Printable View
I have an off question for Tasty, where does magesium lis an the thremal chart in relation to copper and diamond?
The reason i'm asking is I have access to it and was considering making a hs of it.
I'll have temps some time tomorrow or friday. I've been watching the temps on the Opteron all day under load and idle. I'll change the mount inthe morning and watch it all day.
Here's a list of the elements sorted by thermal conductivity, worst to best:
http://environmentalchemistry.com/yo...c/thermal.html
Diamond, while being carbon, is a special matrix thereof. Magnesium is actually not that great. Sorry.
I don't always know as much as I think I know.Quote:
Originally Posted by SiGfever
When I test in the lab on a test I am always at optimmum 90-100 psi. and what I see is a drop of 1-2C after two hours then maybe 1/2 C overnight then stabilizes.
I tried it on a few after market sinks with loadings in the 40-60lb range in systems with similar result and did not think it would be an issue. Recent Beta testing giveaways some people report back 48- 72 hrs, week later and announce another 2C-3 drop. So the time component probably is related with pressure and heat and varies some depending on the combined factors
From memory I believe that Intel recommends something like 43lbs +/-10lbs with a max board loading of around 70lbs. I would assume
from the Intel spec much under 30lbs and you will would not have enough to flatten the IHS, Actually having just filed a cut for a thermocouple recently I am surprised 40LBS does it.
I have run some cursory tests with glass slides and some uncalibrated binder clips. Cold (room temp) and maybe 15- 20lbs perclip typical paste takes about 1 to 2 min. to reach a pretty good spread with 2 clips about 30 sec with four. I took a mic to the glass and checked spread. With ICD7 maybe 5 min. with the 2clip and about a min with the 4 clip configuration. @ 50C ICD7 flows well.
So the scenario should be some spread when cold and a more rapid spread as it heats up balancing out or seeking the bondline thickness with adequate pressure. Colder systems may take a little longer.
To answer your question fire it and see what happens. if it seems high let it set for a day or so and see if trends downwards.
With MX-2 I see the best results after one week of cure time under load. I will give your product the same amount of time before inspecting my mount. I can not test PSI of the mount but it is close to a "Hard Mount" with the springs almost fully compressed.Quote:
Originally Posted by tastymannatees
I could not ask for any thing more. I am getting high quality feedback here, initially it is the best so far.Quote:
Originally Posted by SiGfever
PSI is beyond the scope of this effort, It is one of those things that turn up in real world testing and I am speculating as to cause and could very well be something else. Although I might try and get specs from the manufacturers and/or buy a sample of good and marginal result sinks as/if patterns develop and test them myself.
One thing to keep in mind is that a majority of the guys here use ThermalRight Ultra Extreme(TRUE) heatsinks and to be blunt, the mounts suck but the HS itself is the best available right now.Quote:
Originally Posted by tastymannatees
Mine when "on the stops" still is relatively easy to move and thats with 3/16" of washers underneath the mount in the center so the pressure on the IHS that many will have here will be lower than what you would expect.
Okay, Here are some of the results i got so far after ~24 hours of testing with ALOT of pictures.
My setup:
http://i31.photobucket.com/albums/c3...d/IMG_0264.jpg
Q6600 @ 3.9ghz with 1.6v
Asus P5k Deluxe
Watercooled with D-tek fuzion, PA120.2 etc
Stock mounting kit with TRUE bracket
****Each mount the screws were tighten ALL the way down to make sure the TIM was spread.
BEFORE WITH AS5:
TIM: AS5
Duration: ~1 month
Ambient Temp: ~20C
http://i31.photobucket.com/albums/c3...d/IMG_0270.jpg
Idle Temps (CoreTemp):
core 0: 44
core 1: 43
core 2: 38
core 3: 40
http://i31.photobucket.com/albums/c3...d/IMG_0265.jpg
Load Temps (100% WCG Loaded) for 5 hours
core 0: 60
core 1: 59
core 2: 56
core 3: 57
http://i31.photobucket.com/albums/c3...d/IMG_0274.jpg
This is how the mount were with AS5 after disassmble:
http://i31.photobucket.com/albums/c3...d/IMG_0277.jpg
http://i31.photobucket.com/albums/c3...d/IMG_0278.jpg
AFTER with IC Diamond 7
First try with grain size dot in the center:
Ambient Temp : 21C
Same Setup and bios Configuration
http://i31.photobucket.com/albums/c3...d/IMG_0283.jpg
Idle Temps (CoreTemp):
core 0: 44
core 1: 42
core 2: 38
core 3: 44
http://i31.photobucket.com/albums/c3...d/IMG_0285.jpg
Load Temps (100% WCG Loaded) after 10 minutes
core 0: 65
core 1: 60
core 2: 60
core 3: 70
http://i31.photobucket.com/albums/c3...d/IMG_0287.jpg
As you can see after ~10 minutes of WCG loaded, the load temps were out of hand, so i suspect the spread was not enough and i decide to disassemble the setup.
