 Show Posts
|
|
Pages: « 1 2 [3]
|
|
Promising miner...
I tested for a while with 3 x 1080 Ti and it reports a lot higher hashrate than 2:nd best miner I found.
It was no point trying a longer test to check actual results from the pool because it started generating lots of rejected shares after a while.
This is what it looked like when it worked well. Sometimes, all shares where rejected.
[2018-04-13 19:12:24] GPU #2: ASUS GTX 1080 Ti, 4029.46 kH/s
[2018-04-13 19:12:24] accepted: 17/36 (diff 0.246), 12.42 MH/s yes!
[2018-04-13 19:12:30] GPU #0: ASUS GTX 1080 Ti, 4194.66 kH/s
[2018-04-13 19:12:30] accepted: 17/37 (diff 0.060), 12.41 MH/s booooo
[2018-04-13 19:12:30] reject reason: low difficulty share of 15.28838648248119
[2018-04-13 19:12:46] GPU #0: ASUS GTX 1080 Ti, 4219.58 kH/s
[2018-04-13 19:12:46] accepted: 17/38 (diff 0.023), 12.42 MH/s booooo
[2018-04-13 19:12:46] reject reason: low difficulty share of 5.978946698755245
[2018-04-13 19:12:50] GPU #2: ASUS GTX 1080 Ti, 4135.22 kH/s
[2018-04-13 19:12:51] accepted: 17/39 (diff 0.038), 12.42 MH/s booooo
[2018-04-13 19:12:51] reject reason: low difficulty share of 9.82034601280336
[2018-04-13 19:12:56] GPU #1: MSI GTX 1080 Ti, 4126.38 kH/s
[2018-04-13 19:12:56] accepted: 18/40 (diff 0.079), 12.42 MH/s yes!
[2018-04-13 19:13:08] lyra2z block 82707, diff 27409.834
[2018-04-13 19:13:14] GPU #2: 2008 MHz 28.03 kH/W 146W 51C FAN 43%
[2018-04-13 19:13:43] GPU #0: ASUS GTX 1080 Ti, 4183.70 kH/s
[2018-04-13 19:13:43] accepted: 18/41 (diff 0.061), 12.42 MH/s booooo
[2018-04-13 19:13:43] reject reason: low difficulty share of 15.586001478118957
[2018-04-13 19:13:48] GPU #0: ASUS GTX 1080 Ti, 4218.41 kH/s
[2018-04-13 19:13:49] accepted: 19/42 (diff 0.127), 12.42 MH/s yes!
[2018-04-13 19:13:54] GPU #0: ASUS GTX 1080 Ti, 4220.90 kH/s
[2018-04-13 19:13:54] GPU #0: 2054 MHz 27.80 kH/W 151W 51C FAN 45%
[2018-04-13 19:13:54] accepted: 19/43 (diff 0.050), 12.42 MH/s booooo
[2018-04-13 19:13:54] reject reason: low difficulty share of 12.836641482873395
[2018-04-13 19:13:59] GPU #2: ASUS GTX 1080 Ti, 4119.52 kH/s
[2018-04-13 19:13:59] accepted: 19/44 (diff 0.020), 12.42 MH/s booooo
[2018-04-13 19:13:59] reject reason: low difficulty share of 5.027982789460722
[2018-04-13 19:14:11] GPU #1: MSI GTX 1080 Ti, 4127.66 kH/s
|
|
|
|
|
With all these new cryptonight types, it would be better to keep that parameter in the pools file per pool.
Now it's not possible to keep all coin configurations in the same directory.
Config parameters probably also has to be different for different cryptonight variants. At least, that's the case with XMR-STAK and others.
|
|
|
|
|
Monero needs another fork again in 4 weeks just to cure the disease properly.
As said already, Bitmain is probably already well on the way for V7.
SUMOKOIN has now forked to their new heavy version.
Their hash rate is down to 10-20%
|
|
|
|
|
One can wonder if NVIDIA and AMD are doing the same thing as BITMAIN and the others.
It would probably be tempting for them to set up big farms with the latest hardware. Their cost of doing so would be lower than for anyone else by a big margin. And they can still sell the stuff whenever they want.
It's strange that there is still a GPU shortage at this time. There is basically no ROI at all anymore if you take increasing difficulty into account and the used market is beginning to be flooded with mining hardware.
Who is buying new GPUs to the level that distributors can't even get anything?
