Retro polymod: Abit NF7-S 2.0

[quicknick]

Hi everyone,
Got myself the board in the title, from the local fleamarket. Have really good memories about it (had it back in the day), and as I recently got interested in retro stuff... why not?

Board was in very good shape, except for bulging on 13 (out of 21) Nichicon HM 1000uF/10V, and another bulged 1800uF/6.3V near the AGP VRM. Tried a power-on nevertheless, another one of them (near the BIOS chip) spewed its juices when I applied power. It didn't start at all.

For a quick check, and too see whether I should spend money on a full recap, I took 10 caps with similar specs from another board and swapped them in. I replaced those that were in the "more critical" areas - RAM VRM, or near the chipsets etc. I also replaced the AGP cap with a new Panasonic 2200uF/10V.

After replacing those, board was alive and well! With an Athlon XP 3000+ I had around, I even managed an overclock to 2500MHz@1.9V (from the original 2167@1.65).

Fast forward a few days, replacement caps arrived from Farnell, all solid polymer. Although they didn't look damaged at all, I started with the CPU VRM caps, replacing them as follows:

(4x) Rubycon ZL, 1200uF/16V (22mohm/2150mA) replaced with Wurth Elektronik 820uF/16V (12mohm/5500mA)

(5x) Rubycon MBZ, (3300uF/6.3V) replaced with Wurth 2000uF/6.3V (7mohm/6640mA)

Well... testing after this shows that I lost the overclock to 2500MHz, most it can do now is 2400@1.85V. This comes as a bad surprise for me, as I expected to at least keep the same OC headroom. (OK, I hoped that it would get better ). Any ideas on what might have happened? Is Wurth Elektronik a bad/"stay away" brand? Are the new caps "too good" for the rest of the Abit VRM?

[I don't think it has any influence, but on this first round I also replaced again the cap near the AGP VRM, as the temporary replacement Panasonic was too tall and I also left it with longer legs when soldering, to be able to re-use it when I need. For the rest of the board I bought Kemet capacitors, in this spot I used 1500uF/6.3V instead of the original 1800uF. I guess I should've visited the board to ask about these brands (Kemet, Wurth) before buying quite a bunch of them - I'm planning full recaps to another two boards, one Abit NF7 (non-S) and one KT7A. Quite an Abit fan, I know ]

[RJARRRPCGP]

Re: Retro polymod: Abit NF7-S 2.0

Hi everyone,
Got myself the board in the title, from the local fleamarket. Have really good memories about it (had it back in the day), and as I recently got interested in retro stuff... why not?

Board was in very good shape, except for bulging on 13 (out of 21) Nichicon HM 1000uF/10V, and another bulged 1800uF/6.3V near the AGP VRM. Tried a power-on nevertheless, another one of them (near the BIOS chip) spewed its juices when I applied power. It didn't start at all.

For a quick check, and too see whether I should spend money on a full recap, I took 10 caps with similar specs from another board and swapped them in. I replaced those that were in the "more critical" areas - RAM VRM, or near the chipsets etc. I also replaced the AGP cap with a new Panasonic 2200uF/10V.

After replacing those, board was alive and well! With an Athlon XP 3000+ I had around, I even managed an overclock to 2500MHz@1.9V (from the original 2167@1.65).

Fast forward a few days, replacement caps arrived from Farnell, all solid polymer. Although they didn't look damaged at all, I started with the CPU VRM caps, replacing them as follows:

(4x) Rubycon ZL, 1200uF/16V (22mohm/2150mA) replaced with Wurth Elektronik 820uF/16V (12mohm/5500mA)

(5x) Rubycon MBZ, (3300uF/6.3V) replaced with Wurth 2000uF/6.3V (7mohm/6640mA)

Well... testing after this shows that I lost the overclock to 2500MHz, most it can do now is 2400@1.85V. This comes as a bad surprise for me, as I expected to at least keep the same OC headroom. (OK, I hoped that it would get better ). Any ideas on what might have happened? Is Wurth Elektronik a bad/"stay away" brand? Are the new caps "too good" for the rest of the Abit VRM?

[I don't think it has any influence, but on this first round I also replaced again the cap near the AGP VRM, as the temporary replacement Panasonic was too tall and I also left it with longer legs when soldering, to be able to re-use it when I need. For the rest of the board I bought Kemet capacitors, in this spot I used 1500uF/6.3V instead of the original 1800uF. I guess I should've visited the board to ask about these brands (Kemet, Wurth) before buying quite a bunch of them - I'm planning full recaps to another two boards, one Abit NF7 (non-S) and one KT7A. Quite an Abit fan, I know ]

I had a Barton 3000+ with stepping code AQZFA where I was at 2.431 at only 1.80 V, IIRC. I dunno if I would even need 1.90 V for 2.5. IIRC, Prime95 would pass if I kept blowing cold air into the case.

That was with the Asus A7N8X-X, the one with the dreaded Chemi-con KZG caps! But that was years before the caps did their version of a Hawaii eruption.

[ChaosLegionnaire]

Re: Retro polymod: Abit NF7-S 2.0

I took 10 caps with similar specs from another board and swapped them in.

