Overclock.net banner

1 - 20 of 185 Posts

·
AMDiggity!
Joined
·
1,060 Posts
Discussion Starter · #1 ·
Hello all,

First, I got this information thanks to a couple sparse posts in the Sabertooth 990FX owners club - many thanks to those involved. From what I could gather there was one person who had done this on the original board and one on the Rev2.0 that I have.

Second, I'm posting this separate from the Sabertooth 990FX club thread so that it's easier to find on Google for those like me (and you) in the future that are having a hard time getting the info otherwise. I'll post an ~identical postthere for the benefit of the people who have that thread subscribed and don't check the AMD Motherboard forum regularly.

Third, be aware that this requires some metalworking tools. If you don't have the facility to perform the modifications yourself any local machine shop should be able to do it for you quite cheaply.

Finally, this is for the Rev2.0. Looking at pictures of the Rev2.0 Gen3 this should also apply, but I cannot confirm that. Reading results from the Rev1 example I saw, modifying the NB cooler may be slightly different.

The Problem
Despite the ASUS Sabertooth 990FX being one of the premium AM3+ boards for overclocking there are no pre-made watercooling solutions for it. Improved cooling, whether achieved through air cooling the original heatsinks or pursuing watercooling, is recommended to offset the heat created by . We will be solving this using generic off the shelf components with some modification.

The Parts
Required:
Koolance MVR-100
Koolance MVR-PLT140
Koolance CHC-122
Optional:
10x10mm copper or aluminum heatsinks (and adhesive thermal pad if not included)

Preparation
Removing the original cooler
On the Rev2.0 the Voltage Regulation Module (VRM) cooler and Northbridge (NB) cooler are one item, as the two large heatsinks are joined by a heat pipe. Two screws hold the VRM portion of the cooler to a backing plate on the rearside of the motherboard, which has studded threads. The Northbridge also has two screws, which thread into studs that are themselves threaded into the motherboard. Remove all four screws and you will be able to remove the cooler, being careful to retain all the screws and their springs The thermal paste on the NB is fairly adhesive so it will take some pressure to lift it.

Here's what it looks like when you succeed:


On the reverse side of the motherboard are small threaded studs for the NB and the backing plate with studs for the VRMs. Here's what the underside of the VRM area looks like with the VRM backplate removed:


Modification
Modifying the NB cooler
The only problem with adapting this part to your Sabertooth 990FX is that one of the mounting arms will interfere with capacitors that help deliver power to the NB. I believe the arms are made of chromoly steel. I used an angle grinder to cut an approximately 3/4" off of one of the arms to give a generous amount of clearance. I then finished it with a metal file - don't worry the metal keeps its appearance after filing. The area you have to modify is outlined in the following image:


Modifying the VRM cooler
The issue here is compatibility with the stock VRM backplate and spacing relative to the VRM chokes (the 12x12x12mm metal components under the black portion of the original cooler). You will have to drill through the MVR-PLT140 (made of nickel-plated copper) to match the mounting of the original cooler. To do this I clamped the MVR-PLT140 to the original cooler and used a transfer punch to mark the locations to drill. I believe the size bit required was 3/16", but start smaller and work your way up until the stud on the original backplate can fit through your hole. As it's copper you will probably have to countersink your holes for the studs to insert smoothly. Best practice here is to mark and drill the first hole, then put the drill bit through the stock cooler and the first hole drilled in the plate and use the transfer plate to mark the second hole so that the two holes are spaced perfectly in relation to each other. I used a drill press and a heavy clamp to make the holes and keep the transfer plate in place. The final result looks like this:


Keep in mind that where you drill the holes will affect where it mounts on the motherboard in relation to the VRM chokes. There isn't a lot of clearance so please take care - it may be best to buy two of them to be safe.


