I currently have two of the following setups:
Intel Celeron E3300 boxed with standard heatsink+fan
ASUS P5G41M-T LX
2x1GB Corsair XMS3 DDR3 1333mhz
Both computers experience the same problems, and I'm getting no joy from ASUS regarding one of the two issues which is a result of their BIOS being configured incorrectly, namely the motherboard defaulting to 1T for its command rate setting, which cannot be changed by the user, due to the ram running significantly under-spec.
The other problem, which is significantly more disconcerting, however, is that the motherboards in both systems is being bent significantly by the heatsink's clips, which are pulling the board up towards the clips, around the processor mount. Below is a photo of this problem, taken on the second set of pieces, which is brand new; if the heatsink is removed from the board, it straightens out. The other system, which was built in November of 2010, also straightens out somewhat, but is beginning to permanently deform. This flexing of the motherboard causes both systems to become very sensitive to external shocks, to the extent where a light tap to the case by a single finger can sometimes cause either computer to switch off.
I also cannot mount any 'heavy' graphics cards in the PCI-E slot whilst the computer is standing upright, as this causes counter-flex causing a wave-deformation in the board, due to the mounting hole positions; this is however a design flaw on the part of ASUS, and is remedied simply by using the computer in a horizontal configuration. The CPU's Heatsink's board deformation significantly aggravates that problem, however, causing the system to become immensely unstable. I fear that should these computers be allowed to continue operating like this (they randomly switch themselves off occasionally, currently, with no notable criteria causing a shutdown).
So I need to replace the heatsink+fan for these computers due to a design flaw for the boxed products, and need to know whether I would be compensated for this (I have no special requirements, and the processor also doesn't need anything fancy, I just need these computers to work at stock speed) or if replacement heatsinks+fans will be provided by Intel for what I believe to be a design flaw on their part or their heatsink+fan manufacturers' part.
If someone from the Intel warranty/customer support could contact me regarding this issue, it would be greatly appreciated.
Photo of the board flex below.
Note that this thread is to address concerns regarding the stock Intel coolers, not aftermarket coolers. What you're describing is also not an issue with the motherboard's socket, as there are various mounting methods employed by different coolers, some with and some without backplates fixed to the motherboard and some with and some without 'cradles' that are fixed to the board, and the coolers onto the cradles instead.
In servicing two computers both based on the LGA775 socket, one which was used for four years and the other for about 3, both motherboards are also permanently warped from their stock coolers. The former used an Intel Celeron D 3.06ghz single-core processor while the latter used an Intel Q6600, both with their stock coolers and both assembled by 'label' system builders local to South Africa.
I have, since creating this post, also purchased a third set of the board and processor mentioned in the opening post, and will be employing the same balanced assembly method I've found to help with the other two computers. Both of the aforementioned computers, those mentioned in the opening post, are significantly more stable now, but still occasionally suffer the same BSODs as before. Neither's casing is now sensitive to knocks, at least.
I'll ask again. Intel, please revise your cooler designs as the coolers employed by seemingly all your desktop processors from the old LGA775 socket up to the new LGA1155 socket seem to use exactly the same clip-to-motherboard configuration and it's an *unhealthy* method of bringing the cooler into firm contact with the processor. System builders, consumers, and I'm sure your fabrication lines and service centers, would be far happier to not have to deal with systems meeting untimely deaths due to poor cooler design - the ecological impact of having to recycle broken components that should have easily lasted far longer where their power usage is of a far lesser concern in the long term is also significant, I imagine.
While the standard Intel CPU coolers have always been controversial in the enthusiast community, if you consider how many of them have been deployed in PCs all over the world, literally millions, I would think that if they were all causing mother board bending and damage, that would have been noticed and well documented in the PC world.
Another thing to consider is if the mother boards themselves have the holes for CPU cooler mounting in the correct locations with respect to each other. Variations of a millimeter could cause the bending that you are observing. Also, an out of spec run of CPU coolers would also do this. Not to mention very cheap and thin mother boards that easily flex and cannot withstand the weight of these coolers, which have become lighter since the days of socket 775 CPUs.
I'm not saying the mother board bending does not occur, or that it is insignificant, but simply that there are other potential factors that can lead to this phenomenon. The push-pin CPU cooler design is a compromised solution that is biased towards ease of assembly with minimal intervention on the part of the user. All designs have their tradeoffs that we must live with, but certainly any design that does cause damage (if it does) to other parts of the whole must be addressed.
That would be understandable, yes. The issue here is one of the board literally being pulled towards the 'push pins', while the socket itself is forced in the opposite direction by the heatsink, which is of course pulled towards the socket by the bracket that the push-pins are attached to.
I've noticed this issue on several different motherboards employing the LGA775 socket with various Intel processors and their stock heatsinks, and on my own LGA1356 board with an I3-530 with its stock cooler, however. Reseating the heatsinks with the motherboard completely removed from the system and carefully getting the pins to clip into position allows for a near-stable build, but near and completely are two entirely different words, and I would prefer to have an as per specification completely stable build, as I'm sure many others would be happy to have.
