Poly-Modding: Does it actually work?

IBM NetVista 8305-21A

The original capacitors

Unfortunately, I didn’t get any pictures of this PC’s motherboard before I replaced the capacitors, but I do remember what went where. The more critical areas like the CPU VRM all had Rubycon MBZ series capacitors, which were holding up just fine, which comes as no surprise considering Rubycon’s excellent reputation for reliability. There were three 1500µF 16V capacitors on the CPU VRM in and eight 1500μF 6.3V capacitors on the CPU VRM out. The other capacitors were three 1500μF 6.3V between the CPU socket and the RAM slots, a single 3300μF 6.3V below the CPU socket, near the northbridge, all for which were from Rubycon. There were also five 330μF 25V and a single 680μF 6.3V (which was bulging) from the infamous Taiwanese manufacturer GSC, which are now called Sacon.

The replacement capacitors

A common belief about poly-modding is that you can use about half the capacitance of the original electrolytic capacitors. However, this only applies to the CPU VRM and similar DC-DC conversion circuits where the capacitors are used for filtering ripple. In other parts of a PCs motherboard, they may be used for timing controls, and changing the capacitance there can throw off the timings and make the motherboard unstable or even cause it not to pass POST, (Power On Self Test). I will be sticking as close to the original capacitance as possible outside the CPU and RAM VRMs, since the purpose of this experiment is not to prove what values of caps should be used, but rather, to see whether it is actually possible to completely poly-mod a motherboard.

In this case, the manufacturer made it quite easy to tell which caps were on the VRMs, by using Rubycon there and GSC everywhere else (and yes, I did double check the voltages on them – just to be sure). On the CPU VRM low side, it is safe to use 2.5V capacitors in place of the original 6.3V ones, since they will be filtering the CPU Vcore’s voltage, which will only be around 1.5V, well within the 2.5V maximum. If you’re going to change the voltage anywhere else, though, measure it with a volt-meter to make sure you won’t be overvolting your new capacitors, otherwise they may explode when you first power the board on. In this case, I found that all but the CPU VRM in and the single 680μF and 3300μF capacitors were rated for a higher voltage than required, so I used the following values:

(Attention – as this is a DDR1 board, and DDR1 RAM ran at 2.5V nominal (2.6V for the highest speed), 4V or higher rated capacitors should have been used there.  —LongRunner)

The single 680µF cap had a blank spot right next to it, and a test with my DMM confirmed that it was in parallel. Since I don’t have any 680µF 6.3v polies on hand, I decided to use a pair of 330µF 16V polies – 1 in place of the original and 1 in the blank spot, so the total capacitance would be 660µF, which is very close to the original capacitor.

_

The Stress-Testing

I must say that I think the board looks quite nice with all those polymer capacitors, (if I do say so myself), but this is where I’ll find out whether the board likes it or not. The system will now be subjected to five 24-hour stress tests, all of which it must pass without getting errors, locking up or getting a Blue Screen Of Death (BSOD). Prime95 will be my first stress test utility. It comes with 3 different modes – Small FFTs, In-Place FFTs and Blend. Tests 1-3 will consist of 24 hours under each of these 3 modes. The next test will be Memtest86+, again, for 24 hours. The 5th and final test will be using Stanford’s Folding@home. The computer will be made to process and return at least 1 work packet (which, on an older system, will take around 24 hours for a larger packet).

The poly-modded NetVista passed all the stress testing with flying colours, which proves that you don’t need to reduce the capacitance anywhere except on the DC-DC filtering circuits, but before we can call it a rule of thumb that motherboards can be poly-modded outside the VRMs, we’ll have to try it on a couple of other boards and see what they think, so lets get on to our next test subject.

Pages: 1 2 3 4 5 6 7 8 9 10 11