Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Locked thread
Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

1gnoirents posted:

Also I've been seeing videos pop up about using liquid metal pastes on GPUs to great result on Pascal cards, might be worth looking into. I'm going to be testing that on both a Gigabyte 1080 G1 (pretty poor fan performance overall) and a EVGA 1080ti SC2 ICX today or tomorrow and see what happens. I have some leftover from Haswell delidding days

It does work but like all liquid metal it's conductive so if it spreads out to the discretes and shorts something you're hosed. And some of them are pretty tricky to apply properly. I don't think the gains are remotely worth it.

You see big gains from repasting. Factory thermal pastes are not very good compared to aftermarket stuff and often end up pretty dried out after a while. But there's no reason to use liquid metal instead of just some Gelid GC-Extreme or NT-H1, and some pretty big risks/downsides.

Pretty sure those videos were sponsored content from the companies who make liquid metal paste.

Adbot
ADBOT LOVES YOU

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Cygni posted:

Yeah, especially considering you can get a 3-ish use tube of ceramic ArcticSilver 5 for $6, and a single serving of something like CoolLabs Liquid Pro is $25... and the CoolLabs stuff is conductive and only cools like 2-3 deg better than AS5, ehhhhhhh

This is really what it comes down to for me: like absolute best case liquid metal makes a 5 degree difference vs random thermal paste, and against good stuff ($10 a tube instead of $5 a tube) it should be <3 degrees. It's just not worth even a trivial chance of killing my card over three degrees.

I am very serious about that whole video being sponsored content from CoolLabs, I'm pretty sure it is. It makes no damned sense to use liquid metal on a bare die when they cool so easily on AIOs.

Most people are too chicken to open up their GPU anyway and then on top of that you need to be super careful not to short out your card? Why the gently caress would you not just buy an AIO bracket and call it good? That's going to make like a 20-30 degree difference in your temps, gently caress 3 degrees. And if you're looking to throw money away, get some Noctua fans to go with it. They are expensive as gently caress (literally $20/fan) but they are legitimately super good fans.

(I actually got a Noctua 120mm fan for my G10/H75, will be waiting a few more weeks for the Gelid VRM bracket to show up but... I'm tempted just to assemble it and go. My 1080 temps are getting really bad as the summer heats up, and I think my thermal paste is dried out pretty badly from mining on it for a month. Even at 90% power limit/100% fans I'm at 80C.)

Also, Liquid Pro in particular is supposed to be loving awful to deal with. It's really liquidy, whereas the Liquid Ultra was a little more viscous.

Paul MaudDib fucked around with this message at 20:49 on Jul 10, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
Doesn't liquid metal also dry out pretty quickly and needs to be reapplied on like a yearly basis too?

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

1gnoirents posted:

Keep in mind guys a huge glaring factor last time with the 290 purge in that during the mining craze it went from a $600 card that was worth $600 against the competition (780ti, although not really, but still) to barely being competitive with a $300 card (970's hit $300 when it all went to hell). I do believe the main reason was huge supply of course but I wouldn't hold your breath trying to pick up 1080's for $200 next month.

The pricing should be good to pretty great if it goes down but people are still going to be willing to pay a closer to retail new prices than not, which in reality was all that happened last time too if you adjust for price/performance at the time.

The perfect storm would be, say, a surprise volta launch at the same time but I dont see that one happening anymore

AMD cards will probably be pretty awesome prices because MSRP was already very low to start with

1070s were dropping below $300 at times before they got bought out a month or so ago. 1080s were flirting with $400.

I'd say 1070s are going to end up at $200-225. Since miners didn't buy up 1080s prices aren't going to tank as immediately/drastically, but they are still in the same market. And the 1080 in particular still isn't quite enough for 4K, it's a 1440p card still, so it's very subject to substitution by 1070s if the price is too high. I would bet we see $300-350 for 1080s after all of this hits the market.

A huge amount of those mining cards are 470/570 and 480/580 so those are going to be the prices that tank hardest. I wouldn't be surprised to see 480 4 GB models hit $100 and 8 GB models under $150. That's basically where they were before the mining bubble, but I think there's pretty good demand for them from gamers at the $100/150 price points.

I would be very curious to watch a graph of new units sold over the past 2 years and over the next 6-12 months.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
wow, Michael Cera got fat

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Herman Merman posted:

Otherwise correct, except that earning $ from buttmining isn't economic growth.

Keynes' famous example was burying bottles of money and letting private enterprise dig them up, which is basically the same thing conceptually as mining. You can view the whole scheme as a form of stimulus from the "government" of the coin, as encoded in the rules the client follows regarding generation and distribution of currency.

