Question for fellow XP-120 owners...
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
Question for fellow XP-120 owners...
I'm wondering what you found the base surface on your XP-120 to be like. I recently installed one and found that when I ran my fingernail across the base there was a a marked roughness, and very close visual examination revealed a ridge/valley look to the surface. I am wondering if mine is defective. I am getting quite high temps on idle - 45C - with an FX55 when the MB is only 23C (all temps from Asus Probe). Curiously temps don't rise much at 100% CPU load - up to about 51C in Morrowind with indicated 100% load. I'm using a Nexus 120 at 12 volts as the cooler's fan, and I have tried both airflow configurations (up/down) with no difference in temps. I used Arctic silver 5 paste, following AS instructions, plus I rubbed a tiny amount into the HS base to try and fill in the grooves there.
Could you let me know whether the surface of your Xp-120 was very smooth (if you did a "fingernail test" on it)? And for any FX55 owners out there what are your temps like? I may need to RMA this cooler....
Thanks.
Could you let me know whether the surface of your Xp-120 was very smooth (if you did a "fingernail test" on it)? And for any FX55 owners out there what are your temps like? I may need to RMA this cooler....
Thanks.
-
- Posts: 69
- Joined: Mon Nov 22, 2004 7:43 pm
- Location: Southern California
I ended up lapping mine (which I do with any heatsink I get) and fount it to be pretty concave. With my P4 2.4c OCed to 3.3 with a nexus at 12V, my idle is 31C and full load is barely over 50C. I used AS5 for thermal compound. I think a good lapping would make a noticeable difference. FWIW, I lapped mine right out of the box so Im not sure what I gained over unlapped. There were some machining marks on mine, but it seems that is the case with all thermalright heatsinks so I doubt getting a different one would help much
Mine wasnt as smooth as I was expecting. I was expecting the base to have a protective covering over it or something (I think in a review I read, it mentioned that), but mine had none.
I just applied arctic silver 5 according to the instructions, and let her go. With Ambient temps of ~27C my CPU temp at idle is around 45-47C (Panaflow 120mm at 5V). Under load, it might get close to 60C (My cpu is a P4 2.4c, not overclocked at all). These CPU temperatures are of course read by my Abit IC7 mobo however, so they always will read as a higher temperature than some other manufacturers motherboards.
I just applied arctic silver 5 according to the instructions, and let her go. With Ambient temps of ~27C my CPU temp at idle is around 45-47C (Panaflow 120mm at 5V). Under load, it might get close to 60C (My cpu is a P4 2.4c, not overclocked at all). These CPU temperatures are of course read by my Abit IC7 mobo however, so they always will read as a higher temperature than some other manufacturers motherboards.
Mine certainly wasn't mirror but didn't seem bad in any way. The only significant test is what your temps are compared to your previous hsf. Mine dropped about 5C under load from a TT silent boost lapped to 2000 grit.
How much trouble was lapping the nickle plating, and did you get through it?
Mine didn't have a protector either, probably dropped since the nickle is very hard and corrosion resistant.
How much trouble was lapping the nickle plating, and did you get through it?
Mine didn't have a protector either, probably dropped since the nickle is very hard and corrosion resistant.
-
- SPCR Reviewer
- Posts: 2696
- Joined: Thu Dec 11, 2003 9:53 pm
- Location: Scarsdale, NY
- Contact:
Herb, what mainboard are you using? Have you tried performing a calibration procedure on it to ensure that the CPU temp. report is not off? Mainboards have been seen to be off in either direction by a huge margin; for example, there are ABIT boards that overreport temps by up to 12C.
Your Morrowind test is not the optimal load temp. test, even if it supposedly puts a 100% load on the CPU. Test for maximum heat by using CPUBurn, K7 normal priority mode. You'll get more heat, I guarantee you that.
Judging by the minor difference between idle and 100% load (even factoring Morrowind's lower heat output compared to CPUBurn), I believe your mainboard is overreporting temps, at least a little bit.
45C idle is not good, not good at all, but 51C under load is excellent. However, 51C is not your maximum heat; run CPUBurn for at least 20 minutes to get a better idea how hot it really can get. If it stays under 60C under 30minutes of CPUBurn, there's no reason at all to RMA the cooler or even to be concerned.
Have you tried undervolting at all? Every fraction of a volt counts when it comes to reducing heat output.
Oh yeah, and you've got PM.
-Ed
Your Morrowind test is not the optimal load temp. test, even if it supposedly puts a 100% load on the CPU. Test for maximum heat by using CPUBurn, K7 normal priority mode. You'll get more heat, I guarantee you that.
Judging by the minor difference between idle and 100% load (even factoring Morrowind's lower heat output compared to CPUBurn), I believe your mainboard is overreporting temps, at least a little bit.