After replacing those, board was alive and well! With an Athlon XP 3000+ I had around, I even managed an overclock to 2500MHz@1.9V (from the original 2167@1.65).

Well... testing after this shows that I lost the overclock to 2500MHz, most it can do now is 2400@1.85V.

to determine whether if its really the caps, please state exactly what are the specs of those "10 caps with similar specs" that u used to perform the test recap first.

my initial assertion is that the vrm of the board prefers higher capacitance rather than lower esr but need more info to prove or disprove this.

also when overclocking, dont neglect the cooling of the cpu vrm mosfets. try touching the mosfets during overclocking and at max load while running a stress testing program like prime95. do they get burning hot?

except for bulging on 13 (out of 21) Nichicon HM 1000uF/10V
For the rest of the board I bought Kemet capacitors

for those 820-1000uf caps that are found scattered around on motherboards, these caps merely perform general filtering of the minor rails, 5v usb power, agp and vdimm filtering, 5v sb power etc. these are not electrically stressful areas and u dont need to spend money on expensive polymers there. buying 21 polymers for those areas will make a big hole in your wallet! those areas work just fine with panasonic fr, rubycon zlh or chemicon kzn which are 6000-8000 hour long life electrolytic caps.

lastly, u didnt post an image of your board, so i had to find one from around the web. is the attached image the correct picture of your board? i need a proper image of the board to study the cpu vrm to see how its designed to help u with the overclocking difficulties u are experiencing with the board.

[quicknick]

Re: Retro polymod: Abit NF7-S 2.0

Yes, that's the board. Mine also has SATA, but otherwise it's identical.

The caps that I replaced first were 11 out of 14 or so that were bulged. After this step, the board started working. 10 of them were Nichicon HM 1000uF/10V, replaced them with Rubycon YXG 1000uF/6.3V from a dead Gigabyte board. I'd have replaced the other three but 10 was all I could salvage from the dead board. Coincidentally, my other Abit NF-7 uses the exact same Rubycons instead of the HMs (and none is bulged). The 11th replacement was at the AGP slot, very bulged Nichicon HM 1800/6.3, replaced with Panasonic FR 2200/10. Second NF-7 board has the same HM in this spot. Bulged, no surprise.

I suspect this last replacement was essential in getting the board to work, because with the bulged cap I measured the output of the AGP VRM and it was something around 0.3-0.4V (1.8 or 1.5V after replacement).

After the above, overclock was at least workable to 2500MHz. I had no time for proper stability testing, just OS booting.

Then I started the full re-cap, as described in the first post, starting with the CPU VRM caps and the AGP cap.

Here's an in-depth description of the board's VRM. I'm not going to add the fourth phase because it doesn't look pretty at all, but what do you think about populating the pads on the back of the socket? What would be better, the 20 or so ceramics, or maybe solder 4-6 more polymers there to make up for the lost capacitance?

[Per Hansson]

Re: Retro polymod: Abit NF7-S 2.0

I count 15 turns on the coils for the VRM, that is not going to take well to solid polymer caps.
The design is low frequency and the ESR of solid polymer caps is too low for that...

[stj]

Re: Retro polymod: Abit NF7-S 2.0

lowering the uf on the vrm caps just because the esr is lower is not a good idea IMO.

[ChaosLegionnaire]

Re: Retro polymod: Abit NF7-S 2.0

well if u read the link quicknick posted about the discussion on xtremesystems about the board's vrm, someone mentions the switching frequency on the nf7's cpu vrm to be around 180 khz. not sure if that is considered a fast or slow frequency, so im gonna have to defer to the smps experts here to determine if that is a fast or slow switching frequency.

if its slow, then it wont take well to polymodding. the OP must use the same capacitance as the original or better/higher. in that case, the OP should put the rubycon zl and mbz caps back as they are fine. the only thing to be careful of is that electrolytic ultra low esr caps dont like heat even though they are 105°C rated, so some active cooling with a 40-60mm fan around the vrm area over the caps as well as mosfet heatsinks would be needed to make the board last as long as possible.

[quicknick]

Re: Retro polymod: Abit NF7-S 2.0

Thank you for all your suggestions and advices. Job is done, here it is:

After swapping all the capacitor, board seems to have regained some of the original OC capability. Further tests are needed, but not much time to do that ATM. For one, I clearly need to test with a stiffer PSU.

The coils are not 15 turns, rather 5.5 turns or so, with three wires in parallel.

I chose the lower capacitance because that is suggested everywhere as a rule in poly-mod. Now it's the first time I hear this isn't a good practice. Well, it's done, and I ain't coming back to the original caps, even for the CPU VRM.

[stj]

Re: Retro polymod: Abit NF7-S 2.0

the halving of capacitance was just an experiment done on old apple hardware - that seemed to work.

the esr can be interpreted as the response speed - lower = faster. but if it's too low it can be miss-interpreted as a short-circuit.
but capacitance = storage and the circuit will require a minimum level of stored energy to work.
in the case of a psu, to fill in the gaps when the transistors are off.