Some people believe that government spending inherently displaces private spending in all cases, which probably isn't true, especially when a currency is just being issued and "the government" owns 100% of the money supply.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Peachfart posted:

mowing your neighbor's lawn(you may even get a bonus glass of lemonade).

Look buddy, do I look like someone who wants to get brain damage from heatstroke

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Methylethylaldehyde posted:

Sever, sell your card now before the entire butt-economy collapses.

Come August 1st, this thread will be solely devoted to reposting r.buttminer sob stories.

The window to sell your cards is closing almost immediately if not already closed. 480/580 prices are down 25% from peak bubble prices on eBay. I guess it'll be even lower in another month, but now that prices are plummeting you are at higher risk of buyers backing out of sales or refusing delivery when they see that prices have dropped $100 between when they ordered and when it arrived.

eBay BIN and Craigslist are going to be tougher to sell at, and Auction format is going to push your sale date even farther into the crash zone.

RX 470/480/570/580 cards and "oddball" cards like Fury and 390 are going to collapse sooner and harder than 1060s and particularly 1070s. Might still be a little bit of a window left on 1070s but not much. Party's over.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

MaxxBot posted:

I still don't understand the crypto market fully, what would make the decline pause like this? Looking at the difficulty chart for ZCash it was steadily going up for a while but seems to have leveled off.

https://www.coinwarz.com/difficulty-charts/zcash-difficulty-chart

Profits have declined to the point where people are no longer comfortable buying new hardware, I guess.

poo poo seems very weird right now, there seems to be a lot of uncertainty about the Segwit2X patch (a relatively large conceptual change being deployed onto a 'production' system), to the extent that when it looks like it might kick in the prices dive, which tanks the price enough that miners move to other currencies.

There might be a degree of capital flight from Bitcoin to other currencies, but normally Bitcoin is the anchor and other coin prices move against Bitcoin rather than against USD. When Bitcoin goes up and down, the other currencies move up and down proportionally (in USD) as well.

It also doesn't help that Ethereum is going through a huge change to their core algorithm too, and there's some real questions in terms of security with larger pools (not majority but plurality) who might be able to do stake-grinding rewrite attacks.

Paul MaudDib fucked around with this message at 20:53 on Jul 20, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
IMO now that you're looking at a ROI timeline of 6+ months there is no longer the massive demand to build new rigs, so prices are slowly drifting back to normal. However coin prices have remained high enough that we haven't seen an outright popping of the bubble either, people seem content to crank away at $2 per day or w/e.

(the Ethereum network hashrate has continued to increase at a pretty good clip so 6+ months ROI is effectively "never")

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
there is a super legit taqueria in town, their fish tacos with guac are to die for and their barbacoa tacos are pretty amazing too.

they need to put together a taco truck or something because i'm tired of the taco-truck with lovely white-person tacos

bet they don't take internet fun-money tho

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
I remember reading something once about the programming model of the Ethereum Virtual Machine being really non-intuitive to deal with especially since you can easily run out of gas.

random guess: it's like Java except you can get exceptions thrown out of random bullshit code you don't expect to throw, and most people don't handle the failure cases gracefully.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Splicer posted:

Wasn't there a miner or a wallet a while back where the "source code" didn't actually match the executable on the website?

Wouldn't be surprised, but it's also not necessarily malicious. Most builds aren't actually reproducible unless you're specifically focusing on that as a goal.

It's probably malicious though.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

1gnoirents posted:

And it was found out presumably by somebody checking the source code, which is the whole idea

You can't really do that (easily) with a non-reproducible build. That's the problem. Things like timestamps, the order you iterate through inodes on your FS, or the number of threads you use to compress resources can lead to different checksums on every build or for different people, even if the source is identical.

You have to specifically go after these deltas if you want a reproducible build. It's a worthy goal for security-sensitive software but it does take specific engineering effort to make it happen.

Some discussion on it regarding Debian's builds: https://lwn.net/Articles/630074/

Paul MaudDib fucked around with this message at 18:08 on Oct 2, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
Don't worry I'm sure the Ethereum founders will just roll it back. Again.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
There is no such thing as the cloud, only other people's computers. And in this case the computer belongs to some Russian guy whose last name sounds like some kind of drug.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
Smoke is beginning to pour out of Bitfinex (particularly due to their Tether scheme), if you are only in it for burritos rather than THE MOON you might want to consider cashing out at this juncture.

My passive mining paid for a good chunk of a new NAS so I'm pretty happy calling it quits. Of course who knows with Bitcoin, it'll probably double in the next month just because I cashed out.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Craptacular! posted:

Just curious what that means for the average schmoe with a GPU he isn't much in the winter?