http://i31.photobucket.com/albums/c3...d/IMG_0289.jpg
http://i31.photobucket.com/albums/c3...d/IMG_0291.jpg
Second Try with straight line method
Ambient Temp 20.5C
Same setup and configuration
http://i31.photobucket.com/albums/c3...d/IMG_0298.jpg
http://i31.photobucket.com/albums/c3...d/IMG_0297.jpg
Idle Temps (CoreTemp):
core 0: 43
core 1: 40
core 2: 35
core 3: 39
http://i31.photobucket.com/albums/c3...d/IMG_0300.jpg
Load Temps (100% WCG Loaded) after ~19 hours
core 0: 68
core 1: 63
core 2: 59
core 3: 62
http://i31.photobucket.com/albums/c3...d/IMG_0304.jpg
After disassemble the setup:
http://i31.photobucket.com/albums/c3...d/IMG_0305.jpg
http://i31.photobucket.com/albums/c3...d/IMG_0306.jpg
As you can see, the spread was way better than the first try and so are the temperatures.
Will update later with Third Try Pea size in the center and some aircooling testing as well.
http://i31.photobucket.com/albums/c3...d/IMG_0311.jpg
Quote:
Originally Posted by Movieman
Some things work to my advantage and some against. Whats interesting is that good mounts, bad mounts errors in measurement etc. do average out. I can take averages of particular configurations make some power adjustments and I come within a couple of 10th's of a degree of what I see on the bench.
Nice report Bail....good pics, nice approach on testing patterns! :up: I'll add the best pieces of your report to my next report. I've been minimal with the pics....I like your approach here.
OK, here’s the biggie in my book. Testing against Shin Etsu X23….I’ve have had my best mounts with this paste. It can be somewhat difficult to work with, but very good…..
Specs:
Q6600 G0 VID = 1.275
Abit IP35 Pro bios = 14
Gskill 2 x 1Gb regular old 800mhz stuff
Settings:
FSB: 360 MHz
Multi: 9x
Core Speed: 3.24 GHz
DRAM Ratio: 1:1
PCI-E Clock: 100 Mhz
Vcore bios: 1.335v
DDR2: 1.825v
CPU Vtt: 1.2v
vMCH: 1.25v
vICH: 1.05v
ICHIO: 1.5v
DDR2 ref: 0%
CpuGtlref 0/2: 67%
CpuGtlref 1/3: 67%
Readings:
vcore "actual idle" (CPUz reading) : 1.290
vcore "actual loaded" (CPUz reading) : 1.260
Cooling:
CPU Block/Cooler: Fuzion - Standard install – screws cranked down very hard to the end of spring compression, then about 2 turns more. Call it a hard mount…
Before TIM: Shin Etsu X23
Mount age: about 6 months old.
Paste pattern: 2 rice grains in the middle with a thin square ring around the "ditch", about 3/16 from the outer edge.
Before Temps:
Idle-Before: CoreTemp v0.97.1
Ambient: 26.4C
Core 0: 32
Core 1: 34
Core 2: 31
Core 3: 31
Load-Before: (Boinc-all four cores @ 100%) CoreTemp v0.97.1 @6 months runtime...
Ambient: 26.4c
Core 0: 42
Core 1: 44
Core 2: 38
Core 3: 39
After Paste pattern:
Healthy Pea size in middle. See pic below
Mount: Screws cranked down very hard to the end of spring compression, then about 2 turns more. Call it a hard mount…
http://i149.photobucket.com/albums/s...m04peasize.jpg
After Temps:
After Idle: Coretemp 0.97.1 @5 min runtime
Ambient: 26.4
Core 0: 32
Core 1: 33
Core 2: 30
Core 3: 30
After Load: (Boinc-all four cores @ 100%) CoreTemp v0.97.1 @5 min load.
