Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

[Aetherone]

Rebuilding & resurrecting my first dual-core system

The board itself seems to be in good condition, from testing it's had no troubles with an overclocked Opteron 175 and 4x 1GB DIMMs in it.
The problem is, the PCIe slots don't work. Given the numbers of sad looking caps around the slots, this isn't surprising (one of the caps started smoking in the first testing POST!).

Runs fine with a PCI video card which is how I clocked the CPUs and tested it.
The caps around the CPU socket are all sealed jobbies, so I'm planning on leaving those alone and just replacing the swarm of KZG's everywhere else.

Current count (rev 1.01 board)
3x 16v 1000uf around the CPU aux power socket
2x 6.3v 1800uf at the molex EZ_Plug
24x 6.3v 820uf all over the board (many vented)
2x 16v 470uf under PCIEx4 & PCIE16_2 (x4 right next to a 6.3v 830uf; better not get 'em confused!)
11x 16v 100uf (which I'm presumung don't need replacing)

I'm logging these results for a local forum search here
Replace caps on ASUS A8N32-SLI Deluxe? (2014)
ASUS A8N32-SLI Deluxe (2012)
Help Identify Caps on Asus A8N32-SLI Deluxe (2005)

Keeping in mind whats available in Australia in 2018/2019, what would you suggest I replace these with?
I consider this system to have some sentimental value, so I'd like to do the job properly and am thus looking to avoid whatever no-name cheapie low-ESR caps are available from my local electronics store.

cheers
Aeth..

[Aetherone]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Just trawling around e14 this morning (yes it Christmas morning but I live with severe depression and this is much better than googling noose tying).
Looks like they stock various Rubycon, Nichicon, Panasonic & others. Reasons to pick between each? Pannys seem quite expensive, but they rate much higher runtimes at temperature. Rubys do too, but they're not all available in the right sizes.
Nichicons are pretty cheap. Pros? Cons?

16v 1000uf (8mm x 20mm)

Thinking Rubycon ZLH here

2x 6.3v 1800uf (8mm x 20mm)

Nichicon HM vs Panny FS? Half the listed ESR vs 4x the hour rating? (Panny is also 2000uf)

24x 6.3v 820uf (8mm x 12mm)

Rubycons again. ZLH? ZLJ? Anyone know the effective difference?

2x 16v 470uf (8mm x 13mm)

Lots of options - Panny, Ruby, Nichi, other brands I don't recognise... Just slap in a pair of Ruby ZLHes?

[Per Hansson]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Welcome to Badcaps!
All the brands you listed are good brands, well, the original caps are too, just they fucked up the formula for a quite extended time period.
Usually Panasonic FJ is used interchangeably by the manufacturers in place of UCC KZG.
The Rubycon ZLH & ZLJ might be a tad lower spec than the original, but they should be alright.
Nowdays it's hard to match the original caps, because ultra low ESR alauminium electrolytic capacitors have pretty much entirely been replaced with solid polymer capacitors.
These usually have much lower ESR than even the best ultra low ESR aluminium caps of the past...

[ChaosLegionnaire]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

hi, good to see another interested in retro stuff like me! i recently acquired this board off ebay and thankfully, i was sharp enough to get one that has all rubycon mbz and panny fj caps around the board. glad to assist u in making this board last as long as possible since that seems to be your aim.

Reasons to pick between each? Pannys seem quite expensive, but they rate much higher runtimes at temperature. Rubys do too, but they're not all available in the right sizes.

not really. for example, for small 8x11mm caps, the zlh rates higher in lifetime compared to panny fr. for some other sizes, panny fr lasts longer. u have to check the datasheet and compare which lasts the longest for the size u need.

where the cap is located on the board and what job it is responsible for is also the important reason to pick the right caps for the job.

