My Thermalright Ultra 120 eXtreme

[renozi]

I got my Ultra 120 X on Friday and I thought I did something wrong.
My cpu is lapped and I don't know what I did wrong. ?_?

My Zalman 9500(lapped) w/ AS5 got:
64/64C TAT @ 3.00GHz w/ 1.34 vcore

TR Ultra 120 X (unlapped) w/ Ceramique:
54/53C TAT @ 3.00GHz w/ 1.34 vcore

TR Ultra 120 X (unlapped) w/ AS5:
57/55C TAT @ 3.00GHz w/ 1.34 vcore

TR Ultra 120 X (unlapped) w/ Ceramique:
65/65C TAT @ 3.33GHz w/1.45 vcore

TR Ultra 120 X (unlapped) w/ AS5:
66/66C TAT @ 3.33GHz w/1.45 vcore

Ambient @ 24C, TR Ultra 120 X with Yate Loon D12SL-12, and the Ultra was not mirror finished on the factory-similar to Graysky's

Opinions, comments, questions, answers before I lap this sucker on Monday?

Edit: Temps taken with Core Temp 0.95, SpeedFan 4.32, TAT, and Everest 4

 

[Welshtrog]

Hi My ultra 120 Extreme has a mirror finish when supplied, be very careful when lapping or polishing as at least one forum user has had problems.
see my previous forum report under Thermal extreme.

Good Luck

 

[tylerdustin2008]

If you cpu is lapped why not lap the HSF? That would make sense to me. And those temps might be about* right. Because the e6700 is stock at 2.66, and you oc'ed to 3.33, thats about 600mhz. That is a moderate OC.

 

[graysky]

Well, my system (both q6600 and u-120 x are lapped) gives average temps from x264.exe of [43,44,40,40] in speedfan. I wish I could run coretemp on this machine, since it too has a logging option, but it causes a BSOD/reboot. After about 30 min of 2x orthos (2 cores each) I get the following:

Speedfan (averaged per core over 30 min): [46.9,47.0,45.0,45.0]
TAT (not averaged, I just took the last set of data points): [61,61,60,60]

I would think since you have 2 less cores (2 less heat sources) you'd get better numbers. I'm wondering if you have the base of the HS more or less centered over your CPU. It does seem to make a difference. On my P5B-Del board, you can mount it on there such that the center of the HS base is actually not aligned with the CPU.

The other suggestion is make sure you have good contact. This can involve rotating the HS 90 degrees and bolting it down as such. It makes a difference on my machine (better contact when it's North/South). Finally, make sure you're using enough TIM (not too little and not too much). I was surprised that I needed more than I did after I lapped than before... but the data doesn't lie. I'd recommend you apply it as per the manufacturer's instructions (thin line). To give you an idea, here's a shot of mine before I seated the HS:

Pic

I'd also be willing to bet you 5 bucks that your u-120 x is concave just like mine and about 2 other people who also lapped their u-120 x's. You can see pics of mine from start to finish in this thread to convince yourself that I had a salad bowl of a heat sink base

 

[renozi]

Here's idle temps, ambient was 24C
My Zalman 9500(lapped) w/ AS5 got:
46/46C Idle @ 3.00GHz w/ 1.34 vcore

TR Ultra 120 X (unlapped) w/ Ceramique:
36/36C Idle @ 3.00GHz w/ 1.34 vcore

TR Ultra 120 X (unlapped) w/ AS5:
37/35C Idle @ 3.00GHz w/ 1.34 vcore

TR Ultra 120 X (unlapped) w/ Ceramique:
43/43C Idle @ 3.33GHz w/1.45 vcore

TR Ultra 120 X (unlapped) w/ AS5:
45/43C Idle @ 3.33GHz w/1.45 vcore

It probably is concave, I'll know after I lap it tomorrow once I find my sandpaper. Graysky, from the pictures of your lapping, the remaining nickel around the center of the hs suggests that it's convex rather than concave, because when you started lapping did the center become copper before the outside? If so, I believe it's convex. I could be crazy, though.