Assembly
Moounting the VRM cooler
After cleaning both surfaces with alcohol, use Thermal Interface Material (TIM) like you would on the Integrated Heat Sink (IHS) of a CPU between the MVR-100 and MVR-PLT140 when bolting them together. Remove any thermal pad material remaining from the stock cooler on the VRMs, cleaning them with alcohol and a q-tip if necessary. Some people like to apply TIM to the VRMs; I did not do this. The MVR-PLT140 comes with a thermal pad - apply this to the bottom of the MVR-PLT140 on the opposite face of where you bolted the MVR-100. My stock backplate's thermal pad was intact after removal so I reused this. Install the backplate on the motherboard by sliding the studs through the holes on the motherboard - the sides of the backplate should be parallel to the motherboard's edges. Place the new VRM cooler on the VRMs, with the studs of the backplate sliding through the holes you drilled on the MVR-PLT140. Use the original screws with springs to secure the new cooler to the backplate and motherboard.

Mounting the NB cooler
First clean the NB chip and area of existing TIM with alcohol. Assemble the cooler and mount the studs for affixing the cooler to the motherboard per the instructions, making sure to use the included clear nylon washers where the metal parts meet the motherboard. Clean the base of the NB cooler with alcohol and apply TIM to the NB chip and mount the cooler per its instructions, with the modified arm toward the capacitors and choke for power delivery to the NB as shown in the pictures in this thread.

(Optional) Adding heatsinks to the chokes
As it turns out the original cooler doesn't actually make contact with the chokes, despite appearances. Regardless, they do get warm to the touch. If you wish to, clean the chokes with alcohol and use adhesive thermal tape to affix your 10x10mm heatsinks individually to each choke. It's probably best practice to ensure they do not touch each other or the VRM cooler.

The Results


With heatsinks installed on the 12mm motherboard VRM chokes:


With 1.58V vcore and 5.017GHz on a FX-8350 the VRMs get up to 48*C in my system during IBT AVX testing. The internal construction of the VRM cooler is rather simple so cooling is likely less impressive than with purpose built coolers for other motherboards, but this is to me an acceptable result. The NB temperatures stay in the low 30s even during stress testing with 1.25-1.3V. The most dramatic change I observed was socket temperatures, which have gone instead from being 10*C+ warmer than core temperatures to reliably a couple degrees cooler than the package temperature on the sensor in the processor.

Overall it's a very attractive and acceptable improvement to the original motherboard cooling solutions. I hope this guide helps those considering the modification to pull the trigger and try this.

Update 08/27/2015

I suggest using thermal padding on the North Bridge rather than TIM. I disassembled the motherboard cooling yesterday evening and found scoring on the NB block from the NB die itself:


I put the system back together with Alphacool Eisschicht 14kw on the VRM and Fujipoly Extreme on the NB. The original thermal pads on the VRM were 1mm and that is what was used.


 

·
Registered
Joined
·
265 Posts
Wauw thanks so much for your indepth tutorial!

I already have the FX-9590 cpu in 2 weeks time i should have the same motherboard as you have (Asus Sabertooth 990FX r.20). And i will also water cool the cpu, vrm, and nb with the exact same waterblocks except the cpu which will be the Koolance cpu-380a. The FX-9590 (wil run it at stock settings) does have an impact on the VRM's so i had to buy a vrm waterblock and might as well toss in the NB block as well. All will be cooled with a XSPC 480 rad. And the vga will be a Amd R7-250 no need for water cooling that one. Hopefully my vcore is below 1.5v, 1.5v is generally the voltage for the FX-9590 but ive seen lower. You gave me an impression what to expect in terms of temps while under stress.

Is it a single loop and what rad(s) did you use? I see 2 x 140mm rad correct? is this similar to a 480 rad?
Could you tell me your average temps (cpu, vrm, nb) while idle and for general use ie browsing, videos?

Thanks again for your tutorial plus a rep to you!
thumb.gif
 

·
AMDiggity!
Joined
·
1,060 Posts
Discussion Starter · #4 ·
Quote:
Originally Posted by Cobra26 View Post

Is it a single loop and what rad(s) did you use? I see 2 x 140mm rad correct? is this similar to a 480 rad?
Could you tell me your average temps (cpu, vrm, nb) while idle and for general use ie browsing, videos?

Thanks again for your tutorial plus a rep to you!
thumb.gif
Single loop. A 280x140x30, a 240x120x60 and a 120x120x30 (all alphacool)

Thanks for the compliment and rep
smile.gif

Quote:
Originally Posted by hucklebuck View Post

Awesome tutorial! Wish I had some cash, I would do this.

