Intel QX9650 Drastic Core Temperature differences!

Hi

I am not sure whether this is normal behavior for a Quad Core CPU or if something is wrong, so please pardom my ignorance on this subject as I have just upgraded from a S478 Northwood!

Anyway here is the story, basically I have a QX9650 running on an ASUS P5E3 (non deluxe and no bells or whistles) running (at the moment) @ Stock speed and voltage.

Both Coretemp 0.95.4 and Everest build 4.2.1233 report a huge difference in temperature between cores 1+2 and 3+4

Basically during stresstesting and/or wPrime 1.58 Core temps are as follows (after 20mins of 100% load).

1 = 49C
2 = 50C
3 = 60C
4= 61C

Any ideas as to why there is a huge temperature difference between the cores?

Idle temps are also different

1 = 31C
2 = 30C
3 = 49C
4 = 51C

My motherboard is flashed to the only BIOS which supports the QX9650 (0703) as the shipping BIOS (403) says "Unkown CPU detected" Please update BIOS from ASUS (although system POSTS, it is unstable)

So, is this normal?
If not any suggestsions?

Thanks

John

My suggestion : have a look at cooling fixation

it is either the heatsink seating, the application of thermal paste, or the ihs on the chip.

i think it is a interpolation problem with the DTS output off each core...i have the same problem at idle and maybe a little worse...could be they don't know how to write the bios code to interpret the readings....even the latest beta coretemp 0.95.6 is not reading correctly.

same here, cores are like the following at stock load (@IFX-14) in the first pic attached and on the scythe mugen on the seccond pic.

the heatsink where seated correctly (i have remounted them several times).

maybe the TIM under the heatspreader is the cause of the problem?

here is my 9650 at 4400 mhz for idle and then under wPrime load (1024M bench)

you can see the huge spread in temps for idle conditions but during wPrime 1024M bench they are close enough in my book...it has to be an interpolation problem versus a mounting problem...i am using a TRUE with a strong clamp down.

It is normal even on sub zero cooling. Normal, as in they all seem to do it. Not sure why though. It's probabaly worth one quick check to make sure that you have full coverage though, just to be on the safe side, but they all read waccky like this.

Quote:

---

Originally Posted by xgman

It is normal even on sub zero cooling. Normal, as in they all seem to do it. Not sure why though. It's probabaly worth one quick check to make sure that you have full coverage though, just to be on the safe side, but they all read waccky like this.

---

EXCUSE ME!;)...are you trying to say that i don't know how to mount an air cooler?!:D

Same deal here: 48/46/45/30 @ load :\

http://individual.utoronto.ca/bzhou/...000-1.360v.jpg

coretemp 096 is released...still the same problem:

http://www.xtremesystems.org/forums/...07#post2666507

Quote:

---

Originally Posted by xgman

It is normal even on sub zero cooling. Normal, as in they all seem to do it. Not sure why though. It's probabaly worth one quick check to make sure that you have full coverage though, just to be on the safe side, but they all read waccky like this.

---

I believe the 45nm cpus have some improved power management, that could be it

Quote:

---

Originally Posted by Ace-a-Rue

EXCUSE ME!;)...are you trying to say that i don't know how to mount an air cooler?!:D

---

I'd say break out the guns and inform the terrorists, we'll hit him with everything we got:rofl:

i'll get my pee-shooter out of the closet.:D

nuttin wrong here
(lol just had to show off the nice cool undervolted lappy cpu)

Check the temperatures at stock settings and see if the difference remains.

temps reported by core temp are the same as everest now.

Quote:

---

Originally Posted by Praz

Check the temperatures at stock settings and see if the difference remains.

---

here are default (stock) settings in idle and load conditions (used wPrime 1024M to load cores).

notice core #0 still reads 51c...doesn't even budge under load.

i think you can see if it reads 51 idle and 51 load then that shows either the probe on die is broke, or core temp is just reading it wrong.

contact intel and rma it if it bothers you too much.

you also have core 2 that is way off also.

see if TAT works on that cpu it has an onboard stress test so you can see what it reads

TAT does not work; at least the version that i had (already discarded).

AND...core #0 will register above 51c at 4400 mhz when benching wPrime 1024M.

it does not really bother me but i am inquisitive as to why it is this way...that is what i am trying to find out.

I was same problem my friend, with his Q6600 G0. I lapped cpu and Scythe Ninja too, but problem stayed.:( The first core overheating...
..example 3600Mhz 1,36v on load : 90+/73/72/70. At stock speed was same situation, the first core was lot hotter than others...I never saw one like this before it..


