RAM Problems? Cooling too Tight?

I've often run into these types of problems on several of the older P35 and P45 LGA775 boards as well as newer P55 boards so it's definitely not unheard of.

For instance last year when I got my (totally awesome) Gigabyte EP45-UD3P and put my rig together with the Tuniq Tower 120, half the memory slots wouldn't work. After loosening the thumbscrews on the Tuniq a couple of turns everything was fine.

I don't think the motherboard designs are at fault, I think it's the third party cooler designers that need much tighter tolerances in manufacturing and design. There has to be a better way to design these coolers and their hold down plates (waterblocks included) so that only a certain amount of pressure is applied.

I know with the Tuniq that when all four thumbscrews were tightened hand tight the motherboard would bend quite a bit, almost disturbingly so. Of course once you get the board bolted down into a case or motherboard tray properly that bending tends to all but dissapear, the the stress the cooler holddown mechanism is putting on the board most definitely does not just vanish.
 
Eulogy said:

It's only showing 2GB at POST sir... this has *nothing* to do with OS.
Click to expand...

I'm running vista 64 on a different hard drive on the same machine, it shows 4 gigs. I'm also running Win7 64 retail on a different hard drive same machine. Just want to clearify I'm duel booting with seperate drives.Win7 shows 4 gigs (3.25 usable), reminds me of xp 32 bit and vista 32 bit. If you Google this problem, its on a massive scale. As a side note I have Win7 RTM running on another hard drive on the same machine, and I get the same results.
 
Dataman said:
I figured Kyle uses a BIG Cheater Bar on his Builds. Something like what they use for M1 Tank when removing the Road Wheels.

Kyle is this what you are using?

Ingersoll Rand 295A 1 Inch Heavy Duty Air Impact Wrench
Enlarge Image


Ingersoll Rand 295A 1 Inch Heavy Duty Air Impact Wrench
http://www.shop.com/Ingersoll+Rand+295A+1+Inch+Heavy+Duty+Air+Impact+Wrench+-186998894-p+.xhtml

Just kidding but glad you figured it out.

DM
Click to expand...

I admittedly have a bad habit of NOT using the tensioning springs that come with Koolance water blocks like I am supposed to. But obviously this problem has happened to others as well.

cpu-350ac_p3.jpg


I am using the springs as pictured above and it is working fine.
 
Uncle said:
I'm running vista 64 on a different hard drive on the same machine, it shows 4 gigs. I'm also running Win7 64 retail on a different hard drive same machine. Just want to clearify I'm duel booting with seperate drives.Win7 shows 4 gigs (3.25 usable), reminds me of xp 32 bit and vista 32 bit. If you Google this problem, its on a massive scale. As a side note I have Win7 RTM running on another hard drive on the same machine, and I get the same results.
Click to expand...

This problem has nothing to do with what Kyle experienced.
 
Eulogy said:
This problem has nothing to do with what Kyle experienced.
Click to expand...

I understand that, but until Kyle found the answer I was trying to be helpful as the rest were. Now maybe someone can explain the problem with Win7 and the usable memory.
 
Paladin21 said:
I wonder if this is socket-specific. Anandtech did an article last week (http://anandtech.com/mb/showdoc.aspx?i=3661) on how Foxconn sockets on P55 boards were burning out pins during extreme overclocking due to poor manufacturing. Is it possible this is another problem from the same source?
Click to expand...

Two other people mentioned that link previously in the post. It sounds plausible it's connected to this though. (intermittent connection between processor and motherboard)
 
Last edited:
Uncle said:
I understand that, but until Kyle found the answer I was trying to be helpful as the rest were. Now maybe someone can explain the problem with Win7 and the usable memory.
Click to expand...

This is solely due to the fact that MS limits ram to only 4GB and below on 32-bit windows client versions, even with PAE. Using the memory remap, and windows server OS, and PAE, you CAN get all 4GB in 32-bit mode. Or, use 64-bit windows client (windows 7/vista/etc). This is, however, a totally different issue than what this thread is about, and this issue has been covered a billion times in many other places already.
 
extide said:
This is solely due to the fact that MS limits ram to only 4GB and below on 32-bit windows client versions, even with PAE. Using the memory remap, and windows server OS, and PAE, you CAN get all 4GB in 32-bit mode. Or, use 64-bit windows client (windows 7/vista/etc). This is, however, a totally different issue than what this thread is about, and this issue has been covered a billion times in many other places already.
Click to expand...

Read my thread, I am using 64 bit.
 
Blue Falcon said:
I know with the Tuniq that when all four thumbscrews were tightened hand tight the motherboard would bend quite a bit, almost disturbingly so. Of course once you get the board bolted down into a case or motherboard tray properly that bending tends to all but dissapear, the the stress the cooler holddown mechanism is putting on the board most definitely does not just vanish.
Click to expand...

Only when you bolted down the motherboard, each point where a screw went into a standoff mount on the far end of the board away from the block became an -additional- stress point in order to "bend it flat."

