Ethereum GPU mining?

[reb0rn]

@Elixer
fraud where in the contract? so if there was a fraud why ppl signed/invested??
Also The DAO was a perfect example of tulip and pyramid example (also it was promoted from ALL ethereum dev team!), lock ~11M eteherem from trading in stupid organization which do nothing and vote on stupid projects which as results have pumped ethereum price!

who is that body who will be judge is it man or its the ethereum virtual machine??

you ALL/MOST don`t understand the concept of crypt and decentralization, rules of the administration and ppl DO NOT APPLY here!

Also its not the point here but 72M ethereum was primined woth 0 cent invested, someone can and might have bought 51% in the next iteration of the system called POS they will rule, zero decentralization!

On top of that if ethereum allow forking what is the point and advantage of that system to any old system we already have if you only need to have 10% of ppl voting! (see miners voting most don`t care or understand and rest see only short term profit!)

 

[Madpacket]

With all the stress of the forking and hacks etc. This made laugh

https://www.youtube.com/watch?v=_O5fdMFKEC0

 

[MrTeal]

I would call this exploit fraud, so the initial premise is false.
If you allow fraud, then ethereum is worthless anyway.
It is the same as printing money on a home printer, and then the Feds don't do anything about it when that person goes to make purchases. Watch how fast the currency value will drop.

That's not really the same at all. The Feds expressly forbid people creating extra money, and will go after someone who counterfeits. Here someone wrote a contract, but did so in a way that allowed one of the parties to enrich themselves greatly. This probably isn't fraud although I'm sure you can make a case for unjust enrichment. I think the other owners of the DAO could definitely try taking it to court and it would be interesting precedent, though jurisdiction is obviously an issue.

My problem is that this isn't an Ethereum issue, it's a DAO issue. The DAO needs to be fixed and I imagine it will be in future versions, but the restitution idea being proposed isn't a good thing IMO. It's just that like the big banks, the DAO has been deemed too big to fail. One of the appeals of cryptocurrencies is that what you own, you own. Short of me being careless or someone torturing my private keys out of me, no one can take the Bitcoin I have without my consent. Here, you have the developers of the system pushing to rewrite the fundamental contract in order to bail out a private entity (that they are also part owners of) that was wronged. The mere fact that the central group is influential enough to push through a hard fork and actually make it work in order to save the DAO should give anyone pause before investing in the Ethereum ecosystem.

 

[MrTeal]

Yes my socket 1150 H97 Anniversary board works great after some troubleshooting.

I was having issues with Windows detecting the cards at first but I solved detection this way.

- Stopped using the onboard GPU and Intel graphics.
- Added "one" Radeon GPU using a riser in the x16 slot, this acts a the primary card which I hook up to a monitor for setup.
- Removed older drivers using latest DDU and installed latest AMD beta drivers.
- Disabled crossfire.
- Ensured the female Molex on the board near the x16 slot was being fed power (this shouldn't have any effect if using powered risers but it seemed to help with detection).
- Added one card on a time and moved the PCIe adapters around if a card wasn't being detected.

Eventually all five cards worked. Now I think one my ports is dead, the third 1x PCIe port on my board doesn't respond to any card with a riser. I couldn't get any card to be recognized in that slot so it's possible you have a dead slot as well, but I wasn't even getting an error message.

I also helped a friend put together a similar rig with the same motherboard and CPU and the same refurbished Radeon cards (he's only using four instead of five cards). He's not having any issues with his board or never complained about card detection but I warned him about the issues I ran into. He's also using Windows 10.

Question, what's this 6x GPU mod you speak of?

https://bitcointalk.org/index.php?topic=712228

AMD made some changes to their drivers a few years ago, which seemed to limit it to 4 GPUs. Mine isn't a physical issue, it's strict a Windows driver one as near as I can tell. The slots are on the board from top to bottom
PCIe1 - 1x
PCIe3 - 16x
PCIe4 - 1x
PCIe5 - 1x
PCIe6 - 1x
PCIe7 - 1x

I have one card not on a riser plugged into PCIe3, but it doesn't matter how I arrange the rest. I have one mounted by the CPU that's plugged in to PCIe1, and the other 3 below the main GPU. If I use 1,3,4,5, and 6, the GPU is slot 6 shows the Error 43. If I move 5 over so it's 1,3,4,6,7, the GPU that had been working in 5 and is now in 7 shows Error 43, and the one in 6 starts working. If I just unplug the one in PCIe1, the primary and the three on the bottom of the board on risers all work.