16v 1000uf (8mm x 20mm)

Thinking Rubycon ZLH here.

this is for the cpu vrm input area and yes that is fine. u can pick panny fr too since it has a 9000 hour lifetime for this size compared to 8000 hours on the ruby zlh and chemicon kzm.

u can also try polymodding with 8mm 470uf 16v poly caps. look for nichicon fpcap, chemicon npcap and panasonic oscon datasheets for the right cap series and size.

so if u are going to be overclocking the cpu, choose poly caps as the lower esr and better ripple filtering specs *may* help somewhat in the overclock.

2x 6.3v 1800uf (8mm x 20mm)

these two caps are responsible for chipset filtering. looking at the vrm for the chipset, it wont have any problems being polymodded. replace this with an 8mm 820-1000uf 6.3v poly cap.

24x 6.3v 820uf (8mm x 12mm)

Rubycons again. ZLH? ZLJ? Anyone know the effective difference?

zlj has lower esr and better ripple but zlh has a longer lifespan. these caps are just responsible for general filtering throughout the board. they filter usb 5v power, ram 2.5v vdimm voltage, 5v sb voltage, cmos battery voltage etc. 820-1000uf 6.3v caps like the ruby zlh, panny fr or chemicon kzm would be the recommended series for longest lifespan.

u can also polymod this with 560uf 6.3v poly caps but poly caps are quite costly and buying 24 of em will break your wallet! so stick to long life rated electrolytics instead for better cost effectiveness.

2x 16v 470uf (8mm x 13mm)

Lots of options - Panny, Ruby, Nichi, other brands I don't recognise... Just slap in a pair of Ruby ZLHes?

these are responsible for filtering the pci-e 12v rail. this isnt really critical as video cards already also have filtering caps for the 12v rail on them. just put in ruby zlh or chemicon kzm since they have the longest lifespan for this cap size.

u can also polymod with 270uf 16v caps but it isnt worth the extra cost incurred from paying more for poly caps here. only use em where they matter most so just stick to long life electrolytics again.

and lastly, make sure u use a good quality jap cap psu with ultra low ripple on the outputs like the superflower leadex platform and evga g2/g3 series. it is well known around here that junk psus are responsible for killing the caps on mobos so be very careful what psu u use with the board! unbranded junk psus go out the window...

[Th3_uN1Qu3]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Sorry to bring bad news, but that board uses a nForce chipset which has the dreaded "Bumpgate" failure baked-in. That is the reason why your PCI-E ports don't work.

Removing the chipset heatsinks, blasting some hot air onto the Northbridge and replacing the thermal compound with something high quality will get it back running - for a short while. Unless you add a fan to that chipset heatsink and cool it really well, it will die again. Not worth bothering with IMHO.

[Dan81]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Sorry to bring bad news, but that board uses a nForce chipset which has the dreaded "Bumpgate" failure baked-in. That is the reason why your PCI-E ports don't work.

Removing the chipset heatsinks, blasting some hot air onto the Northbridge and replacing the thermal compound with something high quality will get it back running - for a short while. Unless you add a fan to that chipset heatsink and cool it really well, it will die again. Not worth bothering with IMHO.

It could just be bad caps. I know lots of nVidia stuff from that era has bumpgate issue, but it could still be bad caps and the slots to not work because of them.

Don't get me wrong, but nForce 4 is the toughest when it comes to this issue. I have a K8N4-E that despite heating up like a pan it won't give up at all. Hence why I'd agree on the caps being bad. Before using mine I had to replace about 3-4 KZGs because they were bulging. It did work fine afterwards.

[momaka]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Looks like they stock various Rubycon, Nichicon, Panasonic & others. Reasons to pick between each?

If you got those brands and they are genuine, then go with whatever is cheapest / easily available.

I've done a board full of bulged 820 uF United Chemicon KZG caps with Rubycon ZLH. Works fine, never skipped a beat. Did a few other mobos with them afterwards - all fine too.

