xe on and on

i have a little guy to install

MIDNIGHT PROJECT GOGOGO

before:
sysbench cpu events/sec
1 thread: 660.88
6 threads: 3642.20
12 threads:  5704.65

sysbench memory speed
1 thread: 3503.25 MiB/sec
4 threads: 6024.63 MiB/sec
12 threads: 11021.86 MiB/sec

sysbench threading test total events/10 sec
1 thread:  11258
6 threads: 56184
12 threads: 32456

ok. lets get gloopy

1 AM PROJECT GOGOGOG

new rectangle!!!1!one!!
high end gamer heatsinks: lapped to an unsettlingly crisp mirror finish of minimal optical aberration
OEM server heatsinks: my fingernail gets caught on the linear tool marks of this contact surface. oh well

do u think dell would do me a solid and gimme the BOM/kit bits to solder in a second cpu

seems a fun weekend project right

oh man the dead PERC battery is deceptively turgid
holy shit windows small business server 2011
with a five CAL license

oh its configuring ram

oh its configuring the idrac that doesnt exist

oh the fans are MAXED

fingies cwossed

oOH MY GOD THAT WAS NOT MAXIMUM

IT HAS NEVER BEEN THIS LOUD WGHAT THE DRAC CONFIG FAILED IM SCARED

flashing back to the datacenter days

i think i have to zap the pram or some shitt

oh its going eeee

im gonna need the kilowatt of fans going rn to calm the hell down tjough

why did i think this was a good project to start at midnight

the idrac is cooked. this thing doesnt even have an idrac but it still uses it for system mgmt or some shit. i wouldnt care if not for the GHOSTLY HOWL of an active leafblower

throwing freedos onto a thingy at 3 am yaaay

🎵 idgaf 🎵

do it anywaaayss ~

there are two (2) ways i have seen this situation be recovered

one is: Buy a new mainboard lol
the other is: desolder the iDRAC EMMC
clean pads
install new preballed EMMC
somehow reinstall iDRAC firmware

i could try to put the old cpu back in

eeehhuuhh

NOPe
welp this noisy bastard is getting a shitty little UPS and going in the garage where it ostensibly belongs anyways

im pupset

i ont know if im "spend $250 on a new mainboard and another cpu and a cooler" pupset but im converging on it. the sunk cost weighs heavy rn

ughHHh

ok

i could get an entire new one of these servers for like 60% off rn ($75)

same mainboard, just absolute bottom of the barrel hardware otherwise

EXCEPT.. the cpu is already a v2

this might be perfect. ugh. UGH

fuckit lets do it
qualia's happy fun time sunk cost adventure
i might get an actual proper working LOM out of it tho so i'll take that

CUrRENT PROGRESS:

i took a. nap

alright alright alright

cpu in new mainboard: iDRAC works. front panel works

imma put the old cpu in the old mobo and begin swapsies

ten philips screws, two hex safety bits, 8 cables

wish me luuuckkkk

the security bits do not need to come out loloops
yeah that sure is 36 watts of frightening shit-rippin' air displacement in a 120mm form factor. thanks delta!
success

there are a lot of sas ports in this thing. two cables and a fuckload of sleds and i could pack this mf with 16 2.5" drives

i dont know if i would necessarily want to but the idea is interesting

oh, dang. SSDs are now capable of more storage in this form factor now. i guess that makes sense

anyways i need $22000 ok

meanwhile, the other guy is still like

i might relegate this to running really weird infrequently-used hardware. like that fuckin stupid K80

we're so fucking back

after:
sysbench cpu events/sec
1 thread: 660.88 → 868.10 (31.3% faster)
6 threads: 3642.20 → 4719.33 (29.5% faster)
10 threads: 7749.72
12 threads: 5704.65 → 8729.12 (53% faster)
20 threads: 12799.91

sysbench memory speed
1 thread: 3503.25 MiB/sec → 5076.27 MiB/sec (44.9% faster)
4 threads: 6024.63 MiB/sec → 6917.59 MiB/sec (14.8% faster)
10 threads: 11919.66 MiB/sec
12 threads: 11021.86 MiB/sec → 13013.31 MiB/sec (18% faster)
20 threads: 15993.6 MiB/sec

sysbench threading test total events/10 sec
1 thread: 11258 → 16260 (44.4% faster)
6 threads: 56184 → 81484 (45% faster)
10 threads: 59613
12 threads: 32456 → 55273 (70.3% faster)
20 threads: 46462

29 HOURS LATER YAAAY that was an adventure goddamn. pretty pleased with the results!

maybe the most exciting addition actually is that NVMe drive stuffed into a PCIe slot for the less-critical-but-still-gotta-go-fast VMs

random 4k R/W:

read: IOPS=27.4k, BW=107MiB/s (112MB/s)(3209MiB/30001msec)
write: IOPS=27.4k, BW=107MiB/s (112MB/s)(3215MiB/30001msec)

compared to the current VM drive

read: IOPS=12.4k, BW=48.6MiB/s (51.0MB/s)(1459MiB/30001msec)
write: IOPS=12.5k, BW=48.6MiB/s (51.0MB/s)(1459MiB/30001msec)

or the zpool i started moving stuff onto and then went noooope nope nope nuh uh

read: IOPS=2797, BW=10.9MiB/s (11.5MB/s)(335MiB/30618msec)
write: IOPS=2806, BW=11.0MiB/s (11.5MB/s)(336MiB/30618msec)

i forgog there are two identical intel SSDs in here

what. where are they. fdisk says they're in use. zhey're not in a zpool. What

oh. those ARE the zpool. that explains why it is like 130GB in size

these are not fast SSDs. but they are in a mirror. okay. i thought that still seemed way high for hdds. and it very much is
the boot drive for this server however is an old 5200 rpm laptop drive which is like

read: IOPS=46, BW=188KiB/s (192kB/s)(5644KiB/30040msec)
write: IOPS=49, BW=197KiB/s (202kB/s)(5924KiB/30040msec)

lol,