It's really annoying. Apparently Win8 and Win10 don't have these issues, but I'm hesitant to use the free Win10 update.

 

[Madpacket]

@Elixer
fraud where in the contract? so if there was a fraud why ppl signed/invested??

who is that body who will be judge is it man or its the ethereum virtual machine??

you ALL/MOST don`t understand the concept of crypt and decentralization, rules of the administration and ppl DO NOT APPLY here!

Also its not the point here but 72M ethereum was primined woth 0 cent invested, someone can and might have bought 51% in the next iteration of the system called POS they will rule, zero decentralization!

On top of that if ethereum allow forking what is the point and advantage of that system to any old system we already have if you only need to have 10% of ppl voting! (see miners voting most don`t care or understand and rest see only short term profit!)

Sorry to jump in but you're making some assumptions here about Ethereum but I don't think you quite understand what's happening.

1. You need to separate intent from the expectations of running automated code. The intent of the DAO contract was never to be abused in the way it was. This recursive bug was "exploited" and money was stolen. You can try and justify it anyway you want but this was clearly someone taking something that didn't belong to them. Simple theft.

2. Currently PoW miners are the deciders of forks. This will change once PoS is established. Sure many miners are in it for the money but there many miners (such as myself) that haven't sold a single ether yet. You can't assume to know the intentions of each miner.

2. 51% attacks are basically mitigated with Proof of Stake. It's not in the best interest stake holders to allow abuse in the system else they risk losing their stake. Proof of stake adds more decentralization with regards to the risk of a 51% attack. Think 256 giant stake pools. Do you think you could take over the majority of spread out hashing power of each of these pools with a 51% attack?

3. The 10% of miners is representative. You'll never get everyone to vote on something so you need to take the representative proportion and use that as a baseline. If 10% of miners have voted of which 80% want a soft fork, then we should get a soft fork. We can only work with the votes that we have in front of us. No one is holding a gun to someone's head asking them to vote.

If the rest of the miners don't care enough to vote then they are not representative of either splitting or not splitting therefore they don't count.

This has been repeated ad-nauseam but the soft and then hard fork will "save face" of a system that's worth saving.

Fixes to code will come, more mistakes will happen and most likely additional soft /hard forks will happen. This is disruptive technology and in order for Ethereum to evolve at the pace it is, these types of things need to happen. Failure is how to move something this important forward (Cliche as that sounds).

Ethereum does not need to follow in the footsteps of a failing system (Bitcoin). Ethereum will take the good things about Bitcoin (and other systems) and then move on.

 

[Madpacket]

t's really annoying. Apparently Win8 and Win10 don't have these issues, but I'm hesitant to use the free Win10 update.

Interesting. That must be super annoying. I am using riser cards for all five cards, not sure if that matters. Windows 10 isn't that bad for mining. The only issues I've ran into is sometimes forced reboots due to security patches but overall it's worked well.

 

[EightySix Four]

Fixes to code will come, more mistakes will happen and most likely additional soft /hard forks will happen. This is disruptive technology and in order for Ethereum to evolve at the pace it is, these types of things need to happen.

Ethereum does not need to follow in the footsteps of a failing system (Bitcoin). Ethereum will take the good things about Bitcoin (and other systems) and then move on.

The fear from some of us is that by forking to save the DAO users their Ether that you've introduced a situation where people believe they'll be bailed out if there is an issue, therefore spending less money on security audits for giant projects. By "saving" the DAO we are in fact creating a situation that encourages more dAPPS that need saving.

I'm going to vote to fork, because I think it is in the best interest of Ethereum to just past all of this, but I really hope that solidity is killed off in favor of a more accurate, descriptive language with significantly less abstraction.

 

[poofyhairguy]

Interesting. That must be super annoying. I am using riser cards for all five cards, not sure if that matters. Windows 10 isn't that bad for mining. The only issues I've ran into is sometimes forced reboots due to security patches but overall it's worked well.

Yeah luckily Windows is stupid easy when it comes to running your mining software on boot and overclocking on boot and all that. I still haven't figured it out how to start mining software in a terminal on boot for my Ubuntu machines.