If you can, send back to rma..

it is normal when i wPrime it; as normal as can be expected...it's just not accurate when at idle.

i guess i could put in for a RMA...might get back a better overclocker!;)

At stock/idle all cores appear off unless ambient is that low. Appears as if something is definitely broke.

my shows bad temps too. Cores 0 and 1 never change.

Thanks for the feedback everyone :up:
I checked the mounting of the HSF and all looks well (all for corners are level and the HSF is level, not at an angle).
To me it smacks of 2 things, (having read all the posts), either a BIOS issue (as this processor is relatively new and this is the ONLY BIOS which supports this CPU and even that is not the same version as listed on the ASUS site to support the QX9650) ie I have BIOS 0703 which is the latest version, but ASUS site says that 0704 supports the QX9650. I mailed their Techsupport and they said that the new BIOS was currently in testing.
Idle and Everything Stock
http://img220.imageshack.us/img220/4...etemp96gk2.jpg
The temperature margins do get narrower during wprime or everest stability test, however Cores 3 and 4 are still reading 10 to 15C higher.
I am sure this will be fixed by a BIOS/Microdode update, if not I shall ask for an RMA or Intel's advice (as I don't want a dead CPU at stock volts and stock clocks!!)
I am going to play it safe and try to resist overclocking until I am happy that the temps are more or less making sense (Less than 10C spread between them).
Thanks for the imput.
What motherboards and BIOS revision do you guys who are having the similar problems have?

I have an ASUS P5E3 BIOS revision 0703

John

i tried my current QX9650 on Abit's IP35 Pro and IX38 QuadGT with the same results...supposedly, IP35 Pro's latest beta bios was supposed to accommodate the QX9650...but...it gave the same basic temp readouts.

on my DFI LT X38 which was recently released, it shows the same thing.

i am also leaning towards a need for micro-code adjustment since many of us are experiencing similar conditions.

EDIT:...just found this showing even temps being read which would indicate that the bios micro-code is correct to handle the 9650.

http://www.xtremesystems.org/forums/...&postcount=294

Ace-a-Rue: Core#2 looks like the only one being sensed and calibrated properly in your case.

There is one way to test if your CPU TCC, thermal diodes or this software is working and reporting correctly...

When QX9650 reaches ~84.5℃ it shuts down as a fail safe mechanism (THERMTRIP). :p:

LOL...you want me to melt the core..hehehehehehe

I was just kidding. PROCHOT will be asserted the second you reach ≥64.5℃ which throttles the core down to cool it from nearing the TJunction. ;)

Quote:

---

Originally Posted by Ace-a-Rue

i tried my current QX9650 on Abit's IP35 Pro and IX38 QuadGT with the same results...supposedly, IP35 Pro's latest beta bios was supposed to accommodate the QX9650...but...it gave the same basic temp readouts.

on my DFI LT X38 which was recently released, it shows the same thing.

i am also leaning towards a need for micro-code adjustment since many of us are experiencing similar conditions.

EDIT:...just found this showing even temps being read which would indicate that the bios micro-code is correct to handle the 9650.

http://www.xtremesystems.org/forums/...&postcount=294

---

Hmm interesting stuff Ace-a-Rue

Could still be a potential microcode fix as some batches may have different sensors? maybe?hopefully?
John

ok, just received my replacement 9650...it has a 12/07/2007 pack date...my previous 9650 had a 11/21/2007 pack date...i will put the 9650 in later this evening.

Quote:

---

Originally Posted by Ace-a-Rue

ok, just received my replacement 9650...it has a 12/07/2007 pack date...my previous 9650 had a 11/21/2007 pack date...i will put the 9650 in later this evening.

---

My QX9650 had a pack date of 11/21 to? Hmmmm I wonder :confused:
With anyluck my replacement does not have the same pack date
John

OK...HERE IS THE NEW AND OLD RESULTS.

notice that core #0 is not locked on to 51C any longer under idle or load.

GEEZ...this new processor puts out some heat as i overclock it...i think the cpu has a convex IHS...it is also doing some strange things on overclocking once i go beyond what the voltages can support in mhz...it appears like the PSU is shorting out with short burst of power on and then off in a continous cycling...i am at a little over 4300 mhz and idle temp is 57, 50, 57, 50 from coretemp 096....do i dare to wPrime it to see it go through the roof!:shocked:

i think this new 9650 is a rejected 9770...my PWM temp is almost 10C higher than my old 9650 which had 10-15c lower core temp...i heard about the 9770 had problems with drawing to much current and making it a hot chip.

