Temperature problems with ASUS Rampage Extreme and Intel Q9550

[svrzix]

Greetings

A newly built system has the following components:

ASUS Rampage Extreme
Intel Q9550 2.83 GHZ (45 nm)
4GB OCZ Memory
Intel stock cooler
Two rear exhaust 80 mm case fans, 1800 rpm
Seasonic M12 II 500 W PSU (The PSU fan points towards the CPU but is partially blocked by the video card)

The case is currently cover less. No OS has been installed. The system boots just fine and initially, the BIOS shows a CPU temperature of 49 C. As time progresses, this temperature rises to 62-63 C and stays around these numbers. MB temperature stays steady at 39 C and NB is also nearly constant at 66 C.

Presently the system utilizes Intel's stock cooler. This cooler has been installed, uninstalled and reinstalled several times. Arctic Silver 5 has been applied using two different methods, once according to the instructions posted on its website which is to apply one line across the middle of CPU and another using a plastic spreader to evenly distribute a small amount of AS across the CPU surface. The latter has been the most effective.

Most everyone would probably agree that these numbers are unusually high. Since the case cover has not been installed, there should be no ventilation issues. What is puzzling is the gradual rise in CPU temperature when CPU is idle.

Any help or hint is greatly appreciated.

Regards

 

[sgrinavi]

If you have the newest BIOS, everything is seated and installed properly then Try setting the CPU voltage manually in the BIOS, my ASUS was overvolting the crap out of my E8600 causing overheating issues too....

 

[svrzix]

Thanks for the response, sgrinavi.

Rampage Extreme has a voltage light near the I/O panel. The light is green which means that the voltage is normal. I also checked the BIOS voltage values and they are also in the normal range.

I also flashed the newest BIOS. Initially the temperature seemed to be steady at 48 C but after a reboot, it ramped up to 58 F.

I touched the base of the heatsink and it is barely warm. If the heatsink was not making contact with the CPU, then more than likely the system would have shut down. Hence I wonder if the BIOS readings are accurate.

I think I will go ahead and install the OS then proceed to measure the temps with an app such as RealTemp to see if it reporst the same findings.

Thanks again.

 

[BonzaiDuck]

What sort of chipset is in that Rampage puppy?

I had an E8400 in a Striker Extreme nvidia-680i board until 10 days ago, when I apparently burned out the Northbridge.

Then I replaced it with an eVGA 780i.

Both of these boards lagged behind the new Penryn chip releases in several respects. They had been touted as being Penryn-compatible -- the Striker limited to the Wolfdale dual cores. And BIOS versions from late-summer 2007 would post with the new cores, but three BIOS revisions later -- one or two of which wouldn't recognize the exact identity of the CPU, or they would post a message "Update BIOS to take advantage of new CPU features."

I've said this many times, but the issue keeps coming up. Look for an article on the E8500 among Anandtech articles for March, 2008, and read -- or at least scan -- the material on a revised approach to thermal sensoring.

The 680i wouldn't show "normal" temperatures for an E8400 until a BIOS release in early August this year -- last month.

The 780i board reports a TCase temperature that is high but in the ballpark, and it fluctuates normally, showing an idle-to-load temperature spread consistent with my heatpipe cooler, lapping and thermal paste. (But the idle and therefore load values both seem about 5C or more higher than they should be.)

The core temperatures with the 780i are stuck at 56C/56C -- and only increase slightly under load.

The other gripe I have with the July 08 v.P07 780i BIOS is the fan-control features: THEY DON'T . . . . FREAKING . . . . WORK!! This was similar to a problem with the STriker 680i early-on in late 2006 early 2007, and BIOS updates fixed it. Since I depend on the fan-control features and prefer NOT to get a front-panel automatic thermal controller with sensors or software using a USB connection for $40, I'm freaking sitting here, listening to what seems like the wind-tunnel at the Lockheed Skunkworks at Edwards AFB. :frown:

EDIT: If you check INtel's web-site as you drill down looking for processor specs, there is a FAQ somewhere about "Why is my processor running so hot?"

The Intel gurus didn't think to explain it more succinctly, but item #1 on their list of three or four "fixes" was "Update the BIOS."

Fine -- if the mobo maker and BIOS programmer have addressed the problems . . . . ah, duhhhh!!

