• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Bad Airflow over motherboard due to my tower cooler?

P

panfist

Senior member
So anyone who has been reading this thread knows that I have an 80mm fan that is resting on top of my RAM modules and leaned up against my tower cooler. When I point it at the northbridge on high, Everest tells me that idle and load temps drop from 70C to 60C. On low, it drops as low as 62C idle, 66C load, and it's inaudible.

Well, I was unhappy with the fan resting there because if the case was bumped a little bit sometimes the fan would fall down, and I didn't want that. So, I thought, hey if I cut a hole in the side of the case and mount up the fan there, I can blow cool air from the outside right on the NB, and it will be secure, right?

Wrong!

With the fan on low, it does not drop my temps AT ALL. It's pegged at 70C. With the fan on high, it goes down to 65C idle, 69 load. And it's like twice as loud as anything else in my computer. And those load temps were measured after only 15 minutes, so it could even be higher. And idle temps for my CPU are up 5C. So in conclusion, the side fan is crap crap crap crap.

Tomorrow in the morning I'm going to cut out a bigger hole and try to mount a 140mm fan and see how that works out. I'm not expecting good results, but who knows, I might surpise myself.

If it doesn't work out, I really don't care because the side of my case is usually strapped flush against the bottom of my desk. But I was hoping it my drop my NB temps a little, and then i would leave the computer on the floor next to my desk.

-------------------------------------------
old first post
I have an e8400 paired with an Asus p5n32-e SLI plus motherboard. The motherboard NB and SB coolers both have heatpipes attached. The heatpipes have fins that are literally under my tower, so that air that is flowing through the tower doesn't really move over the motherboard heatpipes at all.

http://picasaweb.google.com/se...to#5184801934713353826
There are a few more pics following that one in the gallery.

My temps are 50 idle, 60 load for my CPU but I don't think that's reporting accurately. I'm not sure how I can get an accurate temperature reading of my penryn. It feels completely cool to the touch all over the heatsink even if I remove the fan for like 5 minutes at full load.

My motherboard temps hover around 43C @1.35V. Is this too much?

I was thinking of getting a downward blowing heatsink and that might help cool my motherboard, right? I was thinking of this one but was hoping I could maybe get something cheaper than $60.
http://www.scythe-usa.com/prod...5/sczp1000_detail.html
 
I guess that is a normal temperature for your motherboard. Don't get another cooler just to have the chipset cooled too. If you don't have any instability issues during overclocking I would say to leave it as it is. However if you like to have a lower temperature, just put a 80mm fan to blow air over the chipset. But, like I said, is not really necessary.
 
If you look at the gallery, I did try to place an 80mm fan blowing over the chipset but it doesn't drop the temps even 1 degree. There is no good place to put the fan because the tower, RAM, and video card all get in the way. Plus, I think airflow directly over the NB would not be as effective as airflow over the heatfins attached to the heatpipe.

pic of my poorly balanced 80mm fan
http://picasaweb.google.com/se...to#5184802076447274642

My highest stable overclock is 3.4GHz @ 375MHz FSB, 1.225V on the CPU and 1.35V on the NB.
 
Well that is a good overclock, whatever cpu you have. I say that you are fine like this so don't worry. 43 C is a good temperature for a chipset. 😉
 
3.0 -> 3.4 GHz is hardly what I would call a good overclock for an e8400. I was hoping for 3.6 at least.

My other issue is not how high can I overclock but are my temperature readings accurate? I can't hold my finger anywhere on the NB cooler/heatpipe/heatfins for more than 1 second and it says 43, and my CPU cooler feels barely above ambient...except where it's near the motherboard coolers.
 
Honestly in my opinion, when overclocking you should never run your nb/pwm passively. Try to get a small fan blowing onto each heatsink.

Your board is probably holding you back though. 3.4ghz on an E8400 is 420mhz fsb right? Probably at the limit for an nForce 4 board.
 
Originally posted by: EarthwormJim
Honestly in my opinion, when overclocking you should never run your nb/pwm passively. Try to get a small fan blowing onto each heatsink.
Click to expand...

He stated that there is no space for that.
 
Originally posted by: error8
Originally posted by: EarthwormJim
Honestly in my opinion, when overclocking you should never run your nb/pwm passively. Try to get a small fan blowing onto each heatsink.
Click to expand...

He stated that there is no space for that.
Click to expand...

Looking at the picture I see some space. Even an 80mm or 120mm fan would be fine suspended from his tower heatsink. Doesn't have to touch his nb heatsink, just get some airflow over it.