Ambient: 26.4
Core 0: 42
Core 1: 44
Core 2: 38
Core 3: 39
After Load: (Boinc-all four cores @ 100%) CoreTemp v0.97.1 @4 HOURS load.
Ambient: 24.1
Core 0: 38
Core 1: 40
Core 2: 36
Core 3: 37
My interpretation of the results….
D7 averaged 0.75C better at idle than ShinEtsu.
D7 was the same as Shin Etsu at only 5 minutes at load.
D7 was 0.7C better than Shin Etsu at 4 hours loaded. (this factors in the 2.3C degree change in ambient temp…)
So, my conclusion is that D7 can beat Shin Etsu. I honestly didn’t think it would…..
EDIT: Another Data Point
Farm-04
After Load: (Boinc-all four cores @ 100%) CoreTemp v0.97.1 @12.5 HOURS load.
Ambient: 21.2
Core 0: 34
Core 1: 37
Core 2: 32
Core 3: 33
Ambient delta is 5.2C from "before load" to this data point. Calculating this in and doing the average across the cores, I now come up with a 1.55C improvement. So, it is improving over time, without the twist and with the heavy compression I have in this block mount. Good stuff. END EDIT
To Tasty, some “brain drizzle”.
1. I got to thinking about the composition of this stuff. It has diamonds as micro particles and the primary heat transfer mechanism. Diamonds are a VERY hard substance. When we twist the mount, are we really screwing it up? We move particles that may have stuck into the mating surfaces, and create micro-gouges in the IHS and HS. Is this better? Should we perhaps not twist and instead apply only downward force? The squish, without the twist? I would think we want to embed the diamond crystal in both opposing sides of the mount. That is the key to diamond’s heat transfer. Transfer is across its crystalline structure. If I twist, do I stack up particles in the grooves and screw up the interface layer? My empirical tests show worse results when I really mash and twist. I blew out the mount on the first machine I reported on in this thread (farm-12). I did that to the second one as well, farm-09. This one (farm-04), I did not do that to, and I won’t. My temps on farm-04 improved over run-time….
2. The Pea size seems golden so far. My best mounts came from that.
3. I’m about to switch gears on the first three machines here. Farm-04, 09, and 12 are going to be playing musical parts for some other testing I’m doing. Therefore, their mounts won’t be able to age. What I’m thinking of doing is exactly some aging tests with them. I’ll do my swapping, record the data on each machine with D7 mounts, then let them run for a week, or month, or whatever you want to see….
4. I have 10 other quads I can test on. I’ll do them all if you want. Most of the rest are Shin Etsu x38, G751 types. I may have one or two AS5 machines. I’ll have to check their logbooks. (For the rest of my teammates, yes, I keep logbooks on each machine. I track the hardware, settings, op systems, clocks, freeze-ups, failure, swaps, etc in the books. Well worth the time to write in them. I can tell you my clocks, for any given config, using the logs….It’s hard to be an Engineer sometimes…..) The question here is on getting more samples. I can cover the 10, but may run out if you want further pattern, application, or aging tests. (And no, I'm not eating the stuff on toast. It really doesn't taste that good...:rofl: I am intrigued by the bacon grease part though, still.....) I'm up to testing different compositions too. I wonder if a bit thinner, no twist, would result in better transfer. The animal you're dealing with is not like metal or ceramic based TIMs. I find this entire exercise fascinating....and yes, if Innovative profits from it, just remember your friends here at XS/WCG......;)
Regards,
Bob
EDIT: This is supplied by my mass water cooler on the farm.....