 

[sgrinavi]

I too had good voltage readings in my bios, but nevertheless it was overvolting the CPU causing heat problems. I learned my lesson, with a new CPU series you need to go manual for your voltages.

 

[svrzix]

Originally posted by: BonzaiDuck
What sort of chipset is in that Rampage puppy?

Hello Bonzai

It is an X48 based mobo.

EDIT: If you check INtel's web-site as you drill down looking for processor specs, there is a FAQ somewhere about "Why is my processor running so hot?"

The Intel gurus didn't think to explain it more succinctly, but item #1 on their list of three or four "fixes" was "Update the BIOS."

Fine -- if the mobo maker and BIOS programmer have addressed the problems . . . . ah, duhhhh!!

Well, I updated the BIOS with the latest version that was downloaded from ASUS' website. Unfortunately the problem still persists.

Thank you for the reply.

Regards

 

[svrzix]

Originally posted by: sgrinavi
I too had good voltage readings in my bios, but nevertheless it was overvolting the CPU causing heat problems. I learned my lesson, with a new CPU series you need to go manual for your voltages.

Hello sgrinavi

Thanks again for the response. There are two voltage values in the BIOS: 1) CPU PLL Voltage and CPU Voltage. I take it both of these should be manually set.

Realtemp shows values of 48,50,56 and 56 for each respective core at idle.

I wonder if the CPU is damaged in some way.

Thanks again.

 

[BonzaiDuck]

Originally posted by: svrzix

Originally posted by: sgrinavi
I too had good voltage readings in my bios, but nevertheless it was overvolting the CPU causing heat problems. I learned my lesson, with a new CPU series you need to go manual for your voltages.

Hello sgrinavi

Thanks again for the response. There are two voltage values in the BIOS: 1) CPU PLL Voltage and CPU Voltage. I take it both of these should be manually set.

Realtemp shows values of 48,50,56 and 56 for each respective core at idle.

I wonder if the CPU is damaged in some way.

Thanks again.

I think -- check me -- that they released that board in fall, 2007? Then Penryn cores were released this January. It's a little late, and ASUS even "fixed" my problem with the E8400 for my Striker 680i board -- in August '08.

See, on the one hand, it's possible that the sensors are "damaged" or "stuck," and Intel themselves said the sensors "shouldn't be used to assess actual temperatures" for building thermally advantaged chassis'. On the other hand, according to that March Anandtech article, they put some effort into revising thermal monitoring for these chips -- both Wolfdale and Yorkfield.

For me, I STILL have to wait for a new BIOS beyond the July 08 release with my 780i board -- hoping they can fix both the thermal monitoring AND the fan control features.

 

[error8]

Oh come on, forget about the bios temperature readings. Just use realtemp or latest coretemp and read the temperature in windows under full load. If you have 100 C then there is a problem indeed. Who knows where Asus puts those sensors...

 

[sgrinavi]

Originally posted by: error8
Oh come on, forget about the bios temperature readings. Just use realtemp or latest coretemp and read the temperature in windows under full load. If you have 100 C then there is a problem indeed. Who knows where Asus puts those sensors...

Umm.. huh:? I think he did.

Realtemp shows values of 48,50,56 and 56 for each respective core at idle.

 

[BoboKatt]

Did you manage to install the OS to thus further test temps using RealTemp?

I recently dropped a new Q9550 into my old and trusty P35 Asus P5KE-wifi. I am using a Tunic Tower, which is certainly better than stock intel cooler but not the best out there. I was so shocked and still am shocked to see that the CPU at stock voltage of 1.185 will run just under 3.6 Ghz and my idle temps in Real Temp after about 10 min are low 30's. I have seen a max of 43 while playing games. Anyhow certainly in your case it has to be the chipset and the new 45's that would cause that.

No way that cpu runs that hot at stock... if your actual stock is manually entered. I mean run Coretemp, take your VID that you see and slap that in your BIOS.

These new CPU's run so cool -- my other q6600 G0 on my 680i with a TRUE for pete's sake runs at low 40's in Realtemp at idle (oc'ed to 3.2)... and get right up there in the 50's after playing for a while. So these new Q9550's (and I can't imagine the new E0 9660's) run much cooler.

 

[sgrinavi]

Originally posted by: BonzaiDuck


I think -- check me -- that they released that board in fall, 2007? Then Penryn cores were released this January. It's a little late, and ASUS even "fixed" my problem with the E8400 for my Striker 680i board -- in August '08.

