<lain>
if anything is clear it's that intel changes their naming schemes and ecc support decisions every freaking generation
<lain>
so who knows :D
<rqou>
why does the pentium/celeron branding still exist?
<lain>
as an excuse to charge more for "atom" cores
<rqou>
isn't it already tainted to mean "shitty oem pc?" :P
<lain>
atom x7-z8700 was like $35
<lain>
the next generation equivalent is pentium n4200
<lain>
$135
<rqou>
but pentium g4400 is a skylake, not an atom?
<lain>
oh I dunno then
<lain>
the only celerons and pentiums I've dealt with are atom cores under a different name
* lain
looks
<rqou>
it's pretty hilarious because for skylake you can have a motherboard that literally supports chips with branding all the way from celeron to xeon
<lain>
g4400 is 14nm, relatively new, interesting
<rqou>
another dumb question: why do some motherboards require gpus and/or monitors to boot? just dumb bios programmers?
<lain>
most likely, yes
<lain>
bios code is terrifyingly bad
<lain>
I don't have as much experience as I'd like in that area, yet
<lain>
but I've heard stories, and seen forum posts... :P
scrts has quit [Ping timeout: 240 seconds]
<lain>
I used to dump bios flash chips for fun, a lot of them contain amusing stuff in the unused area
<lain>
one was filled with the full script of "all your base ..."
<rqou>
oh and btw before i used the d510 as a nas, i was using an atom n270 board as a nas
<rqou>
actually probably less crappy than the d510
<lain>
lol
<rqou>
the n270 is 2.5W tdp except slow as shit
<rqou>
the d510 is 13W tdp
<rqou>
but it's not 5x faster
<rqou>
also the d510 is the really fun "you can run 64-bit but probably don't want to" chip
<rqou>
it has the powervr graphics that never got a 64-bit driver
<lain>
ha
<lain>
hm yeah
<lain>
it seems like they just randomly brand shit celeron and pentium
<lain>
rqou: btw have you seen the "mobile xeon"?
<rqou>
no? why would that exist?
<lain>
for beastly laptops :D
<lain>
oh sorry "mobile workstation"
<rqou>
are they xeon dies?
<lain>
sadly no :<
scrts has joined ##openfpga
digshadow has quit [Quit: Leaving.]
<rqou>
wtf
<lain>
they're skylake
<lain>
basically the highest core i7 skylake, but xeon branded and without ecc disabled
<rqou>
right, the skylake "real xeon"/hedt isn't out yet
digshadow has joined ##openfpga
<lain>
it looks like it might have some xeon technology in it? but it's really unclear
<lain>
I don't see any indication that it has, for example, ecc on the internal caches
<rqou>
from what i understand e3 xeon's aren't "real" xeons
<rqou>
they're just rebadged i7s
<rqou>
but e5 xeons are "real" xeons that also get fused down to "better" i7s
<lain>
amusing
<lain>
yeah I dunno what exactly mobile xeon is
<rqou>
also, apparently skylake broke pcie acs in really amusing ways that makes pcie passthrough not work very well
<rqou>
whereas on the other hand my haswell macbook has a fully working but not useful ability to do working pcie passthrough
<lain>
my server, an older x5560, has a bug in the chipset that prevents using pcie passthru
<lain>
it's one of those really nasty bugs
promach has quit [Ping timeout: 276 seconds]
<rqou>
btw have you watched the 33c3 talk about how marvell decided to do pcie on ps4?
<lain>
something like "A specific sequence of loads and stores at particular offsets along with other behavior results in undesirable behavior from the chipset"
<lain>
yeah
<lain>
you mean how they put everything behind a single device?
<rqou>
and how "everything" is a marvell armada soc :P
<lain>
yeah
<lain>
so weird
<rqou>
but the iommu fixes everything! :P
<lain>
lul
<rqou>
except dma-ing onto the stack :P
<rqou>
or, for iphones, dma-ing from the bootrom?
