this... just like the PhII would throttle at 90C... but anyone who's had one will tell you, there isn't a PhII that will run over 65C... so AMDs thermal throttling is silly), you're really risking your chip... and if it's inaccurate you'll be nearing the thermal ceiling very shortly.
 

Lawrence Orsini

Distinguished
Jun 10, 2013
29
0
18,540


Thanks, that's helpful, I do have a couple more questions though. first, what's the best way to get true temps off the chip if not the core monitor? Second, how high can I take the temp if the CPU is idle or very lightly loaded... I'm assuming it would run at a higher temperature if it weren't actually loaded. And lastly, what physically happens to the chip when it eventually fails on temp?

Thanks for the help.
 

chromic

Honorable
Jan 6, 2013
1,091
0
11,360


Wut o_O
 


ok... when a chip gets too hot and fails typically you'll either blue screen or your system will just reset on the spot. It blue screens when the chip is unstable due to temps (it happens) and errors, it black screen resets when it just straight fails from too much heat; if you pass the MB's safe temps for your chip the mb will beep when it turns itself off.

Generally speaking we're talking about 2 types of failures.
1) instability caused by excessive heat
2) overheat causing motherboard to turn off system OR damage which kills chip.

Generally MBs are good enough to keep your chip from melting. There are two types of wear your chip suffers from overclocking.
1) gradual warping of the chip itself and the and degrading of transistors due to continual excessive heat, eventually leading to chip death.
2) gradual degradation of CPU due to excessive voltage.

Both of these can give the same symptoms... such as gradually needing more voltage to achieve a stable overclock
similar to one you had when the chip was new. ie... if i needed 1.4385 vcore to overclock my PhII to 3.8ghz when i got the chip new... and 6mo later i need 1.4625 vcore to stabilize the exact same overclock, that's a chip which is breaking down due to excessive voltage/heat). Once this starts to happen your chip is on it's last legs.
 

Lawrence Orsini

Distinguished
Jun 10, 2013
29
0
18,540


Things that go ping is a Monty Python reference for the most expensive of whatever.

The radiators are sandwiched between 3 Noctua fans with airflow all pointed the same direction and the 10 gallon thermal reservoir is a liquid/liquid heat exchanger that stores the heat from the computer so I can do other things with it... instead of dumping it into the room.

Makes sense?
 

Lawrence Orsini

Distinguished
Jun 10, 2013
29
0
18,540


Thanks, that was really helpful. Do you know what temps a chip will live to when it is idle - I'm assuming it has a much higher threshold if it isn't creating its own heat. I know I'm likely pushing some boundried here but... thats what they are for!
 


well, my cpu generally runs around 5-10C hotter then ambient temps when idle... that seems to be pretty much standard. though that will start to get warmer the more vcore being pumped into your cpu. for example... this cpu will idle at (ambient temps at 28C) 33C at stock 1.4V, but at 1.45V it will idle much closer to 35C-37C depending, and at 1.5V it idles at about 40C... (of course this is a PhII, and even if i had your FX cpu, i would point out my cooling solution and chip are not your cooling solution and chip, and therefor this is not much to go on)
 

LO3

Honorable
Sep 19, 2013
4
0
10,510
So it turns out, after almost a year of testing and tweaking, that the AMD 8350 Black is actually quite stable and happy up to about 79c as long as you underclock. You can see in this pic I'm running the loop temp at 71c at 3.4 on the clock and have the fans configured to maintain the temps at that level while the CPU/GPUs grind away mining coins... which turns out to be a pretty stable load without some of the bumps that prime 95 seemed to have when it was running. There is no damage to the cpu, if I drive the loop temp down to 21c I can still get a stable 4.8 out of the 8350.

I've checked the loop temp with a Fluke and it is running at the 70c in the pic, you can see when I stop the CPU miner the CPU temp drops right to loop and sticks so thats a pretty good indication that the algorithm AMD is using to calculate temp must be pretty close.

Thanks for the advice, all.

So it turns out, after almost a year of testing and tweaking, that the AMD 8350 Black is actually quite stable and happy up to about 79c as long as you underclock. You can see in this pic I'm running the loop temp at 71c at 3.4 on the clock and have the fans configured to maintain the temps at that level while the CPU/GPUs grind away mining coins... which turns out to be a pretty stable load without some of the bumps that prime 95 seemed to have when it was running. There is no damage to the cpu, if I drive the loop temp down to 21c I can still get a stable 4.8 out of the 8350.

I've checked the loop temp with a Fluke and it is running at the 70c in the pic, you can see when I stop the CPU miner the CPU temp drops right to loop and sticks so thats a pretty good indication that the algorithm AMD is using to calculate temp must be pretty close.

Thanks for the advice, all.