See, on the one hand, it's possible that the sensors are "damaged" or "stuck," and Intel themselves said the sensors "shouldn't be used to assess actual temperatures" for building thermally advantaged chassis'. On the other hand, according to that March Anandtech article, they put some effort into revising thermal monitoring for these chips -- both Wolfdale and Yorkfield.

For me, I STILL have to wait for a new BIOS beyond the July 08 release with my 780i board -- hoping they can fix both the thermal monitoring AND the fan control features.

The CPU is new not the chipset. I had a similar experience with a new CPU on the same board; maybe it's those pesky revised sensors that Intel is working on?

 

[svrzix]

Originally posted by: BoboKatt
Did you manage to install the OS to thus further test temps using RealTemp?

I recently dropped a new Q9550 into my old and trusty P35 Asus P5KE-wifi. I am using a Tunic Tower, which is certainly better than stock intel cooler but not the best out there. I was so shocked and still am shocked to see that the CPU at stock voltage of 1.185 will run just under 3.6 Ghz and my idle temps in Real Temp after about 10 min are low 30's. I have seen a max of 43 while playing games. Anyhow certainly in your case it has to be the chipset and the new 45's that would cause that.

No way that cpu runs that hot at stock... if your actual stock is manually entered. I mean run Coretemp, take your VID that you see and slap that in your BIOS.

These new CPU's run so cool -- my other q6600 G0 on my 680i with a TRUE for pete's sake runs at low 40's in Realtemp at idle (oc'ed to 3.2)... and get right up there in the 50's after playing for a while. So these new Q9550's (and I can't imagine the new E0 9660's) run much cooler.

Hello Bob

I had some problem with my case hence I decide to 'bread board' my build until the new case arrives.

I also replaced the stock cooler with a Zalman CNPS 9700 LED. Using RealTemp, the core temperatures are now 39, 40, 46 and 46 respectively. This is still quite high considering the mobo is sitting outside of the case. Granted that since I live in the Phoenix area where currently the outside and inside temperatures are 100 F and 78 F respectively, the temperatures of cores should be somewhat higher. However I was expecting much more of a better cooling performance from a 45 nm processor.


I will take your advice, that of sgrinavi's and enter the values manually in the BIOS. I meant to take sgrinavi's suggestion sooner but I was distracted with the case difficulties.

To add my woes, now the LCD Poster won't display messages although its back light is on.

If I may ask you gentlemen one more question: I have also purchased XIGMATEK HDT-S1283 and am wondering would this offer any significant advantage over Zalman's cooler?

According to this chart:

http://www.frostytech.com/arti...?articleid=2233&page=5

The difference seems somewhat negligible.

Thanks in advance.

 

[BonzaiDuck]

Originally posted by: svrzix

Originally posted by: BoboKatt
Did you manage to install the OS to thus further test temps using RealTemp?

I recently dropped a new Q9550 into my old and trusty P35 Asus P5KE-wifi. I am using a Tunic Tower, which is certainly better than stock intel cooler but not the best out there. I was so shocked and still am shocked to see that the CPU at stock voltage of 1.185 will run just under 3.6 Ghz and my idle temps in Real Temp after about 10 min are low 30's. I have seen a max of 43 while playing games. Anyhow certainly in your case it has to be the chipset and the new 45's that would cause that.

No way that cpu runs that hot at stock... if your actual stock is manually entered. I mean run Coretemp, take your VID that you see and slap that in your BIOS.

These new CPU's run so cool -- my other q6600 G0 on my 680i with a TRUE for pete's sake runs at low 40's in Realtemp at idle (oc'ed to 3.2)... and get right up there in the 50's after playing for a while. So these new Q9550's (and I can't imagine the new E0 9660's) run much cooler.

Hello Bob

I had some problem with my case hence I decide to 'bread board' my build until the new case arrives.

I also replaced the stock cooler with a Zalman CNPS 9700 LED. Using RealTemp, the core temperatures are now 39, 40, 46 and 46 respectively. This is still quite high considering the mobo is sitting outside of the case. Granted that since I live in the Phoenix area where currently the outside and inside temperatures are 100 F and 78 F respectively, the temperatures of cores should be somewhat higher. However I was expecting much more of a better cooling performance from a 45 nm processor.