The equivalent of a bank run without a FDIC. Bitfinex issued a lot of bearer bonds against BTC without the BTC to back them up, and the bill is coming due.

There is the potential for BTC to drop severely during one of these events - it may recover in the long term but if you intend to cash out quickly rather than hold forever, you may want to cash out quickly. Or it could be nothing - the market can be irrational for a long time (famously, longer than you can stay short).

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Computer Serf posted:

when I'm hearing random idiots on the street try to describe to their friends why bitscoin is a groundbreaking technological innovation that will rival global finance institutions and all you need to do is get some and sit on it... I wonder what it will take, to actually crash the bubble.

quote:

Joseph P. Kennedy, the multimillionaire father of Sen. Edward Kennedy and the late President John F. Kennedy, pulled out of the stock market shortly before it crashed. Why? Because his shoeshine boy was giving him stock tips.
https://www.washingtonpost.com/archive/business/1979/10/28/the-1929-stock-market-crash/70a06059-d5c7-4662-b9a6-38763fd69143/

Paul MaudDib fucked around with this message at 05:21 on Nov 22, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Knifegrab posted:

I am so loving confused now. I thought that my machine averages .5kWh meaning every hour it uses 500 Watts. so that is to say underload my machine is rated at 0.5 kWh.

Think of watts as being a rate of power consumption, while watt-hours are amount of power. Consuming 500 watts for one hour, or one watt for 500 hours, means you have consumed a total of 500 watt-hours of power (and a kWh is just 1000 watt-hours of course).

The other way you can measure power is by current (the voltage is implicit - often either grid voltage or the voltage of the battery). So you could give the rating of a battery in amp-hours - if you have a 3 amp-hour battery it could sustain a 3-amp current for one hour, or a 1-amp current for 3 hours. Power = voltage x current, so you can see that higher voltages let you deliver more power with a given amount of current (and current is what causes heating in connectors/wiring/etc).

Paul MaudDib fucked around with this message at 02:26 on Nov 23, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Risky Bisquick posted:

I'm saying base off max draw, and if that is unprofitable or not terribly so, you should take some time to evaluate which algorithms you check in nicehash. You can safely use average for solo coin mining however because you can control what power you are expending for your 'somewhat' determined reward.

Yeah, mining is not the best for hardware so hopefully you are not riding real close to the edge of profitability such that one coin pulling 10% more power than another is going to push you into the red. Otherwise you are just buying bitcoins in a roundabout fashion, while wasting electricity and burning up your hardware.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
Do people find that risers are actually reliable? I tried them once, I did all the usual stuff (turn PCIe speed down to 1.0x, enable 4G+ addressing, etc), and I was still getting maybe a 48-hour MTBF with 4 or 5 GPUs on a board. Honestly it seems like it makes more sense to go with fewer but faster GPUs and to not use risers.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Risky Bisquick posted:

Gaming while mining on the secondary GPU in the same box :monocle:

QuarkJets is a voice of reason, you can't regret not getting in years ago because it was just as much a bad idea then as it is now. Don't risk what you can't afford to lose.

Regretting not getting in early is the same as regretting not getting in early on penny stocks. Sure would have been great if you knew which one of them was going to spike 10000% this year, and which 99 of them were going to go to zero. Odds are good that even if you had picked it at $0.01, you would have sold at $25, or $100, or $2500, or $8000. Or been hacked, probably.

Paul MaudDib fucked around with this message at 20:21 on Nov 30, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

willroc7 posted:

What's the best way to store them offline?

The general idea of the paper wallet is sound, but there is no reason to prefer a real paper wallet over a properly sandboxed Linux VM instance that you spin up, generate a wallet, and then shut down and encrypt with a strong password/keyphrase/keyfile and back up.

Even then, there have been exploits like the Debian SSH keygen turning out to only generate 32768 different keys (a similar vulnerability occurred on some phone wallets). If something like that happens you are hosed regardless, there is no fix for accidentally generating predictable keys.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Axe-man posted:

Under birdbaths are considered the most secure!