 

[Madpacket]

The fear from some of us is that by forking to save the DAO users their Ether that you've introduced a situation where people believe they'll be bailed out if there is an issue, therefore spending less money on security audits for giant projects. By "saving" the DAO we are in fact creating a situation that encourages more dAPPS that need saving.

I'm going to vote to fork, because I think it is in the best interest of Ethereum to just past all of this, but I really hope that solidity is killed off in favor of a more accurate, descriptive language with significantly less abstraction.

That's one way to look at it sure. But I disagree with that premise

People expecting bailouts in the future for similar situations will get zero sympathy. Fool me once, shame on you. Fool me twice....

You'll never see future proposals earning the trust of the people where it'll reach 150 million dollars. Security will undoubtedly improve. In fact I wouldn't be suprised if the fork enforces a 10 million or less dollar limit (as per Vitalik's auggestion).

This was a huge wake up call not only for The DAO (which is fundamentally still a good idea) but for Ethereum in general. I agree Solidity either needs a major overhaul with giant safe guards or be replaced altogether by a a more functional deterministic programming language. This will all take time (soft fork) but will happen (hard fork).

I work in computer security for a living and have seen this countless times by traditional companies.

A rushed piece of (normally) Agile developed software is released to meet the unrealistic "set in stone" dates by the project managers. Product security is called in the week before release and is told to "scan" the application (they have no idea what we do). We end up finding numerous flaws, flaws that were communicated to the developers early on in the project to watch out for. Simple things like XSS header protection and sanitizing user input are ignored. We communicate risk to the PM's and they decide to take the risk and promise they'll fix it post launch.

Because secure programming best practices are thrown out the window (they don't really teach secure programming in most colleges!), development goes on and a crappy unfinished product is released.

High fives all around. Bonuses for the PM's for releasing the product on time and a week or two later the site gets owned by a script kiddie with simple prebuilt Metasplot modules.

This happens all the time. If lucky a white hat reports to the company via responsible disclosure and things eventually get fixed. If your company is non responsive or doesn't have a decent bug bounty expect your shiny new app flaw to show up on Pastebin.

Anyways this scenario is all too common.

Developers often learn the hard way how critical security is to get right. Secure decelopment is very hard! But it can and will be done.

Add in the complexity of matching intent with autonomy without being able to make changes post launch even with critical flaws disclosed and you'll start to understand just how difficult a position Ethereum is in.

In order for Ethereum to grow into a mature environment, I expect soft and hard forking will continue to be commonplace. Else we'll just end up with problems like 1MB block size limits years after being warned about it being a problem.

And that's no fun

 

[tential]

Yeah luckily Windows is stupid easy when it comes to running your mining software on boot and overclocking on boot and all that. I still haven't figured it out how to start mining software in a terminal on boot for my Ubuntu machines.

All the guides I've seen seem pretty complicated. Hell the one RS sent me for mining on Windows 10 confused me.

Only reason I've never gotten into mining was it wasn't easy to do all of this on boot for Windows 10 efficiently without changing drivers and what not.

 

[EightySix Four]

People expecting bailouts in the future for similar situations will get zero sympathy. Fool me once, shame on you. Fool me twice....

Have to disagree here, too big to fail is too big to fail. I have no sympathy for people who are in the current DOA because the chances of it going well were way smaller than it going terribly terribly wrong, which is why I didn't buy the tokens in the first place. The code and concept was way too new for v1 to be done right.

The incentive is now there to get as big as possible when building something like this so that if it fails the whole community has to go back and fix it.

Anyways this scenario is all too common.

This is the crux of what you said right here - it will keep happening and by hard forking we're creating further incentives for it to keep happening. If we soft fork and lock all the Eth away, maybe the next DAO won't get so oversized.

And I've been both on the PM and Ops side of the exact situations you mentioned in both startups and big companies. They don't care unless the pain comes to their users and customers. Hard forking is pretending the pain didn't happen.

Like I said, the fork will probably go forward either way. But I do think it is a moral hazard and calls into question some of the fundamental premises of the chain.

 

[IntelUser2000]

\On the 1792SP Tonga card, I can even drop down clocks to 850 with no loss of hash rate; it's purely tied to memory speed.

