Can it run Crysis?
The Register doesn't really cover motherboards, but when one comes along with 19 PCIe slots – yup, nineteen, ten plus nine, so it can run 19 GPUs – it's hard not to want to share the news. ASUS this week teased the new “B250 Mining Expert” which boasts all those slots because – as the name implies – its role in life is mining …
You could fill them with NVMe PCIe cards since they appear to be cheaper than some of the 2.5" NVMe drives with similar capacity.
Still think custom case and all.
And you can still have a couple GPUs. Not sure why they didn't go with the 1080Ti unless these aren't full sized slots. (card length)
One thing for sure... you're going to have one heck of an electric bill and you wouldn't have to use your heat during the winter.
I haven't followed the mining stuff, but I thought you would need 10-15 of these in a cluster in order to get a bit coin out of all of those hashes.
Yes, you can. Unfortunately, all the solutions in the particular space I've come across involve using a single PCIe lane per GPU card, so if you are thinking about stuffing other card types in, you have a limitation there. Personally, I'd go with the dedicated compute GPU boards, but that's just me being me.
The only reason I know anything on the subject is that several tech and hardware sites out there have classified me as a miner. Whatevah. I just know a few of them.
Do they really have that many problems of interest? These GPUs aren't going to be working effectively unless the problem is embarrassingly parallel, in which case there is a set of mathematical base vectors in which the problem is linear, in which case the qualitative features of the answer can be cranked out analytically. I won't go further and say that the computer merely draws the pictures, because the quantitative details will be interesting for specific problems, but it's damn close. You might say that this hardware is interesting to engineers, but much less so for pure scientists.
A mobo with 19 CPU slots might be more interesting, though. You could run non-trivial programs on that, solving non-linear problems, as found in Real Life.
Machine learning of almost any kind is sufficiently parallelisable to exploit such a monster. Deep learning neural networks are very popular these days, and they need lots of processor power. It is already running on graphics processors for that reason. DNA analysis too.
Which DNA analysis problems? I ask because I support scientists doing this stuff (and bioinformatics is a bit of a hobby*) and almost everything bioinformaticians do is running on x86. We do use GPUs but for other tasks.
*I think this is quite geeky even by the standards of the commentards!
I'd assume genome wide association stuff and generally areas where you're using bootstrapping type methods which require you to repeat your calculation over permutations. Many of those will run on x86 because turning them into GPU algorithms does take extra work and is almost a research project in itself (one common thing you'd have seen a few years back is comparisons of a slower but existing or formally 'correct' method on CPU versus a GPU version which would make some assumptions to achieve the required parallelism). 3D non-linear image registration is another good case, you can't parallelise the entire problem, but parts can be.
I have to post this anon for a couple of the obvious reasons...
You have multiple problem sets. Deep Learning is one where you need a lot of cores in parallel. A friend over at Nvidia said that the smallest board he'd recommend for some basic stuff is the 1080Ti.
Now imagine 10 of them working together.
But what happens when you have a large data set and you need to work on multiple data points within the set in parallel and separately? (You have a cluster of these machines.)
Or what happens if you have a problem that can use the GPU to accelerate the solution, but need to host multiple tenants working on different problems simultaneously ? So now you can.
Will this run minecraft..??
Is this a craftyminer..yes!
and next on this shopping list: a suitable UPS, capable generator, solar panels, frost free fridge freezer, internal fans, ducting, cuddly toy, hi-res monitor, fibre router, pint glasses, various beers, wines and spirits and an executive swivel seat.
I surely can't be the only one that thinks that the cryptocurrency fad is a colossal waste of fossil fuel... I mean it's marginal demand that otherwise wouldn't have been there. And the marginal generators of choice are almost always coal or gas! Surely a more important stat would be how many fps you get in counterstrike or quake 3?
So the Cryptocurrencies replace central banks and governments with anonymous rich people to control increase of money supply.
How is this a good idea?
Also speculators "hoarding" cryptocurrency causes same problem as cash under the mattress or in Offshore accounts to avoid tax. It reduces the amount of money in circulation available to buy & sell stuff.
Very simplistic explanations. Very introductory texts on Economics explain it better.
"So the Cryptocurrencies replace central banks and governments with anonymous rich people to control increase of money supply."
Not anytime soon. The market cap of all crypto is tiny compared to any of the major traded world currencies. Even with massive expansion we're not seeing the end of central bank fiat any time soon.