 

[graysky]

Oops, right you are: the HS was convex since material came off the center first. The CPU was concave since the sides wore down first. My bad. Let us know how your lapping goes.

 

[GuitarDaddy]

Your temps look about right, you got a 10c decrease from the 9500 which is a good heatsink, can't ask for much more. And they are very similar to mine with an E6400 in a packed case with good airflow under my desk.

You can't compare to other peoples reported temps, way to many variables to consider. Case cooling and placement, other heat generating components, different ambient temps, variance in temp sensors, different temp monitoring progams, different mobos giving different actual volts, different load conditions, and on and on and on

And you certainly can't compare to reviewers(Anandtech) temps because they are testing on a bench with no case and minimal components in a very cool room, with a very cool running cherry picked ES CPU. I belly laughed when I saw their chart, 37c under full load on an E6600 at 3.93ghz, give me a freakin break:laugh:

If your CPU and heatsink are indeed not perfectly flat you will gain a few C by lapping, but don't expect miracles. And don't forget it will get 2-3c better as the compound sets, I did mine about 10 days ago and I noticed just this weekend that my temps have come down a couple of degrees

 

[WoodButcher]

Have good look at your heatsink and chip with a strait-edge before you lap. My 6600 was concave and would have mated nicely with a sink like graysky has posted pics of. (convex before lapping)

 

[renozi]

Thanks for the comments! I'm gonna try rotating the hs 90 degrees so the exhaust points up and change back to Ceramique. Maybe even change the fan to pull because I saw a thread on XS about pull cooling a bit better. Still determined to lap on Monday. Wish me luck!

 

[graysky]

You'll do fine. Just take your time and be careful to keep the water to a minimum and the pressure you exert on the chip to a minimum. Change the paper frequently (say 90 laps or so). Let us know how it goes!

 

[renozi]

update: my hs is CONVEX!

 

[renozi]

well I lapped it today, but the ambient conditions have totally gone out of wack! It's freaking 90F in my room (32C)!! Will resume testing when temps go down.

 

[graysky]

Got any pics of the process? How did it go?

 

[nealh]

Originally posted by: renozi
I got my Ultra 120 X on Friday and I thought I did something wrong.
My cpu is lapped and I don't know what I did wrong. ?_?

My Zalman 9500(lapped) w/ AS5 got:
64/64C TAT @ 3.00GHz w/ 1.34 vcore

TR Ultra 120 X (unlapped) w/ Ceramique:
54/53C TAT @ 3.00GHz w/ 1.34 vcore

TR Ultra 120 X (unlapped) w/ AS5:
57/55C TAT @ 3.00GHz w/ 1.34 vcore

TR Ultra 120 X (unlapped) w/ Ceramique:
65/65C TAT @ 3.33GHz w/1.45 vcore

TR Ultra 120 X (unlapped) w/ AS5:
66/66C TAT @ 3.33GHz w/1.45 vcore

Ambient @ 24C, TR Ultra 120 X with Yate Loon D12SL-12, and the Ultra was not mirror finished on the factory-similar to Graysky's

Opinions, comments, questions, answers before I lap this sucker on Monday?

Edit: Temps taken with Core Temp 0.95, SpeedFan 4.32, TAT, and Everest 4

those temps look ambient...the C2D run warm and heat increases signif. with more vcore

I think you guys are expecting too much

 

[domski]

I know this is a little off-topic, but my Ulta-120 extreme arrived yesterday, and I discovered (at almost exactly the same time the Anandtech retail review was published) that it does not come with the Socket-939 adapter.

If I had known the S939 adapter would not be there, I probably would not have purchased the heatsink. But unfortunately, I pre-ordered it a long time ago, when it appeared that the Ultra-120 extreme would be supplied with the same hardware as the regular Ultra-120.

So anyway, I was wondering if anyone out there might have a Ultra-120, and not be using the S939 adapter. I would be happy to buy it off you for a nominal fee, and pay any postage etc.