I've now got a third build with the above. After installing the heatsink I again observed the deformation of the motherboard as a result of the cooler design.
The second system built using this configuration was moved recently, and since has been experiencing these issues again, 4 shutdowns in the past few hours, 18 this week.
If the third system exhibits these issues as well I'm going to be filing for an RMA of all three processors' coolers, as this is unacceptable behaviour for the stock product.
Having taken some time to look over the heatsink's cradle, or the part that holds it to the motherboard and also houses the fan, there's a small 'ridge' on either side of the four clips that the heatsink rests against. At first I thought the primary purpose of this, besides the designers getting to add some detail to a product that really needs to be more functional than pretty looking, was was to ensure there's sufficient space between the hub of the fan and the top surface of the heatsink. There is, however, a good 2-3mm worth of gap between the lowest point the fan's hub can extend down to and the top surface of the heatsink. At this point I'm sorely tempted to shave down this 1mm ridge all around the heatsink and see how that system ends up working, but I know that doing this voids my warranty on the processor and HSF, so that's not really an option.
I'm going to try partially securing the heatsink and allowing the system to run some stress-tests for a day or two and see how it does if the motherboard isn't bent into oblivion. If that works, I guess it's time to file for replacement coolers.
This is something I didn't notice before; whether it just looks like that in the photo or if it's actually touching the heatsink I can confirm when I later remove the board for inspection from the system I'm again diagnosing. It's currently running Prime95 overnight with the heatsink partially secured (black clips have not been pressed through completely, only enough to separate the gray prongs to hold the heatsink in place, case lying flat so that the heatsink rests flat on the cpu).
I'll report here with my findings once I'm satisfied with the outcome of my testing.
After being aware of your recent posts, I reviewed this thread and have a few comments.
As Mr. Gifford said, it does appear a capacitor is touching the heat sink. That might be an affect of the photo, or not. If not, that is a rather major mistake on ASUS' part, as the CPU cooler "keep out" zone is well defined in Intel's CPU documents. With a stock Intel cooler, that is quite an oversight by ASUS. But, it is also possible an incorrect heat sink element was used in a production run of CPU coolers, possibly due to the many types of socket 775 CPUs being considered to be generally similar.
Frankly, I must say that if the pins of the stock cooler slide through the holes in the mother board without any effort, which IMO would indicate all the alignments and specifications are within their tolerances, I'm having difficulty understanding how locking the push-pins into place would cause the mother board to warp or bend. Yes, I can see the slight curvature of the board in the picture. After reviewing the issues you experience with your mother board, which are major and ridiculous, I am leaning more towards the board itself being the problem. That is particularly due to your experience with the video card, which is really not supported by the PCI-E slot, but the metal mounting bracket that is attached to the PC's case. You said you are aware of this issue with the board, but obviously you believe that the apparent flimsy nature of the board is not the contributing factor in the case of the CPU cooler. Of course, I don't have your components in my possession to see for myself, so these are just my thoughts.
Do you have an idea what is actually causing the PC to shutdown when it is tapped as you described? Poor contact between the CPU and the pins in the socket, for example?
Your first post mentioned a memory setting problem. The DDR2 memory used with your CPU should have SPD data programmed into a special chip on the DIMM, which is used by the BIOS to get the appropriate settings for the memory. The 1T command rate could be in that data. CPU-Z will display this data in it's SPD tab, and the current memory setting in the Memory tab. The higher performance memory settings will be programmed in the EPP profile in DDR2 DIMMs. If your BIOS allows you to choose to use the EPP profile, you just need to do that, and if not you'll need to set them manually, again if the BIOS allows you to do so. Otherwise the BIOS will read the standard/default SPD profile and use those settings. That is really the standard protocol for a BIOS, and since the memory manufactures do not know exactly what mother boards their products will be used in, they must play it safe and set the default settings to values that will likely work the first time the memory is used.
The graphics card in this example was a 9600GT with its original cooler design and essentially reference board. In its case, the heatsink as a whole acts as a support for the card helping to keep it 'flat', but the weight at the outer-most corner, ie the one opposite the mounting bracket, pointing into the case, is still left essentially unsupported, which causes the weight distributed out that way to pull down slightly. The effects of this are most evident in a computer that's had such a 'heavy' card in a case-vertical configuration for many years, but the weight is there pulling down from the moment the computer is allowed to stand upright with such a card.
The memory in this case is DDR3, as the board uses the G41 chipset with a DDR3 memory controller, not G31. The SPD data for the ram has a 2T command rate as its default for all speeds, whether 800mhz, 1066mhz, 1333mhz. 1333mhz is its default speed, while the FSB after multiplication for the Celeron E3300 and E3400 is 800mhz (200mhz); as a result, the memory automatically runs at a lower clock, in this case 1066mhz. The BIOS doesn't allow for changing the timings of the comman rate, specifically, manually. You can change many other timings, such as the tCL, tRAS and tRCD, your standard timings available to users on many motherboards, but not the command rate. Running the memory at 1333mhz, its default speed, is also not possible on this board without pushing the cpu's FSB up to 333mhz. While it may be a Wolfdale chip and may well be able to run at that speed as its default, doing so is wholly unecessary in all three the configurations I have put together, and so they run at their default speeds of 200mhz.
