Safe overclocking of a Phenom II 940 BE on a GA-MA790X-UD4P

[KamenG]

This is my first post on Overclockers, so please forgive me if I haven't chosen the most-appropriate forum or seem to be asking too trivial of a question. I did some amount of searching (both the 'Net and this message board) but it is just too much to digest and really hope you will indulge me and offer some expert advice. I am not new to this - my first overclock was an Intel 486DX33 to 40MHz (no fan!) back in the early 1990's but now I am older, busier professional and really have no time to keep up with the onslaught of new PC technologies. Here's where you guys come in!

So, I have built a PC with a GA-MA790X-UD4P motherboard, in a decent but cheap Ultra case with a Ultra 500W PS, a PNY 9500 GT video card, 2x2GB of Patriot Viper DDR2-800 RAM and other standard components. I initially had a AMD X2 7850 with the stock cooler in there an couldn't overclock much (even at 3.0GHz it wasn't perfectly stable) but a couple of weeks ago I jumped on a SlickDeal for a Phenom II 940 BE, and I had an Arctic Cooling Freezer 7 Pro Rev.2 92mm cooler laying around, so I put this all together last weekend. This CPU seems to have no trouble operating at any speed I throw at it, it seems to pass all stress tests (including the Intel Burn-in at max) without crashing but I had to stop all tests as it seems like the CPU is getting very hot. I have used multiple temperature monitoring software programs but it all seems to come down to the sensors as they seem to report the same values. I tend to favor the CPUID Hardware Monitor (HM) as it displays the most data at the same time. But here starts the funny part.

First, while I still had the F7 BIOS, HM would report all 4 core temperatures. After I updated to the current version (F10A), it started showing 0C on all of those! Then, the MB sensors (which even the current version of HM shows as TMPIN0, TMPIN1 and TMPIN2, i.e., no names), show temperatures that seem a little high. For Instance, at idle, shortly after boot up, the coolest (TMPIN0) starts at 35C, while the temperature of the HDD (for an ambient temperature reference) is about 23C. Isn't one of those TMPIN the case temperature? The current settings are V core=1.375V, the RAM is +0.1V, the multiplier is 18x for 3.6GHz and everything else is automatic. If I try running a stress test, the temperatures, as shown by all sensors, go through the roof. While I had BIOS F7 and could see them, the core temperatures during one test went over 75C (at 3.8GHz), the TMPIN were fairly close (some, a little lower). I know the maximum operating temperature for the 940 should be 62C but how do I accurately read that? I've read that the CPU sensors on those CPUs cannot be relied upon but how about the sensors on the MB? I am not about to instrument the CPU for accurate measurements, so how do I read the values I get from software monitors?

I am kind of concerned about longevity - judging by my other computers (I have four other PC from 6-7 and more years ago) - I'd like to use this one for a few years, as well. I have read other people running the 940 at 1.5V core @ over 4GHz, so I'm starting to think those temperatures readings I am seeing can't be right. I could blame the cooler (it's not the best but it's a good one) but, then again, how could the PC be stable if it was really running at 75C? I would appreciate your input, thanks!

Kamen

P.S. I am attaching an image of a test at the current settings (not pushed much) for some objective look at what's going on.
K.

 

[QuietIce]

to OCF!


TEMPIN0 is usually the NB chipset and TEMPIN1 is usually the CPU - not always but usually. As far as I know there isn't an "ambient" temp sensor. Given the Min and Max temps shown for the NB chipset I doubt it's an airflow problem but you might check for that, anyway, if you haven't already. ((Take the case side off and use a room fan to blow air into the case.)) NB chipset temps generally run 40°C (±5°C or so depending on the system) so those actually look normal. And given the Min temp on the CPU, 40°C, it looks like that sensor is also reporting correctly. Some boards (for whatever reason) seem to have a temp offset but the sensor is usually fairly precise - in other words, your 27°C change from idle to load is most likely correct regardless of the temperature's accuracy. Are you sure the CPU fan is running at 100% all the time? Is SmartFan disabled or do you have the CPU fan plugged straight into the PSU (my preferred method)?? It's possible you've got a bad sensor but with the core temp (when you could read it) more or less agreeing I'd say you've got a heatsink problem of some kind - though the case airflow test is worth a shot.


PS
For comparison my 940BE running 3.52 GHz at 1.36v hits ~42°C load core temp and the board's CPU temp pretty much agrees with that. But I'm using a TRUE Black and a Panaflo fan along with an extra 120mm fan directed near the base of the CPU to cool the NB chipset and MOSFETs. Case airflow isn't an issue for me because there is no case ...

