NO LONGER Very dissapointed with the XP-90 !!

Cooling Processors quietly

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
burcakb
Posts: 1443
Joined: Tue Mar 09, 2004 9:05 am
Location: Turkey

NO LONGER Very dissapointed with the XP-90 !!

Post by burcakb » Wed Feb 09, 2005 4:01 pm

I finally got around to installing the XP90. It went into my Thor rig to replace the Zalman 7000B-Cu there running with a very heavily undervolted fan.

I was looking forward to using my "preciousssssssss" Nexus 92mm without cutting it up and maybe putting on another 100 Mhz or so on the clock.

Right now, I'm VERY dissapointed with the results. With everything the way it was (same vcore, clock speed, etc) my CPU temps jumped 15C !!!!!

From the multitudes of installs, reinstalls and yet rereinstalls I made, I've come to the conclusion that the clips are just not applying enough pressure on the CPU.

When I took out the Zalman, all the excess TIM had already been squished aside. I've tried little TIM, lots of TIM, TIM spread, TIM left as glob per AS instructions, etc, every possible combination, (including no TIM !!! :shock: ) I could never get my temps down to Zalman levels. And every time I took the sink off, the TIM was there, not very spread out

Anybody have any ideas on how to properly adjust this thing? Don't advise RMA as the darn thing is pretty difficult to clip into place (inside the case) and a LOT harder to take out and in the multitudes of installs I managed to bend a few fins and left slight marks on the heatpipes - I doubt my retailer will accept it back now.
Last edited by burcakb on Sun Mar 06, 2005 1:40 pm, edited 1 time in total.

ckolivas
Posts: 393
Joined: Thu Jan 27, 2005 7:16 am
Location: Melbourne, Australia
Contact:

Post by ckolivas » Wed Feb 09, 2005 5:19 pm

This is very disappointing as I have read a lot of good things about the XP90/120 :?. Perhaps your airflow direction as a result of fin orientation and case airflow is counter-productive?

The Instigator
Posts: 69
Joined: Mon Nov 22, 2004 7:43 pm
Location: Southern California

Post by The Instigator » Wed Feb 09, 2005 5:37 pm

lap it. its sounds like you have a concave base. my xp-120 was pretty heavily concaved and ive heard of many others having the same problem with their thermalright xp series heatsinks. Dont worry about RMAing it...youll want to lap that one too.

Pauli
Posts: 237
Joined: Thu May 27, 2004 10:10 am
Location: California, USA

Post by Pauli » Wed Feb 09, 2005 6:13 pm

I agree with The Instigator -- it has to be caused by the mating surfaces between the CPU and XP-90. I replaced my Zalman with an XP-90 and have gotten temperature decreases of about 2-3C at idle and 5-6C at load. There's no other way to explain a 15C increase with this exchange.

ckolivas
Posts: 393
Joined: Thu Jan 27, 2005 7:16 am
Location: Melbourne, Australia
Contact:

Post by ckolivas » Wed Feb 09, 2005 6:16 pm

I've heard this now a couple of times on the forum. Is their (Thermalright's) quality control that much lower than Zalman? :(

The Instigator
Posts: 69
Joined: Mon Nov 22, 2004 7:43 pm
Location: Southern California

Post by The Instigator » Wed Feb 09, 2005 6:58 pm

I think its more of a manufacturing issue than QC. The bases are machined perfectly flat initially, but the process of soldering on the heatpipes and nickel plating warps and tweaks the base. After they are finished, it would be near impossible to machine lap them again without costing us a lot more money in the end. In most cases the base isnt deformed enough to make a significant difference so they are OK with that, but there will be some that are bad enough to cause the problems you are seeing.

burcakb
Posts: 1443
Joined: Tue Mar 09, 2004 9:05 am
Location: Turkey

Post by burcakb » Wed Feb 09, 2005 11:55 pm

Ugh,

Yeah, the TR might be concave. It's usually the CPU that is and I had checked that mine wasn't. Never thought about the XP90 :? The bottom nickel plating is near-mirror though?

As for quality, I was slightly dissapointed. The retention bracket looks crude. The bracket that came with the Abit board is MUCH higher quality.

And be warned - installing/uninstalling this beast INSIDE your case is VERY hard.

Now, to find some nice way of lapping.... I'm open to suggestions. I don't have lapping kits sold here in Turkey and frankly importing one to fix an expensive cooler is just plain wrong.