The system with the partially secured heatsink is still running its tests just fine roughly 18 hours after being set up to do so. At the last time that it had shut down it achieved 4 shutdowns in a space of 1 hour. I'm going to leave it until tomorrow then take the board out for inspection of the heatsink/capacitor locations.
In the meantime, if anyone could direct me to any whitepapers regarding heatsink design considerations put forth by Intel for their processors and/or for motherboard manufacturers for the purpose of ensuring compatability, I would appreciate it, as I'm not sure where to begin looking in this regard.
This is the document you are looking for. Specifically, Appendix A (not shown in the bookmarks/table of contents, it follows Section 7) and Section 6:
Appendix A discusses in depth "Mother Board Deflection" in section A.3.1. This surprised me, as I have never seen this in other Intel technical documents. Just the nature of this beast?
Regarding your memory issue, I assumed DDR2 memory for a board/CPU of this vintage. That's a shame you can't set things correctly, I have an ASUS board that allows setting all kinds of obscure memory timings, and is frankly worthless IMO.
Message was edited by: parsec
Having reviewed that article I can say that it sounds reasonable to have a repload deflection in the opposite direction of the center of gravity for a mounted cooler, yet in this case, deflection is clearly more than what appears to be a 0.5mm limit, as it's visible to the naked eye. I don't think most people would, at a glance, be able to tell there's 0.5mm worth of deflection if it had to exist in an example model.
An example of a product where the cooler designers failed to preload sufficiently would be in the Microsoft XBOX 360's initial model, which used a hand-bendable bracket behind the board to 'pull' the heatsinks' studs towards the backing of the board, with insufficient strength to ensure the mentioned fatigue on the solder joints from occuring, which combined with high heat and thermal cycling caused the initial launch models to have a very short lifespan.
Anyway, I can't make out clearly who's responsibility it is in this case to ensure that the board doesn't suffer excessive amounts of deflection from heatsink installation. The mounting design in conjunction with heatsink height, applied to the top of the cpu package, which is in a metal+plastic mounting cradle placed on top of the motherboard make up the components that come into question, which gives the following possible scenarios:
a) the CPU cradle has a thicker interface between board and CPU package than specification
b) the fan cradle of the stock intel cooler has clips that mount at a 'higher' profile than intended
c) the heatsink height mismatches specification for the fan cradle, causing total mount height to be out of specification
d) missed design flaw in fan cradle causes extra height to be applied to the total assembly, causing the resultant mount to be out of specification
This of course assumes that what's seen here is out of specification.
As I don't have high accuracy calipers I cannot measure the supplied cooler's components to assess whether it's within the design parameters put forth by this document, but I can say that all three these processor packages (PIBs) exhibit the same problem, all three being the same models.
This part from Appendix H is frustrating, however:
"These vendors and devices are listed by Intel as a convenience to Intel's general
customer base, but Intel does not make any representations or warranties whatsoever
regarding quality, reliability, functionality, or compatibility of these devices. This list
and/or these devices may be subject to change without notice."
Does this essentially indemnify Intel against a defective PIB cooler design?
Okay, the computer ran through several days of Prime95 with no incidents, and I've now removed the motherboard from the case to investigate the keepout zone.
There are some components underneath the heatsink, but all of them have a good 3mm worth of clearance from the bottom of the heatsink. The capacitor that was questioned is, in fact, next to one of the securing arms and a good half a centimeter from the edge of the heatsink, so it's in no way making contact.
So at the end of the day this does seem to be a matter of the heatsink assembly causing excessive deformation of the motherboard, which causes it to become unstable.
So, do you want to try a different cooler to solve your problems? Yes, that does not address the who's at fault issue with the stock cooler, but as well as solving your stability problem, it would demonstrate a different cooler does not cause the problem. Dynatron and Thermaltake have several coolers similar to the Intel design, but with backplate mounting. Coolers that fit socket 775 CPUs are very common, and inexpensive ones can be found. One example here:
Personally I feel that mounting the stock cooler with the clips only half secured (ie the gray, transulcent plastic pins are pushed through and the black push-pins are pushed down to a point where they are firmly positioned, but do not go all the way through requiring that one force them through just to get them to click into their secured position) and having the system run stable where it's unable to operate in a reliable, stable fashion when the cooler is properly secured as per the Intel instructions (and design requirement for securing the clips), is proof enough that the coolers are to blame.
Furthermore, mounting non-supplied coolers would void the CPU's warranty, according to the warranty booklet received with each of these processors; removing the TIM supplied on the heatsinks does the same, from what I understand. I know that not all people are so pedantic about such things that they would immediately disqualify a warranty claim due to such things, but I'd rather not take my chances.
*edit* I also don't believe it should be my responsbility (financially) to have to replace a component of a boxed processor package that is supposed to function properly, that clearly doesn't.