I would guess that the others you've tried are your server grade stuff. Interesting comment from my perspective, since I have a lot of 9700s, which REALLY do suck, and many Fuzion blocks, which do not. I find the TRUE mount to be very good from the sample, of exactly one, that I have. Can you twist the server stuff as easily as the TRUE? As you know, I've never messed with the gear you know well. Your perspective is worthwhile here.....Quote:
Originally Posted by Movieman
What I think would be ideal is for Fuzion, or SwiftTech, to adopt or buy Thermalright's backplate. In the upcoming "musical parts" I'm going to play, I may mate the TRUE backplate to the Fuzion block. The problem with this, from the science perspective, is the number of variables I'll be changing. I don't think I can make a conclusion, unless I spend the time changing one variable at a time...:eek:
Regards,
Bob
1.The Twist is something that was added recently on the Anandtech giveaway. Some members with lighter loaded sinks were having some difficulty with spreading the paste by compression alone and only had favorable results after they tried it. I use only the compression pea method. The twist should be a light 1-2. Intense twisting may introduce air bubbles(I Probably should put some slide pictures together on that).Quote:
Originally Posted by 123bob
2. Pea works well and is supported with the slide pictures. Intel recommends that the entire IHS be covered. I thought it was interesting Bail_w's contact pictures mirrored our slide pictures exactly.
3&4. As much data as I can get. same, multiple variations and Identical hardware all useful. Shin Etsu is one of the better compounds,high quality and I do not have much user data on that maybe 1 or 2 out of 80+ results so I would like to see more.
Just to note.Tower sinks and some water blocks sometimes give me(and other pastes) a problem when mounted vertical, weight of the sink, hoses can sometimes lift the sink .001-.002 on the top side and levering heavier on the downside, perhaps even deflecting the IHS some. It is a little subtle but can be detected in the contact pattern with the compound being a slightly thicker in the top side.
""""I find this entire exercise fascinating....and yes, if Innovative profits from it, just remember your friends here at XS/WCG......;) '''''
We can always use working capital partners for expenses, bump us up to first class from business class on those tedious trips to China and Europe, we need to dine at better quality restaurants with our customers and stay @ some of those 5 star hotels to make a better appearance. Send all money to the following PO Box....
And don't forget to send more thermal paste with it to us...
How about a beer and a Big Mac? :rofl: :ROTF: :rofl: :ROTF:Quote:
Originally Posted by tastymannatees
Here's a bit more data. I'll add it to my report above too.
Farm-04
After Load: (Boinc-all four cores @ 100%) CoreTemp v0.97.1 @12.5 HOURS load.
Ambient: 21.2
Core 0: 34
Core 1: 37
Core 2: 32
Core 3: 33
Ambient delta is 5.2C from "before load" to this data point. Calculating this in and doing the average across the cores, I now come up with a 1.55C improvement. So, it is improving over time, without the twist and with the heavy compression I have in this block mount. Good stuff.
Regards,
Bob
EDIT: All of my data has the mobos in the horizontal plane. I use rack mount cases in the farm, not towers.
It is very tough to spread the IC diamond TIM, i mean tough. :) btw, all my mounts were horizontal.
Possible crude mounting test would be first test horizontal then rotate the board to vertical and see if your temps change. Heat pipe performance might be affected although water and stock type sinks I think would be OK.Quote:
Originally Posted by Bail_w
What's awkward about this kind of testing is the minutia that emerges when trying to parse the last couple of degrees.
On OC tests the average of 45 or so ended up to be 2.8-2.9 C - Subjectively a well motivated group
Hardwarelogic with the emphasis (I assume) on the "pea" averages increased to 3.7-3.8C - A highly motivated group
Anandtech is in process with "pea" emphasis with only 8 people reporting so far ( I like 20+ for the averaging) is at 2.5C. - Highly motivated group, one user was experimenting spreading the paste with a vibrator :clap:
AmdZone which is just about finished come in at around 2.3 C with low enthusiasm, low response rate - Most on those that signed on for a free sample it was their first and sometimes their last post on the forum which is why my focus changed to the more stable, experienced dedicated Distributed computing forums.
You have to assume that different "island" community forums will have some different mixes of before compounds and equipment That will move the averages a couple of 10th's of a degree one way or another.
I would like to get the averages up a bit overall, which should be doable.The devil is the detail.
The break in or cure time I do not have nailed and I would like to be able to say something like "2 hours to 2 days depending on sink pressure" or whatever it works out to be. so some attention here would be appreciated. Thanks everybody
I see that all most everyone at Anandtech who tested the IC diamond achieve improvement... hmm interesting.