I will take your advice, that of sgrinavi's and enter the values manually in the BIOS. I meant to take sgrinavi's suggestion sooner but I was distracted with the case difficulties.

To add my woes, now the LCD Poster won't display messages although its back light is on.

If I may ask you gentlemen one more question: I have also purchased XIGMATEK HDT-S1283 and am wondering would this offer any significant advantage over Zalman's cooler?

According to this chart:

http://www.frostytech.com/arti...?articleid=2233&page=5

The difference seems somewhat negligible.

Thanks in advance.

ANSWER: YES. Look for Anandtech's accumulating "cumulative comparison heatsink reviews" between May 5, 2007 and March of this year.

Not sure about the quad-versions, but the dual core Conroes and Penryns are rated at a TDP load thermal wattage of 65W. Over clocking them will push them toward 100W. But we are sure of two things -- not necessarily applicable to your situation, but worth thinking about:

1) OEM/non_Intel BIOS revisions (all chipsets) have lagged behind the various Penryn releases and steppings, with older BIOS' giving ridiculously low AND high BIOS TCASE readings.

2) Several people, including me, have reported "stuck" core sensors, while there appears to be a disclaimer coming from Intel that the tJunction Core sensors are not meant to read "idle" temperatures.

3) At the voltages required by the 45nm CPUs, real empirically measurable temperatures should be as low or lower than for Conroe 65nm CPUs. I'm running a B3-stepping Q6600 which shows idle cores around 38 to 40C at room-ambient 75+F, and the load temperatures will reach 64 to 66C when the room temperature is around 80F. The processor is volted to 1.42V, and I have a custom air-cooling approach with a TRUE cooler. I've over-clocked it between 20 and 25%. The voltage component (squared) contributes more to temperature than anything.

Therefore, the REAL temperatures on a Q9550 quad core should be lower, even at the same percentage over-clock, and the culprit would be a combination of these:

Stock cooler or "good-but-not-top-end" cooler

Sub-optimal case airflow [which wouldn't apply to the OP since his case is open]

Stuck core sensors

Bias in reading the core sensors

You will probably find BIOS-bias in reading both the core sensors and the TCASE temperature. We have placed too much stock in the core sensors anyway, because the thermal limit is defined in terms of TCASE -- not the core values. And although people may have yawned at recent posts I've made, nVidia's software-design folks screwed up with their new "nVidia Tools" that replaces nTune and nVidia Monitor. The old versions read and reported TCASE; the new version only reads and reports the core temperatures. This leaves us with a catch-22 in reference to the "ESA" feature of nVidia mainboard design -- where you could otherwise set fans plugged into devices like the Silverstone Commander so they respond to either TCASE or core temperatures. "No TCASE -- no Buck Rogers."

So I also suspect that Intel has played a bad trick on the enthusiast community. Where do they get most of their revenues? Not from us. They get it from mainstream, business and government users -- and a majority of people who wouldn't know Adam from over-clocking and buy their machines ready-to-go from OEMs.

Someone else at this or another forum told the story of Intel saying that a 17C or higher SPREAD between core values would be sufficient for replacment under warranty, yet RealTemp's author quotes them as saying " . . . . not meant to read idle temperatures."

 

[svrzix]

Thank you Bonzai for the most detailed reply.

Well, I put all of the components inside ThermalTake's Tsunami which has fans on the front, side and rear. I used Xigmatek's HDT-S1283 cooler along with AC5.

I changed the CPU Voltage (VCore) to 1.1 in the BIOS.

RealTemp had the idle temperatures at 44,40,48 and 48. After 10 minutes of Prime95, the temperatures climbed to 61,58,62 and 61.

Xigie was the third cooler that I used which provided better results than the stock cooler and Zalman CNPS 9700 LED. I used Xigmatek's crossbow which attaches much more securely to the CPU than the default snap on screws.

RealTemp's sensor test shows a maximum sensor movement of +3 between the cores which supposedly mean that the sensors are not stuck.

Thank you all for your inputs.

 

[svrzix]

Update: I have reinstalled the cooler and applied AC5 as suggested by this article:

http://benchmarkreviews.com/in...1&limit=1&limitstart=5

[Last picture]

The temperatures have now dropped by 5 C. Hopefully as AC5 cures over time, temperatures will decline further.