 Show Posts
|
|
Pages: « 1 2 3 [4] 5 »
|
I made some tests with a Z270 and it ended with 9GPUs max https://bitcointalk.org/index.php?topic=2204115.msg33127677#msg33127677It's not the splitter card or M2-PCIe adapters, its the BIOS and Hi IO lanes of which there are 16 from the CPU, and the chipset use some for other things than Video Cards (Mining cards without video might be possible to add beyond the 9 for this chipset/bios). That setup have been stable since then btw. |
|
|
|
|
Got a good deal on a pair of 1050 some time ago (two for somewhere between the price of a 1050ti and a 1060 3Gb), and they have been chugging away like champs since.
323 Sol/s at 50W each (so 3.23 Sol/W roughly at +190 core +1300 mem). No Eth of course (2GB) but they are decent at XMR (650 h/s).
ROI at current pace is about 220 days (I use the heat so I don't count the power).
|
|
|
|
|
Flexibility means you can use many more that the special-always-of-of-stock Mobos to use 9-13 GPUs even without M2 adapters.
You are limited to the number of Hi IO-lanes the board uses for other things than graphical cards (remember, Mining expert can handle 13 Video cards and 6 mining cards, not 19 cards with video out).
I did not have an extra card to test my mini-budget rig with QC5000M with a splitter for 6 cards, but it worked nicely with 5 (miniATX with 3 slots). The APU can mine a little as well :-)
|
|
|
|
Why do I get the bootup to stop with the screen filled wth all these [OK] lines and at the bottom "Please use ^D to try to continue boot" ?
The only way past this is the ctrl-alt-del and try again, and after a couple of tries I get into Simpleminer.
I have KVM over IP on the way, just to be able to bypass this remotely.
I've been having the same problem for quite a long time. Maybe usb stick is dying? My rig so stable now that I haven't bothered testing with a new stick though  That rig is currently running on a spinning disk HD, partition extended to the whole drive expect a small swap partition. So i don't think it's drive related (although it's whining about drive validation stuff in the beginning of the boot process). It would really help a lot of there was comprehensive logging that could be enabled (with decent rolling sizes). Thereis plenty of space even on the 4Gb image for the sticks. But logging requires code lines that write the log messages, and it's mostly messages from the logging software that is needed. most mining sw seems to only spit out to stdout, what we watch with screen, but having rolling 24 hours of that (or 10M) would be extremely valuable when you approach a rig that hangs at startup (two issues why did it reboot, and why is the boot process stopped) . The disturbing thing is when you have something that you see as stable for a couple of weeks, and then it suddenly fails. I hope the ancient HP KVM over IP (8port Eo1013) i'm getting isn't too old to be run from a Pi with teamviewer. Edit: Adding the lines on the screen (there could e more lines above this). A simple ctrl-alt-delete made the rig reboot and stratup normally. [ TIME ] Timed out waiting for device dev-disk-by\x2uuid-35E2\x2d1928.device.
[DEPEND] Depandency failed for /mnt/user.
[DEPEND] Dependency failed for Local File Systems.
[DEPEND] Dependency failed for File System Check on /dev/disk/by-uuid/35E2-1928.
Starting Enable support for additional executable binary formats...
Starting Set console font and keymap...
[ OK ] Stopped getty on tty2-tty6 if dsub and logind are not available.
[ OK ] Closed UUID daemon activation socket.
[ OK ] Stopped Daily Cleanup of Temporary Directories.
[ OK ] Stopped target Graphical interface.
[ OK ] Stopped target Multi-User System.
[ OK ] Stopped D-Sub System Message Bus.
[ OK ] Stopped Regular background program proicessing daemon.
[ OK ] Stopped OpenBSD Secure Shell server.
[ OK ] Stopped Accounts Service.
[ OK ] Stopped LSB: daemon to balance interupts for SMP systems.
[ OK ] Stopped LSB: Record successful boot for GRUB.
