Ethereum GPU mining?

[ozzy702]

It seems to be a big problem with Mist and Geth. The software is still in the beta stage so there's a lot of work to do.

Last year when I playing with Mist, it took a long time, but it at least completed. Now, it just doesn't get finished. Even the developers seem to acknowledge the issue.

Mist and Geth are garbage and in need of some serious work. It's a black eye for Ethereum to not have a functioning wallet associated with it's main development.

 

[Madpacket]

There's a ~+-300GB torrent floating around of all blocks since the Genesis Ethereum block. I downloaded that to save time and then ran Geth to get caught up after validation. After 24 hours or so (on an external USB 2.0 drive, spinning rust) it eventually finished. The process is very I/O, RAM and CPU intensive. 12 - 16GB and 4 cores minimum to run a node, or at least to get caught up and then you can xfer the blocks to a slower PC.

 

[Madpacket]

Mist and Geth are garbage and in need of some serious work. It's a black eye for Ethereum to not have a functioning wallet associated with it's main development.

Parity is much much faster than Geth but the team behind it has the stigma of some serious security issues.

 

[DrMrLordX]

Which is why most people just use MEW.

 

[ozzy702]

Which is why most people just use MEW.

I use MEW these days.

 

[Charlie22911]

Got everything downloaded on my main PC in a few hours, working on transferring it over to the NUC to see how it performs. If it doesn’t I’ll just find another project for this little box.

 

[IntelUser2000]

There's a ~+-300GB torrent floating around of all blocks since the Genesis Ethereum block.

300GB? Is it really that big? When I said I played around with it last time, it was in the 20-30GB range. Man, that's huge. SSD setups are out of the running for most folks then.

I hope they do something about it eventually. What happens when it reaches in the TB range? 10TB?

Parity is much much faster than Geth but the team behind it has the stigma of some serious security issues.

Security is a big deal.

 

[Charlie22911]

Even after copying the chain data over it still can’t keep synced. Maybe a faster hard drive would fare a little better but it’s starting to look like a SSD is becoming more of a requirement.
FYI the chain data on my desktop after a full sync is sitting at a hair over 44GB (%appdata%\Ethereum).

Looks like I’m going to be looking for something else to use this Apollo Lake NUC for, oh well.

 

[DrMrLordX]

Damn that sucks. If syncing is that intensive, it may chew up some SSDs and spit them out. My NVMe drive is already reading 92% remaining life and I've only had it in service since March. Bleh!

 

[Charlie22911]

Damn that sucks. If syncing is that intensive, it may chew up some SSDs and spit them out. My NVMe drive is already reading 92% remaining life and I've only had it in service since March. Bleh!

Yeah, my 950 Pro kicked the bucket a couple weeks ago with no warning.
Once things are synced up though drive activity is pretty tame, I’ll probably revisit this in the future if I come across a decent SSD.

 

[DrMrLordX]

Hmm okay. If you trust Crucial's MX300 drives, you can get a 1.1TB for around $240:

https://pcpartpicker.com/products/internal-hard-drive/#t=0&i=25&S=960000,12000000&sort=price&page=1

 

[Charlie22911]

Hmm okay. If you trust Crucial's MX300 drives, you can get a 1.1TB for around $240:

https://pcpartpicker.com/products/internal-hard-drive/#t=0&i=25&S=960000,12000000&sort=price&page=1

Nice! IIRC Crucial is owned by and uses Micron flash, which I trust. I'll likely pick up a couple of those for my unRAID box and steal the SSDs from it for this (among other things).

 

[DrMrLordX]

Only reservation I ever had about the MX300 is that it's TLC NAND instead of MLC like the MX200. MX200 availability has been low and prices have been high, so for the performance it wasn't really "worth it" vs Samsung drives at similar price points after awhile. For example, a 1 TB MX200 drive:

https://pcpartpicker.com/product/tvKhP6/crucial-internal-hard-drive-ct1000mx200ssd1

Ouch.

 

[Charlie22911]

I don't care for TLC, but cache drives don't see very write heavy workloads so I don't mind putting them there, it'd free up a couple of ports too.

Is anyone dualmining? Crunching ETH + LBRY right now, HODL of course because I think LBRY has a lot of potential. I'm quite enjoying the "free" heat right now!

EDIT:

It takes about ~33 seconds to sync each block (averaged over 5 blocks), so yeah HDDs are out. The blockchain is outrunning the HDD, even with nothing else running.

 

[IntelUser2000]

Is anyone dualmining? Crunching ETH + LBRY right now, HODL of course because I think LBRY has a lot of potential. I'm quite enjoying the "free" heat right now!

How long have you been dual mining?

My experience hasn't been too good with them. They are quite hard on the cards.

 

[Charlie22911]

I’ve been dual mining on and off for the last year or so, dual is all I do now that nicehash is kaput though. It doesn’t really hurt my ETH hashrate that I can tell, so it’s just bonus coin.
I don’t notice a difference in temperature or power consumption versus straight ETH, granted other algos may differ.

EDIT:
I have several 1080ti,1080,1070,1060 though, no AMD cards so I’m not sure how hard dual mining is on them versus nVidia.

 

[IntelUser2000]

I don’t notice a difference in temperature or power consumption versus straight ETH, granted other algos may differ.

The temperature is fine, because you can just ramp up its fans. Its the wear that I worry about. I dual mined on RX470/480s for few months and after that the cards were behaving weird for a while. I stopped it right away. I can't see why it would be different with Nvidia, they both use silicon chips. For the gains on profit, which turned out to be 10-15% I didn't think it was worth it at all.

