Sempron 3850 Clocks Down at Safe Temperatures

A

Aaron11

[H]ard|Gawd
Joined
Oct 19, 2009
Messages
1,480
Hey guys. I have a cheap HTPC with a Sempron 3850 and a ASRock AM1B-ITX motherboard. Normally, this works fine as I use it mostly just for streaming videos and browsing the web. Occasionally, however, I use it for some light gaming. I've noticed that after about 45 minutes to an hour or so, performance will decrease because the CPU will underclock to 800mhz as if it is getting too hot; it will stay there until basically it idles again.

I noticed that this behavior happens also when I run Prime95. Within about 5 minutes of running the blend test, the task manager shows the CPU is clocked at 0.79 Ghz and it stays there until it goes back on idle; then the cool-n-quiet stuff ill hit back in and adjust the CPU based on load. I checked HW monitor and it reports that the CPU idles at 31 degrees. It also reports that when running prime95, it reaches 47 degrees before clocking down to 800mhz. It's worth noting that the highest temperature was reported by the motherboard to be 59 degrees - still within safe operating limits.

I have tried disabling cool-n-quiet, but this behavior persists. Also, this is at stock settings with no overclock. What could be going on? Do I have something faulty? Thanks for the help!
 
I'm not that familiar with kabini/am1 but could it be throttling due to power consumption rather than temperature?
Are there any bios options to increase tdp?
 
It currently has the latest bios on it and I couldn't find any options pertaining to increasing the TDP. There are actually options available to OC the CPU and adjust voltages, which I have messed with before, but currently it is all stock. I've noticed this behavior for a bit now. Once I did actually under volt the CPU before and that did solve the problem of it clocking down, but I thought it was due to the decrease in heat and not power. I also did not get it to stay stable. Still, clocking down like that doesn't seem like something it should do at stock settings.
 
What sort of temps are you getting? That sounds exactly like thermal throttling. Is is possible that the heatsink is slightly loose?
 
Darakian said:
What sort of temps are you getting? That sounds exactly like thermal throttling. Is is possible that the heatsink is slightly loose?
Click to expand...
HWMonitor reports that the CPU/GPU is getting to 47 degrees. HWMonitor's highest temperature it reports before throttling down is 59 degrees and that is from the motherboard. Core Temp reports the same temperatures.
 
Go into the Catalyst Control Center and set the minimum clock speed to what every you want. When I had a 5350, I set the minimum to 1.8Ghz and it worked fine.
 
It sounds like a heat/temp issue, so I don't see how any bios changes will help you.
Try to increase air flow to keep it running cool. I mean airflow toward and over the heatsink, plus also airflow through the case.
And check to make sure the heatsink pins are all the way through the board like they should be.
I have an Athlon 5150 and the heatsink was no picnic to install. Instructions sucked too so I could understand people not getting it installed correctly.
 
The heatsink appears to be installed correctly. I will admit that airflow in the case is terrible but that still doesn't change the fact that the reported temperatures are not that high.
 
If the VRMs get hot it will do the same thing, although I am not sure that chip could pull enough to heat them. Most lower cost boards do not have temp sensors on VRMs that you can see.
 
I had a similar problem with my Core i5 2500k on a cheap P67 motherboard with inadequate VRMs. Every 15 minutes or so the CPU would throttle to the classic 1.6 GHz idle speed, and then eventually back up to the spec speed for a time. Was not a cooling issue, as the cores never topped 50C! Was just a case of the chipset telling the processor that the VRM power levels had been exceeded.

Wouldn't be surprised if this happens on these cheap-as-chips AM1 motherboards, which probably sacrifice VRM capacity to hit the low price. I used Throttlestop running constantly to stop the throttling on my 2500k - not sure if the program supports AMD chipsets, but if it does it could help you out!

\http://www.techpowerup.com/downloads/2288/throttlestop-6-00/
 
VRMs are about the only major buying criteria I have for a motherboard nowadays. Everything else is garnish.
 
FWIW I had this problem with an Intel Q9550 (C1 power hog) even on stock settings and numerous BIOS revisions.
This was on a cheap P35 board. Only core0 was affected and only by 20% or so, so I ignored it.

After a year I took the board out of the chassis and LOL'd. The VRM area was greenish and seriously warped. It still worked, overclocked and is rock stable, but the only way to keep it from throttling was to launch Throttlestop.

Now I sometimes play pranks on people and use Throttlestop to set a chipset clock modulator.
Hilarity ensues when friends have CPU-Z showing full CPU/MCH/FSB/HTT/DRAM clocks and yet their PCs take their sweet time on 1M prime :D
 
Have you tried manually setting the clockspeed? Setting the voltage yourself to whatever it runs at when its doing stuff at 1.3ghz? You can use load line calibration if your mobo has it as well to keep the vcore from jumping around. I think that would help you with at least knowing if it is having a problem with temps or just faulty cores? Just food for thought.
 
Well I just thought I would update. I ended up puting in a 120mm fan even though the case doesn't offciailly support it and that seemed to do the trick. What I did is I changed where the hard drive wasa mounted and put a fan where a hard drive would normally be. The reported temps are definitely lower. I ran prime95 for an hour and a half along with furmark at the same time to stress the GPU and the CPU never clocked down. This is the case I'm using. It's a very poorly engineered case with terrible airflow. I managed to get it with a bundle including the CPU and mobo for $90 a while back; that's why I got it.
 
I'm glad it really was just a thermal throttle and you managed to fix it.
Still weird it was THAT bad temps wise. However you still don't know which component went over the limit.
Just out of curiosity what were your core voltages? And other thermals, like, from hwmon?
I'm kind of curious what numbers it'd show during a thermal throttle, so if you ever get it again please post a pic ;)
 
You must log in or register to reply here.
Back
Top