[ OK ] Started Stop ureadahead data collection 45s efter completed startup.
[ OK ] Stopped Daily apt upgrade and clean activities.
[ OK ] Stopped Daily apt download activities.
[ OK ] Reached target Timers.
[ OK ] Stopped Getty on tty1.
[ OK ] Stopped Hold until boot process finishes up.
[ OK ] Reached target Login prompts.
[ OK ] Stopped Login Service.
[ OK ] Closed D-Sub System Message Bus Socket.
[ OK ] Stopped LSB: Set the CPU Frequency Scaling governor to "ondemand".
[ OK ] Stopped Terminate Plymouth Boot Screen.
[ OK ] Stopped Permit User Sessions.
Starting Set console scheme...
[ OK ] Stopped /etc/rc.local Compatibility.
[ OK ] Stopped target Basic System.
[ OK ] Reached target Sockets.
[ OK ] Stopped target System Initialization.
[ OK ] Started Emergency Shell.
Starting Create Volatile Files and Directories...
Starting Tell Plymouth To Write Out Runtime Data...
Starting LSB: AppArmor initialization...
[ OK ] Reached target emergency Mode.
[ OK ] Listening on Load/Save RF Kill Switch Status /dev/rfkill Watch.
[ OK ] Started Set console scheme.
[ OK ] Started Tell Plymouth To Write Out Runtime Data.
[ OK ] Started Create Volatile Files and Directories.
Starting Uptime UTMP about System Boot/Shutdown...
Starting Network Time Synchronization...
Mounting Arbitrary Executable File Formats File System...
[ OK ] Started Uptime UTMP about System Boot/Shutdown.
Starting Uptime UTMP about System Runlevel Changes...
[ OK ] Started Uptime UTMP about System Runlevel Changes.
[ OK ] Mounted Arbitrary Executable File Formats File System.
[ OK ] Started Enable support for additional executable binary formats.
[ OK ] Started Network Time Synchronization.
[ OK ] Reached target System Time Synchronized.
[ OK ] Started Set console font and keymap.
[ OK ] Created slice system-getty.slice.
Starting File System Check on /dev/disk/by-uuid/35E2-1928...
[ OK ] Started File System Check Daemon to report status.
[ OK ] Started LSB: AppArmor initialization.
Starting Raise network interfaces...
[ OK ] Started ifup for eth0.
Welcome to emergency mode! After logging in, type "journalctl -xb" to view
system logs, "systemctl reboot" to reboot, "systemctl default" or ^D to
try again to boot into default mode. |
|
|
|
ONDA B250-D8P ver 2.0or Onda 3865 D8P (the green one with celeron onboard) ? Both appear to be decently priced (looking for one) and have decent space (56.5 mm c-c and no interfering connectors), but is there anyone having experience with the 3865 version (older/newer, any revision)? The use of 6pin is preferred over sata/molex, if they only included the picoPSU on the board as well (as Octaminer). Then there is a mix of the two Zeal-All onda 8 pcie ?? (all components around the PCIe slots is identical to the green one so a copy of a copy of...)
|
|
|
|
Block reward will go from 3 eth to 0.6 eth when casper arrives.
Just calc for kicks: Lets say that this makes 80% of the hash power go to other coins (since the grass is not dead on the other side, yet). Based on 213Th/s that's 170Th/s which equals to 5.7 Million 30Mh/s GPUs If we guess that those GPUs have 600h/s on CryptoNight7 and all of them would choose Monero7. That would add 6,5 time the Hashpower from 519 Mh/s to 3927 Mh/s. But this is playing with numbers, the hashpower will never move as one, and some of the hashpower might not be movable. The number per card is just guess numbers, but they are not far off for several cards (many cards fall within a factor 2). Regardless, there is a boatload of GPUs that will be affected by the radically changed block reward. |
|
|
|
|
When the stocks of GPUs at retailers that were reordered 8 weeks ago (at stupid prices) either get sold or something makes them clear out those stocks due to low demand (or new cards being announced).