<lain>
lol yeaaaahhhhh
<rqou>
so what i wasn't clear on from ramtin's blog is whether the nvme has persistent firmware or if it gets boostrapped every boot
<rqou>
because if it's persistent, you should be able to make a permanent jailbreak once you either open the phone or somehow pwn the nvme once
<lain>
if/when this handheld mobo gets done and the firmware is written and it actually is a useful machine, I'd like to design a mobile workstation next :P
<lain>
using mobile xeon for dat ecc
<lain>
unfortunately that only has up to 16 pcie lanes, that's really sad
promach has joined ##openfpga
<lain>
I'm going to use an MXM module so I can plug in a nice nvidia gpu, but I'll probably have to run it at x8 so I can attach other interesting devices
<lain>
it's probably fine though >.>
<rqou>
hmm i just found the iphone 6s schematic, and it has a "ANTI-ROLLBACK EEPROM"
<rqou>
um..
azonenberg_work has quit [Ping timeout: 240 seconds]
<lain>
lol
<lain>
guessing that is to prevent installing older firmware?
<rqou>
i actually have no idea
m_w has joined ##openfpga
<lain>
off-topic, but fast food is becoming a real problem around here. apparently there's a serious lack of employable workers
<lain>
several mcdonalds used to be 24hrs, now they're closing at midnight
<lain>
food quality at mcdonalds, taco bell, and others is going down
<lain>
the upshot is I'm saving money by only having soylent, since fast food isn't worth buying around here anymore >.>
<rqou>
lain: wait wat
<lain>
hi
<rqou>
do you ever eat things other than fast food and soylent?
<lain>
not typically
<lain>
on rare occasion I'll make some stir fry or order pizza, chinese, or thai
<lain>
but I'm cheap and lazy
<rqou>
aka a college student? :P
<lain>
perpetual college student
<lain>
lol
<rqou>
lol
<lain>
cup noodle is good though
<rqou>
alright, i should actually work on cleaning my room today
<pie_>
idk but mcdonalds here is really shitty
<rqou>
this is week 2 already
<pie_>
idk if its an america/europe thing
<rqou>
the mcdonalds here is shitty too
<pie_>
or mcdonalds really has gotten that much worse
<rqou>
i love hk mcdonalds
<lain>
there's like four mcdonalds aronud here, three of which are generally shitty, but one has always been REALLY high quality, like, as fast food goes it was way up there
<lain>
but they've declined over the past year
<lain>
everything has, because nobody can find good employees I guess
<lain>
wonder what's up with that :/
<rqou>
hmm you would think that that would be easy here, but it seems that no college students want to work for mcdonalds when the university has lots of temp positions all the time
<rqou>
e.g. it's actually hugely competitive to be a CS department grader
<lain>
haha
<rqou>
alright, my computer is hitting the "must reboot or too slow to do any useful work" point
<rqou>
"up 41 days"
m_w has quit [Quit: leaving]
<rqou>
yay, need to update my Windows VM again
<rqou>
I don't get why Windows needs to "prepare" to install updates
digshadow has quit [Quit: Leaving.]
digshadow has joined ##openfpga
<rqou>
so Overwatch finished updating and Windows is still "preparing" updates at 2%
<felix_>
rqou: both cpu and pch have fuses and embedded microcontrollers that talk to each other
<felix_>
rqou: most parts of boot guard don't run on the me, but on the x86 side. the me basically provides the firmware tpm for the acm running on the x86
<rqou>
I thought the CPU wasn't supposed to have fuses because the semiconductor processing needed for efuses vs finfet is totally different?
<lain>
well the socs are all one single die so clearly they're doing it somehow
kuldeep_ has joined ##openfpga
kuldeep has quit [Remote host closed the connection]
<lain>
if someone happens to know what intel documents explain the fuses (if any), I'm interested :P
digshadow has quit [Quit: Leaving.]
<felix_>
rqou: there are also fuses in the uncore on the cpu die. e.g. for the maximum frequency, ht enable, one of the two ecc enables, maximum l3 cache, ...