 

[trents]

Did I miss something or is there a reason HWMonitor is not reporting core temps? Usually, we see this when people unlock cores.

 

[QuietIce]

I know, it is odd.

I didn't think to ask if ACC or something had been enabled by mistake in that BIOS update ...

 

[KamenG]

Thank you very much for the welcome and for the useful response! I was in a bit of denial about the case airflow issue but it was obvious to me it was a problem, even as I was installing the new cooler, since I had to remove the side-panel vent duct it originally had as the new cooler takes up all the space, up to the side panel. I cant imagine how you could fit anything larger than that. On top of that, the cooler can be installed in only one orientation (and even then it blocks RAM slots 1 & 2) and the fan blows the air down towards the PCI cards (although the PSU fan should be sucking air from the other side). The problem, though is that the air is not routed directly out of the case.

Now, the CPU fan's speed seems to be reported correctly and is changing all the time, it's the fan duty cycle that's reported wrong - always 99%?! I can believe it doesn't go up to 100% (DC), as the fan maximum speed of 2500 RPM is never reached, but it definitely goes lower, as you can see from the minimum value of ~1900 RPM. In the BIOS, I had the defaults - in "Advanced BIOS features", the "AMD K8 cool & quiet control" set to "auto", and in "PC Health Status" the "System Smart FAN Control" is set to "enabled". My PSU doesn't seem to have a cable to plug in the FAN but I can look into it - do you recommend the fan to be at full speed all the time? The fan is very quiet, so I could do that, if it would help. I tried turning those two settings off to get the fan spinning full speed all the time but it didn't seem to have any effect - it seem that Windows is somehow taking over, regardless of the BIOS settings. This kind of reminds me - do I need any special drivers, beyond what comes from the Gigabyte website for the motherboard? Like any AMD CPU or cooler drivers or utilities?

Anyway, so I took off the top panel and set a fan to blow inside the case and below are the results of this test. Indeed, the temperatures stayed quite a bit lower - around 10C lower than before across the board! It seems like I'll have to do some engineering to improve this situation.

Kamen
P.S. Could you, please, also explain what is a "true" black, how it differs from what I have and how you can tell?
K.

 

[KamenG]

ACC is set to:
1) EC firmware selection: hybrid
2) ACC: auto
3) CPU core control: auto

Should I set (1) to "normal"?

Kamen

 

[trents]

A black edition CPU is one where the CPU multiplier can be increased over its stock value. a non-black CPU is one whose multiplier can be lowered but not raised. With a non-black CPU you can only overclock with the HT Reference (aka, fsb). With a black CPU you can use both the CPU multiplier and the fsb which gives you more flexibility and potentially more precise control.

 

[trents]

Well, "Auto" is rather ambiguous. Can you set it to "disabled"? That might give you your core temp readings back which are the critical ones. CPU temps are actually measured from the socket area of the motherboard rather than the CPU die. CPU temps and core temps can vary 10+ degrees C.

 

[trents]

Could you put your system information in your "Sig" so as the thread gets longer we don't have to flip back to an early post to get the info we need. Sig: "Quick Links">"Edit Signature".

Did I understand you to say your PSU fan is taking air from the outside an blowing it downward into the case interior? Are you sure about that? What kind of PSU is this. That's not a good arrangement. I would be tempted to dismantle the PSU (after discharging the capacitors) and reverse that fan.

 

[KamenG]

1) Thank you, trents, my question was about the "true" black that QuietIce was referring to, since my CPU is a black edition, I was wondering what he had in mind - as he seemed to be implying that his is different than mine.

2) I set the ACC to "disabled" and the core temperature readings are back. I'm not sure why the default BIOS settings were "auto". I am not missing anything with ACC disabled (since I have not changed the defaults of -2%), right?

Below are the results of the test, again. Hitting 62C with the case open and ventilated - not good, is it?

Kamen

 

[KamenG]

I wasn't clear - the PSU fan is sucking the air from the top of the CPU cooler and blowing it up and out of the case, it's the CPU cooler that's blowing down from the CPU onto to the video card. What I wanted to say is that, other than the PSU, nothing is getting the air out of the case. I should probably put a case fan on that back wall that you see in the picture.
I updated the signature but it hasn't shown up yet under the posts (probably a matter of time?)

Kamen

 

[QuietIce]

The last ones are easiest so I'll start there. If ACC or Core Control has a [Disable] option I'd use that. Otherwise leave them on Auto and hope the BIOS can figure out they're not needed. Do you remember using some kind of unlock option just after you updated the BIOS? Some boards have a little window that pops up at POST that activates unused cores. None of that stuff is need for the X4's.
Oops! Long post and just caught up!


