gigabyte getting into the epyc workstation market, but dunno about it for snowblossom price-wise with just one socket https://www.anandtech.com/print/13566/amd-epyc-for-atx-workstations-gigabyte-mz01ce0-mz01ce1-motherboards AMD EPYC for ATX Workstations: GIGABYTE MZ01-CE0 & MZ01-CE1 Motherboards
Is epyc a bus or a socket or what?
amd cpu
8 ram channels per socket
the name is stupid, but 'amd xeon' is close enough to the truth
they dropped the opteron name at some point for whatever reason
and they're pretty good in the way of dollars per ram channel for just the cpu
the low end 120W models should do rather well for snowblossom https://www.amd.com/en/products/cpu/amd-epyc-7251
i love the coy jab at oracle licensing there, the use case of that processor is to be 'licensing cost optimized' :stuck_out_tongue:
Got two of those. Can't use them till the riser comes tomorrow
neat, I'll be interested in your results.
my mess is all on field 8 now
it sucks how much less efficent it is to read from shm file system over memory allocated in the jvm
but the jvm is a hot mess, so if you want to get anywhere close to system memory limit, you need to use shm
what is shm
shared memory file system? like a ram disk?
Yeah
It is in the linux kernel so works really well, but Arktika can't seem to access it as cheaply
@Fireduck I don't understand what you mean by JVM being hot mess.
for example, if I want the jvm to cache 16gb, I usually have to set the max heap size much larger, like 24gb for it to work
which is fine for some things, but when I want to cache 240gb on a machine with 256gb, I can't quite make that work
Yeah, I wonder if it caches chunks out of order and requires room for defragging or something
that sort of thing is more the kernel's problem
`74% to 41` yay, my code is useful
nice
@Fireduck fscache works as well as shm
@arichnad yeah, it is
:slightly_smiling_face: if anybody has trouble using the Dockerfile, please send them my way. i use it exclusively and everything seems to run smoothly.
also, i'm adrian by the way
yeah, I remember the pull requests
I've never used docker but apparently it is hot
yeah, i was expecting that mining would be slower inside the container. i figured the IO would slow down a little bit going in and out of the container. but no, i get the same hash-rate in both situations.
The container is a fairly lightweight construct right? It doesn't have its own kernel like full virtualization, right?
more of a jail
that's right. it's really just a special process. much like a chroot.
PIDs even show up in top.
depends on how you back the container and/or volume storage
Anyway snowblossom reward system is more to like farming snowfield than mining maybe change the term?
eh, it is flailing at some thing and checking a hash
nothing really to do with "mining" even in the bitcoin case
but it is the term that everyone understands, do some work in hopes of making blocks to get block reward
Hmm. What's the analogy? A scavenger hunt?
!! Scavenger Hunt!
picking up dog turds, putting them in your mouth in the hopes that one is actually eventually a sausage
Eww. I think that would make the documentation more confusing.
ah, today is one of those days where I shoudn't be allowed a keyboard
Yes, you should be forced to code using text to speech.
I do sometimes use speech to text on android
it works surprisingly well even for technical topics
usually
maybe that shouldn't be surprising, it was probably trained by nerds
I almost have pool failover working for PoolMiner and Arktika
probably get that in tonight assuming Lola goes to sleep reasonably
She was really helpful in the design phase (long stroller walks were good for thinking about design)
but less helpful now
@Fireduck https://youtu.be/AF_nfazQaek YouTube Video: Naptime!
@Fireduck just make sure the documentation thereof does not turn political