Galactic Federation mole detected. I'm onto you, Phoenix Person.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

QuarkJets posted:

And even sound ideas in bitcoin get destroyed by the weird sort of cargo cult programming that is prevalent in the bitcoiner community, like asking random.org for random numbers but instead getting back the same error code every time

I mean but what if Intel had compromised /dev/random using RDRAND!? Let's use a trusted source of randomness, like an unencrypted HTTP connection to Some Random Guy's server! :jeb:

Lol what do you want to bet that the NSA is/could be the biggest holder of bitcoins right now

Paul MaudDib fucked around with this message at 06:23 on Nov 30, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

1gnoirents posted:

the safest way to store your bitcoins is as dollars in a real bank

no it's in NAS parts that you have purchased from newegg

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

1gnoirents posted:

hah

look at the bottom

made in china

suckersss

/begins desperately overturning possessions

playing the long con :negative:

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
What miner/algorithm/coin are you getting those numbers on? From my experience whattomine.com is pretty much dead-on with my power and hashrate, whereas things like the "nicehash profit calculator" can be like 30% or more above what I'm seeing. Even if some of you are talking about Canadian Funbux, the numbers are high enough that I'm wondering if they're fudging them a little :chloe:

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
Network power is bouncing around like crazy between VTC, Mona, Electroneum, and to a lesser extent SIB. I'm seeing like 50-70% changes in profitability in a matter of a few hours as the difficulty tries to track.

I think people are getting their miners back online though because the overall trend is down vs yesterday.

edit: graphs



Paul MaudDib fucked around with this message at 17:43 on Dec 7, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Fauxtool posted:

they kept my magic the gathering transactions safe, why cant they handle millions of dollars daily?

Pfft, like they could even handle that much responsibility. These weren't MTG cards traded online, this was an exchange for virtual Magic The Gathering Online cards.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Alzion posted:

While mining with a 1080ti I've noticed that Winminer is doing 55 MH/s mining Vertcoins. But, when I use my own CCMiner2.2r2 I pull in 67 MH/s. Does Winminer have a reduced intensity setting on their miners for default?

With this kind of behavior combined with Winminer's tendency to sporadically remove VTC as a mining option I don't see the point in continuing with it.

It may just have decided that it wasn't a profitable option at that point in time. Coin prices have been spiking around really badly, yesterday I saw Vertcoin as high as $6.50/day and as low as $1.50/day on a 1080.

I get the feeling some of these Captains of Industry are still trying to figure out how to use a regular miner and don't have enough hysteresis set up. Yesterday I saw a pool with 300 MH/s on a network with ~1200 MH/s, and the pool hashrate would periodically bounce up to 2500 MH/s for 10 minutes (i.e. that one person had 64% of the total network hashrate), presumably at that point their miner told them that coin wasn't profitable anymore, and they'd bounce to another coin.

A lot of the small coins don't really have much depth, and like one warehouse full of GPUs can run the profit into the ground.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Busy Bee posted:

How would living in Japan where the voltage is 100 volts instead of 120V in North American and 230V in places in Europe affect my mining?

pretty high, you should open an exchange and get some plebs to mine for you

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

DrDork posted:

If you're a small-fish miner, in most cases you should be appending /0 to your wallet address. This will reduce the difficulty, which has the end effect of increasing accepted shares and upping your payouts. Also don't mine to like 10 different pools at the same time, since it'll just mean you're not earning shares efficiently anywhere.

OCM is a fairly solid program in and of itself--basically any issues you'll have with payouts and profitability is going to be a problem with the pool.

From my understanding, the difficulty you specify (including with a "d=xyz" password) is just a hint and the stratum will adapt to the shares you actually push out.

If you have multiple machines going it would behoove you to run a stratum-proxy for things you are mining on a lot.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

divabot posted:

(The journalist has pinged me for tips on the story.)

(1) to kill the cyberdemon, shoot it until it dies

(2) don't invest in Bitcoin

(why are people so uptight about the term bcash?)

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Nitrousoxide posted:

This Crypto bubble has to be just doing gangbusters for AMD and Nvidia. I wonder what their plan is for the crash and if they are trying to constrain supply some to prevent difficulty curves from getting out of control and allowing ASIC miners to become profitable like they did for Bitcoin.

Pretty sure NVIDIA's plan is "release ampere" and they go back to raking in the bucks as everyone is forced to upgrade to keep their efficiency competitive.

Pascal is already way more efficient at mining than Polaris, Ampere is going to be beastly when the chips are down and profits are low.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

craig588 posted:

I don't know AMD so well, but Nvidia DGAF about mining and kind of doesn't like it. Miners are buying gaming segment cards, but instead of gamers they're only buying cards based on a nebulous profitability rather than every generation to play new games. They'd rather have consistent long term sales from gaming than miners using up all of the supply now. Miners buying cards now makes it harder for them to build mind share with new customers when the narrative changes from "These cards are good" to "These cards are good, but overpriced". When kids are at the 8-13 age they're not buying anything directly but brand affinities are getting built subconsciously that pay off in 5-10 years. That's the Nvidia and Nvidia partner stance, I don't know people at AMD and partners to get their comments.

