Ethereum GPU mining?

[philipma1957]

So I reached roi on my eth gear. I sell off about every coin and 8 cards I had 1 390 and 4 nanos. I figured I wpuld mine and hold the eth for a bit and maybe buy some 480,s in about 9-10 days.

Then boom s hits the fan and I decided to sell all my cards . I am now in a profit zone and could replace the 15 cards i sold with 15 480's or do nothing.

I think I will just keep waiting for some real info on what the 480 wiil do as a miner. Then add some cards back online.

 

[Madpacket]

Wow...

https://medium.com/@hrishiolickel/w...-we-can-do-about-them-119aa2843007#.o4tg5bnei

http://www.comp.nus.edu.sg/~loiluu/papers/oyente.pdf

In a time of crises, we get gifts like this from the community. This is why Ethereum will likely succeed.

I'm humbled.

 

[Erithan13]

Sneaking toward $12.50 as I type, too early to be optimistic?

Fascinating reading on the hack even though most of this stuff is way above my understanding. I can certainly appreciate why the decision to fork eth to 'undo' the hack or not is causing so much trouble, seems to be a damned-if-you-do and damned-if-you-don't situation. Personally I think allowing it to go uncorrected will set an extremely dangerous (and lucrative) precedent for future hackers.

 

[Thala]

Personally I think allowing it to go uncorrected will set an extremely dangerous (and lucrative) precedent for future hackers.

Does not really matter much from the viewpoint of a potential hacker. A hacker would typically short-sell the coin before the attack. The variety of potential attacks is also much larger compared to those, where the attacker directly benefits from "stolen" value.

Doing a roll-back (hard fork) will set much more dangerous precedent. It also sets the margin for acceptable errors in the code/contract higher, as error seem to be "easily" corrected by roll-backs.

 

[DrMrLordX]

Doing a roll-back (hard fork) will set much more dangerous precedent. It also sets the margin for acceptable errors in the code/contract higher, as error seem to be "easily" corrected by roll-backs.

I disagree. The DAO is fairly unique . . . it's sort of the first smart contract to achieve public recognition. It's a mascot for Ethereum. Now that it has already been shown to be flawed once, it's unlikely that people are going to be willing to pile so much value behind a smart contract that is essentially in alpha, nor are people going to have quite the same reaction if it gets looted yet again.

People ought to know the risks now. Really they should have known them before hand, but . . .

 

[philipma1957]

Sneaking toward $12.50 as I type, too early to be optimistic?

Fascinating reading on the hack even though most of this stuff is way above my understanding. I can certainly appreciate why the decision to fork eth to 'undo' the hack or not is causing so much trouble, seems to be a damned-if-you-do and damned-if-you-don't situation. Personally I think allowing it to go uncorrected will set an extremely dangerous (and lucrative) precedent for future hackers.

Well the coin is over 13 as I type. and here is a link to the new 480 cards


http://wccftech.com/amd-rx-480-1500mhz-overclocking-tool-voltage-2/

seems to be encouraging if you want to mine with the cards like I do.

 

[poofyhairguy]

Doing a roll-back (hard fork) will set much more dangerous precedent. It also sets the margin for acceptable errors in the code/contract higher, as error seem to be "easily" corrected by roll-backs.

That doesn't look likely now as a Robin Hood attack might fix everything just with soft forks. Most of the DAO funds were already secured that way:

http://www.coindesk.com/ethereum-developers-draining-dao/

 

[Erithan13]

It's the moment we've all been waiting for: 480 hashrate. If this source is to be believed we're looking at 24MH/S at stock frequency up to 28MH/S with a memory overclock. No word on whether this could be pushed higher with a core overclock or the appropriate localwork/farm recheck etc commands although power consumption looks very good regardless (100W stock, 120W overclock).

That's basically 290 level mining performance at drastically reduced power. Certainly can't say I'm disappointed.

Edit: Link doesn't seem to work properly. It's on the front page of crypto mining blog.

 

[Mercennarius]

How important is CPU power in Eth mining? Building my first mining rig and focusing on getting a very low power CPU to reduce power consumption and want to make sure this won't cripple my performance much if any.

 

[EightySix Four]

How important is CPU power in Eth mining? Building my first mining rig and focusing on getting a very low power CPU to reduce power consumption and want to make sure this won't cripple my performance much if any.

