… replacing a south bridge. Or PCH. However You want to call it, they essentially do the same thing. As a bonus I’ve acquired 2 GPUs “not tested”.
Mhmm.
Besides that, one of todays subjects is a previously restored motherboard, which died on me, again. On that sad note, let’s start with this one.
Case 1: ASUS SABERTOOTH 990FX R2.0
So, what happened?
00 happened. Again.
And this time it wasn’t bad BIOS.
I don’t have any pictures from diagnosing the issue, but basically I googled “ASUS SABERTOOTH 990FX power sequence” and got all the way down that list untill the oscilloscope showed no communication from the NB to CPU socket (obviously I had to pull down the CPU_PRESENT_L to ground so that SB wouldn’t get upset).
Desoldering dead NB.
Cleaning pads.
There is again a bit of a problem as I didn’t take a picture with the new NB installed.
Anyway, I think it’s believable that I replaced it, I seriously doubt that anyone does BGA rework just for the pure fun of it.
Well, it works now.
With this motherboard I even managed to find an impostor among my DDR3 RAM sticks 🙂
If You are perceptive then You might have noticed a CPU swap – I also stress-tested it with Prime95, no problems found after a few hours of testing, so I believe this motherboard is again fully functional.
Case 2: ASUS Maximus IV Extreme-Z
I also don’t have any pictures from diagnosing it ( :/ ), but I can more or less recreate my thought process using a boardview.
When I got it onto my bench this board was DEAD. Nothing. Nada. Not even a gust of life. I could be pressing that “ON” button till now and it wouldn’t fire up.
First thing to do then – check for shorts.
12V fine, 5V also OK, same for 3.3V, 5V_STBY was good.
After that I consulted the boardview.
Power-on signal, after resistor, goes to the Super I/O controller, gets processed and is spewed out on the previous pin.
Where does it go then?
Yep, right to the PCH.
As a bonus the silicon core itself has two chipped corners, smaller one is probably just a damaged protective layer, but the bigger one… It’s a different story.
Let’s not waste any more time and just simply replace it.
OK, PCH is replaced, but does it work?
Surprisingly yes!
However, during testing I noticed a big problem.
It works fine, but only for a few minutes, after that it just freezes.
After depleating “sane” troubleshooting options like a bad CPU, RAM, GPU, PSU I decided to flash it a clean ME region BIOS, same as You would do in a laptop.
Did that help?
NO.
But then it hit me like a truck.
This board has a “NEC TOKIN” proadlizer capacitor, known from laptops and PS3 consoles to be excellent when new, but now “IT SUCKS MONKEY BUTTS LIKE ALL THE F*****G REST” in words of, the one and only, AVGN.
That’s right, I replaced it with 4x330uF 2.5V tantalum capacitors, all measuring above 400uF on my tester with below 0.14Ohm of ESR.
Did all that help?
NO. It still hangs up.
At this point I don’t really know what’s wrong with it, for a second I thought that there might be another BGA defect, but no amount of pressure neither on the PCH, CPU nor Nvidia bridge chip does anything. Unfortunately I deem this board partially fixed with no prospect for fully working.
Case 3: Nvidia GTX285
Unfortunately there are SO MANY resistors, capacitors and other SMD components missing, that I deem it not worthy of my time. If I had a working one of those, then maybe I could think of something, but I don’t.
Case 4: MSI Nvidia 8800GTX
Same situation as with the 285. Many, MANY SMD components missing, or even snapped in half. These two pictures show only about 30% of absolute carnage this card went through. Obviously GPU is not recognized, same as the 285. Moreover, this GPU made a literal dust cloud when the fan started spinning.
I had a spare dead GPU to steal these parts from though.
After about 2 hours of replacing resistors this is what I was greeted by:
A working display! Well sort of. It’s very obvious that VRAM is broken, as per usual in 8800 series, BUT THE MOTHERBOARD NO LONGER DISPLAYS “2b” POST CODE!
Fortunately MATS shows only one dead VRAM IC. Also fortunately I had a spare one already desoldered and reballed. 8800 series is my most worked on GPU lineup, this is going to be my fifth working one. I generally try not to get multiple same components, but 8800s are an exception.
I can see where the dust cloud came from :/ Same goes for the broken VRAM.
The date code got me confused for a second. 8800 GTX wasn’t revealed until november of 2006 and the manufacture date of the itself says mid-september of 2006. This GPU then very likely is an early sample, sold at launch of this SKU. Pretty nice find!
One VRAM IC later GPU is fully working! Time to rebuild it and fix the PCI-E capacitors.
As I mentioned previously, the 8800 series had these weirdly soft thermopads that are about 0.7-0.8mm thick, making ‘normal’ 1mm thermopads unable to squish enough, which consequently rises the temperature of the core. That’s why I opted for thermal putty on the RAMDAC and VRAM, while using standard 6mm pads on the VRM.
Before You say that it looks bad, first of all I know, second – I wanted not to use hot air to preserve as much of the sticker as I could, third – I don’t have a microscope and having a hot soldering iron a few centimeters away from my face makes me quite nervous.
Forget about that though, check out this thing:
All 16 lanes are working!
Also, during stress test sqeezing the pcb and cooler makes no difference on temperatures, which means that that thermal putty is better than traditional pads.
There is a small issue though. Temperatures are kind of high. I know that Furmark is unrealistic, and that’s exactly the point of it, but there is a problem with the GPU core and it’s called IHS. After almost 20 years thermal paste between silicon and copper very likely turned into a rock. Delid is possible, but it’s ULTRA difficult. By that I mean that capactiors are almost touching the epoxy, one wrong move and it might be dead. I might try it on my dead PCB, but just thinking of doing that to one of the working ones makes me sweat.
Case 5: ASUS P5Q-E
OK, time for some less exciting work. I bought this MoBo as not working, and when it got to me it worked right out of the box. It’s going to be a restoration then.
Free CPU as a bundle.
Yes, it’s trash, but who cares, IT’S FREE!
As I said, it was supposed to be dead. For some reason I don’t think it really is.
Standard P45 NB. Good for Xeons, good for OC. Probably the best choice for a 775/771 machine.
I wouldn’t be me, If I didn’t sneak in a 771 mod, along with modded BIOS.
That’s it. It just works. Even testing didn’t reveal any problems.
Case 6: MSI K8N Diamond
I wanted a SLI 939 build for the longest time and as soon as I saw an approachable price on one of those boards I jumped the gun and bought it.
Who would have thought, bad capacitors. Unfortunately I don’t have any pictures after I replaced them, but let me tell You one thing. This PCB made me reget taking them out. Its ground plane is VERY thick, making my 80W soldering iron NOT able to clean the VCC holes, and GND were almost untouchable, even at 480 degrees.
Broken fan, interestingly bearings are like new.
A funny anegtode about the NF4. You see the resistors on the substrate, on the side closest to RAM slots? These are straps, meaning that if You wanted You could have made a SLI version of NF4 just by adding one resistor. Link to full article: https://www.anandtech.com/show/1590/2
Anyway, even after my war with capacitors this board works fine.
Obligatory BIOS flash.
I couldn’t find anything more wrong with this board, so it goes back to the pile of dreams waiting to happen 🙂
Thanks for reading!