4960X stable at 4.8Ghz 1.46v, 1.1v VCCSA - ok for long term?

[Intel_of_Borg]

I just repaired the 2x 12v pins on my PSU connector and the board. 12V PSU ATX pins were melted and dropping to 11.4V under load.
Since the 24pin was repaired I can now get the 4960X stable at 4.8Ghz 1.46v high LLC which reads as 1.468v under load. (mem at 1866 1.65v)

12v ATX now reading at 12.12v under CPU load (12.288v idle) with 64bit linpack and 11.7v with OCCT powersupply test (which is a brutal test running 4way GPUs and linpack together!) hottest core was 88c and coolest 84c after 15 mins. I think im pretty happy with that considering nothing I do pushes the CPU that hard so Ill probably end up with 75c max.

What do you think is that safe for long-term use?

Also I have 4way 3GB GTX580's with VGA hotwire pulling 1.3v @ 1000/2200. Is there anything I can do to supplement the 12V lines to stop the ATX pulling so much current and melting pins? Im using the 8-pin and 4-pin CPU 12V and the PCIE 12V ezy-plug. Also using 2x 1500W max revo PSUs with load split across ATX, GPUs and accessories. From the wall @ 240V I'm pulling 1400W from one and 900W from the other PSU under OCCT load test. Any recommendations?

[Raja]

Simple answer: The best thing you can do is stop running OCCT to load the system. 🙂

Were you using the EZ plug when the initial meltdown occured?

If you are not bothered by DIYing and warranties, then I'd look up a PCIe connector diagram and augment the "high current pins" (see which rail is pulling the most) directly by soldering a reasonable gauge wire from the PCIe AUX (EZ) for 12V or the requisite source (if 3.3V).

-Raja

[Intel_of_Borg]

Raja@ASUS wrote:
Simple answer: The best thing you can do is stop running OCCT to load the system. 🙂

Were you using the EZ plug when the initial meltdown occured?

If you are not bothered by DIYing and warranties, then I'd look up a PCIe connector diagram and augment the "high current pins" (see which rail is pulling the most) directly by soldering a reasonable gauge wire from the PCIe AUX (EZ) for 12V or the requisite source (if 3.3V).

-Raja

Thanks Raja;
Yes I know, no need to "play" OCCT 🙂 but its a good indication of system stability when you run the powersupply test. I find running Prime or linpack on its own doesn't pick-up other system instability issues like that beast of a test does.

Yes I always use the 8pin+4pin and EZ-plug with such a high load, I wouldn't think it would be safe or stable otherwise!
Its possible it started melting when I was running a long series of Bluray rips over a few days and the CPU was pegged at 100%.

Thanks, I'll consider soldering another molex to the high current pins underneath then for long-term safety and run another dedicated 12V line from the lower stressed PSU. I think I still have one 12V rail spare.

This old board has a failed HD audio section (analog doesn't work, optical does) and another new board I just bought has DOA intel Sata ports. So I'll wait till I can get the new board replaced and then send the old away for RMA before I mod it.

BTW what's the process to RMA, should I do it direct with ASUS and then give to supplier or just give to the supplier to handle RMA?

Cheers

PS. Here's some pics of the beast in my build-log. http://forums.overclockers.com.au/showpost.php?p=14863959&postcount=165

Close-ups of the RIVE fully dressed. 😉
WARNING LARGE IMAGES http://forums.overclockers.com.au/showpost.php?p=14071652&postcount=81

[DooRules]

I have been trying to decide on whether to get the 4930 or 4960 chip to replace the 3960x I currently have. My original thought was it would give me a stronger IMC most likely but it seems that high chip clocks are simply not there to be had. I mean 1.46 Vc for just 4.8 is terrible when compared to my current 3960x

I have read Raja's post about higher clocked ram not mattering much except in ram benching so just chasing ghosts there, thanks for that info Raja.

