GA-890GPA-UD3H 3 long beeps continuous

[Stern007]

I got pretty much the same problems with a similar setup, using 2x2Gb Corsair XMS3  1600, Phenom 965 BE. Known good high quality really powerful PSU, SSD, graphics card. I've built a few dozens of computers over the years but this turned out to be one of the harder challenges...

Problem was the "three beep syndrom" right from the start, no sound and no video at all, possibilities to get into bios now and then and a certain time beore freezing... booting into win 7 with random bluescreens after 0 - 5 hours/minutes... horrible. Tried one memory module at the time, looked good for a while then *bam*: back to zero. Tried with onboard video, worked fine for a while then *bam*. Thrown back to stoneage no matter what I did. Newest bios FF, 2.1 board revision.

Latest glimpse of success was after a complete disassembly, reassembly of everything from the processor up, clearing bios, now I got no input from the usb keyboard all of a sudden. So I used an old PS2-connected keyboard instead. Everything worked fine for a few hours, perfect gaming through the pci-e videocard etc. Then a new BSOD, then revert to zero again. Not even the keyboard works anymore. The three beeps again...

This is really getting quite annoying. Wrong: it is ridiculous. No way I'm going to invest another hour in getting this crappy hardware to work. I'm suspecting serious problems with the MB, in my experience I have yet to see a CPU that is faulty so this piece of POS MB is going back to the dealer now ASAP. That is, if nobody has a really good idea how to solve this system builders's nightmare.

It was a *very* long time since I've seen such serious problems with a new build. Maybe I just used to be lucky...

[demonslayer]

man i know its frustrating.
i cant tell you how.but with reseating everything i never ever had the problem again.(no matter which dimm slots i used)
but i shot 2 gtx460 graficcards meanwhile without OC without nothing.
First one died.Second one has the famous driver crash which one billion users suffer from.
And my soundcard was the next problem - making noises all day long (creative forums filled with this threads)
all i can assume is that hardware got more crappy over the years.

because i never ever had such problems on a built in the past.

i cant help you out with the error.maybe you should try to flash to the newest bios?
which seemed to help me?

i even run my memory on the correct timings of 7.7.7.21 now without boot or crash problems :-)

[Dark Mantis]

Hi

Although I would probably agree that the moptherboard is the most likely culprit I wouldn't be so quick to dismiss the CPU. There are many members who have had faulty/failed CPUs recently and AMD normally just changes them without any fuss.

[Stern007]

Thanks for your support! Here is some more details what I did and also additional observations, tests and results (positive results!).

First of all, before yesterday's post I upgraded to FF bios version which I believe is the latest except for a beta version that I don't want to try at this point.

I tried a complete disassembly/assembly, re-mounting the excellent CPU cooler  Noctua NH-C12P SE14 with fresh NT-H1 thermal compound. I chose this model over the even more powerful tower model specifically for its extended cooling capacity for chipset and other mainboard components around the CPU. It is truly a work of art, this cooler: 140 mm fan, controlled by the MB's fan control to like 600 rpm normally. Dead silent, super efficient cooling.

During this the second assembly, I noted that the heatpipe chipset cooler was quite loose, kind of wiggly. It is fastened with two screws right through the board so I tightened them slightly. No effect on the slight wobble though. After reassembly it worked fine for a few hours, even ran the memory at 1600 MHz without problems, DMM voltage at 1.65 according to spec for Corsair XMS3. Then after trying some serious gaming the problem reappeared exactly the same... and I wrote yesterday night's post out of frustration.

Today I woke up seeing a pattern behind the observations. Keyboard problems (and sometimes sound glitches from integrated sound) following after gaming-induced crashes. Those are functions that can be associated with chipset functions... and that chipset cooler turned out to have a rather poor contact as I said. I should have seen that pattern yesterday.

So I got an idea now to minimize the thermal load on the 890GX chipset. I don't even know if the internal graphics HD 4290 is activated by default if you are using a standalone video card but the logic was to cut off chipset functionality that I didn't need at the moment. Disabled integrated graphics since this should have the greatest effect on the thermal load of 890GX.

As a matter of fact it seems to have helped, up and running now for several hours including gaming!

Diffuse problems like we are talking about in this thread can be produced by many things, including weak PSU's, but it is conceivable that the chipset cooler could randomly seat in a more favorable position after a complete disassemly/assembly and this is why it sometimes seems to help.

I have requested a RMA on this board, let's see what they say but I'm pretty certain my problems boils down to that poorly mounted chipset cooler. Of course I could take it off and remount with great care using NT-H1 and eventual extra stabilizing spacers etc, but it is too much work. It would also void warranty which may be needed if the chipset will turn out to have been damaged by overheating. However, I'll ask them if it is ok that I ensure the chipset cooler on the new MB is properly mounted without affecting warranty.

[Dark Mantis]

The chipset cooler is often a bit of a problem on these boards and many people haven't been happy with the mounting/performance.

It is even more likely however that the cPU is damaged as you have been running it at 1600 mhz. The recommended safe speed is 1333 mhz by AMD. It is very common to burn out the memory controller part of the CPU if you exceed this speed.

