archive snowblossom slack mining
prev 2018-11-11 next
2018-11-11 01:19:49
Clueless
@Fireduck @Rotonen ``` Single 760p NVME SSD - Intel(R) Xeon(R) CPU E3-1241 v3 @ 3.50GHz threads, hashrate, cpu unset - 3K/s - ? 8 - 10.993K/s - 52-60% 16 - 14.792K/s - 60-70% 32 - 18.293K/s - 88-92% 64 - 21.860K/s - 108% 128 - 22.942K/s - 120% 256 - 23.206K/s - 120% ```
2018-11-11 01:21:39
Fireduck
That is shit
2018-11-11 01:24:17
Fireduck
No improvement with more threads?
2018-11-11 01:28:16
Clueless
not much improvement
2018-11-11 01:30:22
Rotonen
details of setup?
2018-11-11 01:31:27
Clueless
that cpu, https://www.asus.com/us/Commercial-Servers-Workstations/P9DM/ The ASUS P9D-M is a micro ATX form factor serverboard featuring the latest Intel® Xeon® E3-1200 v3 processor family, dual Intel® server-class I210 LAN ports, SATA 6Gb/s, and digital power delivery. These make the P9D-M a great serverboard for small form factor servers, data centers, and SMB printer/file/mail/exchange servers.
2018-11-11 01:31:32
Rotonen
output of lsblk as a starting point
2018-11-11 01:31:56
Clueless
``` root@miner:/var/snowblossom# lsblk NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT sda 8:0 0 894.3G 0 disk ├─sda1 8:1 0 878.3G 0 part / ├─sda2 8:2 0 1K 0 part └─sda5 8:5 0 16G 0 part [SWAP] sdb 8:16 1 7.5G 0 disk ├─sdb1 8:17 1 291M 0 part └─sdb2 8:18 1 416K 0 part sr0 11:0 1 1024M 0 rom nvme0n1 259:0 0 238.5G 0 disk └─nvme0n1p1 259:1 0 238.5G 0 part /mnt/miner ```
2018-11-11 01:32:21
Rotonen
oh, 256GB, yeah, less channels
2018-11-11 01:32:45
Rotonen
get the 2TB model to reach vendor promise
2018-11-11 01:33:23
Clueless
I have a second one, I was going to try them in software raid.
2018-11-11 01:34:03
Rotonen
double minus kernel overhead
2018-11-11 01:34:16
Clueless
that's what I figure, just want to prove it
2018-11-11 01:34:54
Rotonen
i think mdraid has been proven a few times over
2018-11-11 01:35:43
Rotonen
what is interesting is if striping sizes make a dent on the overhead
2018-11-11 01:36:09
Rotonen
the difference will not be a pragmatic difference, but i’m curious
2018-11-11 01:36:16
Clueless
I'll try both configurations, raid0 and 1
2018-11-11 01:36:48
Rotonen
mirroring will be faster
2018-11-11 01:37:21
Rotonen
but within striping lies configurability and nuance
2018-11-11 01:37:35
Clueless
I think I'll only bother testing mirroring then
2018-11-11 01:38:09
Rotonen
striping lets you fit a bigger field
2018-11-11 01:38:39
Rotonen
thus mirroring makes no sense for snowblossom?
2018-11-11 01:39:10
Rotonen
the difference should be measurable and consistent, but marginal
2018-11-11 01:39:36
Clueless
I'm confused. mirroring nets more speed, right?
2018-11-11 01:39:45
Clueless
I think the increase over striping would be more than marginal
2018-11-11 01:39:55
Rotonen
yes, but not meaningfully more
2018-11-11 01:41:07
Clueless
well, now I have to test raid 0 and 1
2018-11-11 01:41:29
Rotonen
many have for decades
2018-11-11 01:41:37
Rotonen
just do raid 0
2018-11-11 02:49:15
Clueless
``` Raid 0 threads, hashrate, cpu 512, 46.314K/s, 262-278% ```
2018-11-11 08:35:09
Rotonen
by hashrate you mean..?
2018-11-11 08:36:03
Rotonen
(same for cpu, but not seeing that too relevant, and that gets complex)
2018-11-11 20:06:19
Fireduck
Anyone have one of these: https://www.amazon.com/gp/product/B0772XB3MY Intel Optane SSD 900P (280GB, AIC PCIe 4.0, 20nm, 3D XPoint)
2018-11-11 20:06:24
Fireduck
seems like it might be alright
2018-11-11 20:28:01
fydel
yes. i have one of these.
2018-11-11 20:28:26
fydel
85 khs
2018-11-11 20:29:19
Fireduck
Cool
2018-11-11 20:30:58
fydel
i was a little bit disappointed.
2018-11-11 20:35:02
Fireduck
Seems in like with the 550k random read iops
2018-11-11 20:46:13
Rotonen
optane is all about write density, the 1TB sandisk extreme should have the same read specs
2018-11-11 20:47:15
Rotonen
@Fireduck these might be interesting https://www.corsair.com/us/en/Categories/Products/Storage/M-2-SSDs/Force-Series-MP510/p/CSSD-F960GBMP510 The CORSAIR Force MP510 NVMe PCIe Gen3 x4 M.2 SSD provides extreme storage performance with blazing fast read speeds of up to 3,480MB/sec.
2018-11-11 20:47:37
Rotonen
untested as to is the steady state dependent on cooling
2018-11-11 21:47:33
Clueless
@Fireduck twice as expensive, but 4 times fast.
2018-11-11 22:51:19
Clueless
@Rotonen ;) `Raid 1 - 512 - 46.385K/s - 281%`
2018-11-11 22:52:25
Rotonen
n=1
2018-11-11 22:52:40
Rotonen
and even if you measure proper the difference is just random jitter
2018-11-11 22:53:24
Rotonen
or i expect you'd need over 10k samples to be certain of that being consistent
2018-11-11 22:53:35
Rotonen
but the diff is in line with what i expected, not meaningful
2018-11-11 22:53:51
Rotonen
but more importantly, how do you measure?