Most caps on the board aren't exactly critical on the ESR and not in stressful spots. So I wouldn't worry about the "endurance" rating too much. The only caps you'd want to go with lower ESR are probably the 16V 1000 uF caps around the CPU and the 6.3V 1800 uF (which I suspect are either for the RAM or Northbridge and likely on a buck-regulated rail). So for these, Nichicon HM/HN/HZ might be the better pick. But Panasonic FR, FM (and possibly FS) should be okay too if HM or higher class are not available. (Truth be told, even ZLH would work - anything is better than bad KZGs really. )

That said, consider not using the board until you get it recapped. Otherwise, the bad caps could do damage to some circuits (namely any buck regulators).

Sorry to bring bad news, but that board uses a nForce chipset which has the dreaded "Bumpgate" failure baked-in. That is the reason why your PCI-E ports don't work.

Removing the chipset heatsinks, blasting some hot air onto the Northbridge and replacing the thermal compound with something high quality will get it back running - for a short while. Unless you add a fan to that chipset heatsink and cool it really well, it will die again. Not worth bothering with IMHO.

Yes, it's possible the O/P may have a bad NB.

But if the board has bulged caps, there's also a big chance those could be the issue. So I think it's best to first fix the caps first and see if that gets the PCI-E slots going. If not, then perhaps attempt a reflow. And either way, if the board gets up and running, I agree with your suggestion - a bigger, more powerful fan + heatsink for the NB should be considered.

[Aetherone]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Thanks guys, I appreciate the feedback. Looks like element14 has everything I need to get this baby recapped.

polymodding

Thanks for the info, I'd been debating about polymodding it but have been head-scratching over the right values and thinking it falls into the "too hard" basket.

and lastly, make sure u use a good quality jap cap psu

I've got an original Corsair HX620 tucked away for it already. (y) If I recall correctly, 105 degree all-japanese caps was their big selling point way back when. The 5 year warranty was simply unheard of at the time
12 years on, I still have four of the original HX'es in service, two of them running modern hardware on a daily basis!

That said, consider not using the board until you get it recapped.

Oh yes. I picked it up for a song a few years back, sorted the CPU for it and ran it just long enough to be happy that the rig was OC stable (albeit with a PCI video card).

Sorry to bring bad news, but that board uses a nForce chipset which has the dreaded "Bumpgate" failure baked-in. That is the reason why your PCI-E ports don't work.

This is true, but I'm leaning towards the vented caps being the more likely culprit. The A8N32-SLI board has a MASSIVE copper heat-pipe cooling solution that cover the NB, the VRM and the PCIe switch so they ran dramatically cooler than other contemporaneous boards.

The older A8N-SLI boards just had crappy little aluminum coolers that used to run crazy hot with dinky 35?mm sleeve bearing fans that died so regularly it used to be a big part of my day job swapping those suckers out. Asus distie used to send us replacement fans in packs of 20.

I've literally just finished baking a GTX275 that I suspect was suffering from bumpgate (what else does one do on the first day of the new year?).
Just have to replace the solid caps that didn't like 220*C
Thread to confirm cap choices is over here: https://www.badcaps.net/forum/showth...569#post871569
Good thing I haven't placed my Element14 order yet...

[ChaosLegionnaire]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Sorry to bring bad news, but that board uses a nForce chipset which has the dreaded "Bumpgate" failure baked-in.

I know lots of nVidia stuff from that era has bumpgate issue

Yes, it's possible the O/P may have a bad NB.

lol yea, haha! in my excitement to reply, i forgot this has the dreadful bumpgate defect. ah well... thats life! lol

The A8N32-SLI board has a MASSIVE copper heat-pipe cooling solution that cover the NB, the VRM and the PCIe switch so they ran dramatically cooler than other contemporaneous boards.

yes but im afraid to say that despite that, the a8n32-sli still runs quite hot. i have this board and i've already stress tested it before. i believe it has something to do with the dual pci-e x16 sli function of the northbridge so it dissipates a lot of heat due to that versus the dual x8 pci-e sli on the older a8n-sli.

since it looks like u want to run dual gtx 275 cards in sli which run at over 200w tdp *each*, momaka's advice still holds true. u better ensure active cooling on the nb, sb and vrm. try to ghetto mount some 50-70mm fans around the area and make sure the case is adequately ventilated with lots of case fans. i'd imagine your computer should sound like a boeing 787 dreamliner taking off! this means its properly ventilated to handle over 500w of tdp from the gpu, cpu and chipset.

