Saturday, December 22, 2007
Saturday, December 15, 2007
Recapping an Albatron motherboard
Albuquerque sent me a pm asking me to take a look at his board and determine if it warrants a recap.
By visual inspection, I can say that none of the caps are leaking or bulging at this point.
The caps are as follows:
All the crap GSC's are going to be replaced and the empty spots near the inductor are going to be populated. I'll be including O-scope signal screens and if I can finish building the probe, you may see some spectrum analyzer shots as well.
O.k, this one is a bit puzzling. There are no visible signs of bulging or leakage on these caps, yet the owner reported a huge drop in overclockability. Frankly, there could be n number of reasons, from aging FETs (first suspect) to those GSC crapacitors going bonkers.
In his own words:
This is what I suspect regarding the GSC's. They may be o.k at this point, but only because there is less chance of a cascading failure due to being paired with some low ESR caps, which would take the bulk of the transients/inductor ripple current. This is a better way to cut costs than say the Epox board I fixed last month. However, eventual failure seems inevitable with garbage caps like GSC, G-Luxon or Chissi. I would never leave any of those crapacitors on my board!
Phase I: 14 Dec 2007
In this phase, I just plug the board in and see how it does on a basic memtest run. This is done in order to have a bottom line on what the heck to expect. So far it is not looking good.
Something is definitely wrong with the board. Memtest errors out with a register dump loop. I tried different floppies, different FDD's and different cables. I also cleared the CMOS, beefed up the voltages a little bit, tried different slots on the MoBO, tried a single stick of RAM, swapped them around and also tried a generic crap Value RAM stick I had laying about. No go.
I'm gonna try a CDROM drive, but unless I see memtest running, at this point, I'm calling the board suspect and will proceed with the next phase.
Phase II : 16 Dec 2007
Well, now I'm beginning to suspect that there may be something else afoot. I tried threee different versions of memtest and they all failed to run beyond test#2.
Setting this aside for a moment another thing I'd like to bring out is the act that this board runs HOT HOT HOT! Even at stock settings, the Intel Presshot lives up to its name. The underside of the board was very warm to the touch. More alarming was the fact that the DRAM sticks were hot to the touch as well. Pictures will follow after I experimentally determine if my conjecture (using available documentation) regarding the choice of capacitors is correct or not.
For anybody trying to take O-Scope readings, let me tell you at the outset that what I am doing is NOT ACCURATE. In fact I would almost say it is the wrong way to measure signals/transients.
Where shall I start? Hmmm....the probes are just two wires, no shielded cable, no capacitive terminations, no impedance matching...In brief, the idea was to show how bad EMI can be if the probe is not grounded. Note that I'm not trying to measure transients, because I need a DAQ for that, but I'm just trying to see if anything crazy is going on with the voltages. I used an old scope. I did take a few pictures and save waveforms, but this is nothing like hooking up the scope to a GPIB interface for some real time data acquisition. I won't do it because it is too much work for some pretty un-interesting DC signals. :)
The point I'm trying to make is, the maximum amplitude I see here is about 3.1V, so I can use an OS-CON at 4V (5V surge rating) with no problems. I connected a wire to the Drain/Source and another to the Ground). I should have used the casing of the FET, but the spot was a bit too tight for me to spend time on.
If the ground is left floating, RFI is picked up and you see the familiar 60 Hz signal from the fluorescent lamps/other devices interferer. When grounded, this noise signal disappears.
Phase: III
Now I'm going to replace all the GSC's in the DRAM area with OS-CONs. Next update with more O-Scope shots will be available tomorrow.
By visual inspection, I can say that none of the caps are leaking or bulging at this point.
The caps are as follows:
- CPU VRM side: Sanyo WG (K-Vent) and OS-CON (Purple)
- Chipset: GSC and Panasonic (T-Vent)
- DRAM: Panasonic (T-Vent) and GSC
All the crap GSC's are going to be replaced and the empty spots near the inductor are going to be populated. I'll be including O-scope signal screens and if I can finish building the probe, you may see some spectrum analyzer shots as well.
O.k, this one is a bit puzzling. There are no visible signs of bulging or leakage on these caps, yet the owner reported a huge drop in overclockability. Frankly, there could be n number of reasons, from aging FETs (first suspect) to those GSC crapacitors going bonkers.