Introductory macro economics is more about why personal/business accounting and government accounting are utterly different beasts, and that when the right magic words are spoken a Ponzi scheme is not, debt is money, and 100 gold coins from the queen of thorns is less than you might expect. You'll notice that in many textbooks even just a few years old that negative interest rates are not possible, because *reasons* and Keynes said so. After ~10 years of effective negative rates and ~4 years of actual negative rates, they just leave that section out now.
Explaining the Fed (and other central banks) ability to be their own biggest creditor, whilst continuing to issue more money is always fun. I'll always tell my students that if you expect any aspect of the economy to be fair or just, you picked the wrong subject.
Mobo is nice if you like going the whole rig route. Personally I find it easier to just chuck a card or two in an old/budget box and leave it somewhere I'd like some regular heat.
I still cannot understand why anyone with half a brain would ever buy a P106. The 470 4Gb mining version maybe, but you can compare a P106 directly to a 1060. They cost roughly the same, although you can usually find a 1060 cheaper, the 1060 has full consumer guarantee (2-3 years) versus 4 months, 1060 is likely to have better resale value due to is greater utility. Worst of all they aren't any better at mining than a 1060 made by themselves.
A bog standard 1060 6Gb made by Asus* will do about 19 MHs out of the box. Using Asus GPU tweek you increase the memory speed and fiddle the power target based on what numbers make you happier. I have always got ~22 MHs, with most having a sweet spot about 21.5 MHs for about 80 watts on the card and 110 watts at the wall with a silver rated PSU. 22 Mhs is usually closer to 100 watts.
19*21.5 = 408.5
So 19 P106 in an Asus built enclosure hash at the same rate as 19 1060 in my shiteboxen space heaters, those P106 would cost me more, and I strongly suspect that that funky mobo plus big ass PSUs are going to end up costing more than my shiteboxen.
I think they should have started making riser cards, since they are easily the shittiest component in most mining rigs and should have a pretty high margin. They tend to make me hulk level mad, hence why shiteboxen rather than milk crate rig or other such innovations.
* there's some chip binning and cooling differences between the cards, mainly not significant for mining
"Power draw. You are asking the wrong questions. For the extreme miners it is all about power draw. I dont know the specifics but possibly form factor too if it allows cheaper or better cooling."
It doesn't draw less power. In fact, it draws almost exactly the same power as a 1060 when running at ~22Mhs. It can hash slightly faster, but at the cost of more power. I've seen 24Mhs @ 120w per card, 150w at the wall. If your power cost is over US$0.17/KWh, then it's a money loser to run it at higher power.
The P160 has better fans than the entry level 1060. Same heatpipes. Same form factor. If you want a lower or shorter form factor, then there are a range of 1060s for that, my preference is for Zotac. Fans run about 35% for most boxen here, less for the ones in unheated spaces.
The issue remains that it gives a very marginal gain on max speed, no gain on Mhs/watt, and has no real resale value and no warranty. I've had mining cards that ran for 4 years that I could still sell for 30% of their retail price* plus even fairly ancient cards that came with the shiteboxen can still fetch a fiver for someones minecraft box.
If my mining card needs to RMAd as long as I've:
- run it within it's performance envelope
- not buggered with the BIOS settings (encrypted on 1060 and presumably P106)
- run it on compatible and tested hardware (no bloody risers)
- run it in an ESD safe enclosure (case rather than rig)
- used only the manufacturer supplied OC SW
- run it under a "tested" OS
Then my suppliers will replace it anytime within 3 years. Why I'd want to give that up for 4 months warranty for a small potential performance gain is beyond me.
The 1060 was already an amazingly efficient mining card. The 470 has more grunt (and draw) for mining, is in more need of cooling (+5-10 C over a 1060), thus would probably benefit more from a specialized mining only card. But even then, a new 570 4Gb is about 250 euro retail, so any mining card would have to come in under 200 for me to even think about it.
It feels like the responses to the demand for GFX cards are quite odd. Most companies make 3-4 versions of the same chipset and memory combo, with better binning, cooling and more power requirement on the higher end models. Adding another model of fully functional GFX card, with low binned processor, high binned memory and the best cooling, and then pricing it accordingly would seem more logical than making a new range without any cost savings. Or just make more, in response to demand.
Instead it's been used as a price hike. Shortage of supply, price rise, supply resumes, prices kept high.
* x is about 90 cents at the moment. Between 50 cents and 5 euro over the last 6 months
** as an example RX270 2Gb for 50 euro
"Why would anyone buy a traditional CPU/GPU system to mine cryptos?"
Because certain of the crypto currancy algos are designed to be ASIC resistant. You can usually guess which algo it is because of the quite different speeds. And the fact that a GPU is being considered at all implies it's ASIC resistant. BTC and LTC are both only viable to mine on ASICs, ETH and XRP are only viable on GPUs and CPUs.