I don't recall if I asked you before so here goes again. Do you just have 1 pump running all that? If so what pump? Look at my rig, if I add another rad and this to my rig would I need another pump?
Thanks. Yeah, one pump - an EK DCP 4.0 (800L/hr). Your pump it looks like is 1200L/hr so it should be more than adequate - I'm no pump enthusiast though
 

·
Registered
Joined
·
153 Posts
hi,
i have just these 2 on the mobo
Koolance MVR-100
Koolance MVR-PLT140
and the vcore1(i guess that's vram1) temps are hitting over 80 after 30 min of prime95
NB still has the original heatsinks on with the heatpipe and the part that covers chokes
had the same going on on sabertooth 990fx rev1 with similar results
could the NB block and the indyvidual choke heatsinks make such a difference?
 

·
AMDiggity!
Joined
·
1,060 Posts
Discussion Starter · #6 ·
Quote:
Originally Posted by 3m3k View Post

hi,
could the NB block and the indyvidual choke heatsinks make such a difference?
The choke heatsinks didn't make much a difference - I don't think it will solve your problem

The only thing I can think of is the TIM between the MVR-100 and most importantly the thermal pad between the PLT140 and the VRMs. I used the thermal pad that came with the PLT140 though so there's no black magic there

What voltage are you running?
 

·
Registered
Joined
·
153 Posts
Quote:
Originally Posted by Archea47 View Post

The choke heatsinks didn't make much a difference - I don't think it will solve your problem

The only thing I can think of is the TIM between the MVR-100 and most importantly the thermal pad between the PLT140 and the VRMs. I used the thermal pad that came with the PLT140 though so there's no black magic there

What voltage are you running?
was using this block on 2 different mobos(rev 1 and rev2) with similar temps - used 2 different thermal pads and and always reapplied thermal comp.
the only difference i would say is the mounting - i did drill the holes in PLT140 in order to match original backplate
was even thinking to throw in a slim fan behind the mobo to help with the airfow
the temps are ok when gaming only go bananas when stress testing
which is not allowing me to do proper stability tests and go further with clock/volts
i have 8350 @ 4.7 on 1.4V so i think it's not to much as i saw some go above 1.5
 

·
Registered
Joined
·
153 Posts
Quote:
Originally Posted by 3m3k View Post

was using this block on 2 different mobos(rev 1 and rev2) with similar temps - used 2 different thermal pads and and always reapplied thermal comp.
the only difference i would say is the mounting - i did drill the holes in PLT140 in order to match original backplate
was even thinking to throw in a slim fan behind the mobo to help with the airfow
the temps are ok when gaming only go bananas when stress testing
which is not allowing me to do proper stability tests and go further with clock/volts
i have 8350 @ 4.7 on 1.4V so i think it's not to much as i saw some go above 1.5
sorry just had a second look on the pics - mounted the same way
 

·
AMDiggity!
Joined
·
1,060 Posts
Discussion Starter · #9 ·
Quote:
Originally Posted by 3m3k View Post

was using this block on 2 different mobos(rev 1 and rev2) with similar temps - used 2 different thermal pads and and always reapplied thermal comp.
the only difference i would say is the mounting - i did drill the holes in PLT140 in order to match original backplate
was even thinking to throw in a slim fan behind the mobo to help with the airfow
the temps are ok when gaming only go bananas when stress testing
which is not allowing me to do proper stability tests and go further with clock/volts
i have 8350 @ 4.7 on 1.4V so i think it's not to much as i saw some go above 1.5
Take a look at this run I did just now for you:



That's after 5 rounds of IBT AVX, and with my fans at ~70%.

I hate to say this, but something is wrong or misconfigured if you're seeing those kind of temperatures at 1.4V. I don't know what, but yours are wildly different from mine. I figured you were running 1.6+V when you said 80*. My sig rig is up to date if you want to see what I have for cooling - some people say it's a lot for just a CPU + motherboard but that's what it takes apparently

You should make a sig rig so we know what you're working with
 

·
Registered
Joined
·
153 Posts
Quote:
Originally Posted by Archea47 View Post

Take a look at this run I did just now for you:



That's after 5 rounds of IBT AVX, and with my fans at ~70%.

