2nd FX-8120 still running hot, now what?

[BJB]

So I exchanged my first FX-8120 due to it running too hot even at stock settings. I now have a 2nd FX-8120, and it still runs too hot IMO.

I purchased a Corsair H100 a couple days before I got the new processor. It easily cooled my X3 720 at 3.8ghz. So it seems to me the cooler isn't defective.

The first FX-8120 had a vid of 1.2675, and the new one has a vid of 1.4125. Based on what I have read the higher vid processors are supposed to have less leakage, and that seems to be the case for me. The only problem is that both processors hit a 55c socket temp at stock speeds. After 10-15 mins on p95 processors start to throttle down to keep the temps around 55c.

I see some success on this board and others, so it really makes me wonder if something is wrong with my H100 or if I just can't mount it well. I am not a fan of the H100 mounting system. It is easy to get on and off which is nice, but it also seems like the mounting design could be inherently prone to bad mounts. Maybe I am overthinking it, but it seems like it might not be perfectly flush if one side is tightened more than the other side. Originally I tightened both sides simultaneously until both screws bottom out. After that I tried alternating full turns on the screws until it is tight. I seem to get similar results both ways.

I am really frustrated and I don't know where to go from here. I have been building computers since the pentium 133 was around, and I have never experienced this level of frustration trying to cool a processor before. So does anyone have any suggestions besides buying an i5/i7 processor?

One more thing I forgot. The case is open, so no air flow issues, and the ambient temp is 21c.

 

[I.M.O.G.]

Have you tried adding more powerful fans on the H100? I would like to get more airflow on it to see if it has a large effect on temps - if it does, that means you have a good mount, if it doesn't that could mean you have a mounting issue.

Have you checked the TIM to see what your contact looks like?

 

[BJB]

Have you tried adding more powerful fans on the H100? I would like to get more airflow on it to see if it has a large effect on temps - if it does, that means you have a good mount, if it doesn't that could mean you have a mounting issue.

Have you checked the TIM to see what your contact looks like?

I have not tried more powerful fans, but I did try a push/pull config. I had some Yate Loon D12SM fans laying around, so I put those on to see if it helped. I think it dropped temps around 2c with the fan setting on low.

I have taken the waterblock off a few times, and there is always a nice shaped circle, but there does tend to be more TIM toward the top than bottom.

I also just put the stock heatsink on there just to have something to compare to. I am getting around 70c socket temps with the stock heatsink with the bios set to "optimized defaults". So the H100 is running about 15-16c cooler than the stock heatsink with the H100 in push/pull and set to low.

 

[EarthDog]

One thing to note... Hex vs X3... MUCH more heat to dissipate with the Hex. Temps look in the ballpark to me.

 

[I.M.O.G.]

Same idea, but the 8120 is an "eight core"/octo, not a hex. Your point is very valid about it being a whole lot more heat to handle.

@BJB: Thats a good test, not sure if 15C lower than stock is good or not for the H100. Sounds alright to me I think. I would be concerned if the spread on the paste doesn't appear even - if its thicker at the top that may mean its not getting good even pressure.

 

[BJB]

As for the x3 vs x8 comment, I didn't expect to get the same level of cooling. I know the thermals of the processors are different, and the 8120 generates a lot more heat due to a lot more power consumed by the processor. What I wasn't expecting is for the processor to run this hot on water cooling even if it isn't a custom kit. I just can't believe that AMD puts these things out when they get up to 70c with the stock cooler. It reminds me of the p4 days.

On the bright side, I remounted the water block after trying the stock heatsink, and I am getting about 2-3c lower temps than before. My assumption is this time the paste is a little more evenly distributed than last time. So now the H100 is about 18c cooler than the stock fan which seems about right.

Anyone running this thing at 4.6 either has crazy loud fans, a huge rad, or is doing some sort of chilling. I wish Corsair would come out with a pre-built kit with a 360 rad. I had a custom kit in the past, and I just don't have the time for all the tinkering these days.

Are the Scythe GT fans the best a person can do for quiet fans that still perform well?

Thanks for the feedback, but I think I am at the point where this is as good as it is going to get with what I have. I will either have to deal with it or get an 1100T.

 

[ssjwizard]

Sounds like you should invest in some stronger fans. The H100 has pretty dense fins from what I recall. Those low speed yates could be choking out. You should consider the yate mediums or as you pointed out the scythe GTs. I imagine that should make a substantial difference in the cooling. You also mentioned using low speed setting on your H100 have you tried setting it to one of the higher modes.

 

[oldbrave]

Your talking 55c socket temps, I wouldn't worry about socket temps, what are your core temps ??

 

[bassnut]

I have running my8120 running @ 4.6GHZ with 1.42v CPU. Temps with my Antec 920 are still down around 53*C on prime after 12 hours of stress testing this is with the stock fans ussing the custom cooling settings which starts ramping my fans up after coolant temps reach 30*C (I think). Its still a pretty quiet build even when running X264 do the fans rarly get to the point I would call them anoying. The only temps I have seen that really get pretty toasty is the Socket temp which can get up to 65*C on my board Crosshair V mind you I have an insane amount of airflow through my Phantom case to help with board temps. I have tried to push a bit farthur but this seems to be my limit with the cooling I have at the momment.

 

[marjamar]

Kind of in the same boat with high cpu temps. I'm running my 8150 in a CHIVE using the Kuhler 920. I'm pretty much static at 4.876 GHz with core temps at 68 C. during Prime95. Socket temps are through the roof at 78~79 C. peak, which is pretty insane, but still stable.