This stocking issue is probably why the orders of GPUs dropped so much in April. If you have unsold cards, why stock up on more?
I don't expect prices to get close to MSRP until late June or so, retailers don't like to sell stuff at loss.
But if the 1170 and 1180 will be impossible to get hold of after release the MSRP+50% might continue... If they will be plentiful all other cards will crash in price. That would require all GPU mineable cryptos to have crashed IMO.
|
|
|
|
|
Make sure you know which (physical) card it is that is flakey, or is it random?
Move that card to a quarantine rig and see if the situation get better.
Does the issue follow the card, riser, USB cable, power, or just the fact that 13 suddenly is an unlucky number with the bios?
|
|
|
|
|
Mining as a hobby is never to late.
i have never had an hobby with an suggested ROI, they are all black holes for money.
The safest way to (continuously) make money of mining, is to supply hobby miners with equipment. Not that many miners that have quite their day-job. The volume needed to do so can be quite large. And I'm here referring to the "mine directly to FIAT and pay your bills" type of mining.
If you lack basic computer building experience, going from a bunch of boxes to a constant payout from a mining pool can be quite a difficult journey (but a rewarding exercise, for the experience, you might get some coins in scattered wallets as well).
I would suggest buying a working rig without GPUs (where someone have chosen a decent mobo, cpu, mem, cables, risers, rig hardvare, ...), and then add (or buy with) a couple of Nvidia Pascal GPUs (heck, even 1030 works as a start, but you need 3GB for Etherium, for now).
The reason to stay away from AMD is to not "need" to bios mod etc.
A 6GPU rig without cards should be possible to find for $400.
Such a low investment rig will be easier to learn with since you can ignore the ROI, the power consumption etc, and focus on learning the management of a rig and make it stable.
Then you can place a preorder of 6x1180...
|
|
|
|
can you share your smos overclock settings for your 1060? with ethash -100 core +1600 mem 75W claymore-eth-v11.7 -wd 1 -r 1 -epool stratum+tcp://europe.ethash-hub.miningpoolhub.com:17020 -ewal username.$rigName -epsw x -eworker username.$rigName -esm 2 -allpools 1 Higher mem makes it unstable (meaning 1-2 reboots a week). And for kicks it's most probably hanging at startup now, when I'm away for 4 days... Why do I get the bootup to stop with the screen filled wth all these [OK] lines and at the bottom "Please use ^D to try to continue boot" ? The only way past this is the ctrl-alt-del and try again, and after a couple of tries I get into Simpleminer. I have KVM over IP on the way, just to be able to bypass this remotely. And this rig is not on a smartplug, but that does not always help either, since some MoBos go into Bios instead, so the KVM over IP is needed anyway... |
|
|
|
All gpus are available in my country, but the prices not changed... That's because the GPUs are not tradeable commodities (which have spot values), they are things in boxes that are stored in several hands before they are sold. This means that someone have filled their stock at a certain price, and will not sell to a loss until they "have to". Prices must drop until the 1170 and 1180 are announced, or at least after they are getting available. On that note, when will it be possible to pre-order 1170 and 1180 at MSRP? With that in the equation, who think they can sell a 1080Ti for $600+ after the release date of 1180? If demand will be high enough (which it can be at current profits/difficulties) the new cards will be impossible to get you hands on and they will be sold at 50% over MSRP... Then the value of current cards will not drop by much, we just got even faster more efficient cards that noone can get their hands on. |
|
|
|
So what combos are people finding most profitable right now? 9x 1060 3G Etherium 205 MH/s at pool (24.5 peaks reported on all cards at 75W) 2x 1050 (yeh really) on QC5000M (AMD APU) with XMR-Stak for Monero 678 h/s at 50W I'm a lazy bum using miningpoolhub (autotransferring to LTC and then to FIAT) which seems to hurt Equihash since the coin flippityflop switching is too slow, sensitive or something. And the 1060 3G i have seem to have decent memory (no windows in sight to check which), not the 19MH/s crap. A 1080Ti needs a boner pill to beat a pair of these 1060 on Etherium |
|
|
|
|
Is your miner exposed to the internet (especially port 22)?
|
|
|
|
most likely this limit is caused by chipset or processor limitations.
