Separate names with a comma.
Discussion in 'Great Deals' started by chinesestunna, Aug 17, 2018.
Yeah, lucky I have 32gb from my old server, might buy just because it so cheap
Haven't seen it better than $35 per stick for this week: SAMSUNG M378B1G73DB0-CK0 1x8GB DDR3-1600 UDIMM PC3-12800U Dual Rank x8 Module | eBay
Careful, PC3-12800U is unbuffered non-ECC memory, aka regular RAM. Unbuffered ECC memory is usually designated with an "E". i.e. PC3-12800E.
Thanks man. I have been keeping an eye on the udimm ecc prices for a while eventually picked up 2 x 8gb Hynix HMT41GU7BFR8A-PB this week for $35 each.
Also been watching these for $42 each, which isn't that bad:
Crucial CT102472BD160B.M18FED 8GB PC3-12800E 2Rx8 ECC UnBuffered RAM | eBay
HP 669239-581 8GB PC3-12800E 2Rx8 ECC UnBuffered RAM | eBay
Hope you are able to finish the build soon !
2 sticks of Crucial CT102472BD160B.M18FED 8GB PC3-12800E 2Rx8 ECC UnBuffered RAM | eBay are now on the way.
Since I'm silly, and it was a good excuse to play with new things, I spent a while trying to use photogrammetry to measure the hole positions, rather then just manually measure them (because manual measuring is sooo 2017).
It, ah, kind of worked?
I've one of those LGA1150 boards with single SFP+ port and 2x SAS ports. Didn't post with i5 4570/4590 but works with Xeon 1230L v3 using both ECC and non-ECC udimm. Going to test with G3220 next weekend when one as I get a loaner from a friend this weekend.
Going to use that board on a tape backup machine either with low wattage xeon or pentium if the latter works.
Also got one of those together with an i3-4160t. Haven't reached yet but fingers-crossed, hope there is no issues with the i3.
From some pix, I think it comes with a LSI-2308 as opposed to the 2008 in fake name's board.
Wonder if the x8 slot does bifurcation?
Atleast on the PCB's silkscreen it says LSI2308.
G3220 boots with ECC UDIMM, I might have recalled wrong with the non-ECC part. Unlike the silkscreen says the HBA-controller is LSI2008.
So it doesn't boot with the i3-4160T. Will consider next options.
So I'm back poking at this.
After some careful measurement and a bit of poking on digikey, I managed to actually figure out what the connector model number for the "FAN_SIG_CONN" connector on the motherboard is. I picked up connectors for both interfaces:
At that point, the challenge was figuring out what the connections actually do. Poking with an oscilloscope and DMM yielded 8 apparent PWM outputs, 8 pins sitting at ~10.8V, and 4 grounds.
Since I assume there isn't a external controller that runs the fan, my guess was that the 8 mostly-high pins were supposed to be connected to the open-collector tachometer output on your common 4-wire computer fan. As such, I set up a signal generator to generate a fake ran signal, put a 2.2KΩ resistor in series so I wouldn't blow up anything if I were wrong (hopefully), and stuck that on the probable fan inputs:
And Bam: IPMI is reporting the fan is no longer missing!
(This was supposed to be a screenshot of the IPMI interface, but apparently I can't hold two oscilloscope probes AND take a picture with my cell-phone at the same time)
Anyways, the IPMI interface is reporting the fan running at 1200 RPM, and I can change the fan speed by adjusting the signal generator.
I also finally made a proper drawing of the hole locations:
Damn! Wish I had an EL-skilled friend like you!
Worning on an pin out chart
Also ejere did you buy the plugs
The place all lazy engineers do their shopping: Digikey
So this is happening:
Boards came in. I managed to be completely boneheaded while designing AND assembling them. I managed to mirror the pin-1 location of the small connector (P1). Fortunately, I was able to wire the cable to fix the layout without issue. It'd be a nasty trap if someone was building one without carefully checking the wiring, but considering the one I built will probably be the only one ever built, I think I can ignore that.
I also somehow put all 8 of the fan headers on backwards initially, despite the fact that I even put giant white bars on the silkscreen to show the correct orientation. It was easy enough to fix, but annoying enough. I guess I didn't have enough caffeine when I was doing assembly.
(Assembly, with headers on backwards)
Anyways, the end-result works nicely:
I get 3-wire-fan RPM feedback, and 4-wire fan speed control for all 5 channels the BMC is configured to support.
Interestingly, the actual hardware has 8 channels of PWM output/tachometer feedback. The last 3 channels are wired up mostly because it seemed silly to not place them, but unless I actually need more then 5 channels (and can figure out how to convince the BMC to report them, it doesn't at the moment), they'll probably never be used.
Throwing one of those lga1150 boards into 4U atx case was way easier than i thought.
I spent far more time then I'd like to admit before I realized they're basically using ATX hole positions.