What I have noticed though is that the thermal transfer rate on this particular water cooler is not good. I see approximately a 40 C. differential between CPU socket temps and coolant temps. I would have to think there's a problem either in the conjunction between the water block/CPU or within the water block and the coolant itself. I've used both the stock TIM and Arctic Silver 5 and although a bit better with AS5, there's still way too great a differential to allow any higher overclocking.

-Rodger

 

[BJB]

Doesn't the processor start to throttle with socket temps so high? Anytime my socket temps get to 60-61c it starts to throttle the CPU until the temps drop down to 55c. It cuts the multiplier down to 7x and the volts down to 0.9. The core temps are low based on what AOD reports, but AOD has to be at least 10c off from an accurate reading. I have seen it go as low as 9c idle, and ambient is around 20-21c. Anyhow, AOD shows 29c at stock settings, and I have got them as high as 50c at 4.6ghz. I may try placing a fan over the socket to see if that will help keep the socket temp cooler.

It seems as though part of the problem is with leakage, because the core temps will hold steady while the socket temp just keeps climbing until it eventually levels off. I assume the board and socket reach some sort of equilibrium with the ambient air and the amount of heat leaking into the socket.

 

[Woomack]

You can disable throttling by disabling APM ( sometimes you can find it in BIOS but for example you won't find it in Gigabyte boards ) or just turn on and next off Turbo Core in AOD. I'm not sure if it's safe for longer work but so far I have no problems running @4.5GHz+ with full cores speed ( F@H, games etc ).

 

[mjw21a]

Heck, couldnt get a stable overclock with Turbo and all that jazz turned on with my 1055T, switch all the automatic overclocking/clock throttling stuff off in the BIOS and you get better results. I know dozer is a different beast but I doubt it'd be that different.

 

[Woomack]

Heck, couldnt get a stable overclock with Turbo and all that jazz turned on with my 1055T, switch all the automatic overclocking/clock throttling stuff off in the BIOS and you get better results. I know dozer is a different beast but I doubt it'd be that different.

It's different mainly because of APM and LLC while overclocking that causes voltage problems or throttling under full load ( even on stock clocks ) that you can't see on Phenoms.
I can enable C1 and other power saving features and I still can set 4.5GHz+ stable. I'm always disabling Turbo in bios.

 

[mjw21a]

Bugger, so now AMD's gone down the path Intel has? You no longer know its overheating because it falls over?

That was always something I liked AMD better for. Made troubleshooting so much easier.

 

[BJB]

You can disable throttling by disabling APM ( sometimes you can find it in BIOS but for example you won't find it in Gigabyte boards ) or just turn on and next off Turbo Core in AOD. I'm not sure if it's safe for longer work but so far I have no problems running @4.5GHz+ with full cores speed ( F@H, games etc ).

I already have that disabled in the BIOS, and I have done the AOD Turbo Core trick many times with no luck. That was all with BIOS 720 though. Asus recently came out with a new BIOS for my board, and I cranked up the processor again today to see if it would throttle itself over 60c. It appears as though the new BIOS has solved the issue.

An update on the socket temps, I placed a fan over the waterblock and saw a 4c drop in socket temp while the core temp stayed relatively the same. I then placed the same fan behind the socket and saw a 6c drop in socket temp and a 1c drop in core temp. I then decided to change my radiator fans from blowing out of the case to blowing into the case to see if the air blowing across the socket area would help without the aid of the fan blowing directly on the socket. I saw no difference in temp doing that which surprises me. I thought there would at least be a slight difference.

To answer someone else's earlier question about increasing the fan speed on the rad. I see around a 5c difference in socket temp and 4c difference in core temp after changing the fan speed setting to medium. The fan noise is borderline intolerable though. It really sucks that Corsair charges $100 for the part that allows you to control the fan speed yourself when Antec includes the ability to do that for free.

 

[Bobnova]

Is this a gigabyte board?
Disable "Core Performance Boost" and set LLC to "low" or "Medium".

Core Performance Boost gives you a bunch of extra vcore, as does LLC on AUTO.
No BD has a vid of 1.4v, they're all in the 1.2xxx end of things.
Try manually setting 1.25vcore.

 

[BJB]

Is this a gigabyte board?
Disable "Core Performance Boost" and set LLC to "low" or "Medium".

Core Performance Boost gives you a bunch of extra vcore, as does LLC on AUTO.
No BD has a vid of 1.4v, they're all in the 1.2xxx end of things.
Try manually setting 1.25vcore.

Thanks for the suggestions, but that isn't the issue. With my first processor I undervolted it until it became unstable, and it still was mid 50's at the socket.

I am using an Asus M5A97 EVO board, and my eyes must have been playing games on me the first time I checked the VID. The VID is 1.3125, reported by Core Temp and AOD. I have seen plenty of BDs over 1.3, and I have seen reviews of BDs as high as 1.4125 (it was an 8150).

 

[WhitehawkEQ]

I'm waiting for the FX-8170, I plan on using the V6GT cooler on it, MB I will be using is the GA-990FXA-UD5.
My current system is Phenom II 1090T on a GA-880GA-UD3H MB with V6GT cooler and I run at stock speed with very nice temps. It also helps with a HAF 932 case





note: the TMPIN1 temp is the socket temp

 

[KalaKava]

Man dude you got some high socket temps. I am running my FX 8120 at stock volts on H100 full max. Since the stupid thing don't adjust the speed of the fans at stock my loads were 40-44C core depending on how many programs I run on each core. Then oc'ed 4.6ghz 46-50C load idle 36C. Then my socket is near 50C because I put back plate to transfer heat. That I took off my Thermal Take heat sink backing mount off Intel cooler and used some thermal tape to backing with some krazy glue to stick to metal part. My volts are 1.42v mobo info and hardware in sig. yea and this thing does not thottle down at all. I even tried to pump a lot of volts into to it to burn it in max I hit is 55C at 1.46v.