[Stern007]

Dark Mantis, I have read and understood the potential issues with running the memory at 1600 MHz. I did so in order to stress different subsystems in order to isolate the fault. As it turned out, the problems were unrelated to memory speed, i.e it ocurred with the same likelihood regardless of memory speed and even memory voltage (default 1.5 V, changed to 1.65 V due to memory spec).

Once it started to act up on me again yesterday, I changed the memory speed back to default since the problem seemed unrelated to memory speed and I see no reason running at 1600 MHz.

Now the system boots and start nice, running stable also under heavy load with disabled integrated graphics. So it seems the 890GX chipset or its heatpipe cooler was the culprit.

[Dark Mantis]

I would agree and expect that you are right about the chipset cooler being the problem. I just wanted to point out that any time spent running at a faster speed than 1333 could incur damage to the CPU even for a short time. Anyway if it turns out later that you get a problem with the memory controller AMD are very good about replacement.

[Stern007]

I would agree and expect that you are right about the chipset cooler being the problem. I just wanted to point out that any time spent running at a faster speed than 1333 could incur damage to the CPU even for a short time. Anyway if it turns out later that you get a problem with the memory controller AMD are very good about replacement.

Thanks for the advice. I doubt the CPU had any problems whatsoever during this diagnostic testing but who knows, you might be right that even a very short period of exceeding 1333 MHz leads to problems in the future. Time will tell. CPU temp was kept etremely low at all times of course.

Can you recommend a good test for CPU-memory that would reveal any problems with the CPU MCU now and in future tuning? Extremely important to have good tests as I'm planning to OC this system somewhat as soon as it is confirmed to be a stable build. That is, I'm planning to use a bit of the headroom provided by the 1600 MHz memory while keeping a certain measure of moderation of course. Nothing excessive, maybe take it up to 3.8 GHz or so.

It was a while since I last pushed a rig to the limits to reveal weaknesses after OC'ing, applications like Prime 95 springs to mind but I'm sure there are more suitable alternatives.

Many thanks for any suggestions!

[Dark Mantis]

Prime is still one of the favourites along with IBT and SuperPI. Have a look here for some others:

http://forum.giga-byte.co.uk/index.php/topic,420.0.html

You might want to check out here for some useful proggies:

http://forums.tweaktown.com/gigabyte/30530-latest-overclocking-programs-system-info-benchmarking-stability-tools.html

[Stern007]

Thanks, Dark Mantis! Will give it a thorough spin with those.

Just got a response from my dealer, they want me to send in the MB for repairs... could take forever I guess (it usually does), then all the hassle with swapping in and out old spare MBs to have a working system while waiting.

To save precious time for me, I asked them if it ok if I attempt to fix it myself instead without affecting warranty. Obviously the chipset cooler is problematic since it is loose and wiggles/wobbles. If it helps, fine, if it doesn't then it is also another problem (such as consequences of chipset overheating due to inadequate cooling in the first place) for which I would need more help.

While awaiting their response, do you have any ideas how to obtain a professional result here, any experiences from reattaching loose chipset coolers?

All I can think of is to carefully dismount the chipset cooler(evaporator) and the radiator over the power bridge connected with the heatpipe, removing the remains of the supposed thermal pad with stuff like acetone, check to see if there is a sufficient supporting structure around the actual chip and if needed make sure there will be, then apply some good ceramic thermal paste (I have some Noctua NT.H1), then slap it together carefully.

Please do not be shy to contribute with good ideas here! I need to get this permanently fixed as soon as possible.

[Dark Mantis]

If they allow you to try, which I somehow doubt, don't use acetone for cleanng the old TIM off the heatsink/chip. Use Isopropanol (Isopropyl alcholhol) as it is not so aggressive if it touches something it shouldn't.

If possible I would suggest using nylon bolts to hold down the cooler as they are a bit more solid than the usual fixings.

[Stern007]

If they allow you to try, which I somehow doubt, don't use acetone for cleanng the old TIM off the heatsink/chip. Use Isopropanol (Isopropyl alcholhol) as it is not so aggressive if it touches something it shouldn't.

If possible I would suggest using nylon bolts to hold down the cooler as they are a bit more solid than the usual fixings.

Yes, acetone is really harsh and melts some plastics. Got isopropanol as well so I'll use that instead, thanks for the heads-up. The existing cooler/evaporator unit over the chipset is attached with metal screws through the mainboard so there is potential for a really firm fit. The radiator over the power bridge is attached with ordinary push pins, this is sufficient and not a critical component to have perfectly secured. I guess some thin spacers could be useful for fine tuning the fit of the evaporator, preferrably in non-conductive material such as nylon or even cardboard.

The main problem (remembering other chipset cooler attachments I worked with) is the relatively small chipset cooling surface of the actual chip, much like the early AMD Barton CPUs where a supporting frame is needed as spacer around the perimeter of the chip PCB to get a secure mount without wobble. That is, no heat spreader. I suspect they didn't care enough about that little mounting detail which is what I believe produced the problems I see now.

[Dark Mantis]

Yes I think that you are right about the mounting problem. Just a case of bad design. Revision 9.9 will be better  Some spacers would indeed help I am sure.