One explanation could be that both NVIDIA and AMD are prioritizing the more profitable AI and super computer segment at the moment.
If they set up their own farms, it should be possible to detect them. I don't know if their financial reports are transparent enough. If so, maybe one could identify a large increase in power consumption somewhere.
|
|
|
|
|
The current POW is doomed in the long term. The main reason is that it's not energy efficient. All this power is wasted just guessing numbers with artificial difficulty. Crypto will never handle mass market adoption and high volumes of transactions without being orders of magnitude more energy efficient. All this is really stupid if you look at the big picture and something better is clearly needed. Asics are more efficient than GPUs for guessing numbers and all would be fine if not the ASIC manufacturers were hostile. BITMAIN & co covertly use the hardware themselves until the market is sucked dry and then sells to big farms mostly. A small amount of ASICS goes to enthusisasts and they always get late to the party and serve as a dump sink for obsolete hardware. Crypto is centralized and put under chineese control... Not much of an improvement
The next logical step in the market evolution is for NVIDIA, AMD, Intel and Samsung to get into this business. I wouldn't be surprised if we soon see NVIDIA and AMD GPUs with onboard ASICS to boost crypto mining. They will be able to build superior units compared to BITMAIN because they already have huge amounts of RAM, bandwidth and efficient software platforms. Think of how easy it would be for AMD and NVIDIA to integrate some Daggerhashimoto ASICs in current GPUs and use existing memory. All they have to do to secure a new generation of very profitable mass market hardware is to strike a deal with the Etherium project so the hardware is not rendered useless due to some defensive forking. They certainly want a bigger piece of this market and BITMAIN is the enemy.
However, current POW is stupid and must evolve.
It think a viable long term solution is to monetize distributed compute power and use is as a "proof of compute work".
This is already done, sort of, but not properly, by BOINC, Folding@Home and a handful upcoming AI projects.
It would be much better to use all mining power to perform real, useful, mathematical calculations, AI training or whatever instead of guessing numbers. Such computations can only be performed by a general computer or GPU and the power is actually used for something useful in the end. Exactly how to implement this to secure a blockchain, I don't know but I'm sure it can be worked out. The ever changing code doing the computations perhaps must be dynamically validated and hashed into the blockchain together with the output. The one who figures this out first will probably drive many Lambos...
|
|
|
|
|
3 slots is not enough space for cooling.
Advanced GPUs all use 2.5 - 3 slots of space so it would only work for blower style cooling solutions.
With a spacing 4 - 6 slots and some provisions for compartmentalizing the GPUs so exhaust air is not cross-fed, it will be much more useful.
Cooling is the big problem to solve and a solution that takes this into account will be worth more on the market.
|
|
|
|
|
Is there a mining pool somewhere?
|
|
|
|
|
On VEGA, only the blower fan is powered from the PCI slot.
The fan uses 28W at full rpm.
Typically, you will see around 15 - 20W during normal use.
|
|
|
|
|
Have you thought about finding a way to feed the DC from the solar panels directly into the 12V consumers and bypass the two transformation stages?
That could potentially give you another 10 - 15% more net power by reducing losses. I know they do things like that in large datacenters sometimes. Unfortunately, 12V PSUs are still not mass produced and therefore expensive. However, compared to high end consumer PSUs and solar panels, it may be very profitable. You could perhaps implement it by using some properly dimensioned battery bank like an UPS and then add some custom wiring to the GPUs.
|
|
|
|
I've got one card that just always runs 10deg hotter, no idea why.
any tips?
Can be the silicon lottery. That also happens if there is the slighest anomaly with cooler alignment or TIM application. You could check it by taking the cooler apart but then warranty is void. Initially, I saw reports of some VEGA cards having very bad TIM and users getting 6 - 10 C improvement by applying some good TIM. So I took one of my cards apart and installed a high end TIM. I didn't see anything wrong with the stock TIM and it looked exactly the same as the one I applied. Afterwards, I saw a slight increased temperature on HBM compare to before and compared to my other cards. If you are serious about getting to the bottom of it, you could swap coolers between two cards and see what happens. |
|
|
|
|
I can confirm that one reason for performance drop is that the monitor (if any) connected to a mining card, turns off according to power settings.
