@Rotonen @Fireduck I'm ready to build that mining machine. I'm thinking 2-3 of these 4x pcie nvme card Link: http://a.co/d/gN0TzDq 8x samsung 970 evo 500GB or 1TB Link: http://a.co/d/6zQmata Total so far $2100 Need to pick out motherboard(s), cpus, ram
don’t get high power SSDs, dont’t get SSDs with lesser channel counts, that hyperx only works with intel nvme, but then you should do VROC
in miner what does this mean: INFORMATION: Probably code EBCAK ?
That was me being sassy
Error between chair and keyboard
that's really sassy.
naughty
i like your humor. but that's not really helpful.
Agreed
But it tells you that you have a pool
So it is a local problem
could it be a broken snowfield? i am just setting a new server.
Arktika or PoolMiner?
poolminer
That will test all the fields on start up and it will tell you which ones are working and which are not
ah. okay.
@Rotonen I wasn't going to bother with tattoos, as I figured it would add overhead. In which case. I thought it'd still work with other NVME?
Face tattoos ?
@Clueless reports seem to vary, read <#CB4CYRTPG|mining> from discord from over the weekend
@Rotonen Well, I can always send it back and try something else if that doesn't work.
since the larger drives are faster, I think I'd go with those.
at least with the 960 pro, only the 2TB model is actually sustainably fast - i’m a bit sceptical of samsung here
from intel, the 760p looks pretty good, though, especially if on VROC
that seems to mostly ride on bypassing pcie switching with some bus level hardcoding
https://www.intel.com/content/www/us/en/products/memory-storage/solid-state-drives/consumer-ssds/760p-series.html The Intel® SSD 760p Series combines high PCIe* performance and high capacity with low power consumption.
no idea what that 8GB span disclaimer is
no two vendors measure the same way the thing snowblossom mining rides on
350k random read
So 56kh/s
I'm confused, should I use VROC?
and should I be thinking 56KH per NVME SSD?
maybe I should just do a DDR3 ram machine
You can get 128GB (4x32) for $800
I was looking at this, I already have 1 1TB Ram Snow miner, thinking about a second one
https://www.ebay.com/itm/Dell-R910-16B-32-Core-2-26GHz-X7560-1TB-2X-600GB-2-5-SAS-H700-512MB-8MR/173462848991 Our technicians specialize in Dell models ensuring that your unit will arrive on your doorstep with quality components and performing at optimal efficiency. Model: PowerEdge R910 4B. Hard Drives 2x 600GB 2.5" 10K. | eBay!
@SidGrip are you making the memory bus of your existing one?
Might want to try adding arktika worker nodes to it if not
I did have it setup with another 148GB server using arktika but the other server doesn't make up the difference in hash rate loss when using arktika on the main miner
@Clueless as far as i can tell VROC uses your cpu as the ssd controller, directly driving all the nand channels, thus skipping a middle man
@Fireduck i’m having a hunch VROC can squeeze more than the dinky little arm on the ssd, but seeking verification
@SidGrip how many sockets is that? seems like a quite the winner
mobile ebay sucks, dual socket quad channel
It's 4 socket quad channel, the one I have running gets over 3 M/hash and does 2.8 M/hash running arktika
oh?
@SidGrip any guesses as to what’d limit arktika? that gap is sadly close to 7/8
I don't know if I understandthe question, I remember asking @Fireduck a while back about arktika and he said there was performance loss. I am currently running snowfield 9
@SidGrip tcp/ip has 1/8 of the packet as overhead
Artika has a bit of I'm memory enquue and dequeue operation
Which involves a bit of locking
that sounds more plausible than the clown college i suspect
So not surprised there would be perf impact, especially on multicore
Multisocket I mean
The lock stuff will probably always be in cache of other prcoccesor or something
if that lock exists in an another numa zone, the access penalty on acquiring and releasing can be rough
I should make a thread local optimized queue implementation
please do, that’d play beautifully into cuda
i’m thinking of gpu memory as a sacrificial cache in that case
I'll have to think about it
Not as easy as the thread optimized atomic long I wrote last week
hi i need to set up the configs file mining for start mine?
What sort of hardware do you have?
notebook hp
os ubuntu