Pretty much doesn't matter. I noticed a little bit of CPU work (mainly caused by the I/O controller) when syncing up the blockchain the first time, but otherwise it does pretty much nothing on my i3 rig.

 

[poofyhairguy]

It's the moment we've all been waiting for: 480 hashrate. If this source is to be believed we're looking at 24MH/S at stock frequency up to 28MH/S with a memory overclock.

Wow. I am glad I didn't buy a second 290 yet. I am going to get two of these 480 cards this summer and crossfire it up.

 

[Mercennarius]

Pretty much doesn't matter. I noticed a little bit of CPU work (mainly caused by the I/O controller) when syncing up the blockchain the first time, but otherwise it does pretty much nothing on my i3 rig.

Thanks.

 

[thilanliyan]

It's the moment we've all been waiting for: 480 hashrate. If this source is to be believed we're looking at 24MH/S at stock frequency up to 28MH/S with a memory overclock. No word on whether this could be pushed higher with a core overclock or the appropriate localwork/farm recheck etc commands although power consumption looks very good regardless (100W stock, 120W overclock).

That's basically 290 level mining performance at drastically reduced power. Certainly can't say I'm disappointed.

Edit: Link doesn't seem to work properly. It's on the front page of crypto mining blog.

At least in Canada the prices look like they will suck at first:

http://www.hardwarecanucks.com/forum/video-cards/72847-rx480-up-pre-order-ncix.html

The pre-order prices shown are way more than the converted price from USD. If these cards are nearing $400 CAD, even for the 8GB model, I will not be buying unless the hashrate can be taken well over 30MHs with an overclock.

 

[Madpacket]

At least in Canada the prices look like they will suck at first:

http://www.hardwarecanucks.com/forum/video-cards/72847-rx480-up-pre-order-ncix.html

The pre-order prices shown are way more than the converted price from USD. If these cards are nearing $400 CAD, even for the 8GB model, I will not be buying unless the hashrate can be taken well over 30MHs with an overclock.

Well that's disappointing. Makes me happy with my recent 290 purchases for $229

Hopefully it's just NCIX and NewEgg has better pricing as even with USD to CAD conversion this is not even close.

 

[Madpacket]

It's the moment we've all been waiting for: 480 hashrate. If this source is to be believed we're looking at 24MH/S at stock frequency up to 28MH/S with a memory overclock. No word on whether this could be pushed higher with a core overclock or the appropriate localwork/farm recheck etc commands although power consumption looks very good regardless (100W stock, 120W overclock).

That's basically 290 level mining performance at drastically reduced power. Certainly can't say I'm disappointed.

Edit: Link doesn't seem to work properly. It's on the front page of crypto mining blog.

Hmm, looking a little better than I guessed which is nice to see.

However I'm a little surprised of the overclocked memory results - it shouldn't help that much so it shows again how picky bandwidth or latency with DAG is.

Sad thing is this power efficiency (so far) is actually worse than a Fury Nano with a custom BIOS but it's a bit early to call this off one set of results.

Sad thing is if these cards are $400 CAD each the Fury Nano's for $100 more may make more sense for mining. Heck even the refurb 290's for $229 CAD undervolted and underclocked only pull around 150W each for approx 25Mh...

 

[n0x1ous]

How important is CPU power in Eth mining? Building my first mining rig and focusing on getting a very low power CPU to reduce power consumption and want to make sure this won't cripple my performance much if any.

its not important at all. Like zero. get the cheapest cpu you can get.

 

[MrTeal]

Hmm, looking a little better than I guessed which is nice to see.

However I'm a little surprised of the overclocked memory results - it shouldn't help that much so it shows again how picky bandwidth or latency with DAG is.

Sad thing is this power efficiency (so far) is actually worse than a Fury Nano with a custom BIOS but it's a bit early to call this off one set of results.

Sad thing is if these cards are $400 CAD each the Fury Nano's for $100 more may make more sense for mining. Heck even the refurb 290's for $229 CAD undervolted and underclocked only pull around 150W each for approx 25Mh...

I was actually going to post an estimate for the 480 around that based on my experience with the 380 in the last couple days. 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. I get 21.3NH/s at 1500MHz, so if the 480 is 2GHz clock as expected that would be 28MH/s with perfect scaling and more likely in the 26MH/s range. I'm hoping it's similar to Tonga too where core can be clocked way down since it has 30% more shaders.