The older A8N-SLI boards just had crappy little aluminum coolers that used to run crazy hot with dinky 35?mm sleeve bearing fans that died so regularly it used to be a big part of my day job swapping those suckers out. Asus distie used to send us replacement fans in packs of 20.

yepp i have two if those a8n-sli boards. one came with a dead chipset fan which i had to replace with a zalman nb cooler! not sure how long it will last since it already had a dead chipset fan and im not sure how long the previous owner ran it like that!

not only that, many of those small fans of 40mm and smaller are just plain crappy and dont last for some reason... ah well... thats electronics for u lol!

[Aetherone]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

yes but im afraid to say that despite that, the a8n32-sli still runs quite hot.

Ja, the extra PCIe switch chip adds something like 25w over a standard NF4 set. The A8N-SLI Premium was probably the sweet spot board if you could live with 8x8.

u better ensure active cooling on the nb, sb and vrm.

I've got several BNIB of the special blower fans Asus made to fit their cooling system; although I'm planning on modding a 240mm AIO to fit the 939 bolt holes and mounting a 120mm over the block/VRM area to push air around there. The first gen (2000) Lian-Li PC-60 its going in has been modded to have two 80mm exhausts above the I/O.

try to ghetto mount some 50-70mm fans around the area and make sure the case is adequately ventilated with lots of case fans.

Awww, but period correct means running at 65*C at idle right?

Always amazing to look back and see how far we've come. It was always so hard to deal with this kind of thermal load in those days, needing weird esoteric (noisy!) parts that are now just off-the-shelf.

The only bummer is this board is one generation too old to have 4-pin PWM fan control, although the CPU header is good for 2 amps which is a pretty hefty fan.

[Aetherone]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Success #1: no smoke!
Success #2: It posts. No dramas. Opteron 175 and 4x1GB DDR400 all recognised fine. Memtests fine.
Success #3: two cards, a 640 and a 610 (both bus powered) are recognised and negotiate x16 lanes each.
Success #4: Runs furmark on the 640 all night long.
Failure #1: Socket 939 chips don't support anything newer than Win 8.0.
Success #5: found my Windows XP x64 deployment image, phew!
Failure #2: 3Dmark06 on graphics looping as my favourite 3D stress test locks the system up after 3ish hours. Unsure as to exactly why.

Next on the docket: nick the matching GTX770's out of their systems and LOL test them in SLI. Find the Windows 7 x64 deployment image. Prime load it for 24 hours. Find the 939 hold-down kit for the Corsair H100 AIO cooler and clock the balls off it. Play Battlefield 1942 or Medal of Honour...

[Dan81]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Success #1: no smoke!
Success #2: It posts. No dramas. Opteron 175 and 4x1GB DDR400 all recognised fine. Memtests fine.
Success #3: two cards, a 640 and a 610 (both bus powered) are recognised and negotiate x16 lanes each.
Success #4: Runs furmark on the 640 all night long.
Failure #1: Socket 939 chips don't support anything newer than Win 8.0.
Success #5: found my Windows XP x64 deployment image, phew!
Failure #2: 3Dmark06 on graphics looping as my favourite 3D stress test locks the system up after 3ish hours. Unsure as to exactly why.

Next on the docket: nick the matching GTX770's out of their systems and LOL test them in SLI. Find the Windows 7 x64 deployment image. Prime load it for 24 hours. Find the 939 hold-down kit for the Corsair H100 AIO cooler and clock the balls off it. Play Battlefield 1942 or Medal of Honour...

An perfect example that even if the chipset is known to have the "bumpgate" issue, it's not always the problem

[Aetherone]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

An perfect example that even if the chipset is known to have the "bumpgate" issue, it's not always the problem

Pretty sure it's claimed one of the matching pair of GTX275's I had for this build but yes, its not always bumpgate.