With Scrypt, people report that there's optimal clock speeds for the memory before you start losing hash rates. I did experiment a bit with my 280X.

1500MHz-Highest
1350-1375MHz-2nd Highest

I went much lower to 1100MHz or so and started ramping up. If you graph the performance versus clock speed I'd imagine it looking like a pulsed DC wave with full rectification: https://en.wiki2.org/wikipedia/comm...ication.svg/im370-640px-Rectification.svg.png

But of course there are caveats. If its entirely based on latency then 1500MHz and 1375MHz should be the same, but the former was somewhat higher, maybe 5%.

I wanted to play around with core clocks at the same time as I was with memory clocks but my card doesn't allow me to do that. Hopefully reference 480 cards will.

As of right now I am running it 1051MHz with 1500MHz memory at 112mV undervolt(from 1.2V to 1.088V) with Decred at 50. It uses 235W with Decred at 50 and 192W with Decred at 1(my hash rates are few % higher with Decred active and it only adds 2W so...). I have to clock it 1MHz higher because for some reason at base 1050MHz it runs it at lowest core/mem clocks.

(Samsung memory on a Visiontek R9 280X = no fun)

I am wondering if the reason RX 480 performs better with memory clock increase is because its getting close to the multiple of 1350-1375MHz.

Default: 2000MHz(Multiple of 1000MHz): 8GT/s
Optimal assuming 1375MHz: 5.5GT/s, OR 11GT/s

So Ethereum performance is very complex to figure out. I am guessing the reasons are due to following:
-Architecture: Obviously AMD vs Nvidia, and Tonga being a later version of GCN. Not very much of course
-Latency
-Relation between core clock and memory clock? Optimal ratios?
-Drivers?: Though for my 280X it makes zero difference, 15.2, 16.5, 16.6, 15.7...

 

[EightySix Four]

So Ethereum performance is very complex to figure out. I am guessing the reasons are due to following:
-Architecture: Obviously AMD vs Nvidia, and Tonga being a later version of GCN. Not very much of course
-Latency
-Relation between core clock and memory clock? Optimal ratios?
-Drivers?: Though for my 280X it makes zero difference, 15.2, 16.5, 16.6, 15.7...

Part of the answers can be found here: https://github.com/ethereum/wiki/wiki/Ethash-Design-Rationale

IO saturation: The algorithm should consume nearly the entire available memory access bandwidth (this is a strategy toward achieving ASIC resistance, the argument being that commodity RAM, especially in GPUs, is much closer to the theoretical optimum than commodity computing capacity)

The hash algorithm is intentionally I/O limited. It appears the Tonga line of GPUs can hash very very quickly, but the VRAM just cannot keep up with the I/O saturation goals of ethash.

 

[DrMrLordX]

Yeah luckily Windows is stupid easy when it comes to running your mining software on boot and overclocking on boot and all that. I still haven't figured it out how to start mining software in a terminal on boot for my Ubuntu machines.

Once I have something better put together wrt on-boot mining in Linux, I'll post here. Right now my solution sucks. Maybe I'll finally get around to improving the situation this weekend, or if not I'll have more time next weekend . . . well probably.

Right now I am abusing .profile successfully, but that's a pretty crappy solution overall.

 

[reb0rn]

@Madpacket
Still i agree with you but you look it at from the human perspective, in mathematical computation systems there could be no frauds, code (smart contract) run as it was writen, its the main selling point of ethereum, there is NONE value if that concept is broken
If that is so, they should let the DAO rot, it would be only option... who ever invested in stupid dao, let them learn
My main concern with ethereum is also huge 90% premine which is 72M, security?? almost like no one care for it, POS system is not even coded and its hyped??

The DAO was clear ponzy in mind, lock ethereum some 15%?? for long time, pump the price, main reason ppl invested is to gain something!
I believe most contracts would lock ethereum on long time on that way making ponzy for price to go to 5x, 20x, after all who would pay hard cash for smart contract when I see no gr8 value at them!

 

[DrMrLordX]

reb0rn, it seems that you do not like and never have liked Ethereum. Why do you even bother?