[img]https://plus.google.com/u/0/photos/105529742818195922083/albums/5925456913180480657/6014569991763320082?pid=6014569991763320082&oid=105529742818195922083
 

Andreas Vollen

Reputable
Aug 16, 2015
5
0
4,510

i under clocked the cpu to 3.4ghz and i installed a bigger exhoust fan and inverted the old exhoust fan to intake, 2 small intakes and 1 big exhoust, now it sits at about 60*c under stress :D instead of up til 90*C

 

Lawrence Orsini

Distinguished
Jun 10, 2013
29
0
18,540
Here is the solution - we've been running this 8350 at 67-80c at around 3.5 to a peak of 4.8 MHz reliably for almost 2 years. Along the way we have talked with several high temperature computing researchers and gotten a real understanding of how and why most CPU/GPUs fail under prolonged, high temperature use. As CPUs wear the experience something called electromigration. Electromigration is the process by which the electrons moving through the core pick up and carry pieces of the conductor with them during use. Electromigration is accelerated with high temperatures and increased voltages - the hotter and harder you run the chip, the easier it is for the electrons to move move the conductor. The more conductor you move the more errors that are generated by the chip.

Along the way these increased temperatures also add instability to the gates - the actual switches inside the processor. You can increase the conductor durability by using harder materials (tungsten instead of copper) as well as use a Silicon on Insulator (SOI) design instead of the standard CMO S design used in the majority of high speed processors. These SOI designs are more expensive as well as being slower, largely due to the increased resistance of the conductor. We're proposing some high temperature computing demonstrations to the Department of Energy along with the Fraunhofer Institute's high temperature computing lab to determine the worldwide market for high temperature computing. Fraunhofer already makes SOI chips that operate to 300c.

As for the little AMD 8350, it is still chugging along with little sign of degradation after all the abuse. It seems as long as the temperature rise is stable and sustained, (we're liquid cooling all the chips) the CPU hasn't had a problem running near 80c.

To answer the 'why would you underclock it and make it run hot' question, take a look at projectexergy.com
 

OpalSerPenT

Honorable
Jan 6, 2014
6
0
10,510


Do you use a proper gaming case? Even with liquid cooling? The liquid cooling is all good for your cpu or gpu but not the rest of your motherboard. I run my fx 8350 locked to a constant 4.4 ghz and I know I could go to around 4.6 with just my air cooling but I won't risk it for the sake of a few hundred 3dmark.
I'm running a cooler master huf case with 5 fans and one high velocity fan on the top with a pull on the back and front configuration with blowing out the top. I get 50 degrees at the most in normal weather in 3dmark cpu stress test.
Maybe you need an infrared thermometer to see how hot your motherboard and components are really. They are cheap on ebay. Seriously that does seem like an abnormally high temperature. Are you using enough thermal paste? I am using a giant cheap gammax 400 air cooling radiator that barely fits in my case, it has leds to boot. It provides good cooling in my pull push configuration. If I was getting over 55 degrees I'd already be alarmed. Taken in about 29 degrees QLD early summer weather temps.

Seriously if that's water a water cooling is doing for you I'm shocked. Or my cpu is sweet ass.
 

Lawrence Orsini

Distinguished
Jun 10, 2013
29
0
18,540
[/quotemsg]

Do you use a proper gaming case? Even with liquid cooling? The liquid cooling is all good for your cpu or gpu but not the rest of your motherboard. I run my fx 8350 locked to a constant 4.4 ghz and I know I could go to around 4.6 with just my air cooling but I won't risk it for the sake of a few hundred 3dmark.
I'm running a cooler master huf case with 5 fans and one high velocity fan on the top with a pull on the back and front configuration with blowing out the top. I get 50 degrees at the most in normal weather in 3dmark cpu stress test.
Maybe you need an infrared thermometer to see how hot your motherboard and components are really. They are cheap on ebay. Seriously that does seem like an abnormally high temperature. Are you using enough thermal paste? I am using a giant cheap gammax 400 air cooling radiator that barely fits in my case, it has leds to boot. It provides good cooling in my pull push configuration. If I was getting over 55 degrees I'd already be alarmed. Taken in about 29 degrees QLD early summer weather temps.

Seriously if that's water a water cooling is doing for you I'm shocked. Or my cpu is sweet ass.[/quotemsg]

Hey there, the rig is actually designed to purposefully run at these temperatures. You can check out the full build on ModsRigs or ASUS ROG forum

Incidentally, it was just featured in September's Popular Mechanics:
abf89654d880.jpg


8e395b540c12.jpg


f090dd088f7f.jpg


567ba6374fc0.jpg


Here is a video of the thing going through its paces:
[video="https://www.youtube.com/watch?v=6Awi_Bwts00&feature=youtu.be"][/video]
 
Status
Not open for further replies.