The Instigator
Posts: 69
Joined: Mon Nov 22, 2004 7:43 pm
Location: Southern California

Post by The Instigator » Thu Feb 10, 2005 12:09 am

I bought a $10 3M wet sanding kit at my local auto parts store with paper from 250 to 800 grit. I also bought 2 sheets of 1500 for final polish. I got some glass from my local hardware store who had some scraps lying around for a couple bucks. Less than $20 total and works as well as any "lapping kit" you may find for sale online.

And just because the base is near mirror, doesnt mean its flat. The plating process alone has the possibility to warp the base. The plating can even vary in thickness across the base making an uneven surface.

meglamaniac
Posts: 380
Joined: Thu Jul 15, 2004 12:44 pm
Location: UK

Post by meglamaniac » Thu Feb 10, 2005 2:49 am

I too have heard these warnings of the XP90/120 not being terribly flat.
While I haven't gotten around to installing mine yet, I did take the precaution of getting a lapping kit to give it a good going over with first.

If you're in the UK, Kustom PCs provide a couple of excellent options (scroll down) that come with complete instructions for even the most cautious of people :)

TheWesson
Posts: 101
Joined: Tue May 25, 2004 9:41 pm

Post by TheWesson » Thu Feb 10, 2005 1:42 pm

even if the base is pretty flat and temps are good, taking off the nickel can give you a little temp drop - 1 or 2C. That was my experience.

just as a side comment: there are those who will point out that you can't make a base as flat by hand as you can by machine. In fact, after I lapped mine, I checked it against a straightedge and it seemed to be convex by a small fraction of a millimeter at the edges.

but
a) better convex than concave (for contact in center)
b) that degree of convexity means maybe 0.01 mm gap at the edges of the core area which is apparently trivial enough to harm my temps very little.

The Instigator
Posts: 69
Joined: Mon Nov 22, 2004 7:43 pm
Location: Southern California

Post by The Instigator » Thu Feb 10, 2005 6:59 pm

I find that you will always end up with beveled edges after a good lapping, but the edges dont contact the cpu so no biggie right?

Tiamat
Posts: 84
Joined: Thu Dec 02, 2004 12:27 pm
Location: Washington DC

Post by Tiamat » Thu Feb 10, 2005 7:46 pm

When I modified an old slot Pentium heatsink (those black heatsinks from years back) for use as South Bridge cooler, it was enough for me to "Kiss Grind" the surface down. I dunno how accurate Kiss Grinding is compared to lapping. The grinder has accuracy to 0.003". Is this something that you might be able to try, or is .003" not good enough for your purposes? This was done in a machine shop. Im not sure what the machine was called. Its as big as a small milling machine and has a suspended spinning grind wheel that is changeable (to different coarseness). The controls are similar to a milling machine in that you move around the vice on x and y coordinates.

burcakb
Posts: 1443
Joined: Tue Mar 09, 2004 9:05 am
Location: Turkey

Post by burcakb » Sun Mar 06, 2005 1:46 pm

I suppose I owe an apology to Thermalright.

It seems removing the CPU from the socket messes up the CPU sensor diode on my motherboard. The mobo had an initial temp reading problem that was later solved by a BIOS update. But I never thought removing the CPU from the socket would render it FUBAR.

I had the XP-90 on, despite temps in the 68C level under load which I knew to be incorrect. The heatsink was getting very hot with the fan turned off, heatpipes were working and a fan calibration reported +22C overreporting. However, at 68C, uGuru wanted to beep loudly, turn the CPU fan faster and all that sort of noisy things. I kept disabling them but was still very uncomfortable.

A few days ago, Abit released another BIOS and voila!!! Temps are down to 42C level, I even told the Radeon to pay a visit to hell and overclocked all the way to 2400 GHz without breaking 50C (calibrated). The thermalright is definitely a winner !!!

A big sorry to Thermalright and a big BOO to abit

meglamaniac
Posts: 380
Joined: Thu Jul 15, 2004 12:44 pm
Location: UK

Post by meglamaniac » Sun Mar 06, 2005 1:53 pm

Wow, thats some BIOS issue!
Congrats on finding the root of the problem, and good to see it's not TR's fault.

I did lap my XP-120 anyway, and will be installing it shortly so we'll just have to see how that works out...

ckolivas
Posts: 393
Joined: Thu Jan 27, 2005 7:16 am
Location: Melbourne, Australia
Contact:

Post by ckolivas » Sun Mar 06, 2005 1:56 pm

Thanks for clearing that up!

Post Reply