*edit*

Looking at the pic of your 80mm fan, are you sure there's no room for like a 40mm fan directly on the nb heatsink?
 
I tried a (i think) 30mm fan directly on the NB and I got a 1 degree drop. This was both with and without the 80mm fan blowing in that direction also.

My RAM can't possibly be limiting my overclock because I'm UNDERCLOCKING it by 50mhz. The FSB is 375, not 420. The RAM is running in a 1:1 divider because I found that if I run in a 1:1 multiplier the temp on my NB drops about 3 degrees. I am able to achieve 960MHz in 4-4-4-12 overclocking the RAM by itself. This was with the NB at 1.35V, 46C, CPU at 3.0GHz.

So to recap:

max overclock of CPU:
3.4GHz@1.225V (max rated voltage)
375MHz FSB, RAM at 750MHz, "linked" mode
50/60C idle/load
NB@43C, 1.35V

max overclock of RAM - unlinked mode meaning the motherboard will set whatever divider is necessary based on the clock.
3.0GHz@1.115,
333MHz FSB; RAM at 960MHz, "unlinked" mode
50/60C idle/load
NB@46C, 1.35V

I'm sure I could be stable at 376MHz, 377MHz or whatever. It's not worth testing at 1MHz increments.

390MHz FSB was not stable at any combination of CPU voltage 1.225 + the next 3 increments, or NB voltage 1.35 + the next 3 increments. During these tests the CPU temp maxed out at 63C. Again...I doubt the CPU is getting that high because it feels much much cooler in comparison to the NB.
 
Originally posted by: panfist
I tried a (i think) 30mm fan directly on the NB and I got a 1 degree drop. This was both with and without the 80mm fan blowing in that direction also.

My RAM can't possibly be limiting my overclock because I'm UNDERCLOCKING it by 50mhz. The FSB is 375, not 420. The RAM is running in a 1:1 multiplier because I found that if I run in a 1:1 multiplier the temp on my NB drops about 3 degrees. I am able to achieve 960MHz in 4-4-4-12 overclocking the RAM by itself, but I think setting an uneven multiplier somehow increases the load in the NB, so for now I'm underclocking.
Click to expand...

Then increase the CPU voltage further.
 
Originally posted by: error8

Then increase the CPU voltage further.
Click to expand...

Read my edited post above, in particular the bold part. Should I try to push it higher than 1.27V? I'm afraid of pushing the voltage higher without getting a truly reliable temperature reading. I don't trust what I'm getting now.
 
Originally posted by: panfist
Originally posted by: error8

Then increase the CPU voltage further.
Click to expand...

Read my edited post above, in particular the bold part. Should I try to push it higher than 1.27V? I'm afraid of pushing the voltage higher without getting a truly reliable temperature reading. I don't trust what I'm getting now.
Click to expand...

Yes, it's quite strange. I don't have that CPU so I won't know what to say about what high voltage means for it. Maybe someone with experience in overclocking this cpu can help you further.
But I still keep my part, as I don't think the chipset temperature is limiting your overclock.
 
Originally posted by: panfist
Originally posted by: error8

Then increase the CPU voltage further.
Click to expand...

Read my edited post above, in particular the bold part. Should I try to push it higher than 1.27V? I'm afraid of pushing the voltage higher without getting a truly reliable temperature reading. I don't trust what I'm getting now.
Click to expand...

Chip is rated to run at 1.372v according to intel (highest VID). For some reason I had thought it had an 8x multi.

Try lowering your cpu's multiplier to the lowest setting, and find out what your board/memory max out at.

Remember you're running an nForce 4, it was designed for chips with a 200mhz FSB, you're at nearly double it's rated FSB speed. I'm honestly surprised it works with an e8400.

*edit*

Disregard my nForce 4 comments, didn't see the plus at the end of your motherboards model number.

Not really sure why you're having problems overclocking then, definitely lower your multiplier though and find out where your memory/motherboard max out at.
 
Here's a link to a thread where I've been measuring my PSU's voltage regulation.

So I was running Orthos at 6x multi / 2.4GHz / 400MHz fsb / 800MHz RAM and everything looked good, temps were a lil bit lower than full 9x multi so I unplugged my fans so I could take proper voltage measurements and it crashed. I dunno if it crashed because of some kind of power fluctuations caused by unplugging the fans or applying the multimeter so I just did some cable rearranging and now I can measure voltage without unplugging anything.

I'm going to run Orthos and edit this thread with the results, I'm going to post now anyway just in case I crash again I don't want to type this out a 3rd time.