 

[Madpacket]

I was actually going to post an estimate for the 480 around that based on my experience with the 380 in the last couple days. 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. I get 21.3NH/s at 1500MHz, so if the 480 is 2GHz clock as expected that would be 28MH/s with perfect scaling and more likely in the 26MH/s range. I'm hoping it's similar to Tonga too where core can be clocked way down since it has 30% more shaders.

Interesting. The 290's / 390's have a closer relationship between core and memory speeds . I find for 1080Mhz Core you want at around 1175Mhz memory to keep around 95% bandwidth utilization. If you drop down to 980Mhz core you can drop down to 1100Mhz or even 1075Mhz memory to maintain similar utilization rates.

Hawaii is quite the power hog with the memory bus being 512bit so underclocking memory makes a noticeable difference.

So if Tonga operates as you say with such a large discrepancy between core / memory there should be ample opportunity for core undervolting / underclocking and memory overclocking reducing the power consumption closer to the 100W mark or even less. I guess we just need to find the proper ratio given the extra shaders. These preliminary results look decent enough that 26Mh - 28Mh @ 100W may be possible with the right tweaks, we shall see .

Anyways I will be picking up a 480 to replace a MSI 390 sitting in one of my ITX cases. It's just a matter of finding one without getting gouged. No way I'm paying $400 CAD for one. $329 is max for an 8GB model with a backplate or $299 for the 8GB stock unit (which I would rather have anyways for mining).

 

[n0x1ous]

Anyone notice some of the shots of the card from the reveal event have the same Radeon font as the Fury X. That must have been a late change to change that font/logo. I liked the old one better, but maybe this will grow on me. Im sure I'll pick one up next week

 

[poofyhairguy]

Interesting. The 290's / 390's have a closer relationship between core and memory speeds . I find for 1080Mhz Core you want at around 1175Mhz memory to keep around 95% bandwidth utilization. If you drop down to 980Mhz core you can drop down to 1100Mhz or even 1075Mhz memory to maintain similar utilization rates.

I have read somewhere that 1125 is the best for memory on Hawaii to optimize the timings.

 

[thilanliyan]

^how do you actually get that low? I can't go below 1250 mem on my 290s using afterburner.

 

[codyray10]

It's the moment we've all been waiting for: 480 hashrate. If this source is to be believed we're looking at 24MH/S at stock frequency up to 28MH/S with a memory overclock. No word on whether this could be pushed higher with a core overclock or the appropriate localwork/farm recheck etc commands although power consumption looks very good regardless (100W stock, 120W overclock).

I'm torn.. Im currently mining with an overclocked + undervolted 290 @ 28.5 Mh/s. A guy local to me is selling a reference 290 for $150, which im supposed to pick up this weekend. OTOH, I kind of want to try and get one of these 480's at launch, for the power efficiency improvements. Just not sure of the availability as well as initial pricing for an 8GB model.

 

[MrTeal]

^how do you actually get that low? I can't go below 1250 mem on my 290s using afterburner.

Do you have extended overclocking limits enabled?

 

[Feld]

It's the moment we've all been waiting for: 480 hashrate. If this source is to be believed we're looking at 24MH/S at stock frequency up to 28MH/S with a memory overclock. No word on whether this could be pushed higher with a core overclock or the appropriate localwork/farm recheck etc commands although power consumption looks very good regardless (100W stock, 120W overclock).

That's basically 290 level mining performance at drastically reduced power. Certainly can't say I'm disappointed.

Edit: Link doesn't seem to work properly. It's on the front page of crypto mining blog.

Your link doesn't work.

 

[DrMrLordX]

Interesting, I can get 25 MH/s @ ~150W with a 290. So 24 MH/s @ ~100W would be an improvement. If there were a better selection of reliable 6-slot PCIe boards out there, mass-buying 480s would be a good play. Makes you wonder what the 470 can do?

Also, nanopool.org is now letting people vote on the soft fork by "voting with their hash power". They are not voting on the hard fork, and the people running nanopool seem hostile to the idea of the hard fork.

Personally I think that if they are going to do any fork at all, they might need to go all the way, but what do I know . . .

So I'm conflicted as to what to do? I certainly don't want to vote "no" and just leave things as they are, but I'd rather support the hard fork? Not sure if they will have another round of voting for that option?