I hate to say this, but something is wrong or misconfigured if you're seeing those kind of temperatures at 1.4V. I don't know what, but yours are wildly different from mine. I figured you were running 1.6+V when you said 80*. My sig rig is up to date if you want to see what I have for cooling - some people say it's a lot for just a CPU + motherboard but that's what it takes apparently

You should make a sig rig so we know what you're working with
Thanx for the tip. Updated the signature:D

now the missconfig part.. 1.4V that i mentioned earlier was the CPU voltage.

the cooling i have is 360 UT60 and 240 Monsta both in push-pull ( the rest of the temps is great - just the vcore)
honestly i have no idea what voltage do i have on vram - most likely set to auto...will check that when i'm at home and see what volts come up.
if the voltage is the case would you know by any chance what voltage range should i be starting with (i can see you have it on 1.55V and goes to 1.57) - and how much does it affect stability?
i' m aiming at around 5 - 5.1Ghz with the CPU as yours is set but i know it is most likely an individual case just like CPU volts

i found somewhere that vrams are rated for 100000hrs at 105 degrees, but what's the point of WC if the temps are almost the same as on air....
....i bet it's the volts:eek:

i will have that checked at home when back from work and college and post the results that i got with the volts - and if that's the case will set them manually
thank you very much for your help:thumb:
really appreciate that
 

·
Registered
Joined
·
153 Posts

here's the voltage readings
doesn't seem to be the voltage on vcore - appears i had it set manually on 1.35V
i think i will have to tear this apart and install the block once again when i have some time
did you use included thermal pads or had something else?
not sure what else can i try

forgot to mention - above is done after 5min in prime95 blend
 

·
AMDiggity!
Joined
·
1,060 Posts
Discussion Starter · #12 ·
Quote:
Originally Posted by 3m3k View Post

honestly i have no idea what voltage do i have on vram - most likely set to auto...will check that when i'm at home and see what volts come up.
if the voltage is the case would you know by any chance what voltage range should i be starting with (i can see you have it on 1.55V and goes to 1.57) - and how much does it affect stability?
i' m aiming at around 5 - 5.1Ghz with the CPU as yours is set but i know it is most likely an individual case just like CPU volts
The 1.55+V (increases from what I have it set at under heavy load due to LLC) you see is my CPU voltage for 5.0GHz. That's the absolute lowest I've been able to get it and remain stable with that clock - one tick lower voltage and the system freezes with IBT AVX and Core 6 (shown as 7 in Prime) fails.. Vcore/CPU voltage has a very direct relationship with stability. Voltage needs seem to raise exponentially in relation to GHz
Quote:
Originally Posted by 3m3k View Post


here's the voltage readings
doesn't seem to be the voltage on vcore - appears i had it set manually on 1.35V
i think i will have to tear this apart and install the block once again when i have some time
did you use included thermal pads or had something else?
not sure what else can i try

forgot to mention - above is done after 5min in prime95 blend
Both your vcore-1 (VRM) and vcore-2 (I believe this is NB) are significantly higher than mine at 5GHz and beyond.

Yep I just used the included thermal pad on the bottom of the Koolance VRM heat plate and re-used the original thermal pad that was on the VRM backplate

Looking at your system my first thought is your 700RPM <1mmH2O fans, including on a UT60, but you say your temps are OK elsewhere in the loop. Seeing your 44*C (I assume at 4.7GHz) you might be at the thermal limit (62*C) of the chip @ 5GHz
 

·
Registered
Joined
·
153 Posts
Quote:
Originally Posted by Archea47 View Post

Looking at your system my first thought is your 700RPM <1mmH2O fans, including on a UT60, but you say your temps are OK elsewhere in the loop. Seeing your 44*C (I assume at 4.7GHz) you might be at the thermal limit (62*C) of the chip @ 5GHz
Yep i think i will go in that direction
was thinking of noctua nh-p12 as i already have 2
or CM r4's (seen these have around 3mmH2O)
or Sythe GT AP-15's

But to do that i will have to wait till the end of march unfortunately :-(
If you have any suggestions regarding the fans it would be much appreciated
As soon as i have the results i'll post'em in this thread
Maybe even put the NB and RAM under water as well.....
 