-----------------------

This is with CPU voltage at 1.1063...that's what the mobo set as "auto" for 2.4GHz. Memory at 2.025V (this voltage is good through 960MHz). NB is at 1.5...again this was "auto."

Temps are at 61C CPU and Motherboard 41C and climbing. Takes a while to get to 43.

Holy crap my CPU temps just hit 65C? Even though I'm 1GHz under my max stable overclock? My load temps of 60C were measured after running Orthos for 1 hour!! Shit it's 66 now! Hmm it seems to be fluctuating between 61 and 66. What the hell is going on...
 
I just did that a couple of days ago. The amount of arctic silver 5 I applied seemed maybe a little on the heavy side so I cleaned off the paste from the HSF but left what was still on the CPU load plate. I got a 1C drop in temperature.

The mounting mechanism on the heatsink isn't the greatest, but the case is laying on it's side so I didn't think it would be a problem. If the mounting mechanism is applying uneven force to the CPU might it be better to test it -without- a mounting mechanism, and just let the heatsink rest on the CPU? I was running my computer like that for many months. Basically the mounting mechanism is a curved spring rod that you screw in either above or below the CPU, then push down and screw it in on the other side. Months ago when I was first mounting it, the second screw kept shooting out after I screwed it in. A few days ago I un-bent the rod a little bit so there wouldn't be so much tension on the 2nd screw. It seems secure now, but not secure enough that I trust turning the case upright. I don't really care about turning the case upright because I usually have it strapped under my desk sideways, anyway.
 
It does not have a load plate. The mounting mechanism can be seen in this pic. The two leftmost metal strips are screwed in on top of the motherboard, and the rightmost bowed metal strip is screwed into the ones on the left. Does that makes sense? All those other pieces in the middle are for mounting with different CPU sockets.

http://www.rbmods.com/Bilder/A...e/Sonic_tower/pic3.jpg

Edit maybe this is pic is better. The leftmost 2 metal bits screw into the top, the next one over is what pushes the tower down onto the load plate.

http://insanetek.com/reviews/c...er/images/PICT0237.JPG

Edit 2: I do believe that since I un-bent the mounting spring/rod/whatever so that the screws wouldn't pop out, it's not exerting a huge downward force. I can slide the heatsink around fairly easily even though it's screwed in place. It's secure enough that if I bumped the case on accident I don't think it would come loose. Maybe if I bent the spring/rod unevenly, it's applying an uneven force down? Or it could be bowed. How would I test this out?
 
Ok I tried to raise my FSB from 400 to 410 and that didn't work, so I'm back to 400 now and I'm a little confused.

In the BIOS I set my Vcore = 1.31.

Now that I'm in windows, nvidia monitor is telling me my Vcore is 1.44. CPU-Z reports it as 1.36. ASUS PC Probe reports it as 1.31. Temps seem to indicate that it is probably not up at 1.44V, because it's idling at 50C. Let's see what happens when I orthos...

Edit: crap...my multiplier setting doesn't like to stick. It reset itself to 9x...and it didn't crash right away at 400MHz FSB. That's a good sign but I'm worried as to why my voltage ratings are all over the place. Rebooting.
 
Originally posted by: panfist
Ok I tried to raise my FSB from 400 to 410 and that didn't work, so I'm back to 400 now and I'm a little confused.

In the BIOS I set my Vcore = 1.31.

Now that I'm in windows, nvidia monitor is telling me my Vcore is 1.44. CPU-Z reports it as 1.36. ASUS PC Probe reports it as 1.31. Temps seem to indicate that it is probably not up at 1.44V, because it's idling at 50C. Let's see what happens when I orthos...

Edit: crap...my multiplier setting doesn't like to stick. It reset itself to 9x...and it didn't crash right away at 400MHz FSB. That's a good sign but I'm worried as to why my voltage ratings are all over the place. Rebooting.
Click to expand...

50c idle seems high to me still. If you can move the heatsink around, it's not held on tight enough. It should be tight enough such that it can't move around, but not so tight that you you strip the threads on the screws.

As for your paste, what compound are you using and how are you putting it on? For dual core CPUs with an IHS, just put a very small drop (rice grain sized) in the center of the cpu. Don't spread it or anything. Apply your heatsink and gently twist it a little (10 degrees maybe).

To tell if it's bowing, just look at your motherboard pcb, I didn't mean the heatsink but your motherboard itself. Here's a picture http://members.shaw.ca/virtualrain/rainmaker023.jpg

I have heard of FSB holes where the computer just won't work at them, but usually they're at higher speeds.
 