That leaves me thinking there is no way I will get a 4930 that will clock anywhere close to my current chip, which can with relative ease clock and bench at 5.3 on the rive and 5.4 with some tinkering, on water. Looks like my best bet is to stay with the 3960 and 2133 ram for RIVE BE.

Maybe I will retask the chip money to some of them new gpu's floating around. 😄

[Intel_of_Borg]

DooRules wrote:
I have been trying to decide on whether to get the 4930 or 4960 chip to replace the 3960x I currently have. My original thought was it would give me a stronger IMC most likely but it seems that high chip clocks are simply not there to be had. I mean 1.46 Vc for just 4.8 is terrible when compared to my current 3960x

I have read Raja's post about higher clocked ram not mattering much except in ram benching so just chasing ghosts there, thanks for that info Raja.

That leaves me thinking there is no way I will get a 4930 that will clock anywhere close to my current chip, which can with relative ease clock and bench at 5.3 on the rive and 5.4 with some tinkering, on water. Looks like my best bet is to stay with the 3960 and 2133 ram for RIVE BE.

Maybe I will retask the chip money to some of them new gpu's floating around. 😄

Yeah I would probably put the money into GPUs if I were you. I only upgraded to the 4960X because I thought my 3930K was degrading and causing hard shutdowns. In hind-sight it looks like the shutdowns were triggered by under voltage on the 2x 12V+ lines on the ATX connector that melted. Now I have two RIVEs and two CPUs I'll be able to build an encoding server and workstation once the RMAs are sorted. 🙂

Though with the 15%-20% clock for clock increase in performance, in theory the 4960X @ 4.8Ghz is equivalent to ~5.5Ghz. Don't think it quite translates that way but it's certainly at least equivalent to 5Ghz without the high volts my 3930K needed at 5Ghz. (1.5v)

[Raja]

Running these kinds of loads, no matter how good you feel they are might damage the board, so I'd take that into account in future endeavors (it's enough for RMA refusal).

I don't know the exact procedure of RMA in your region to be honest - Google ASUS service Australia and see what comes up. You might be directed to the global service portal, which will take details from you and someone should pick up the case and take it from there.

[Intel_of_Borg]

Raja@ASUS wrote:
Running these kinds of loads, no matter how good you feel they are might damage the board, so I'd take that into account in future endeavors (it's enough for RMA refusal).

Really? Im not doing anything the board isn't designed to do. It is an extreme overclockers board after all and I haven't made any mods other than a waterblock to cool the VRMs!?!?!?!

Thanks ill check online locally for RMA process.

[Raja]

Intel of Borg wrote:
Really? Im not doing anything the board isn't designed to do.

If that's the case you would not need to mod it. These kinds of loads are not even generated under LN2 use because the competitive benchmarks don't pull this kind of insane current. If such a case hits RMA it would probably be deemed customer induced damage (misadventure) - and as you know it can happen, it only compounds the verdict.

[Intel_of_Borg]

Raja@ASUS wrote:
If that's the case you would not need to mod it. These kinds of loads are not even generated under LN2 use because the competitive benchmarks don't pull this kind of insane current. If such a case hits RMA it would probably be deemed customer induced damage (misadventue) - and as you know it can happen, it only compounds the verdict.

Thanks again Raja! Sounds reasonable when you put it that way... I guess I am insane then... It was designed for 2way and 3way VGA hotwire though. 🙂 I think the 580's are pulling 400W each under load along with the 350W for the CPU. 😉 It's lasted a year or more like that without problems and I just realised it was Bioshock Infinite that probably pushed it over the edge because of the huge combined workload running 5870x1080 res.

[Myk_SilentShado]

Hey fellow Aussie!!!! love the screen name btw 😉 wow pulling 1400W from a 1500W PSU is pretty insane and I dare say you are smashing way through the PSU's rated capacity, most push 80% before things start to take their toll and start to kill the PSU.