Uncle said:
I'm running vista 64 on a different hard drive on the same machine, it shows 4 gigs. I'm also running Win7 64 retail on a different hard drive same machine. Just want to clearify I'm duel booting with seperate drives.Win7 shows 4 gigs (3.25 usable), reminds me of xp 32 bit and vista 32 bit. If you Google this problem, its on a massive scale. As a side note I have Win7 RTM running on another hard drive on the same machine, and I get the same results.
Click to expand...

I'm still trying to figure out if you're a troll or an idiot. 5 years on this site, not banned... hrmm.

The post you quoted specifically states that the amount of memory being shown was only 2GB *AT POST*, which means... BIOS screen. You would still get this screen if there is no OS installed at all. If there was no hard drive on the system. This has absolutely ZILCH to do with the OS... any clarification on your end still isn't relevant to the issue.
 
Seems like there are been more head aches since intel went with this method of attaching coolers. The classic or AMD way is so much easier and give you the same end result.
 
What kind of backing plate does Koolance use, maybe youre getting some board warping that's causing the dim slots to distort/loose contact in places.
Have you tried mounting the block on the motherboard and putting a straight edge on the back where the dim slots are?
 
Doesn't this sound like more of a defect, then a minor issue with users over tightening their screws?
 
does that board not support 4 dimms dual channel? it looks like only one i blue and the rest are black.
 
Yeah I had a similar problem, but you mention it to most people asking for help and they think your bonkers or a complete tool.

I have the Gigabyte P55-UD4 using a i5 750 and get conflicting results with my overclocks. Gigabytes mobo program says one thing, the bios says another and cpu-z says another for ram and cpu frequency. I dont know which to trust. So ive held back on my OC (cant seem to get a decent speed stable) until theres some more updates and see if that fixes the problems.
 
So it was a contact problem. Why is that at more than 50% of electrical problems are always contact/shortcircuit related. I guess Intel's inverse design along with the extrapressure made some pins not connect well.
 
Hmmmm, Has anyone ever had a problem like this caused by a weighty air cooler and fan like an ultra 120 exreme mounted in a tower case. Seems to me that being "sideways" with such a heavy sink would also put a lot of "bending" force on a mobo.
 
TomFool said:
Hmmmm, Has anyone ever had a problem like this caused by a weighty air cooler and fan like an ultra 120 exreme mounted in a tower case. Seems to me that being "sideways" with such a heavy sink would also put a lot of "bending" force on a mobo.
Click to expand...

Not really been an issue, the backplates help hold the motherboard rigid. Early days one guy ripped his socket off the board and moving the machine/transport roughly (lay the machine on its side) is not a good idea. A lot of the air heatsinks and some of the water blocks now use spring loaded screws with limited thread length so you can only go so tight (apply only so much clamping force) and the threads bottom out.


Doesn't this sound like more of a defect, then a minor issue with users over tightening their screws?
Click to expand...


With the standard "old school" threaded rods and thumbscrews and a WCing noob like Kyle, you can apply ungodly amounts of clamping force limited only by how much the threads of the rods/bolts will take. It is possible to rip the bolt heads on the underside of the board completely through the holes in the board if you just keep cranking down. Normally water blocks do not use backing plates to help spread the force and keep the board flat because there is not the huge weight and torque due to length/height of the big air coolers. Not sure about the koolance block but doubt there was a backing plate. If there was it is also possible so much force was applied the motherboards solder side connections penetrated the typical layer of insulation on the backing plate shorting something out. I am not convinced this has anything to do with the overheated contacts or socket contacts at all without more investigation and that is not warranted. The correction is to use the springs and not crank them all the way compressed either. If you do that you just turned the spring into a washer and the same issue could occur. As Kyle mentioned, this is not an uncommon problem with newcomers to WCing, many old posts on "my machine will not boot/post/etc" after I installed/updated a waterblock. I am looking forward to "put a spring on it" becoming the pat answer to every issue Kyle mentions in the future. /snicker.
 
Last edited:
drescherjm said:
Could this be due to poor pin contact problem with socket 1156? What I mean is too much tension causing the board to twist a little and this combined with a minimal pin contact..
Click to expand...

Actually yes this is a problem with poor socket connection or a missing pin connection. I have also seen this same issue on the MSI P55 forums. After he reseated his CPU he found 2 pins that were not making contact.

It also may be related to the socket problems that they were seeing bad p55 sockets on the DFI boards.
 
So HardOCP incorrectly installs its wateblocks? Incorrectly as in, not according to the instructions supplied by the manufacturer and not using the parts they supplied?

Doesn't that invalidate all the testing done when the blocks were incorrectly installed?
 
cybrsage said:
So HardOCP incorrectly installs its wateblocks? Incorrectly as in, not according to the instructions supplied by the manufacturer and not using the parts they supplied?

Doesn't that invalidate all the testing done when the blocks were incorrectly installed?
Click to expand...