I'll see if it changes over the week :)Quote:
Originally Posted by tastymannatees
I'm working 12 hr night shifts right now and haven't had time to fight with the TRUE's on my quads. I did try the ICD7 on an E6700 using a Zalman 9700 as that didn't take much time. After ~24 hrs the temps are running approx. 1° C cooler than with AS5 running WCG. I'll start testing on my quads this weekend.
I'm reading this with great interest, but I notice there's no Australian distributor listed on the site. Is there going to be?
We have been more focused on the white box side of the market than the retailers. it'll happen maybe in the next month or so it just has not been an immediate priority. We will be selling a 24 carat size tube into retail that we have been only offering to OEM/system builders in the next couple of weeks, I will be interested as to how that will be received.Quote:
Originally Posted by D_A
Turns out it makes good metal polish, too. ;)
Since I don't qualify for the trials <sniff> <wipes tear> I have a friend chasing me some down in the states. :D
let me explain how I did my little test. I know, I know that it really is not a great test but still gives me an idea. I put a compound on started the pc up for 5 min at idle watching temps and then loaded it for 10 min and took the temps. I only tested 4 different compounds. I have the ICD7 on the cpu now after my testing just to see how things go with it. Also keep in mind that my room temps are going to be very high because I heat by wood. Let me know if there is any other info I forget to put on here because I took all kinds of notes. :D
I did all test with a pea sized drop.
Specs:
Board: ECS 680I PN2-SLI2+
VGA: sli 8800 GTS 320mb w/ Stock cooler
RAM: 4 GB Balistix pc2 6400
Settings:
Stock everything.
CoreTemp v97.1
Core Speed: Q6600 2.4ghz
CPU Voltage: 1.24
PCI-E Clock: stock
Cooling:
Water loop on cpu only
Dtek Fuzion with nozzle and pro mount
D5 pump
120.3 rad
Before testing starts
ceramique about 2 weeks old and I put it on in 2 lines like Bob did earlier but fatter lines
Temps:
Room temp at pc. about 12.8c
Idle
Core 0:25
Core 1:22
Core 2:26
Core 3:17
Load with Boinc
Core 0:32
Core 1:32
Core 2:35
Core 3:28
First Test
Room temp up to about 31.1c
Compound used for this test is put in a pea sized dot. The compound is Shin-Estu
Idle Temps after 5 min of running:
Core 0:32
Core 1:28
Core 2:32
Core 3:25
Load:
Core 0:43
Core 1:42
Core 2:42
Core 3:39
Test 2
room temp down just a little to 29.4
compound AS5
Idle:
Core 0:37
Core 1:37
Core 2:35
Core 3:31
Load:
Core 0:47
Core 1:48
Core 2:45
Core 3:43
Test 3
room temp down just a little to 29.4
compound ceramique
Idle:
Core 0:37
Core 1:37
Core 2:35
Core 3:31
Load:
Core 0:47
Core 1:48
Core 2:45
Core 3:43
Test 4
room temp down just a little to 29.4
compound IDC7
Idle:
Core 0:38
Core 1:38
Core 2:35
Core 3:32
Load:
Core 0:47
Core 1:47
Core 2:45
Core 3:43
Also I don't think I put the nozzle in my fuzion the right way and I need to check that. For the most part I think the IDC7 is right on par with the others. I would like to see more of the great testing you guys are doing! Keep up the good work guys. I will keep on testing. :D and I will be able to take my clocks back up a little here soon. :D I took pics but my camera is really really old and most of them are not very good. If you want me to post them tasty I will!!??
Haha, I thought this was just some April fool's day joke. Good stuff though, Now that its still going, I'm assuming it's quite legitimate.
IC Diamond Seven Carat Diamond Extreme Performance Thermal Compound
* contains 92% pure diamond by weight :eek::D
Test setup:
Maximus Extreme X38
E8400 @ 4.1Ghz 1.34V
2x1GB ddr3, PSU Zippy 700W
Cooling: TR Ultra 120 Extreme, coin added for better pressure
Thermal compound:
* Arctic silver 5 (7 days old) *
* IC 7 diamond *
20 min. prime large FFT, average temp. from coretemp 0.97.1. Idle both 46C (coretemp bug or smth) and ambient is 22C
http://www.eoz.lv/attachment.php?att...1&d=1207311840
Well, diamond won by 1C but maybe it needs some time to settle like AS5. So I will see how it performs after 3+ days :)