 

[techie81]

@Madpacket
Still i agree with you but you look it at from the human perspective, in mathematical computation systems there could be no frauds, code (smart contract) run as it was writen, its the main selling point of ethereum, there is NONE value if that concept is broken
If that is so, they should let the DAO rot, it would be only option... who ever invested in stupid dao, let them learn
My main concern with ethereum is also huge 90% premine which is 72M, security?? almost like no one care for it, POS system is not even coded and its hyped??

The DAO was clear ponzy in mind, lock ethereum some 15%?? for long time, pump the price, main reason ppl invested is to gain something!
I believe most contracts would lock ethereum on long time on that way making ponzy for price to go to 5x, 20x, after all who would pay hard cash for smart contract when I see no gr8 value at them!

Please take your agenda somewhere else.

 

[DrMrLordX]

Looks like nanopool is accepting the soft fork (2% vs .3% yes/no). Their latest update:

We are starting to mine according the voting results. The results are reviewed once a day. Now we are ACCEPTING the fork.

 

[reb0rn]

reb0rn, it seems that you do not like and never have liked Ethereum. Why do you even bother?

I do not like scam coins in crypto world, and etereum is definitely that, ppl here investing should know better, invest only what you can lose, and beware of ppl selling hype

Sure I am miner and from that point I am interested in mining and profit but primary from dumping it from real decentralized and fair crypto as BTC, etc

@techie81
I have no agenda

 

[RussianSensation]

RX 480:

"24 MHS for Ethereum Dagger-Hashimoto mining with about 100W, we are also told that with a memory overclock 26-28 MHS are possible with about 120W of power usage."
http://cryptomining-*********/8008-possible-ethereum-hashrate-for-the-new-amd-radeon-rx-480/*

RX 470 might be also be a sleeper mining card at $149.

*Link doesn't work directly. Search "Possible Ethereum Hashrate for the New AMD Radeon RX 480"

 

[Erithan13]

RX 480:
"24 MHS for Ethereum Dagger-Hashimoto mining with about 100W, we are also told that with a memory overclock 26-28 MHS are possible with about 120W of power usage."
http://cryptomining-*********/8008-possible-ethereum-hashrate-for-the-new-amd-radeon-rx-480/*

RX 470 might be also be a sleeper mining card at $149.

*Link doesn't work directly. Search "Possible Ethereum Hashrate for the New AMD Radeon RX 480"

Posted a few pages back but it got a little overtaken with the discussion about the ethereum fork. Very good news if that turns out to be true, I don't see any reason why a little overclock shouldn't get that clear above 30MH/S. On the other side I imagine a good bit of undervolting will give us outstanding MH per watt - might be able to scrape 380 level hashrate (~20MH/S) with under 75W? Can't wait either way given that summer has finally arrived here and my cards aren't happy with the heat.

Now, in the interests of balanced discussion I feel obligated to include a dissenting opinion:

We already know Polaris is a terrible miner card due to low compute performance.

You heard it here first folks

 

[PhonakV30]

We already know Polaris is a terrible miner card due to low compute performance.
http://cryptomining-*********/tag/rx-480-hashrate/

Note something? Yep, lack of memory bandwidth is making it hard for both.

You heard it here first folks

I surprised that he said it.His hate toward AMD is controlling him.If he knew what r9 290X's Score is, he wouldn't say those nonsense words.

If Rx 480 = 24Mh per 100W and 4 CF = 96Mh/400w then 2 R9 290X = 60mh/600w !! damn It's freaking good mining card.

 

[DrMrLordX]

290X doesn't have to use 300W mining, though. You can easily get 30 MH/s out of a reference 290 @ 200W if you don't mind the blower (and the VRMs haven't started to act up, and the TIM isn't dried out and cracking, and and yeah).

The 480 still looks pretty good by those numbers, though. The 4 GB version for $199 MSRP should do just fine. Plus you'll be able to use the latest drivers with it.

 

[SimianR]

Yeah I think the 470 & 480 will do just fine as mining cards, especially at half the power use. Shintai doesn't really know much about mining it seems, but I think he was trying to use that post to prove a completely separate point.

 

[Thala]

Shintai doesn't really know much about mining it seems, but I think he was trying to use that post to prove a completely separate point.

Wouldn't it be reasonable, in case you have not the slightest clue what you are talking about, to just shut up instead of posting bold statements like RX480 is terrible at mining?

In any case what's the MHS/W (or MH/J) with your current setup?