Official IC Perihelion Test Results Thread

[JoeC]

mailed out more samples today should be up to date.

Going to ask you guys to extend burn in or cure time to see if we missed something (this is why we do these tests)

Some tests run on a PC test bed with a lab sample versus packed tube sample and we were surprised to see a burn in time of 4 days. Mix is similar to ICD with lower viscosity so running more tests along side yours to see what's up.

Thermal conductivity: 4.4 W/mK


Sample Thermal Impedance (deg C/W) Avg. CPU Temp.
(ICD) 0.009 64 C
Perihelion (lab sample) 0.012 67 C
Perihelion1 (sample syringe) 0.014 66C (after about 2 hours)
Perihelion (sample syringe) 0.012 60C (after about 4 days)

Thermal impedance as a function of pressure @ 120 Watts, 60 C:
@ 10 psi: 0.202 oC-cm2/W
23 psi: 0.193 oC-cm2/W
37 psi: 0.186 oC-cm2/W
50 psi: 0.182 oC-cm2/W
64 psi: 0.171 oC-cm2/W

 

[baditude_df]

My subsequent application has been on under 24/7 load since I posted my results in this thread. Applied in the same ~7mm dot as my previous test. I reserved a post under my results to report my findings after a "burn in period" of 1 week.

 

[Spawn-Inc]

crap, i forget what i am running right now. i think it's still Perihelion, but just to be sure i will reapply and do some more stressing for the next week or so.

 

[Conumdrum]

Ambient is a key part to the calculation for performance - some measurement like wall thermometer, meat thermometer, medical thermometer etc. I do not expect 100% accuracy some will report higher some will report lower, if enough data points are obtained it should give us a reasonable final number. but some effort should be made to provide a number with tools at hand.

Amen. Standardization. Somewhat. Glad to see it and looking forward to final results.

You set/demand a standard load test program?

 

[JoeC]

Amen. Standardization. Somewhat. Glad to see it and looking forward to final results.

You set/demand a standard load test program?

I do not think it adds anything as far as part of the objectives here is to trouble shoot problems.

C/W will not change much whether it's 55 or 105 W. Nobody can report a power figure and nobody knows how much power their sink is dissipating (C/W)It's the nature of the beast, generalities--- We need ambient temps to show the differences.

Either a user sees a plus result or a negative relative to something else. If I have 6 tests with a negative and 6 with a positive that the number that is what is important to us, using a particular stress program will not change that number.

Troubleshooting is simple but not easy to do in a lot of cases.

Assuming the application is OK, the basic set up is some TIM between two flat surfaces. The two primary factors affecting performance are pressure and how good your contact on the two mating surfaces is. This is the reason for focusing on mounting, if there is a problem past history is a good place to start.

Problematic is getting all the ducks in a line without exhaustive reworking, remounts and careful attention/observations to past patterns, thickness, thermal results etc. Non trivial on that score and good execution might not be the final answer but will point to another avenue of investigation...burn in or some other unanticipated variable

With IC Diamond 20% of the people who have tested our paste do not see a Positive result out of the gate, 80% do.

We sampled the 20% group with some contact/ pressure tests and found that 80% of those had either light contact or light pressure or some combination of the two. 4% we could not account for.

With Perihelion we are following the same track to find out what we do not know about the compound, user application and systems and make adjustments as required. That's why we do the beta field testing.


A good result might be great contact but low pressure or great pressure and low contact will give positive results.

Best is both good C/P.

Burn in or curing time is not some something we experienced in our normal course of testing and has forced to test outside our normal range of testing. If here in subsequent testing we find it is a factor we will be able to validate it in our own tests and perhaps develop a spec for it or to more closely align the paste to market.

As applied results come in we test in parallel backwards and forwards, double and triple check. Forum testing works well for us and as ever we appreciate everybody's effort to get it right, thanks for taking the time and effort

 

[moocow]

Test Bed:
Intel i7 860 @ 4.2ghz HT/on 1.336v
MSI P55-GD80
Gskill Ripjaws 1866mhz @ 2000mhz
Thermalright Venomous-X (San Ace @12v- 2318rpm +- )
Lian Li K62 (fans @ 12v/)
Silverstone ST70F power supply
2x Diamond Radeon 4890s

Room Temperature- 20*C

Testing Method:
Fresh Application of paste, then run Prime95 Small FTT for 15mins, idle for 15mins, Prime95 Large FTT for 15mins, idle for 15mins, Prime95 blend for 15mins, then 30mins of Prime95 Large FTT.

