TELVM wrote:

Most PC fans are electrically 'noisy' and come without any filtering. It's possible that those particular 140mm fans are creatin […]
Show full quote

dottoss wrote:

... having a little problem powering everything with the MOLEX connectors. Nvidia control panel sometimes complains that the card is downclocking as it is not getting enough power. The 6800 Ultra is powered by two molex connectors ...

dottoss wrote:

...I removed the two 140mm fans from the SATA pwr cable and plugged them into the MB fan outputs and now the error message is gone! ...

Most PC fans are electrically 'noisy' and come without any filtering. It's possible that those particular 140mm fans are creating ripple, which messed with the 6800 Ultra when they were on the same peripheral wire that feeds the GPU.

Now that they're plugged directly to a mobo fan header, the mobo's onboard filtering caps may be getting rid of their ripple.

Though long in the tooth the HX620 is a fine PSU, Seasonic guts & Hitachi/Chemicon caps.

Interesting! The thing though, they are not on the same "wire" but who knows how it is wired inside the psu..

Another idea I got now, perhaps I can try to find an adapter from 6pin pcie power cable to molex and run the card from those two rails. Anyone know if this is a potential voltage/current issue using the pcie pwr 6 pin on a AGP card?

dottoss wrote:

Another idea I got now, perhaps I can try to find an adapter from 6pin pcie power cable to molex and run the card from those two rails. Anyone know if this is a potential voltage/current issue using the pcie pwr 6 pin on a AGP card?

I guess that in theory one could even build an Athlon XP without P4 plug and use an adapter (not sure such even exist) to convert that P4 plug to a molex. (just to clarify, it's for 12v only so it may also just be some fan adapters or something)

Afaic some of the older PSUs that had like 30A on 5v and 2 12v lines used one of the 12v lines solely for the P4 plug (which might go unused when the P4 plug is not used). I'm talking about the era when no PCI-E plugs existed on PSUs.

But I'd like to hear from someone more knowledgeable about PSUs first though. This is just a thought I had a while back.

dottoss wrote:

... who knows how it is wired inside the psu..

It was advertised as three +12V rails, but reviewers found it was single rail in practice:

"... So it is my opinion that we essentially have a single 12V rail PSU here. Certainly there is nothing wrong with this given the problems high end video cards have had with getting enough power from a single 12V rail when the OCP is set to the typical 240VA limit. But we do lose the advantages of multiple rails such as protection from damage to one rail from a short on another and the simple "filtration" of noise introduced from one rail to another."

dottoss wrote:

Another idea I got now, perhaps I can try to find an adapter from 6pin pcie power cable to molex and run the card from those two rails. Anyone know if this is a potential voltage/current issue using the pcie pwr 6 pin on a AGP card?

Not worth thr hassle IMHO. Thing is ultimately you have to plug two Molex into that GPU. Molex is a weak design of connector conceived just for low draw peripherals, and is hopeless for anything with higher draw (voltage drops, arcing, fried connectors, etc). That's why ultimately they introduced PCI-e connectors for GPUs.

I'd just clean the metal pins inside all Molex connectors involved, make sure they're all firmly connected, and if possible use two Molex from two separate wires to feed the GPU.

If we're handy with a soldering iron, we can add extra filtering capacitors close to the connectors of the device with voltage/ripple problems.

^ This example is on PCI-e wires but the same can be done on Molex.

These extra caps help with ripple filtering (the HX620 is a bit rippletastic on +12V), and with handling relatively huge power transients typical from GPUs.

TELVM, may I say that adding of extra caps this way to reduce ripple locally seems pure genius to me, even the thought of this idea never crossed my mind! Cheers! 😁

🤣 Decoupling caps at the end of wires is an old and not so uncommon trick amongst PSU makers. Just some examples:

^ 2009 Antec Truepower Quattro 1200W

^ 2013 Silverstone CP06 multi SATA wire

^ 2014 EVGA Supernovae G2

^ Corsair RMi 'Type IV' wires

^ Bitfenix Whisper M

They're not as common as (IMHO) they should because they cost more and make for bulkier and 'uglier' wires.

But who cares if it makes cables more ugly as long as it works?
Bulkier cables are obviously not always practical but I'd still much rather have the extra stability.

^ Me too, no question, but apparently we're a minority. Most people couldn't care less about the electrical benefits (if they know about) and just see the bulkier and uglier cables. Some 'experts' even believe that 'cable caps are the sure sign of a poor quality PSU', poor devils 😁 .

And then of course there are the 'sleeving fundamentalists', they just get rid of the caps for aesthetic reasons 🤣 . Google 'remove cable capacitors' for a laugh.

(Apologies to dottoss if we're hijacking his thread).

TELVM wrote:

Apologies to dottoss if we're hijacking his thread

Please don't, I find it highly interesting 😀 Also, you have helped me understanf the problem with the fans, and if it is indeed asingle rail as reported in the review of the psu, it will not matter where I connect these fans except if I go through the motherboard or get handy with some caps 😀 Thank you!

A small update to the build;

While this is not aesthetically the best, it is very functional:

I used a Zalman CPU Cooler CNPS9500 bracket and then used a low profile pci bracket to even out the pressure a bit. I'm currently on the hunt to find a thin circular or square metal plate of the size of the S36 Kelvin CPU cooler to spread out the pressure a bit more (I'm not sure how happy the s36 pump is about the current pressure) and make it look a bit nicer.

CPU temps: Zalman CPU Cooler CNPS9500:
Idle: 38C
Load: 49C

S36 Kelvin:
Idle 24C
Load: 27C

Wow, great build. I am curious how you found the 3.4EE? They are awfully hard to come by these days it seems.

ATi_Loyalist wrote:

Wow, great build. I am curious how you found the 3.4EE? They are awfully hard to come by these days it seems.

Thank you 😀

I have one strategy finding a specific cpu that is not easily obtainable on ebay etc (or any other item that goes for premium prices now days)

Look for complete computers with advertised P4 3.4GHz and if it "feels" relatively high-end for it's time then ask for pictures of the cpu and/or make sure you get the S-spec number: SL7CH.

If you're lucky, it is a Extreme edition, then proceed to buy the whole computer. Sometimes, computers are sold without the sellers knowledge of what "type" the cpu is but they know what "GHz" is so they advertise this. It is time consuming but is what i did.

That's a great tip! Will keep hunting 😁