NVIDIA makes enough cards that miners can't really buy them out the same way they can do AMD. It's a fringe benefit of their crazy market share.

Dunno, if they wanted to fix prices they could just ramp production a little more. Mostly they just need more 1070s, 1060s are a little on the high side but not terribly far out of whack, and the 1070 Ti is bang-on at MSRP.

I kinda think NVIDIA likes this new world where they can sell at/above MSRP for an entire generation at a time. That probably goes for AMD too, but they are clearly afraid of repeating their mistake with the 290 overproduction when the bubble popped on them. But again, the fix is just "release a newer and more efficient card", boom nobody cares about your old cards anymore. A big part of the problem comes back to AMD's habitual rebranding and lack of R&D spending.

AMD definitely gets the poo poo end of the stick though, they are just perpetually out of stock on all their products and Vega is a turd apart from mining (bearing in mind its huge die and insanely expensive VRAM subsystem) so they are extremely vulnerable to mining popping if prices collapse. At least they've been able to keep 580s in stock this time, even if prices are out of whack again.

Lisa Su said a "single-digit percentage" of their sales are from mining, I think there is some serious lying-with-statistics going on there. Maybe single-digit percentages of AMD's total revenue, or including OEM sales and poo poo. It makes up a huge percentage of their retail add-in-board sales and probably >80% of retail Vega sales (easily 99% at the moment, given current pricing).

Paul MaudDib fucked around with this message at 21:41 on Dec 20, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Craptacular! posted:

Is TSMC even going to make more 1070s because I figured the ti was going to be a defacto replacement. It's basically a repurposed 1080 that didn't quite pass QC, in the same way that 1060 3GBs were chips that had one CUDA core fail and were repurposed instead of being thrown away. The 1070 was basically it's own unique manufacturing, so I figured it was effectively retired in manufacture and the plant would focus on making 1060, 1080, and ti fabs.

Yields probably favor going with the 1070 Ti nowadays but to be honest die-harvesting is always a racket, literally everyone gimps functional chips in order to fill demand for downmarket SKUs for both CPUs and GPUs. Once you get past the initial launch phase yields are always great, but people want lots of the cheap chips, and you don't have nearly enough actually broken ones to fill demand.

If they want to retire the 1070 then fine but then don't turn around and bitch about losing mindshare after you've left a giant void in the extremely popular midrange of the market. Or, drop the price of the 1070 to fill that space. Right now there is just nothing between $260 and $450.

(also, the way it works is that 1060s are GP106 and 1070/Ti/1080 are all on GP104, it's the same die, you don't try to make 1070s it's just what is left over when you try to make a 1080 and it is partially broken. 1060 die-harvests become 1060 3 GBs.)

Paul MaudDib fucked around with this message at 21:56 on Dec 20, 2017

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Nitrousoxide posted:

Aren't 1060 and 570's the better mining cards at the moment though? You loose ~20% performance but cut your capital costs by almost 50%. They have a payback of 2-3 months instead of 6 months. This reduces your exposure to the risk of the bubble popping and you can just buy more 1060 or 570 cards to get the same hashing capability at a lower initial expense but slightly higher physical space requirement and a tiny increase decrease in margin due to higher electricity costs

The rest of the system costs money too, though. At current prices yeah, the 1060 and 570 make more sense, but if you could get a 1070 for $375 or so there would be a reasonable argument for it.

Frankly the 1080 and 1070 Ti are not a bad deal for miners either, just not for ethereum. They're roughly twice as fast as a 1060 at Neoscrypt and Equihash, and basically cost twice as much, so performance-per-dollar is the same and you are getting more bang out of the rest of your system. And if the market dumps they will be better for gaming, and probably less susceptible to market flooding.

It's basically just the 1070 that is drastically out of whack.

Paul MaudDib fucked around with this message at 22:26 on Dec 20, 2017

Adbot
ADBOT LOVES YOU

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

QuarkJets posted:

I don't know about AMD but NVidia's bread and butter these days is high performance computing, as in datacenters full of Tesla cards. This cryptocurrency mining poo poo is barely even a blip.

You can buy a 2U system that costs $100k and is basically just graphics cards, and datacenters are being built that are just full of these kinds of systems.

Everyone talks about datacenter as if it's the best thing since sliced bread, because it's a growth market and the margins are fantastic, but NVIDIA's bread and butter is absolutely the consumer graphics market by a country mile. That's where they make the vast majority of their revenue and it's still growing at a pretty decent clip all by itself.

  • Locked thread