Average temperature across 4-cores is taken for results.

Idle Results:


Load Results:

 

[[AK]Zip]

So I started running some tests. I am testing an i7 920 CPU running at 4.2GHz (200X21) at 1.329v on load. All 4 cores are active and HT is on. I am using the TRUE heatsink in a push/pull configuration. Below are the results I have so far. All testing is done in a closed case.

For load I run Prime95 large FFT's for the highest amount of heat and power output for 30 minutes. Real Temp 3.4 is used to then record the highest possible temperatures. I then let the computer idle for 40 minutes and record the idle temperature. 40 minutes is more than enough time for the computer to cool down. I then run the Prime 95 large FFT test again for 30 minutes and record the temps again. So far the temperatures have been consistent. The temperature posted below is the average temperature of the 4 cores.

TRUE HSF Unlapped
Compound:Arctic Silver 5
Ambient Temp: 19.8*C
Idle Temp: 36.75*C
Load Temp: 77*C

TRUE HSF Lapped
Compound:Arctic Silver 5
Ambient Temp: TBA
Idle Temp: TBA
Load Temp: TBA

TRUE HSF Lapped
Compound:IC Perihelion
Ambient Temp: TBA
Idle Temp: TBA
Load Temp: TBA

TRUE HSF Lapped
Compound:IC Diamond
Ambient Temp: TBA
Idle Temp: TBA
Load Temp: TBA

 

[MooMasster716]

E8500, 1.28(1.35bios) /3800, xigmatech s1283/Silverstine fan @7V, Gigabyte EP43-UD3L

Pre-Perihelion
Compound: AS Ceramique (6mo+ old application)
Ambient Temp: 21C
Idle Temp: 48C
Load Temp:60c (spiked to 61 for a few seconds) Orthos large FFT 10min
Heat Sink Fan Speed: Silversone @ 7V

Perihelion (heat cycled it and ran it for a week.)
Ambient Temp: 21c
Idle Temp: 48c
Load Temp: :60c (spiked to 61 for a few seconds) Orthos large FFT 10min
Heat Sink Fan Speed: Silverston @7v

I think I'm going to keep it for a while and test again in about a month using Occt then test the Ceramique again with Occt. Then the the IC7

Nvidia XFX 8800GS Stock clocks 680/1700/800

AS Ceramique - load-82c Idle- 52c

Python Load- 81c Idle- 54c

 

[Ancient_1]

Here are some results a got on Friday.
Ci7 920 @ 191x21 1.36v load 191x12 1.14v idle (my 24/7 settings with HT)
TRUE Cu w/TR-FDB-2000 push Sanyo Denki 1011 pull
Windows 7 RC Loading with OCCT and using core 1 temps giving the aprox avg high.

Ambient taken from an instant read thermometer probe about 1" in front of my main intake fan.

TR CF 2
Ambient aprox 59.5°f
Idle ~27°
Load ~60°
Graph temp

Graph v-core


IC Perihelion
Ambient ~61.5°f
Idle ~28°
Load ~64°
Graph temp

Graph v-core


IC7D
Ambient ~62.5°f
Idle ~28°
Load ~61°
Graph temp

Graph v-core

 

[JoeC]

Initial results -I will edit and add data. not complete 1-2 points I am still noodling through & 3 Misc compounds will be added soon also. XS and OCF separate and combined. Note that combined chart should read Ceramique should read +0.217 not minus will fix next go around.

 

[JoeC]

NOTE: WE ARE PAUSING TESTS FOR A COUPLE OF WEEKS


We have been testing alongside you guys with some interesting data . I only have one set of data points and have to run more tests. I ran an impedance/psi curve on the ASTM tester @ 120 W. Then I ran another curve keeping psi stable and scaled power. I suspect the red line is the mysterious "cure" (at least to me) or the "0" result category. At some point the constant psi curve should merge with the Impedance curve and stabilize. So the problem we believe has reverted to a pressure problem. With thermal compounds 99% of the problems are mostly due to contact or pressure in this particular case the pressure problem we believe is on our side with the paste.

