ground

currently have 3 dead/damaged rex here. #1 - dead vcore fujitsu + some conductive thermal paste near NB shorting the nb vrm. Board appears to be working fine again. #2 - dead vcore fujitsu, vcore shorted to 12V. Found dead mosfet, replaced it + driver + controller, it blew straight up again. with lab psu it goes beyond memory training during post, so I assume with proper epower it should work properly. #3 - dead vcore fujitsu, vcore not turning on. lots of corrosion. With controller removed + epower it doesn’t attempt going beyond 00. other unnecessary chips have been removed. tried powering SB with lab psu to check current draw, it only draws 15mA at 1.05V which seems very little. Suspicion is dead SB but beyond that no idea. resistances appear normal on all 3 boards after faulty parts have been removed/fixed. Board #2

https://en.wikipedia.org/wiki/Sandbag

Same here.

Congrats! Great Score!

will try if I can get it working again, today it decided to be dead. Maybe it works again next week who knows.

nope, pcie seems to wall too early. need >135 pcie for >300.

Corsair 1800c7 is probably the most available strong D9GT* bin in that case, though still not easy to track down.

Never had a stick die on me, though I’ve had several DOA sticks. Hypers (MGH-e, avoid MNH-e) ran at sane volts tend to last a long time and should be faster then you can get D9GT. No point handicapping yourself with a weaker IC.

Hyper still easily beats those, why do you insist on using what was best when the platform was released instead of what turned out to be best a couple months later?

I've never ran into that issue, I've ran 2.3V Vdimm with 1.5V VTT without consequences. Also 2.1V vdimm with 1.4V VTT for many hours, with at least a dozen different Chips. I suspect it mostly affects ES Nehalem chips and maybe C0 stepping which I never bothered testing since they tend to be significantly worse then even average D0s. The only source with actual statistics on this I've managed to find seems to confirm this as well. I've heard some people get better stability at a <0.5V delta but never managed confirming this for myself. January 2009 review of Hyper: http://www.xtremesystems.org/forums/showthread.php?214317-A-DATA-Triple-channel-2133X-quot-ELPIDA-quot-ICs-2000-CL7-8-7-20-1T-1-55V-Memtest

another thing of note - avoid mixing pcbs. the entire day yesterday I was fighting a kit of mine. First, these are the top 4 sticks out of ~20 sticks or so, they are from 3 different kits. I swapped everything (tried 2 boards, 2 PSUs, 6 CPUs, 3 OS), and it just made no sense at all. Each stick on its own is capable of 1100+ 7-7-6 32M. Booting 1100+ and running waza with any 3 of the 4 sticks is also easy. However, they do not pass 32M under any circumstances at 1050+ with decent timings. Turns out - corsair is using several different PCBs for their hyper based sticks. I have 2 50-00187A and 2 50-00218A in this kit. Looking back, the earliest sticks with the 50-00218A pcb I’ve seen are 10122374; I’ve also seen it on 10122507.

Added PTCA-01 Adapter

I would bet so, I cant think of any other IC capable of 5-8-5 at those clocks on ambient. Also the 2200c9 Ripjaws in the rigpic kinda give it away

You gotta check ozzies quote, he properly quoted them

price lowered on Hyper. Edit: Selling the hypers elsewhere.

45nm bios - meaning all microcode but 45nm removed I guess?

edit: nevermind

Added a triple kit of Elpida hyper + a dead stick

nobody mad just yet I hope. Rather see stuff said now then mid comp as we had before. Couple suggestions what could be considered - none of these were neck breaking as is, but changing them could make some stuff more interesting. -DDR2 freq is currently pretty obvious classic d9gmh + p45. If you want more action maybe disallow one of the two (though I bet most wouldn’t mind keeping both allowed, I am not opposed, just throwing in the idea) -DDR3 freq same problem, just instead mfr + haswell. Alternative idea: Dual channel required with no capacity lock so other ics might be viable. -DDR4 3d03 no dx12 cards? odd choice but okay

for ddr2 32M, x38=x48 maybe? Its the same chipset, don’t see why to exclude it

https://forums.evga.com/DIY-low-cost-subzero-chiller-for-benchmarking-m2951571.aspx https://forums.evga.com/DIY-directdie-phasechange-cooler-system-35C-at-400W-m2952116.aspx

What heatsink are you using for the NB? Sweet score! :D

I mean... pinouts aren’t available (yet...), can’t 100% sure say that its impossible with cutting notches off. If anyone finds lga1200 pinout tables would be nice to have...

if you are thinking of the 2000c7 hyper ones, good luck finding them, the only ones I heard of being still in circulation are 2 or 3 dead or badly damaged kits. Pretty sure they were only available with MNH-E.