Is there any way you can measure power consumption Ace-a-Rue?
That's a quick way to tell if your chip is drawing more power hence running hotter than the previous one. Chips do obviously differ, some are better bin than others though sold at the same TDP as I'm sure you know.

Whats your stock idle voltage?

i got it up to 4400 mhz with way less vcore...i am at 1.48v (not sure it is wPrime stable at the 1024M bench)...previous cpu i had to use 1.56v to hold 4400 mhz (wPrime 1024M stable)...but...i couldn't boot at 1.48 on the old cpu at 4400 mhz.

i am not sure how to measure the draw on current...if it requires a meter, that i don't have.

droop will tell you power draw(loadline calibration disabled). droop increases as draw increases to offset PWM spiking.

Quote:

---

Originally Posted by JohnZS

Hi

I am not sure whether this is normal behavior for a Quad Core CPU or if something is wrong, so please pardom my ignorance on this subject as I have just upgraded from a S478 Northwood!

Anyway here is the story, basically I have a QX9650 running on an ASUS P5E3 (non deluxe and no bells or whistles) running (at the moment) @ Stock speed and voltage.

Both Coretemp 0.95.4 and Everest build 4.2.1233 report a huge difference in temperature between cores 1+2 and 3+4

Basically during stresstesting and/or wPrime 1.58 Core temps are as follows (after 20mins of 100% load).

1 = 49C
2 = 50C
3 = 60C
4= 61C

Any ideas as to why there is a huge temperature difference between the cores?

Idle temps are also different

1 = 31C
2 = 30C
3 = 49C
4 = 51C

My motherboard is flashed to the only BIOS which supports the QX9650 (0703) as the shipping BIOS (403) says "Unkown CPU detected" Please update BIOS from ASUS (although system POSTS, it is unstable)

So, is this normal?
If not any suggestsions?

Thanks

John

---

heres my crazy temps :rofl:
cores 1 & 2 never change (bios 903)only 3 and 4 fluctuate

Quote:

---

Originally Posted by KTE

Is there any way you can measure power consumption Ace-a-Rue?

Whats your stock idle voltage?

---

with C1E enabled it sits at 1.07v...when i open wPrime it jumps to 1.24v...i just flashed with a new DFI bios...i'll see what it will allow me to get without some absurd core temp...although, my usage is not demanding so i could accept a higher temp.

i am going to switch the processor over to my abit IP35 Pro board to see what happens there...with my 1st 9650, i had similar results as i did on the DFI X38 board, although with far less hassles.

Interesting results Don. It doesn't really seem like that stuck 51 deg was effecting things though. If I see an asus MB in your sig anytime soo, I'll really know then that you have thrown in the towel. :D

hey allan, i guess my fall back position for cooling is the coolit elite cooler:cool: ...just received it about an hour ago...sitting in the box...i not sure if i can use it outside of the case...i am going to yank the 9650 and install it on the pro...probably use the TRUE to see how it runs on that board with temps.

look how hot this baby runs with only 1.48 vcore...i'll look to see how the thermal paste pattern is when i yank the cooler; to see if the IHS is convex or concave.

Those cpu temps can't be right unless that mount is barely touching. 89deg at only 1.48v? Doesn't sound right.
Anyway there does seem to be alot less consistency from one 9650 to the other than what I would have thought.

Quote:

---

Originally Posted by xgman

Those cpu temps can't be right unless that mount is barely touching. 89deg at only 1.48v? Doesn't sound right.
Anyway there does seem to be alot less consistency from one 9650 to the other than what I would have thought.

---

i'm not sure myself...i will yank the HS to see how the thermal paste is spread...i had to yank one time before since i did not put enough...i am going to have to pull the board out of the case and set it up on my bench to further evaluate...i'm guessing that the IHS is not flat or is close to it.

I'm not sure here. 1.488V/4300 is way higher than stock, and voltage throws your temps high very quickly from stock, so would definitely be a high temp anyway. If it's 89C, I'm not sure but I doubt it. Penryn QX9650 is designed to automatically throttle at 85C (20C below Tjunction_max (105C)).

You can try running a timed benchmark with high and low temps to see if the performance increase (scaling) is as you expected or not. That would give you a good indication if the cores are throttling as soon as they hit ~85C.