<felix_>
i don't know about the socs or atoms
<rqou>
how exactly was this determined?
<felix_>
?
<rqou>
how do people know these fuses exist? or is it just guessing?
<felix_>
i don't have documentation on that, but i know that only very few dies exist. the dies are tested at the end of production to get the maximum specs and then the dies are fused down to get all the different chips they want to sell
<lain>
makes sense
<felix_>
like the flir cameras or highly expensive measurement equiment: it's much cheaper to just make one or very few hardwares and then add some limitations later for market segmentation
<lain>
especially given the investment for 14nm
<felix_>
or even some fpgas
<felix_>
yep. that also improves the yield (limit maximum frequency or fuse off broken stuff)
<rqou>
apparently Windows update jumps from "2% preparing" to "done, please reboot"
<rqou>
I guess they gave up on progress bars after being mocked too much when copying files? :P
<felix_>
oh, for sandy brigde i might even have some proof that it has fuses in the uncore
<rqou>
a lot of old files that i would really prefer to have
<rqou>
i guess i'll get a second one
<rqou>
raid1?
<lain>
do you plan to expand it in the future? and do you care about bitrot? what filesystem?
<rqou>
expand: maybe
<rqou>
currently running a definitely-not-unsupported btrfs-on-luks :P
<lain>
lol
<lain>
well, I don't know anything about btrfs other than "not advised for production"
<rqou>
with a separate ext4 (iirc) root partition
<rqou>
which isn't supposed to even work, but with linux it does
<rqou>
booting from a wd red is not supported for windows and is "not advised"
<lain>
raid1 is fine, but hopefully your fs has checksumming and scrubs to prevent silent bit rot
<rqou>
something about 4k sectors
<azonenberg_work>
whitequark: around?
<rqou>
btrfs has checksums
<lain>
rqou: windows boots fine from 4k sector drives, unless yuo're running some really old windows???
<rqou>
eh idk
<rqou>
this machine won't be running windows
<rqou>
anyways, regardless of checksums in the fs, my nas never had ecc ram before
<rqou>
so it might already have latent disk corruption :P
<lain>
ecc ram is less of a concern when your fs is checksummed
<rqou>
unless you flip a bit before computing the checksum
<lain>
people throw hissy fits over "zfs without ecc is pointless" but that's a fallacy - the odds of corruption hitting disk in an undetectable manner are exceptionally slim even without ecc
<rqou>
anyways, this nas has been used as a dumping ground of ancient files from ancient machines, so those may very well have flipped a bit before
<lain>
no, because there's a whole chain that has to fail
<lain>
at least with zfs
<lain>
I can't speak to other fses
<rqou>
what i meant was that this drive has been a dumping ground for files that originated on a win2k-era ntfs
<rqou>
or similar
<lain>
there's also the write hole
<lain>
personally for anything I care about, raidz2 or better (raid6 if you're not in zfs-land)
<lain>
eliminates the write hole
<rqou>
i heard that raid6 was how you got btrfs to eat your data :P
<rqou>
it's one of the least-tested features
<lain>
>_<
<lain>
yeah I just wouldn't touch btrfs
<lain>
I like my data
<rqou>
"The parity RAID code has multiple serious data-loss bugs in it. It should not be used for anything other than testing purposes."