This point of stabilization on the pressure/resistance curve is @ a higher pressure than IC Diamond and is why some only see 1-2C difference with great pressure and some see +5C at pressures that are slightly less but still in a good operating range as indicated with the control group who still had good results with ICD . Not a big issue and requires some minor tweaking on our part at which point we will issue everybody a new set of test samples.

These field tests validate the process and you guys have done a great job (Thanks) probing for weaknesses by providing the info we needed to improve the product for more finely tuned market fit. Bottom line is what you guys see at the desktop is what matters.


Overall, there is a change in the impedance - around 1 deg C for a 1 sq. cm. or about 2.5 C for a 1 sq. inch die @ 100 W. Pretty well lines up and explains results at this point.


Here is the chart



Here is the data

Some tests run on a PC test bed with a lab sample versus packed tube sample and we were surprised to see a burn in time of 4 days. Mix is similar to ICD with lower viscosity so running more tests along side yours to see what's up.

Thermal conductivity: 4.4 W/mK


Sample Thermal Impedance (deg C/W) Avg. CPU Temp.
(ICD) 0.009 64 C
Perihelion (lab sample) 0.012 67 C
Perihelion1 (sample syringe) 0.014 66C (after about 2 hours)
Perihelion (sample syringe) 0.012 60C (after about 4 days)

Thermal impedance as a function of pressure @ 120 Watts, 60 C:
@ 10 psi: 0.202 oC-cm2/W
23 psi: 0.193 oC-cm2/W
37 psi: 0.186 oC-cm2/W
50 psi: 0.182 oC-cm2/W
64 psi: 0.171 oC-cm2/W

here are the results from the ASTM tester when I varied the hot bar temperature.

Sample: Perihelion
Pressure: ~36-38 psi (the pressure increased slightly with temperature...not exactly sure why; the same mass was being applied in all cases)
Cold bar temp: -5 C (all cases)

Hot Bar Temp (C) Average Temp (C) Pressure (psi) Impedance (C-cm2/W)
50 22.5 36.6 0.213
60 27.5 37.5 0.209
70 32.5 37.8 0.206
85 40.0 37.9 0.203
100 47.5 38.2 0.205
120 57.5 38.4 0.202

 

[Spawn-Inc]

will the pressure samples be sent out with the new samples of thermal paste for those who applied?

 

[eobard]

NOTE: WE ARE PAUSING TESTS FOR A COUPLE OF WEEKS

Does this mean you don't want my numbers from the samples I've already received posted? I've got number from my CPU testing, just putting the finishing touches on structuring the results (excel line graphs).

 

[g0dM@n]

I'd like to know too b/c I was trying to make time today to do some quick testing.

 

[Ben333]

Does this mean you don't want my numbers from the samples I've already received posted? I've got number from my CPU testing, just putting the finishing touches on structuring the results (excel line graphs).

Also wondering this... And, darn it, I can't seem to find a spare combo that I can monitor the CPU temp of using software. Might have to break my main rig down...

 

[JoeC]

will the pressure samples be sent out with the new samples of thermal paste for those who applied?

I will send out the pressure samples with the reconfigured paste.

 

[JoeC]

Does this mean you don't want my numbers from the samples I've already received posted? I've got number from my CPU testing, just putting the finishing touches on structuring the results (excel line graphs).

No, wait for new reconfigured sample

 

[deadlysyn]

I was getting ready to start my testing against AS5 and ICD7 when I had a chance tomorrow. How long will the wait on the reformulated sample be?

 

[MattNo5ss]

Updated my post with the AS5 spread pattern.

______________________________
Compound: AS5
Ambient: 23C
Idle: 37, 37, 37, 37, 37
Load: 46, 50, 55, 62, 71
Fan Speed: 100%
Contact:

 

[muddocktor]

So Joe, does this mean that you don't need me to post an OCCT run on my test rig after 3 weeks of the Perihelion being installed and used? I ran the test yesterday, but I guess if you are reformulating and sending out new samples soon that test is pretty much meaningless, right?