Corsair H100 Push & Pull

[trihedral]

Hey!

I have a Corsair H100 for awhile now in push and pull settings but kept getting sort of high temps (In my opinion).

Thermal Paste: IC Diamond | Antec Formula 7

I use pretty much the top of the line thermal paste, but still get up to 62C at a 4.5Ghz (1.36v) overclock.

A friend of mine, with all the same parts, including a Corsair H100 fitted into a NZXT Phantom, gets up to 54-55C at 4.6Ghz (1.39v) overclock.

*We Also use the same thermal Paste*

To measure the temperature we ran Prime 95's Blend test, and between the two of us, his room ambient temperature is 2C Higher.

Why is it like this? He doesn't use the top 200mm on the phantom, only a push setting with two 120mm fans that came with the Corsair H100, while i use the same push system, i also have a pull system with two 200mm NZXT Performance fans (166CFM each) at the top blowing out.

Hope someone can shed some light :lol:

 

[Delirious788]

There's a few reasons why this might be.

1) All CPUs are different...some can be OC with little voltage increase, while others need a bigger voltage increase, and the greater the voltage increase the hotter it is going to get.

2) The thermal paste may not have a good spread, or not enough thermal paste.

3) The air flow through the case may not be the same. More HDDs and GPUs are just going to increase the air temp in the case may be higher.

4) How are you getting the temps? Software temps are not always accurate. Are both H100s reaching equilibrium? It takes awhile for that to happen.

The only true test to see how well your H100 is preforming is using the same case, cpu, and everything else, and just switching the H100 in the systems. And then running Prime95 and letting the H100 achieve equilibrium.

 

[legendkiller]

Either the contact of the CPU and the Base of the heatsink might not be good enough, if you lapped it, you should get 10-20*C drop since it's water cooled... The #1 thermal paste i think is the best is the Noctua Thermal Paste:
http://www.newegg.com/Product/Product.aspx?Item=N82E16835608008&Tpk=noctua%20thermal
but what i actually think is the best is the 100% Liquid Metal thermal Paste:
http://www.frozencpu.com/products/10740/thr-77/Coollaboratory_Liquid_Ultra_100_Metal_Thermal_Interface_Material.html#blank
If you want to, try lapping it B4 complaining your H100 isn't working well because sometime the CPU(HeatSpreader) isn't making good contact with the heatsinnk base...

 

[beenthere]

In addition to the ^^^ above points, the top fans may be causing turbulence around your H100 fans reducing their efficiency.

 

[legendkiller]

There's a few reasons why this might be.

1) All CPUs are different...some can be OC with little voltage increase, while others need a bigger voltage increase, and the greater the voltage increase the hotter it is going to get.

2) The thermal paste may not have a good spread, or not enough thermal paste.

3) The air flow through the case may not be the same. More HDDs and GPUs are just going to increase the air temp in the case may be higher.

4) How are you getting the temps? Software temps are not always accurate. Are both H100s reaching equilibrium? It takes awhile for that to happen.

The only true test to see how well your H100 is preforming is using the same case, cpu, and everything else, and just switching the H100 in the systems. And then running Prime95 and letting the H100 achieve equilibrium.

Make the H100 push(intake) the air "into" the case which is better than exhausting the air our awhile pulling all those hot air onto it..

 

[JackNaylorPE]

Would be good to know what CPU. The temps you are getting are comparable to what I am getting in air w/ a air cooler and fans at lowest possible rpm. I'm using a 2600k w/ HT enabled.....so you are about what I'd expect from the H100 if you have a 2600k.

If it's a 2500k, I'd expect temps about 6-7C lower due to no HT.

 

[legendkiller]

Would be good to know what CPU. The temps you are getting are comparable to what I am getting in air w/ a air cooler and fans at lowest possible rpm. I'm using a 2600k w/ HT enabled.....so you are about what I'd expect from the H100 if you have a 2600k.

If it's a 2500k, I'd expect temps about 6-7C lower due to no HT.

10-20*C lower, H100 cool better than AIR

 

[trihedral]

@ Delirious788

Looks like you only glanced shortly at my post.

1) I know all CPU's are different, my friend and i have an i5 2500k, his uses more voltage along with higher frequency (4.6Ghz) while i use .030v less, but still get 6-7C higher temps at a lower frequency (4.5Ghz)

2) We spreaded our CPU's the same way with the Thermal Diamond Paste.

3) We both have 6970's in crossfire, air flow is no issue.

4) I am using Core Temps & Real Temps.

It is a true test like i already mentioned prior, it is the same exact setup, difference is his phantom case is red, mines white, and that he doesn't have a pull setup.

@ LegendKiller

Not really complaining when the H100 works fine for my i7 950. I doubt i'd want to get something like the liquid metal which is highly conductive.
How can i lap the CPU? should i use 40 Sandpaper?


@beenthere
I will remove the 200mm fans and give it a try.