Damn that sucks. If syncing is that intensive, it may chew up some SSDs and spit them out. My NVMe drive is already reading 92% remaining life and I've only had it in service since March. Bleh!

Yeah, my 950 Pro kicked the bucket a couple weeks ago with no warning.

What the heck?

What's 92% mean? How much writes is that? 92% is still a lot left. 8% in 9 months means it would take you 100 months to make it go to zero. That's 8 years. Plus SSD testing has shown most of them go past 0%.

If it died without warning and in just couple of years I'd put the blame on the controller failing. Failures aren't just about NAND writes being used, failure points exist elsewhere. For example, DRAM can fail, even though write cycles are technically infinite. GPUs don't have "write cycle limits" either but they can fail. There are normal failure besides usage failures like with NAND.

 

[Charlie22911]

I’ve got my power target set to 70% at the moment, fans are auto at 33% @63C; VRMs are within a couple degrees of that. Power draw from the wall for this rig is 1160w +\- 10w or so according to this elcheapo belkin power monitor thing I have. Those numbers don’t really change much whether I dual mine or not.

The cards are all EVGA ACX based, the coolers are quite good and the fans are all running at a low speed, I wouldn’t think cooler wear to be much of a concern. I try to take care of my money tree .

EDIT: words are hard.

 

[IntelUser2000]

I’ve got my power target set to 70% at the moment, fans are auto at 33% @63C; VRMs are within a couple degrees of that. Power draw from the wall for this rig is 1160w +\- 10w or so according to this elcheapo belkin power monitor thing I have. Those numbers don’t really change much whether I dual mine or not.

The cards are all EVGA ACX based, the coolers are quite good and the fans are all running at a low speed, I wouldn’t think cooler wear to be much of a concern. I try to take care of my money tree .

EDIT: words are hard.

The temps are pretty decent, so are the fan speeds. How many cards are you running? 1160W is quite a bit for one system.

I meant wear on the cards themselves.

 

[DrMrLordX]

What the heck?

Sorry, no context. SMART was reading drive life at 100% solid 'til a few months ago. Not sure how it goes about reading these things.

It seemed to decline after I tried running a full node 24/7 for awhile. I got sick of it chewing up all my available SSD space and stopped running the full node. Checked the drive on a lark, and it was in decline.

Maybe I'm misinterpreting the data, but my general impression is that NAND wear is eventually going to reduce the effective available space on the drive. Maybe the SMART system isn't quite tracking at that level. In contrast my MX200 drive that has been in service for a year longer than my NVMe BPX is still showing 99% life on it. Same basic usage patterns, but I haven't run a full node on that MX200 much.

 

[IntelUser2000]

Sorry, no context. SMART was reading drive life at 100% solid 'til a few months ago. Not sure how it goes about reading these things.

It seemed to decline after I tried running a full node 24/7 for awhile. I got sick of it chewing up all my available SSD space and stopped running the full node. Checked the drive on a lark, and it was in decline.

Yea? And how much has been written? This I believe is the important part.

Sometimes computers behave like that. A device might report 100% for quite a while but suddenly drop 5%. I'd figure though the meter isn't that far off as 8% might reflect realistic usage for the device over a 8-9 month period. In any matter, anything being used(batteries, SSDs) staying at 100% is incorrect.

The 950 Pro is rated at 2.5x higher endurance rating than the MX200. Theoretically, the cells on the 950 Pro should last longer as well because its using vertically stacked NAND which goes on older lithography, while the MX200 is on 16nm planar NAND.

 

[Dice144]

This thread got me back into mining in late June. With the crazy raise of coins, would of been better to straight invest the cash but now got 2 x1080 TI "free" and some more gear. Best part is, love building computers.

The GF kept saying I was crazy and her family said it is not real income until you cash some out. Cashed out just enough to pay off the cards. Then in nerd fashion ordered two more. Still want a Vega for my freesync monitor just to test the tech.

 

[IntelUser2000]

Only reservation I ever had about the MX300 is that it's TLC NAND instead of MLC like the MX200.

TLC being worse than MLC should be somewhat countered by being vertical NAND on an older process versus an MLC on a latest 16nm process.

 

[Charlie22911]

The temps are pretty decent, so are the fan speeds. How many cards are you running? 1160W is quite a bit for one system.

I meant wear on the cards themselves.

6 1080tis in this one, I’ve still got the 6970s and my 5870 from the “early” bitcoin mining days. Mined on those cards through the FPGA days up to the point ASICs started to show up; they all work fine. Also have done folding@home for a while after that on my 1080s before the recent mining boom with no issues... I think you could have just had a weak card?

 

[DrMrLordX]

Yea? And how much has been written? This I believe is the important part.

The BPX reports total host writes of 69TB (69,109 GB). The MX200 reports total host writes of 22TB (22,187 GB). If I had kept closer track of drive health, I could tell when the BPX accumulated the majority of those writes.

I still suspect it's geth that did the majority of that writing though.

The 950 Pro is rated at 2.5x higher endurance rating than the MX200. Theoretically, the cells on the 950 Pro should last longer as well because its using vertically stacked NAND which goes on older lithography, while the MX200 is on 16nm planar NAND.

TLC being worse than MLC should be somewhat countered by being vertical NAND on an older process versus an MLC on a latest 16nm process.

Wait, is that 3D NAND really on an older process? I thought both the 950 Pro and MX300 were newer drives than my MX200.