16 PCI (HiIO or something) lanes from the CPU and these are the ones that get mapped to video outputs, if I have guess correctly based on scattered info. It would be interesting if BIOS modding on GPUs could disable video output (to fool the MoBo BIOS) to allow more cards (on any MoBo). |
|
|
|
|
Ethash peeks at noticeable less power than equihash on my 1060's (most probably the same for all Pascal cards).
Equihash like more core clock and that apparently cost more power then boosted memory (which benefits ethash).
|
|
|
|
|
I'd choose the 1000W version of the HP DL380 PSUs. They are a lot cheaper (than 1200w) due to less demand and put a little less strain on the breakout boards.
Two of them + breakout boards with cables can be had for less than $100, making 2000w max from 20 6pin cables. Add picoPSU for completeness.
But Server PSUs are not for living areas, the high pitch whine can be heard through several closed doors.
I'm playing around with adding stacked 60mm fans to fanless DPS700GB, these PSUs cost close to nothing but have a different connector, so cables are custom.
One single data point for reference, this 700W PSU pulls 400W from the wall when feeding 5x75W GPUs, so 93+% going to the cards (assuming the cards really pulls 75W when restricted to 75W).
|
|
|
|
|
Water cooling a a good technique to move heat somewhere else, but the heat must be shed somewhere (but heat in water is easily and silently exchanged to other fluids). Never worth the effort in ROI sense, but could be interesting from an engineering point of view.
I solved my localized heat in my basement by adding a simple fan to make the whole basement the same temp. This does not change the heat output of the rig, but it makes the whole basement absorb the heat and transfer it more efficiently to the rest of the house.
But when the summer comes it will be a question of lowering the power to the cards and increasing efficiency (per W, not per s).
There are some videos about using a grow tent and effectively make the rig use filtered outdoor air (byt having flex hose through a window.
|
|
|
|
My 1060s running very well. Getting 25 hashes consistently. No problems as of yet. STill runs hot though.
mine is running at 24mh/s at around 68 degrees Celsius which is pretty good to me I have 9 1060 3GB OC that is averaging 210 Mh/s day after day (all of them peaks at 24.5). SMOS with -100 core +1600 mem and max 75W. They can give a little more at 80W or above but this is so stable. Temps are 48-55 C depending on which end of the rig. All cards bought to extreme overprice Jan-Feb -18. A friend have some "identical cards" bought in -17 which only can get 19.5. No windows available to figure out the memory, but that does not matter, the result speaks for itself (ROI in 2020...). At "similar" setting the same 9 cards give 2640 Sols/s (+140 core +1400 mem, 80W). A little hotter, but equihash want more core. |
|
|
|
|
Wait, what??
$40 (0.076 ETH) a day with one 1060??
|
|
|
|
|
I have a small test rig that have two leftover 1050 (2GB) cards in it now.
Using Minigpoolhub to autotrade to Litecoin, these two cards give about $0.9 per day now, alternating between Cryptonight7 (Monero) and swapping equihash (Zcach, ZenCash, Bitocoin-Gold, Zclassic). This when set at 50W each (combined about 320 Sol/s equihash and 630 kh/s Cryptonight7).
Ethash is not possible due to the DAG file size .
I use the minor heat generated (not Scandinavian summer yet) so I fool myself by disregarding the power cost.
1050 is low rate GPU, but much more power efficient than older midrange with same hashrates.
|
|
|
|
|