Pretty please...

XOX DOM

 

[aigomorla]

Originally posted by: renozi
well I lapped it today, but the ambient conditions have totally gone out of wack! It's freaking 90F in my room (32C)!! Will resume testing when temps go down.

i assume you live in cali??

ahahahaha yeah 90F ambients is BS. My Quad hit 70F for longer then 15min for the first time in her life. And i wasnt happy.

Draw a X on top of your CPU b4 you start. Goto an autostore and pick up sandpaper there. You'll have a much larger selection.

My favorates :

400 if the ihs is really off
700 <--- you can stop it here if you like
1000 <-- starts the mirror finish
2000 Mirror finish

After 1000 i noticed you dont get any big improvement. And at 1000 you'll only see at most a 1-2C difference at MOST. BUt you ge the nice shine:

http://i125.photobucket.com/albums/p73/aigomorla/IMG_0569.jpg

Use vegi oil on the paper all you need is a little drop in the middle. It will help it slide more evenly.

 

[graysky]

Originally posted by: renozi
well I lapped it today, but the ambient conditions have totally gone out of wack! It's freaking 90F in my room (32C)!! Will resume testing when temps go down.

Do you have any results to share?

 

[renozi]

sorry graysky another 95F day today in southern cali! damm global warming!! damm!!

and i didn't bother taking pics when i was lapping, but i can tell you that the heatsink was convex.

i also moved the top exhaust fan to my 3x5.25" empty bays as an intake fan, so now i have a wind tunnel feeding air to the ultra 120 and drawing air out the back.

[]---->[]HS-->[] []=120mm YLs Fans
i know, don't be jealous of my illustration!

Edit: I'm idling at 32C cpu, 44/43C cores at 6x333 with speedstep on, so thats 3.0GHz under load therefore, ambient is too hot to test!

 

[aigomorla]

wow i can already tell thats somewhat of a big improvement.


Expect another hot day tommarrow. 87's >

i had to have my AC on at peak hours. To prevent my cores from loading over 70C.

 

[graysky]

@renozi - looks like you'll have to do your testing after midnight when it cools down, eh?

 

[renozi]

or really really early in the morning! i sleep without a shirt and no blanket!

 

[graysky]

That sucks... no A/C?

 

[renozi]

yea i have AC but things are starting to get cooler in the next few days, so i dont bother turning it on. This summer, however, you bet AC on full blast!

on another note, just to make sure i'm on the same page. the temps i gave are tjunction temps of both cores, that's why there are 2 temps.
however, in the recent issue of Maximum PC one of the editors asked Intel which is the correct temperature reading of the cpu.
"There are two different temperature measurements going on inside a Core 2. Each individual core has its own digital thermal sensor (DTS) that writes a value to a register in the core. The number the DTS reports, however, is not the temperature of the core, it is a value that counts down to zero. When it hits zero, the core should throttle down. It's also intended to be used for controlling fan speeds, not as a direct temperature reference. Core Temp is guessing (incorrectly, according to Intel) what the offset is.
There's also an old-fashioned thermal diode inside the CPU, just off to the side of the two processors, under the lid. Intel says the temperature the diode reports is probably more indicative of the CPU's actual temperature (and it's probably what the BIOS and more OEM motherboard utilities report). Still, that number isn't necessarily correct."

it goes on to say to accurately measure the temp you have to mill a groove in the IHS and put a thermal couple in the exact center of the IHS. so what i got from this is that Core Temp and all the other programs used to report core 0 and core 1 temps are wrong and that tjunction minus ~ 15 is more indicative of actual CPU temp. Hey that only makes me feel better that my CPU idles at 30C and loads at ~50C! ahhaha

 

[graysky]

Hmm.. I dunno man

 

[renozi]

http://www.xtremesystems.org/forums/showthread.php?t=143171

seems right graysky, i'm going with my cpu temps from the diode and not DTS. Currently ambient 28C 3.30GHz @ 1.41 in bios 1.39 vdroop idle 32C load tat 46C

pic