45C idle is not good, not good at all, but 51C under load is excellent. However, 51C is not your maximum heat; run CPUBurn for at least 20 minutes to get a better idea how hot it really can get. If it stays under 60C under 30minutes of CPUBurn, there's no reason at all to RMA the cooler or even to be concerned.
Have you tried undervolting at all? Every fraction of a volt counts when it comes to reducing heat output.
Oh yeah, and you've got PM.
-Ed
-
- Posts: 69
- Joined: Mon Nov 22, 2004 7:43 pm
- Location: Southern California
Thanks to all for the advice and sharing of experiences. I will definitely run the CPU burn utility (although I know I will be sweating that one out). I will also run the temperature calibration procedure described on this site. As to the motherboard, which I should have noted, it's the Asus A8V rev.2.
One cursiosity is that on my other system (3Gig P4 OC'ed to 3.3, Asus P4C800E, 7000AlCu, 2 Panaflo L120's running at 7v in Sonata) idle temps are MUCH lower - 33 (vs 45) - but gaming temps come in about the same - 50ish.
One cursiosity is that on my other system (3Gig P4 OC'ed to 3.3, Asus P4C800E, 7000AlCu, 2 Panaflo L120's running at 7v in Sonata) idle temps are MUCH lower - 33 (vs 45) - but gaming temps come in about the same - 50ish.
-
- Posts: 1424
- Joined: Sun Dec 12, 2004 5:00 am
- Location: New York, NY
I have an XP-90 with the same problem
I have an XP-90 that I am lapping. The base is exactly as the Instigator described - there is still a lot of nickel right in the middle with copper at both ends. The base was convex. I did find the original finish a little wavy, but not really rough, and definitely not flat. Actually it looked good. I am lapping for the first time and it is a lot easier than I thought it would be. Before I start sanding I spray a guide coat or light coat of paint to the contact area. As I sand I occasionally look at the guide coat to see where the high and low spots are. Paint remains in the low spots until the surface is flat. This way I don't take any more metal off than needed, and I can adjust my technique until the whole process becomes quick and accurate. Using a guide coat is an automotive painters technique to make sure body panels are block sanded flat. - FG
By the way, I wouldn't lap the cpu. There is no way you can damage the heatsink, but I'm not sure about the cpu.
By the way, I wouldn't lap the cpu. There is no way you can damage the heatsink, but I'm not sure about the cpu.
-
- Posts: 69
- Joined: Mon Nov 22, 2004 7:43 pm
- Location: Southern California
Permanent marker, usually used to write disc titles on recordable optical discs.Herb W. wrote:And what exactly is a sharpie?
http://www.sharpie.com
Use google to find more information about the product and, erm, I guess "culture" is the word.
Great tip about the sharpie. Definitely much less messy than spray paint.
Last edited by lenny on Mon Dec 27, 2004 2:14 pm, edited 2 times in total.
-
- SPCR Reviewer
- Posts: 8636
- Joined: Sat Nov 23, 2002 6:33 am
- Location: Sunny SoCal
-
- Posts: 69
- Joined: Mon Nov 22, 2004 7:43 pm
- Location: Southern California
CPU Temperature calibration results....
Following the CPU temperature calibration procedures given in the SPCR guide, I got some interesting results - the caveat being that I could not test over as wide a range of CPU speeds as might be desireable due to the A8V's inability to either let the FSB go below 200, or to reduce the default 13X multiplier for the FX55 below 12. Ran the CPU at same voltage at 2.4, 2.5, 2.6, and 2.7 GHZ, all with FSB of 200. With these data points I get a calibration correction of -9C, indicating an overreporting of temps by 9C, and giving a true baseline at idle of 36C rather than 45, and a peak temp at default CPU speed of 2600 MHZ during max CPUBurn load of 45C rather than 54. Thanks Ed for the suggestion to try this - it makes me more comfortable about trying some OCing.
Unfortunately there are accuracy issues with this method of calibration - merely adding one degree to the highest temperature of the CPU in the highest speed run reduces that 9 degree correction factor to 1.2 degrees, and since MBM only reports temps in full degrees there is rounding error which means the real difference could be as little as 4.5 degrees. At least the correction is in the right direction.
Unfortunately there are accuracy issues with this method of calibration - merely adding one degree to the highest temperature of the CPU in the highest speed run reduces that 9 degree correction factor to 1.2 degrees, and since MBM only reports temps in full degrees there is rounding error which means the real difference could be as little as 4.5 degrees. At least the correction is in the right direction.
-
- SPCR Reviewer
- Posts: 2696
- Joined: Thu Dec 11, 2003 9:53 pm
- Location: Scarsdale, NY
- Contact:
Remember, too, that XP-120 is a heatpipe cooler. Those heatpipes may not kick in until a certain temperature is reached at the base, which means thermal efficiency of the cooler changes with heat! I forgot to mention that if possible, do the calibration with a static efficiency cooler, such as AMD's oem sink.