I have had 4 E8400 Penryns now looking for the magical 4GHz on silent air.. 2 of them had a 15 degree differential and 2 of them were normal

And of course the only one to be 4GHz prime stable is one of the ones that has the 15 degree difference :mad:

In wonder if these temps are really all over the place per each core, or is it just a defective read error or some sort of Intel errata?

Quote:

---

Originally Posted by Ace-a-Rue

i'm not sure myself...i will yank the HS to see how the thermal paste is spread...i had to yank one time before since i did not put enough...i am going to have to pull the board out of the case and set it up on my bench to further evaluate...i'm guessing that the IHS is not flat or is close to it.

---

IHS looked fairly flat....it's just the cores running in pairs have a large displacement in temps ...the replacement 9650, under idle conditions with the Freezone Elite, cortemp is showing core #0 & 1 at 43C and core #2 & 3 at 36C

the chip I have is having the same issue with 2 of the cores temps being alot lower at idle then the first 2 cores. Should i go ahead and rma this chip??

I'm still waiting on my RMA :(
Hopefully my replacement will behave as they should (all temps less than 5C difference in spread).
There does seem to be something odd going on with the relatively large amount of 45nm users on this forum having temperature reading/diode related issues.....perhaps some Q/C or 45nm issues at Intel?
John

Quote:

---

Originally Posted by joe.denice

the chip I have is having the same issue with 2 of the cores temps being alot lower at idle then the first 2 cores. Should i go ahead and rma this chip??

---

No, 90% of them are like this and I don't think it really hurts anything. You would be wasting your time, unless you have a bad clocker anyway.

Quote:

---

Originally Posted by Ace-a-Rue

IHS looked fairly flat....it's just the cores running in pairs have a large displacement in temps ...the replacement 9650, under idle conditions with the Freezone Elite, cortemp is showing core #0 & 1 at 43C and core #2 & 3 at 36C

---

So what is the Apples to Apples compare on the elite at similar cpu speed and vcore to the Xtreme air cooler you were on before in temps or overclockability?

Quote:

---

Originally Posted by xgman

No, 90% of them are like this and I don't think it really hurts anything. You would be wasting your time, unless you have a bad clocker anyway.

---

Whereas that is probably true, I had stability problems on stock clocks when I used anything which stressed the cores 3 and 4.
So you might want to see if you are stable and if so, as xgman said it might not be worth the gamble as so many people are having these problems
John

Quote:

---

Originally Posted by xgman

So what is the Apples to Apples compare on the elite at similar cpu speed and vcore to the Xtreme air cooler you were on before in temps or overclockability?

---

the elite is NO better than the Ultra Xtreme air cooler (TRUE) when loading all four cores at 100%...NOW, i can get cooler idle temps out of the elite....when i load all four cores up, it matches fairly even with the TRUE...that is based on lapping the TRUE heatsink base....the base on the elite looks good...i am not planning to lap it since it is more cumbersome since the liquid tubing is attached to the heatsink base...i guess that i could take off the permanent clamps but that is something i really do not want to do in case i need warranty on the elite later on.

Is it any quiter?

i'll need to put the side cover on it to determine that.;)...i still have a lot of fans whirling inside the case.:D

Fans certainly do make a noice without a controller and stuff :eek:
What is a TRUE?
I'm still waiting on my RMA, this is the latest from the e-tailer

Quote:

---

Hi,

Thank you for your email.

We can confirm your return has been received back and we are currently awaiting stock to replace.

Please advise if we can be of any further assistance.

Regards

---

The fustrating thing is that they have no official ETA from Intel when they will be getting their stock alocation, it could be anything from days to weeks. (to months even). The Good News is that I will automatically get a refund if I have to wait more than 14 days from when they said my chip was faulty (earlier this week).

I wouldn't be surprised if they get stock next week as rumour has it the 21st is the UK launch of the 45nm Dual Cores.

John

Quote:

---

Originally Posted by JohnZS

Fans certainly do make a noice without a controller and stuff :eek:
What is a TRUE?
I'm still waiting on my RMA, this is the latest from the e-tailer

The fustrating thing is that they have no official ETA from Intel when they will be getting their stock alocation, it could be anything from days to weeks. (to months even). The Good News is that I will automatically get a refund if I have to wait more than 14 days from when they said my chip was faulty (earlier this week).

I wouldn't be surprised if they get stock next week as rumour has it the 21st is the UK launch of the 45nm Dual Cores.