I have been testing this for 2 days now and the problem goes away if monitor is set to "never turn off".
|
|
|
|
3 vega 56 flashed with the 64 vega rom
max i can get is 1500 on all three using CAST-XMR , on top of that the hashrate keeps dropping after a high start and settles in at 1400ish
using 1110 core and 1075 mem 20% + power lim and 900 mv on core and mem
Any help ? i see 1200 before i do the disable enable trick , and I do see the error that says blockchain drivers not detected
but I am using the robin hood blockchain drivers , the aug 23 drivers seem to constantly crash my system
you need the blockchain drivers aug 23. no way around it. if desperate, start with a fresh windows install. yeah even on a fresh install BSOD when i run it ok thanks tho The fact that you start strong and go lower after a little while suggests temperature problems and card backing off. You need manual voltages too and the registry hack because it lowers the floors on the voltage on P1-P6 which you cannot touch in wattman. The bsod (in my case) means that I'm pushing a card beyond what it can do end it either locks up in the driver (bsod) or if I'm lucky it stops in userland and it cannot be "connected to" anymore. It's kernel will show 0 hashes. Best way is to use registry hack + overdriventool to set your per-card profiles. Test them one at a time. And write down everything. Atm it's a pita I know. Takes time to stabilize if you're not lucky and have cards that manage all settings with ease. I also have this strange problem with decreasing performance over time. I verified it's not related to temperature. It also goes away immediately if I restart the miner. Same here. I have posted twice about the problem and tried everything that was shared as a solution (clean install, max RPM, high power settings, etc.). But no go. Waiting for someone to share a proper solution. I have 12GB RAM and nothing else running on the rig. Very annoying to restart the miner every 18-20 hours after performance deteriorates. Thanks in advance for whoever shares a constructive solution. I think I now figured out why performance degrades after a while with CAST_XMR. It seems it's related to the power setting and the monitor turning of after a period of idling. Performance goes down when the monitor turns off. This is true if monitor is connected to one of the mining cards. This explains why many don't see this problem because they use IGP. I have set monitor to never turn off and now performance stays at ~1900 indefinitely. |
|
|
|
|
I have some weird problems with Vega.
I had 3 of them running on a rig but couldn't get the blockchain driver working. I clean-installed everything and then the driver finally got detected by Cast-XMR.
However, power usage was through the roof and before I managed to get the powerplay mods installed, the PSU got overloaded and somehow damaged the setup when it shut down. Couldn't get the stuff running again but found no faulty components when trying them separately in other systems either...
So I moved the 3 amigos into another system and repeated the procedure. Started with 2 cards and this time, I applied powerplay. Now blockchain driver is recognized by Cast-XMR. Hashrate is around 1000 per card regardless of how much fiddling and toggling with devices or HBMC I try. Hashrate for DaggerHashimoto is a steady 44.1 so blockchain driver is definitely in use.
When I try dual mining, I get an instant hard crash no matter the settings with clocks and voltage.
When I connected the third card, the entire system bugged out and wouldn't even post. Had to tear out all GPUs and reset BIOS.
Hard to see what that has to do with the driver but I previously had all the cards running in this machine with newer drivers.
The question is why CAST-XMR gets such bad performance. I got better without the blockchain driver.
|
|
|
|
3 vega 56 flashed with the 64 vega rom
max i can get is 1500 on all three using CAST-XMR , on top of that the hashrate keeps dropping after a high start and settles in at 1400ish
using 1110 core and 1075 mem 20% + power lim and 900 mv on core and mem
Any help ? i see 1200 before i do the disable enable trick , and I do see the error that says blockchain drivers not detected
but I am using the robin hood blockchain drivers , the aug 23 drivers seem to constantly crash my system
you need the blockchain drivers aug 23. no way around it. if desperate, start with a fresh windows install. yeah even on a fresh install BSOD when i run it ok thanks tho The fact that you start strong and go lower after a little while suggests temperature problems and card backing off. You need manual voltages too and the registry hack because it lowers the floors on the voltage on P1-P6 which you cannot touch in wattman. The bsod (in my case) means that I'm pushing a card beyond what it can do end it either locks up in the driver (bsod) or if I'm lucky it stops in userland and it cannot be "connected to" anymore. It's kernel will show 0 hashes. Best way is to use registry hack + overdriventool to set your per-card profiles. Test them one at a time. And write down everything. Atm it's a pita I know. Takes time to stabilize if you're not lucky and have cards that manage all settings with ease. I also have this strange problem with decreasing performance over time. I verified it's not related to temperature. It also goes away immediately if I restart the miner. |