Dont worry about it. 500 ft-lb makes for great contact.
 
I am glad that Kyle and the HardOCP (mistype this and it's HardCOP) let us know instead of covering it up and spreading BS. Take a Big Guy to do this. I Plan on sitting out the i5 until 12 months from now. Hopefully Intel will go back to some other (ie better) hs mounting system.

Of course I don't Water Cool or such. Most of us Don't. But it's always dicey buying new hardware until the bugs are out.

100 Dollar Q is if he used Stock HS/Fan would it have happened also? Ie cant this be duplicated with stock. Memory Problems. Then it's major heads up for builders.
 
I had the same problem: it's a problem with pin contact with the CPU. I, stupidly, bent some pins and "lost" half my RAM. Just get that old mechanical pencil out and straighten them SOBs are all is well. But strange how pressure would cause pin contact problems - you'd think it'd be the opposite?
 
Just another reason why this site rocks. Kyle has the balls to ask for a little help. Why not, surely someone out there has run into it before and knows the answer. And they did; some great answers here. The rest of you give the guy a break.
 
Kbomb84 said:
I had the same problem: it's a problem with pin contact with the CPU. I, stupidly, bent some pins and "lost" half my RAM. Just get that old mechanical pencil out and straighten them SOBs are all is well. But strange how pressure would cause pin contact problems - you'd think it'd be the opposite?
Click to expand...

Well, the processor doesnt have pins, so you can't use the mechanical pencil trick on it, but the socket has little fingers that if pressed too hard may not contact the right spot, causing issues.
 
Back when the LGA 775 sockets where first out there was a big stink about how this new style "pad and finger" contact was only good for 20 cpu insertions. No clue if that is still in the spec. Point is I would not dismount the cpu from the holder unless I had a good reason. hmm note to self, look that up again on a slow day.
 
DukenukemX said:
Doesn't this sound like more of a defect, then a minor issue with users over tightening their screws?
Click to expand...

Enduring_Warrior said:
So it was a contact problem. Why is that at more than 50% of electrical problems are always contact/shortcircuit related. I guess Intel's inverse design along with the extrapressure made some pins not connect well.
Click to expand...

cybrsage said:
So HardOCP incorrectly installs its wateblocks? Incorrectly as in, not according to the instructions supplied by the manufacturer and not using the parts they supplied?

Doesn't that invalidate all the testing done when the blocks were incorrectly installed?
Click to expand...

Yes, I purposely "incorrectly" install water blocks during motherboard testing. I have done this for 4 years now. Way back when, we saw a problems with boards not being able to handle an incorrectly or poorly installed cooling solution. I put the test boards under great stress then we incubate the boards to put them under what we think very hot case temps. We usually stress test at 45C or above for no less than 12 hours hours under full CPU, RAM, and GPU loads.

We are here to TEST and REVIEW motherboards, not handle them with kid gloves. If that invalidates our testing, I suggest you are at the wrong website. Barbie's Dream House is down the block. ;)

Ualdayan said:
I wonder if this is Kyle's normal rig, or does he just use Koolance on all of his stuff?

http://www.hardocp.com/image.html?image=MTI1NjA1ODY4MTQ1eXJHams2UUVfMV8xX2wuanBn
Click to expand...

That is a normal test rig. The board in that picture has been Torture Testing for 24 hours as of writing this.

delahaya said:
Just another reason why this site rocks. Kyle has the balls to ask for a little help. Why not, surely someone out there has run into it before and knows the answer. And they did; some great answers here. The rest of you give the guy a break.
Click to expand...

I like the haters. They make easy targets, then they sulk back to the forums that they frequent with a big story about how they told me off and put me in my place. ;)
 
interesting. my water cooling gear is arriving tonight for my p55 build. something to keep in mind.
 
BillParrish said:
Back when the LGA 775 sockets where first out there was a big stink about how this new style "pad and finger" contact was only good for 20 cpu insertions. No clue if that is still in the spec. Point is I would not dismount the cpu from the holder unless I had a good reason. hmm note to self, look that up again on a slow day.
Click to expand...

I recently read that intel data sheet or whatever its called (forgot why), but I remember saying for socket 1366 the socket has to withstand 30 insertion/removal cycles or something to that effect.

EDIT: found it

2.5 Durability

The socket must withstand 30 cycles of processor insertion and removal. The max
chain contact resistance from Table 4-4 must be met when mated in the 1st and 30th
cycles.

The socket Pick and Place cover must withstand 15 cycles of insertion and removal.
 
Last edited:
See the p55 users beware thread, many p55 foxconn socket boards are suffering the same issues if the waterblock/cooler is tightened down too far as the processor sits uneven on the pins and some of the memory controller pins arent making contact. I would think these manufacturers now being aware of the problem would extend an RMA to replace said sockets but i have yet to see an official manufacturers reply. I had a similar issue with the sabretooth rig i built.
 
You must log in or register to reply here.
Back
Top