John

---

sorry, john, a TRUE is a "Thermalright Ultra Extreme" heatsink:)

as far as noise from three 120mm fans are minimal, at least for me, since i am a former US Navy and Airline pilot (retired) who listened to fan noise in the co*ckpit non-stop!;)...listening to it right now sort of reminds me of being there right now.:D

the Freezone Elite is a nice unit but not competitive with the TRUE, especially in price!

So that's why an airplane is in your avatar....

How did you like flying anyway? Why did you retire?

Quote:

---

Originally Posted by Falkentyne

So that's why an airplane is in your avatar....

How did you like flying anyway? Why did you retire?

---

LOVED IT!:up:

navy was due to reaching a desk job after you promote up the chain...it is hard to remain in a flying seat once you become a senior officer.

in commercial airlines, you are forced to retire at age 60 during my era...the FAA just recently up the retirement age to 65...can't go back and i wouldn't either.

if you click on my name to the left, i have my airline and navy career highlighted on my home page.

That page is really awesome, Ace. A very storied history in pictures. +1 :)

thank you:)

Yeah there are some ace planes :up:

Hi as Ace already knows I got my replacement CPU back last week. It has a package date of 01/18/08 and is a L740B120T batch
My idle temps are a bit....interesting, however this time the CPU can sustain a large amount of load (100%) and can survive Intel's Cluster Stress Testing which stresses a CPU more than TAT and more than Prime and even more than Orthos!
There is a thread here on how to get this utility from Intel and how to set it up.

My new idle and load temps

http://img132.imageshack.us/img132/8...andloaddy7.jpg

As you can see it gets quite hot, but at least it does not crash out! For those interested it can run up to 3.6Ghz on stock voltage, but requires a slight bump upto 1.38v (1.36CPU-Z) to get 4.0Ghz
Sadly my system cannot run 1600Mhz at any multiplier (even @ 6) due to ASUS not supporting my 1600Mhz RAM. I am stuck @ DDR3 1333 or DDR 1200 if I try to set 1600FSB.

I have made ASUS aware of this issue and am hoping a future BIOS update will rectify the issue. FYI I cannot use their latest 0808 BIOS as that breaks support for my RAM @ 1333Mhz (leaving me stuck @ 1066!)

Anyway I am happy....ish with my replacement, it would have been nice to see slightly lower load temps as I believe 65C is the maximum these CPU's are supposed to get. However it is hard to say what temperature it is really at as I do not have a lot of faith in the DTS/DTS reading software on the new 45nm cores.

John

be advised, intel specifices a Tcase of ~65°C, not a core temp of 65°C, as recommended max.

Tcase is measured in the middel of the IHS not in the core itself. ;)

Ahh thanks Hornet
But that is hard to measure? as most BIOS' have crap temperature readings (well I think my one does anyway if you see the readings it gives above).
John

i dont know if its possible to measure the exact tcase temp without modifing the IHS.

http://download.intel.com/design/pro...x/31873401.pdf

As you can see on page 91 in this document (its for the E8000 series but it doesn't matter since Tcase measurment point is the same since the P4) you can see how the measurment is performend and that you need a carved IHS.

JohnZS; what were your ambient/case temps there?

The difference between core1 and core2 is very abnormal idling.

it seems that is the norm for these 9650's...

Quote:

---

Originally Posted by KTE

JohnZS; what were your ambient/case temps there?

The difference between core1 and core2 is very abnormal idling.

---

The ambient temps are around 20C/70F. I also noticed that one of the cores in particular reads very off the mark @ idle, however it is not quite so bad @ load (at least I can get 100% load this time!)

Sadly I think Ace is right, it seems that abnormal temperature readings are to be expected with this generation of chip. :(

I might be on to something though which may or may not help.....check out this thread Heat Problems on 45nm CPU's

John

Thanks. In the UK, those temps are basically wrong and impossible even with the low ambients we're having. This again shows bugged DTS. For a >$1000 product, you expect much better. :(

Have you tested at what temp your chip shuts down and throttles? It may help...

Quote:

---

Originally Posted by KTE

Thanks. In the UK, those temps are basically wrong and impossible even with the low ambients we're having. This again shows bugged DTS. For a >$1000 product, you expect much better. :(

Have you tested at what temp your chip shuts down and throttles? It may help...

---

my chip shuts down at 4285 mhz over 80C in coretemp

EDIT:...that's running wPrime 1024M stability bench.