@JackNaylorPE
It's only a 2500k so no HT, the cpu cooler should be like 1-3C lower than that air you have though, so it doesn't add up.

 

[legendkiller]

@ LegendKiller

Not really complaining when the H100 works fine for my i7 950. I doubt i'd want to get something like the liquid metal which is highly conductive.
How can i lap the CPU? should i use 40 Sandpaper?

If you want to lap, buy it from office depot or where they have all these tools which should have sand paper... Get these grit sand paper with at least 2 set of each:
400(Start scrapping aluminium off)
600(Still scraping)
800(Still scraping)
1000(Most aluminium should be gone)
1200(All aluminum should been gone already after you finish 1200 which should be flat)
1500(you'll start to make it smooth from 1500 and up)
1800(Stop here if you don't need it to shine like a mirror)
2000(Shine like mirror)
2500(Stop here because any higher will make you put it in the restroom to look at yourself XD...)
3000+(Keep going on for higher and make it the best reflecting mirror ever made lol)
Note: you can go up to 12000 I think..

 

[trihedral]

Thanks a lot Legend, i saw a video and it looks pretty cool.

Hopefully doing this will not damage the processor.

The Office Depot near my home doesn't sell sand paper so i will stop by a tile and granite store, hopefully can grab some for cheap.

So i need just about 12 inch of each sheet, is that correct?

Grit 400 (Per Sheet) $1 :
http://www.amazon.com/3M-Wetordry-Paper-413Q-weight/dp/B000HTPJ2I/ref=sr_1_2?s=industrial&ie=UTF8&qid=1320629280&sr=1-2
Grit 600 (Per Sheet) $1 :
http://www.amazon.com/3M-Wetordry-213Q-weight-sheets/dp/B00063UM9U/ref=sr_1_1?s=industrial&ie=UTF8&qid=1320629362&sr=1-1

Those right? I do some construction so i trust 3M more, will stop by a store since shipping is extremely expensive and time consuming.

 

[legendkiller]

Yeah, those are fine.. Get two set of each so if one get all useless, there's the other one... CPU dont get damaged as long as your cpu's heatspreader doesn't comes off which i never heard off lol... Just dont let it scratch the green part too much and dont let it scratch the back side of the cpu where the sockets are which will make your cpu 100% useless... your 2500k must not be flat lol OR you got a bad CPU which creates more heat than other CPU lol...
some tips:
Don't press down on it, if you do, it'll be like a hill... If you do want to make it quicker or finish faster, press it down during the 400-800 grit SP but when you do it, make sure you even out the 4 corner weight with your 4 fingers... IF you want to do it smooth and make it actually flat, let the weight of the cpu itself do the job awhile you just push it which takes about 3 hours average..

EDIT: I Just recently found something about their discussion and it fits your needs lol...
http://forums.overclockersclub.com/index.php?s=524672e6c9f142236fa30734016ada4e&showtopic=187691&st=12
Scroll down and read those replies lol and look at pics...

 

[trihedral]

Damn that is hella long.
Found some lapping sheets at Pepboys today and got two of each.
Will give it a go tonight and hopefully finish within 2-3 hours.

If this gets 5-10C off my CPU, then i'll shed some tears.

I hope the CPU Cooler will still install on it fine since the heatspread is going to be shorter afterwards.

Do i need to lap my H100 copper area as well?

 

[beenthere]

Actually unless the heatspreader is warped, you're only gonna see a couple degrees C change in temp. Going beyond 3000 grit is a waste of time and may actually result in less conduction.

 

[legendkiller]

Damn that is hella long.
Found some lapping sheets at Pepboys today and got two of each.
Will give it a go tonight and hopefully finish within 2-3 hours.

If this gets 5-10C off my CPU, then i'll shed some tears.

I hope the CPU Cooler will still install on it fine since the heatspread is going to be shorter afterwards.

Do i need to lap my H100 copper area as well?

Look at this guy's review, he did the exact thing that you are going to do...
http://forums.anandtech.com/showthread.php?t=2181357&page=5

 

[trihedral]

4C isn't significant :O

Oh well i'll do it just for fun, thanks for the fun link Legend.

 

[trihedral]

Best answer selected by trihedral.

 

[legendkiller]

4C isn't significant :O

Oh well i'll do it just for fun, thanks for the fun link Legend.

4*C drop on air cooling(Highest i seen was 10*C drop on air) but around 10-20*C drop on water cooling... The temp drop is depending on if your CPU is making good contact... If you get around 20*C drop that means that the Contact b4 the lapping was really bad... If you get 3-7*C drop from lapping both CPU and Cooler means that the contact was good enough but the Aluminium on the heatspreader was just stacking heat by few Celsius(3-7 in water cooling) IDK if you read the whole thread but it's pretty interesting lol, First time seen a lapped water cooler and they even use toothpaste which i haven't heard about it being useful...