·
Registered
Joined
·
265 Posts
Im currently assembling my rig and i have a question regarding the Koolance MVR-PLT140.

Koolance only stated to remove the CLEAR cover of the thermal pad which i did obviously but...there is another cover which is pinkish in colour i tried to peel it of but had no luck maybe because its supposed to stay on place...?

So do i need to remove the pinkish cover on the thermal pad or not?

Ps: those holes i had to drill where a PITA to do seriously i almost gave up until some one in a works space for metal workings drilled it for me it turned out hopefully good it doesnt touch any metal parts like the chokes but i will triple check everything.

Thanks in advance.
 

·
Registered
Joined
·
153 Posts
i guess so.. the thermal pad is only the gray part..secured from both sides with plastic so you need to use the spongy part in the middle...
dont forget to apply some thermal compound as well ( on the vrams and between the plate and watterblock)
 

·
Registered
Joined
·
265 Posts
Quote:
Originally Posted by 3m3k View Post

i guess so.. the thermal pad is only the gray part..secured from both sides with plastic so you need to use the spongy part in the middle...
dont forget to apply some thermal compound as well ( on the vrams and between the plate and watterblock)
Thanks for the reply,

Grey? well its white and it comes with a clear cover that should be removed but the pinkish cover on the other side i have no clue whether to remove it or not ...since your guessing it made things a bit more confusing for me.
redface.gif


So i also need to use thermal compound ie Artic MX-4 (since MX-4 is better i believe...) on the vrams as well as on top a thermal pad? Isnt the thermal tape provided with the Koolance MVR-PLT140 sufficient enough?

Just to be clear thermal compound is like artic silver, MX-4 that you have to spread.
And thermal PAD is that tape you get in this case white.

Please be more clear and im sorry english is not my native language so bare with me.

Thanks
 

·
Registered
Joined
·
153 Posts
you need to remove protective cover from both sides of the thermal pad in order to use it ...so use the middle part only regardless of the color
as per the thermal compound people say different things...some use it in addition to the pads and some don't
i used both pad + paste and works fine for me
...the worst thing for me was to run the tube from the top rad to this block

edit:

PAD is the spongy tape
and thermal compound or paste is MX-4 for example

good luck, in case of any questions just write
 

·
Registered
Joined
·
152 Posts
this is slick. I will have to do this with mine as i build my loops. still want to see the ek-fc bridge configured for this board, i want to use as few fittings as possible. I am getting thermal runaway in just about every piece of my machine except the cpu, but i always knew this h110 would be a temporary fix.
 

·
Registered
Joined
·
265 Posts
Quote:
Originally Posted by 3m3k View Post

you need to remove protective cover from both sides of the thermal pad in order to use it ...so use the middle part only regardless of the color
as per the thermal compound people say different things...some use it in addition to the pads and some don't
i used both pad + paste and works fine for me
...the worst thing for me was to run the tube from the top rad to this block

edit:

PAD is the spongy tape
and thermal compound or paste is MX-4 for example

good luck, in case of any questions just write
Thanks for the reply,

Ive asked Koolance about this issue and guess what il post the exact response from Koolance quote:

"Hello,

You only need to remove the clear side. You'll then stick the white side to the block itself and the pink side will contact the motherboard.

-Dylan

Koolance Technical Support"

Bit of different stories ie to remove or not to remove maybe its fine with both....? But that pink part is almost glued on the thermal pad i cant peel it of suggesting it wasnt meant to be peeled of....*sigh* well il try to peel it of again if it doesnt il leave it as it is with the pink cover.
 

·
AMDiggity!
Joined
·
1,060 Posts
Discussion Starter · #20 ·
Absolutely remove the paper/plastic/etc. from Each side of the thermal tape - that part's far more critical than optionally putting TIM on the VRMs in addition to the tape
 
1 - 20 of 185 Posts
Top