Success #6: both onboard ethernet NICs link and pass traffic at 1Gbps (well, most of 1G, I'm CPU limited).

Running benchies now with the first GTX770. Gah, I'm horribly CPU limited. Less than half the score this card gets in its hexacore 1366 Xeon @ 4.6Ghz normal home :p

[Aetherone]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Unexpected hiccup. NF4SLI is 2005ish? GTX770's are 2013ish? Seems that NV have dropped support for SLI'ing them on such an old platform, at least under XP x64. Time to find that Win7 disc and see if anyone knows when NF4 is/was dropped from the SLI approved listings.

[momaka]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

Success #1: no smoke!
Success #2: It posts. No dramas. Opteron 175 and 4x1GB DDR400 all recognised fine. Memtests fine.
Success #3: two cards, a 640 and a 610 (both bus powered) are recognised and negotiate x16 lanes each.
Success #4: Runs furmark on the 640 all night long.
Failure #1: Socket 939 chips don't support anything newer than Win 8.0.
Success #5: found my Windows XP x64 deployment image, phew!
Failure #2: 3Dmark06 on graphics looping as my favourite 3D stress test locks the system up after 3ish hours. Unsure as to exactly why.

Well, great to hear its running again!!

As for failure #1: I wouldn't call that a failure. I would call it a HUGE SUCCESS! - Anything that refuses to run Windows 8/10 is GOOD in my eyes

Awww, but period correct means running at 65*C at idle right?

LOL! You're not too far from the truth, really.

Running benchies now with the first GTX770. Gah, I'm horribly CPU limited. Less than half the score this card gets in its hexacore 1366 Xeon @ 4.6Ghz normal home :p

Yes, socket 939 CPUs were great for their time and for once a good bit ahead of Intel. But the Core architecture was a game-changer for Intel again. Even the AMD Athlon FX-60 (the fastest dual core s939 chip) is kind of slow compared to most mid/high -range Core 2 Duo CPUs. If I am not mistaken, an E6700 performs somewhat better than the Athlon X2 6000+ for socket AM2. And that X2 6000+ performs a slight bit better than the FX-60. So in essence, even if you are running close to or above FX-60 speeds, your performance would still likely be somewhat similar to that of a mid-range C2D at stock speed.

Thus, what I have figured (as I like to play with retro systems like these as well) is that it's best to stick with era-appropriate hardware. For that s939 platform, something like the GeForce 8800 GT or 9800 GT or Radeon HD3870 is where I would stop. Anything beyond that, I don't think you will see much added benefit. A pair of GeForce 7800 GTX cards in SLI would have been a dream-machine back then. If you look around / wait patiently on eBay, sometimes they popup for less than $10 a piece.

BTW, what CPU did you end up putting in that motherboard? (Can't see it being mentioned anywhere... or perhaps I am blind? )

[Aetherone]

Re: Blast from the past: Asus A8N32-SLI (Socket 939 AMD glory days)

If I am not mistaken, an E6700 performs somewhat better than the Athlon X2 6000+ for socket AM2.

The real killer for the Core2 gear was it's thermal and power advantage, plus the massive overclocks you could pull with such headroom inside the severely limited cooling of the time.

A pair of GeForce 7800 GTX cards in SLI would have been a dream-machine back then.

When I first built mine way back in the day it had just one of those. Just so happens that I have a BNIB 7900GTX stashed away for this, although I reckon I upgraded it to SLI 275's before finally jumping ship to Core2 gear.
The GTX770's are just because I happen to have two of them laying about in other machines...

BTW, what CPU did you end up putting in that motherboard?

Opteron 175. 2.2Ghz of dual core heat production

Now to crack on with rolling the Win7 image to see if that does any better with SLI ... or SLI hax

make sure u use a good quality jap cap psu

Scored 1 lightly used Corsair HX620 off Gumtree for $10 yesterday. Five in the stable now...