|
|
|
Jag hjälper Håfreströms Företagspark (HÅFAB) med att exploatera deras anläggning för datacenter-verksamhet. Detta är antagligen en av de bästa existerande anläggningarna i Europa med avseende på kostnadseffektivitet. Detta är ett pappersbruk som byggts om till generell företagspark. Här finns 20000 m2 golvyta samt kontor. Anläggningen har ca 32 MW outnyttjad kraftförsörjning färdig. Den rörliga delen av elpriset är självkostnadspris som gäller på den nordiska spot-marknaden så det kan knappast bli lägre såvida man inte har eget kraftverk. När vi får vi igång verksamhet över 0.5 MW så inträder energiskatterabatten. Lokalerna har mycket god takhöjd så man kan dumpa höga effekter i kylluft utan att investera i aktiv kylning. Anläggningen hanterade som mest 52 MW inomhus på sin tid... Golvytan kostar 400 :- / m2 och år. Kontorsutrymmen kostar något mer. Anläggningen ligger här: https://www.google.se/maps/place/H%C3%A5frestr%C3%B6ms+F%C3%B6retagspark/@58.8055193,12.4248188,17z/data=!3m1!4b1!4m5!3m4!1s0x4644ea6cc45de4db:0x4e176b36d32b2d63!8m2!3d58.8055165!4d12.4270075?hl=enKontakta mig gärna för mer information. |
|
|
|
 I live and work in Sweden at the moment. One of my assignments is to find datacenter tenants for a business park, which is a converted paper factory with 32 MW power available. The power price is the spot market price used in the official Nordic power exchange (i.e. the lowest possible). Free cooling and 20000 square meters of cheap industrial floor space to exploit. I have been thinking of starting up large scale mining there myself for a while. If you are looking for a site, this is probably one of the best in Europe at this time. In Sweden, datacenters above 0.5 MW don't have to pay energy taxes. This threshold may drop in 2018 to 0.1 MW. As you can see in the table above, it doesn't get much lower than this. This is the map coordinate (Google maps) https://www.google.se/maps/place/H%C3%A5frestr%C3%B6ms+F%C3%B6retagspark/@58.8055193,12.4248188,17z/data=!3m1!4b1!4m5!3m4!1s0x4644ea6cc45de4db:0x4e176b36d32b2d63!8m2!3d58.8055165!4d12.4270075?hl=en |
|
|
|
|
I have a few questions and thoughts that you could perhaps elaborate on..
1.
Who are going to write your code? Looking at your team, I see no one labelled as developer. In my world, the developers and software engineers are perhaps the most critical resources.
2.
If I understand your concept correctly, the training data is to be produced by advanced video rendering, which is well suited to GPUs. There is also a huge amount of rendering power distributed in the mining community just as you say. However, the mining rigs are mostly ill suited for such workloads since rendering requires considerable bandwidth between computer and GPU. Today, the typical mining setups use cheap USB bandwidth between CPU and GPU. Any rendering benchmark will show this being detrimental to performance.
Sure, rigs can be rebuilt but it's very hard to fit more than 3 or 4 GPUs with proper PCIe bandwidth. Only higher end systems will work well and the mining rigs mostly use the cheapest possible hardware (except for the GPUs). Lack of CPU and memory resources may also be a bottleneck depending on how the code is written. So, if you are serious about tapping into the distributed GPU resources, you need to explain the requirements so the community can adapt in time.
3.
Massive amount of images (or perhaps even video) hints at massive amounts of data. Can you present any calculations on needed internet bandwidth and storage requirements for this. It looks like you have some experimental projects to draw experience from?
5. The distributed computing part of the platform looks a lot like it would match the BOINC architecture. (I'm sure you know of BOINC).
Why not just piggy back on BOINC and get immediate access to an open platform for distributed computing tailored to sending compute jobs to a large number of users (with GPUs). BOINC handles all those issues with getting back scientifically correct data, users who don't complete their work packages etc.
I noticed they even have a crypto currency these days meant for rewarding the contributors on a voluntary bases.
Personally, I have been waiting since the middle of the 90:ies for someone to come up with a business concept for truly distributed computing. It must be worth something not having to build large datacenters for compute workloads for those cases where you don't need short latencies. BOINC evolved from SETI@home and is now ground breaking and record setting. However there is no reward for contributors and the previously "free" computing power is not at all free any longer with the power hungry hardware we have today.
I sincerely hope you are for real!
|
|
|
|
|