EDIT: Anandtech just did a review and im 100% sure you should get lower temp than they are, but if not, than you got a bad CPU which creates lots of heat than other lol...
http://www.anandtech.com/show/5054/corsair-hydro-series-h60-h80-and-h100-reviewed/5

 

[trihedral]

Holy Cow.

Thank you so much for the information. I finished the lap for the cpu cooler (H20) last night in like 1 hour because i had help.

That's wonderful news, hopefully my chips aluminum sinks are very cruddy thus i can't seem to fill the micro spaces. Will put a straight ruler on it and measure if there are any cracks.

I will get back to you guys on the results .

 

[legendkiller]

Holy Cow.

Thank you so much for the information. I finished the lap for the cpu cooler (H20) last night in like 1 hour because i had help.

That's wonderful news, hopefully my chips aluminum sinks are very cruddy thus i can't seem to fill the micro spaces. Will put a straight ruler on it and measure if there are any cracks.

I will get back to you guys on the results .

If you can, take pic or videos lols.. BTW, I tho you have a H100 but a H20???

 

[drfreak640]

i have a h100 debate's on witch fans to go with, doing a push pull config have four coolermaster sickle flow red led fans wondering if i should sleave them up and throw them on or purchasing 4 of these phobya 120mm 25mm fans

http://www.frozencpu.com/products/13534/fan-887/Phobya_120mm_x_25mm_Nano-2G_1500rpm_PWM_Fan_-_Red_LED.html?tl=g36c365s936#blank

 

[trihedral]

I mean H100, sorry for the mistake. :]

Finished lapping my H100 Cooler, I5 2500k CPU, 2x 6970's

Note:
1. I scrubbed off the aluminum starting with 400 Grit, then 800 Grit, then 1000 Grit. Moved to 1500 Grit and 2000 Grit to try and polish it. Weird thing was i had a better mirror with the 1500 Grit than i did with a 2000. I kept it 2000 though.
2. I took the CPU Cooler (H100), underneath is had vertical streaks so i started with 800 grit and moved onto 2000 to remove the streaks.
3. The GPU Chip didn't need to be lapped, it was already mirror finish, i lapped the MSI R6970 heatsink that touches the GPU chip to remove the nickel coating and revealed the copper.

The H100 feels a little more intact with the Processor now, i did notice some idle temp differences at 4.5Ghz.

Before temps Idle temps were 28|32|37|34, now 26|24|34|33

Seems as though it barely went down, but it did a good job evening out my temperatures both on idle and load.

Load temps before lapping : 59 | 61 | 64 | 62
Load temps after lapping : 54 | 58 | 61 | 55

Not too much of an improvement, perhaps i'd have to wait for my IC Diamond to cure.

Should i buy this? would it help lower temps? http://www.newegg.com/Product/Product.aspx?Item=N82E16835100010 to remove old solution and prevent oxidation air bubbles?
i didn't use any alcohol removers to remove old thermal paste, i only use paper towel and rub it vigorously.

 

[legendkiller]

You should buy it(Depend if you got any thermal paste left... Use alcohol 70% because it's an easy way to clean those thermal paste, not as good as those artic cleaner... Im not sure if it's ganna lower temp tho... Did you OC at all? Stock should be under 40*C if on stock, but OC to 4.9GHz will make it high around the 60s and probably the mid 70s*C... Over all the time, i tho you were using stock clock lol... Get the 100% Liquid Metal Therm Paste and see, probably lower temp lol... Metal is the best Heat Transfer which should be better than 99% silver, i think it's better because in this video i saw at school(yesterday...) about conduction, convection and radiation, they said metal is the best source for heat transfer so im guessing it's better than silver lol... If you really want the top best thermal paste in the world, get the $30 thermal paste and probably a 2.5g or 3g or so...

 

[trihedral]

oh you thought i was using stock clocks.

Oh jesus no lol. At Stock 3.3Ghz i get 51-53C Load. @ 1.232v

4.5Ghz i get 62C Load @ 1.35v

5Ghz i get 75-77C Load @ 1.54v

All using Antec Formula 7 & IC Diamond which is tons better than 99% Silver, it's around 92% concentration of micro diamond particles that is more thermally conductive than silver. Not as good as metal though, but those always seem to cause problems.


I think i've given up. It's probably the cpu, it takes such high voltages just to go to a decent clock.

 

[XMSYellowbeard]

4) I am using Core Temps & Real Temps.

Are you using them at the same time? Running any 2 monitoring programs at the same time can cause erroneous readings.

 

[rubix_1011]

If you are going to lap a block or heatsink, you will also need to lap the CPU to mate them perfectly. Most coolers are designed to work with a CPU as it comes stock, so a slight convex or concave design is often used. If you lap it flat, you run the risk of causing problems with how it will seat against the processor.

XSPC Rasa RS240 or RX240 kit would have been better for the price of the H100, but too late for that.

BTW- I got a shipment of 12 different Phobya fans from AquaComputer yesterday for testing...the one you linked above is included with the set.