<rqou>
maybe i should migrate to zfs? :P
<lain>
come to the dark side
<rqou>
the dark gpl-violating side? :P
<lain>
well, it'd be a learning curve, because I can't advise zfs on linux
<lain>
I haven't tried zfs on linux myself, but various people who are probably experts on it seem to consider it not really any better than btrfs at this point, in terms of likelihood of suddenly forgetting all your bits
<lain>
but I don't know what different flavours of zfs on linux exist, etc, or which that may apply to
<rqou>
so (knock on wood) i've never had btrfs eat my bits
<rqou>
but i never ran it in any raid configurations
<rqou>
or any configurations beyond one partition on one disk
<lain>
I'm sensing a serious lack of concern for data integrity here
<lain>
:P
<rqou>
yeah, i've been trying multiple times to clean up my data
<lain>
what's the old saying
<lain>
there are two types of people in the world: those who have backups, and those who have never lost any data
<rqou>
i haven't even gotten around to the step of unpacking old "backups" that were disk images
<rqou>
or worse, Ghost files
<lain>
lol
<lain>
oh joy
<rqou>
i do recall at one point "recovering" old ntfs encrypted files
<rqou>
as in, i still had the decryption key but had mixed up some disk images
<rqou>
at one point my sister and i had nearly identical pcs with same-model OS disks
<rqou>
and they both had similar user accounts
<rqou>
so i managed to mix up the disk images for the OS disks
<rqou>
and wondered why windows claimed it didn't have the key to decrypt data on the data disk that was only in my PC
<rqou>
:P
<rqou>
reassembling the disk images in the correct order in vmware workstation fixed the problem :P :P
<azonenberg_work>
Sooo hmm
<azonenberg_work>
$abc$95$auto$blifparse.cc:375:parse_blif$96 is mapped to LUT2_0
<azonenberg_work>
Which is supposed to be the exact complement of $abc$95$auto$blifparse.cc:375:parse_blif$97 aka LUT2_1
<azonenberg_work>
In my JSON, the inputs are both nets 9 and 10
amclain has quit [Quit: Leaving]
<azonenberg_work>
So gp4par is screewing up horribly
<azonenberg_work>
b/c in the post-PAR netlist, LUT2_0's input is coming from pin 10
<azonenberg_work>
which... isn't even used in my design
<azonenberg_work>
IN0: net vin_too_high
<azonenberg_work>
that looks right at the initial netlist
<azonenberg_work>
cell cmp_vinhi port OUT
<azonenberg_work>
and drives
<azonenberg_work>
cell $abc$95$auto$blifparse.cc:375:parse_blif$97 port IN0
<azonenberg_work>
cell $abc$95$auto$blifparse.cc:375:parse_blif$96 port IN0
<azonenberg_work>
And that looks ok
<azonenberg_work>
so what gives...
<rqou>
offtopic: is it safe for nfs and samba to export the same directory?
<rqou>
or will locking or something be messed up?
<azonenberg_work>
So looks like bitstream[0] is beging written as 0 instead of 1
<azonenberg_work>
hmm
<openfpga-github>
[openfpga] azonenberg pushed 1 new commit to master: https://git.io/vMhrK
<openfpga-github>
openfpga/master b0d5b34 Andrew Zonenberg: Greenpak4Comparator: Don't clobber random bitstream addresses when configuring ACMP_5 (which has no input mux)
<davidc__>
yeah, that seems to be for some weird low profile slot
<rqou>
poking around suggests it is for the "Intel® I/O Expansion Modules for Intel® Platforms based on Intel® Xeon® Processor E5-4600/2600/2400 Product Families"
<rqou>
i can't figure out what connector that thing is using though, but a pinout is otherwise available
<rqou>
great, there can be up to about 300 different candidate connectors
<rqou>
i guess i'll wait until i physically have this in my hand to see if i can make a better guess
<rqou>
hmm so i was looking through that ebay seller to see if he had any other interesting items
<rqou>
and it seems it's just some guy who flips surplus electronics and has no idea what any of them are
<rqou>
e.g. he as a listing with a 2m outdoor-rated ethernet cable for $50
kuldeep_ has quit [Quit: Leaving]
Bike has quit [Quit: neurosexomorphic]
<lain>
lul
<rqou>
so far most of his other items seem to be way overpriced rather than underpriced
<rqou>
the type of stuff this ebay seller has is overall _weird_
<rqou>
vhf/uhf radio stuff, PoS/other retail stuff, random industrial stuff, random life science measurement stuff, random unusual networking/wireless stuff
<rqou>
old cell phones
scrts has quit [Ping timeout: 255 seconds]
<rqou>
hmm poking around ebay mellanox 56g fiber channel/40g ethernet cards show up for reasonably cheap ($200)