In his own words:
The original reason for sending this out was two things:Fair enough mate, no arguments to the contrary from me. I would agree with him on both counts. The board is well built. Apart from the highly dodgy GSC, I could not spot any glaring inconsistencies. Of course, I'm not somebody who cares about outward appearance too much or the board layout (except if it interferes with my ghetto cooling mods!). I was particularly impressed by the use of high quality Sanyo WG's and OS-CONs on the VRM transient filtering side. Odds are, I would not have to touch those bad boys.
First thing: The S478 Presc"Hot" processor in there is good for 4.2 GHz under my water cooling, but the overclock has been slowly dwindling over the last six to eight months. At first I assumed it was the processor degrading, with the 1.575v and all. It's most recent stable OC has been a skimpy 3.6GHz. However, I have another of these exact boards that has almost zero use, and the CPU overclocked right back to 4.2 GHz as soon as I swapped boards.
Second reason to send it in? Because it's a fantastic board. I don't know why people never paid attention to Albatron, especially their PX845PE Pro IIs or this PX865PE Pro II. This board ran a 2.4C at 292FSB with memory at 1:1 (yes, DDR584 memtest stable for days) three years ago. All at 1.6VCPU, 2.85VMEM and 1.8VAGP. Dual BIOSes, dual RAID implementations (one for SATA, one for PATA), firewire, USB, gigabit LAN on Intels' CSA bus (250mbps connection to northbridge for full gigabit transfer capacity, unlike most gigabit adapters which are connected to the 133mbps PCI interface). Also came with Envy24 7.1 audio -- the same basic chipset used in the M-Audio Revolution and the like.
It's an excellent, no, phenomenal motherboard in my opinion, and there simply isn't anything out today that has the same amount of quality features no matter the price tag. And since this puppy (when running at full speed) does everything I want at speeds that are quite reasonable to me, I don't have any reason to upgrade yet.
This is what I suspect regarding the GSC's. They may be o.k at this point, but only because there is less chance of a cascading failure due to being paired with some low ESR caps, which would take the bulk of the transients/inductor ripple current. This is a better way to cut costs than say the Epox board I fixed last month. However, eventual failure seems inevitable with garbage caps like GSC, G-Luxon or Chissi. I would never leave any of those crapacitors on my board!
Phase I: 14 Dec 2007
In this phase, I just plug the board in and see how it does on a basic memtest run. This is done in order to have a bottom line on what the heck to expect. So far it is not looking good.
Something is definitely wrong with the board. Memtest errors out with a register dump loop. I tried different floppies, different FDD's and different cables. I also cleared the CMOS, beefed up the voltages a little bit, tried different slots on the MoBO, tried a single stick of RAM, swapped them around and also tried a generic crap Value RAM stick I had laying about. No go.
I'm gonna try a CDROM drive, but unless I see memtest running, at this point, I'm calling the board suspect and will proceed with the next phase.
Phase II : 16 Dec 2007
Well, now I'm beginning to suspect that there may be something else afoot. I tried threee different versions of memtest and they all failed to run beyond test#2.
Setting this aside for a moment another thing I'd like to bring out is the act that this board runs HOT HOT HOT! Even at stock settings, the Intel Presshot lives up to its name. The underside of the board was very warm to the touch. More alarming was the fact that the DRAM sticks were hot to the touch as well. Pictures will follow after I experimentally determine if my conjecture (using available documentation) regarding the choice of capacitors is correct or not.
For anybody trying to take O-Scope readings, let me tell you at the outset that what I am doing is NOT ACCURATE. In fact I would almost say it is the wrong way to measure signals/transients.
Where shall I start? Hmmm....the probes are just two wires, no shielded cable, no capacitive terminations, no impedance matching...In brief, the idea was to show how bad EMI can be if the probe is not grounded. Note that I'm not trying to measure transients, because I need a DAQ for that, but I'm just trying to see if anything crazy is going on with the voltages. I used an old scope. I did take a few pictures and save waveforms, but this is nothing like hooking up the scope to a GPIB interface for some real time data acquisition. I won't do it because it is too much work for some pretty un-interesting DC signals. :)
The point I'm trying to make is, the maximum amplitude I see here is about 3.1V, so I can use an OS-CON at 4V (5V surge rating) with no problems. I connected a wire to the Drain/Source and another to the Ground). I should have used the casing of the FET, but the spot was a bit too tight for me to spend time on.
If the ground is left floating, RFI is picked up and you see the familiar 60 Hz signal from the fluorescent lamps/other devices interferer. When grounded, this noise signal disappears.
Phase: III
Now I'm going to replace all the GSC's in the DRAM area with OS-CONs. Next update with more O-Scope shots will be available tomorrow.
Subscribe to:
Posts (Atom)