If the Freezer Pro, which is a 92mm cooler, is hitting the top (side) if the case then the 120mm heatsinks like the TRUE (all caps = ThermalRight Ultra Extreme) won't fit. That will limit your cooling to some extent but your case airflow test seems to have found a big part of the problem so that's some good news! I've been looking around for a good 92mm review list but can't find one. The couple of good lists I know of have all the sinks stuck together and singling out the 92mm sinks is a lot of work. Here's a good list of coolers but you'll have to do some research to find the ones that will fit your case.
http://www.frostytech.com/articleview.cfm?articleid=2521&page=4
(FYI - The Freezer Pro is 127mm tall.)


I don't know what you're looking at to resolve the case airflow problem or what your budget might be but if you're considering replacing your case I'd look for one that's a little wider - generally 8" is enough (barely) to fit a 120mm heatsink ...

 

[trents]

No, temp are not good. Your vcore is only at 1.376 which is not high and shouldn't be responsible for those kind of temps. What is you CPUNB voltage set to?

Even though your HSF is not the latest and greatest state of the art air cooling solution it is a good cooler and should be giving you lower temps than 61 C.

Maybe you already know this but what kind and how the TIM (thermal interface material) is applied can make a huge difference in processor temps.

1. A good quality TIM like Arctic Silver 5 or Arctic Cooling MX2/MX3 can knock 2-3 degrees off temps
2. How it is applied is a huge factor. Most people apply it too thickly with the philosophy that if "a little bit is good, a lot should be better". Too much paste insulates the HS from the CPU. The "dollup" method is the recommended way to apply TIM these days. With a paper coffee filter (they don't leave lint) and isopropyl alcohol, clean the HS base and the CPU face. Apply a dollup of TIM about 3 mm in diameter (BB size) in the very center of the CPU face. Lower the HS onto the CPU, being careful to center it from the start. Rotate it to the left and right once for about 20 degrees of turn. Then clamp her down! It might be a good idea to reapply the TIM and reseat your HS using this method.

 

[KamenG]

Thanks again, QuietIce. I'll keep ACC disabled as there seem to be no disadvantages, as the last test is showing. I did not use any unlock option as the CPU is a 940 X4 and is supposed to have all 4 cores. As for the airflow problem, I'll slow down the CPU for now (what a pity!) and look into getting a case fan first, and then maybe a whole new case with better cooling (I got this one for free with the PSU after rebate - SlickDeals!)

Kamen

 

[QuietIce]

Still catching up.

ACC is built into the Phenom II's so it's not needed at all for your 940BE.

What size is the fan opening at the back? Maybe I can recommend a good fan for it ...

 

[KamenG]

Thanks, trents, my CPUNB is +0.1V (to 1.2V) and the DDR2 are +0.2 (to 2.0V), but the NB is running relatively cool and, most importantly, the core temperatures were running high when all voltages were stock!
I did not touch the TIM that came with the cooler - I figured, Arctic will put there what's right. I have a tube of OCZ Freeze, I could redo it, if that would help?

 

[trents]

If you buy a new case, I strongly recommend getting one that is 8" wide or better. All the best air coolers use a 120mm fan and are too tall to fit in a case less than 8" wide.

Having said that, if case air flow were the main problem I would think taking the side off and blowing a house fan into it would take your core/CPU temps down considerably.

 

[KamenG]

The opening at the back is 100mm.

 

[trents]

Thanks, trents, my CPUNB is +0.1V (to 1.2V) and the DDR2 are +0.2 (to 2.0V), but the NB is running relatively cool and, most importantly, the core temperatures were running high when all voltages were stock!

CPUNB voltage is appropriate. Don't confuse NB (dhipset, think onboard video) with CPUNB, however. Its' easy to do. Bios writers and programmers confuse that issue by using the terms interchangeably.

I did not touch the TIM that came with the cooler - I figured, Arctic will put there what's right. I have a tube of OCZ Freeze, I could redo it, if that would help?

What do you have to lose?

 

[KamenG]

What do you have to lose?

Well, I guess, I could make it worse by not doing it well enough... I'll consider it, though. Come to think of it, that paste on the heat-sink of the cooler had been sitting there for a while, and it didn't even have a foil tab for protection (like other coolers I've seen). It wouldn't be impossible that it was hardened and, thus, ineffective, could it? I think TIM in tubes also has a shelf life but it should be better (I think I got the OCS Freeze 6-7 months ago).

For now I have closed the case and set all voltages back to stock, and the multiplier to 16x for only 3.2GHz (+0.2GHz). After running the test, the core temperatures are at 64C... I think this case definitely needs work.

Kamen