-Ed
-Ed
Interesting point Ed - I was not not aware of that property of heatpipe coolers, and that might help explain why the idle temp is high while the load temp is quite good. It might also explain why the measured efficency of the cooler (the C/W ratio, checked during the calibration procedure) was not fully linear - it improved somewhat as the temperature load went up at increasing CPU speeds.
As to trying calibration with the stock cooler, I am reluctant to remove the 120 after having a CPU stick to it during one removal from a DOA MB. The megabucks FX55 was trashed as a result! [Luckily the vendor replaced it.] I'm wondering if it would be safe to use a hair dryer to warm the HS up and so loosen the thermal paste prior to removal. Has anyone tried that? Loading the CPU to heat it, shutting down, and immediately popping off the HS might work but you would have to be very fast....
As to trying calibration with the stock cooler, I am reluctant to remove the 120 after having a CPU stick to it during one removal from a DOA MB. The megabucks FX55 was trashed as a result! [Luckily the vendor replaced it.] I'm wondering if it would be safe to use a hair dryer to warm the HS up and so loosen the thermal paste prior to removal. Has anyone tried that? Loading the CPU to heat it, shutting down, and immediately popping off the HS might work but you would have to be very fast....
A tip for bringing load temperatures down that may apply with your motherboard (check your BIOS): My A8V by default sets the CPU speed and voltage at the BIOS's "auto" setting. When I discovered this and set the voltage manually to the nominal 1.5v (the CPU spec), my reported voltage dropped from the 1.58-1.6 range to about 1.52. At full load with an FX55 this is a gives a reduction of about 8 or so watts in heat dissipation. As a result my T(load) - T(idle) fell 2 degrees.
-
- Posts: 13
- Joined: Wed Oct 08, 2003 11:14 am
You might want to keep something else in mind that I learned when researching Pots that go on the stove. That's right cookware. Many pots are designed to be concave and when they heat up, they flatten out to make a nice flat surface on the burner. I happen to have such a pot and it does indeed flatten out when it heats up.
It is quite possible that the XP120 will also flatten out when heated up, especially considering the amount of heat that many processors are putting out these days.
Another thing to consider, is that the XP120 may be a far better cooler for hotter processors than for processors that put out considerably less heat due to the heatpipe's nature, as mentioned by Edward Ng 4 posts up.
I don't have the resources to do the emperical testing of these cases, but would love to see some results.
It is quite possible that the XP120 will also flatten out when heated up, especially considering the amount of heat that many processors are putting out these days.
Another thing to consider, is that the XP120 may be a far better cooler for hotter processors than for processors that put out considerably less heat due to the heatpipe's nature, as mentioned by Edward Ng 4 posts up.
I don't have the resources to do the emperical testing of these cases, but would love to see some results.
My Pentium 4 Prescott 3.6GHz @ 3.73GHz idles in the mid 30's and highest it's ever been, with my room warmer than it usually is, is 60C with two instances of Prime95 SmallFFT. Most recently with my room at normal temperatures my max was 57C with 2x Prime95 SmallFFT.
Thermalright XP-120 w/ Papst 120mm fan, at 1130rpm even at full load/57-60C.
3.73GHz (14x266MHz), 1066MHz effective FSB, 711MHz memory at 4-4-4-12. Haven't tried overclocking further, I think I'm satisfied with this for now.
Thermalright XP-120 w/ Papst 120mm fan, at 1130rpm even at full load/57-60C.
3.73GHz (14x266MHz), 1066MHz effective FSB, 711MHz memory at 4-4-4-12. Haven't tried overclocking further, I think I'm satisfied with this for now.
-
- Posts: 1424
- Joined: Sun Dec 12, 2004 5:00 am
- Location: New York, NY
google "heat sink lapping"
There are many articles on this, but is no definitive approach. Try this one.
http://www.systemcooling.com/heatsink_lapping-01.html
http://www.systemcooling.com/heatsink_lapping-01.html
-
- Posts: 69
- Joined: Mon Nov 22, 2004 7:43 pm
- Location: Southern California
Id go down to the auto parts store and buy a 3M sandpaper kit for color sanding. It has 400, 600 and 800 grit wet dry paper. I also pick up some 1500 for final sanding. That costs about 10 bucks. You can usually find some glass for cheap at your local hardware store. You only need a 10" x 10" square piece of glass and they will usually sell you some scrap for under $5. Follow the steps in the systemcooling article and you will probably see a difference in your temps. Another benefit is that you will need far less thermal compound after lapping.
-
- Posts: 1424
- Joined: Sun Dec 12, 2004 5:00 am
- Location: New York, NY
ocforums discussion about lapping
If anyone is still interested in lapping take a look at this thread in ocforums w/photos.
http://www.ocforums.com/showthread.php? ... ge=1&pp=30
http://www.ocforums.com/showthread.php? ... ge=1&pp=30