My motherboard PCB looks fine from up here but i really don't want to take the whole thing apart to check for a little bow.

Even when the screws were flying off I could still move the heatsink around if applied the right force at the right spot. It's just the design of the mechanism being so crappy. And I was getting perfectly fine performance with it just sitting on the load plate under it's own weight for many months, so I'm not worried about how tight it is.

50C is probably an inaccurate reporting. Many older motherboards that weren't designed with support for Penryn in mind don't read the temps accurately. Depending on who you ask, NO motherboards read the temps of Penryn accurately. It says 50C, but it feels completely cool. Even when it's at load...it feels completely cool.

I am using arctic silver 5. I used arctic silver 2-part cleaner (the first part smells like goo gone and the second smells like isopropyl alcohol) to clean both the heat sink and the load plate, then I applied a very small amount like a grain of rice. And then a couple days ago when I reseated my heatsink, I cut that small amount in half by cleaning off the heat sink with 2-part cleaner and reseating it with whatever AS5 was left on the load plate. Yes, I know some people recommend cleaning both surfaces any time they lose contact, because you can get air bubbles or whatever...but it went down a degree (according to whatever wacky scale exists between the penryn core and my motherboard) so I think I'm fine.

Also days ago before I started messing with my overclock again, and my comptuer was strapped up, I loaded up the cores with orthos and felt the case under where the CPU is located, just in case there was some situation were the heat sink didn't have good contact and heat was accumulating under it. It felt as cool as ambient except under my PSU.

So...in conclusion, considering all of the above, I'm not worried about my CPU temps as much as I'm worried about other things, like my motherboard being a bad sample, or my PSU dieing.

As I write this I've been running Orthos@3.6GHz, 400MHz FSB for 23 minutes. Both nvidia monitor and PC probe report a load temp of 65C now, but again...feels totally cool. The only thing that feels hot are the NB and SB.

Since things seem pretty stable I'm going to try to run the Supreme Commander performance test while Orthos is going to really blow out my system and see if it's truly stable at 3.6GHz now.

---------------------------------------------------------

I think after all of this back and forth overclocking I'm going to settle on returning to 3.4GHz, which I can achieve at pretty much stock voltage. I mean how much of a difference in games am I going to notice going from 3.4 to 3.6ghz? I should maybe even go back to undervolting and put my time and effort into seeing how much further I can OC my video card.
 
Orthos with small fft should be just fine for testing stability. I use prime95 (same thing) but I'm not sure if orthos defaults to show rounding errors, make sure it has that enabled.

If your voltage is below 1.3625v, you're not stressing your CPU, that's about within spec. So why not keep 3.6ghz, or even higher.

Here's the product page for an e8400.http://processorfinder.intel.c...tails.aspx?sSpec=SLAPL
 
Originally posted by: EarthwormJim
If your voltage is below 1.3625v, you're not stressing your CPU, that's about within spec. So why not keep 3.6ghz, or even higher.
http://processorfinder.intel.c...tails.aspx?sSpec=SLAPL
Click to expand...

Well...on the box it says 1.225V max. Also...just because I'm not stressing to the max doesn't mean I'm not stressing it at all. I'm also using a lot more power, which in turn also means I'm stressing my PSU, which isn't the greatest. Also, I'm not applying a ridiculous +30% over voltage to the NB.

As I'm writing this I'm stable at 3.4GHz, 1.200V on the CPU, 1.35V on the NB, all of the other values that the motherboard fills in as [auto] are reading a lot lower. Why would I sacrifice all that overvolting and power wasting for a mere 200MHz, which I doubt I would even notice. I doubt I can even notice the 3.0 -> 3.4GHz OC. The Supreme Commander performance test gives me a score of 10,000, the max, whether I'm at 3.0 or 3.4GHz. Granted a real multiplayer game can get more stressful than the benchmark, but how many random people I meet online can keep up with a 3.0GHz penryn anyway?

edit: oh btw, Orthos does report rounding errors, and I'm using the large FFT benchmark. It loads both cores 100% and fillls up the RAM. First it does a round of FFT length 1024k, then 512k, then repeat.
 
Do small FFT for testing CPU stability. Then if that passes do large FFT. To test your memory system use HCI memtest.

Trust me, I've had overclocks pass large FFTs with no problem, but fail small in under a minute. Small FFT is the hardest test on a cpu.

Also that 1.225v max, is your VID. It's what your default voltage should be when set to auto in your bios.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top