It is not at all helped by the fact that people refer to it as "mining bitcoin" when in fact they are mining something else and getting paid in bitcoin. Or that many people are still throwing their hands in the air about the concept that currency unbacked by an army and a tax collector cannot exist.
I mine on normal kit because the traditional aspect enables me to more easily resell the used parts, and get business licences etc. Saying to the bank/council that I build specialized cryptographic workstations that I then remotely rent out (for legal purposes), and these boxen look suitably like a normal computers, fall under fire safety et al using an existing enclosure and thus are easy to classify. Having a bunch of open rigs would make at the very least my fire/electircal safety come into question.
Oh, and expect a few visits from the cops too. Large power draw + fan noise + heat + business income from BTC = probable cause. They are quite polite about it, and it's not unusual in my hood for such things.
"This Asus motherboard is for researching stuff as it is general purpose."
My rather long winded posts above are that the *specialist* mining cards it proposes using are no better than the general purpose ones.
> The result? 407 megahashes per second, if the planets align
That math doesn't look right. The antminer s9 is allegedly good for 14THs (call it 35x faster for somewhere around US $2500). If we're comparing apples and apples then you are going to want a pretty special price or at least another zero in the hashes per second stakes.
"People spending $10k on a PC to mine fictional cryptocash"
I assure you crypto is real.
Real enough that the tax department* would like it listed among your assets.
"Or I'm in the wrong line of business."
It's very easy to get in to. I'd assume that about 90% of the reg readers should be able to knock up a shitebox miner from parts laying around plus a GFX card. If you have cheap power, or want a heater in a particular space, it's a no brainer. It's a server with GFX card in it. Set up right, have some sort of monitoring and remote access, leave it alone.
Make sure you pay applicable taxes, and that it's OK for you to run the boxen in the location of your choice.
*the belastingdienst does anyway
knock up a shitebox miner from parts laying around plus a GFX card. If you have cheap power, or want a heater in a particular space, it's a no brainer. It's a server with GFX card in it. Set up right, have some sort of monitoring and remote access, leave it alone.
Not for Bitcoin. You might make a few pennies per year with a GFX.
You would need dedicated hashing ASICs for Bitcoin since, a few years now.
"Not for Bitcoin. You might make a few pennies per year with a GFX."
When replying to a thread that is not of excessive length, it is often helpful to have read the previous comments. Therein you may discover that the only people suggesting mining bitcoin on this, or my shiteboxen, are people who also wish to point out that it was be very poor at this. That is a strawman argument, and beneath you.
Perhaps a reply assuming I'm mining ETH would make more sense. For the record I dual mine ETH and Decred on AMD, ETH and Sia on nVidia.
There are plenty of mining calculators around. A 1060 single card shitebox will be about 350-400 euro, will do 21 Mhs @ 120 watts assuming terrible parts. Plug in your power cost and away you go.
If you think my shiteboxen are bad at mining BTC, you should see them mining for coal. Disgraceful.... :)
People spending $10k on a PC to mine fictional cryptocash, for profit! Something's seriously wrong with reality. Or I'm in the wrong line of business. Or both.
Because of the late surge in value, I was able to take out some of my BTC and take a 2 week family holiday in Florida with money left over. The remaining half is creeping up again and looks like it will double since.
It was all acquired a about 3 years ago when BTC was worth next to nothing. So fictional maybe, maybe not, but rewarding, definitely.
"So fictional maybe, maybe not, but rewarding, definitely"
Every penny anyone makes from crypto currency creation and speculation is someone else's loss or will be.
It is rewarding for some like all pyramid/ponzi schemes, but, overall it is just a waste of silicon and electricity.
There is a difference between the physical connectors and the number of PCIe lanes a component uses when plugged into any PCIe slot (though an x1 slot can only use 1 lane - and x16 slot can use up to 16 lanes)
Nice pic of connectors and max bandwidth here https://i.stack.imgur.com/AMn2Z.gif
"There is a difference between the physical connectors and the number of PCIe lanes a component uses when plugged into any PCIe slot (though an x1 slot can only use 1 lane - and x16 slot can use up to 16 lanes)"
This is touted as a crypto mining machine.
It has 19* PCIe x1 slots.
How many mining cards come with a PCIe x1 slot?
The answer might be 'lots', but the example card they listed is a x16 board, so you can only plug one in.
It was estimated that 5% of the worlds power goes towards this, that was in 2014. It really saddens me that it's being wasted like this, especially that the most common behaviour is hoarding, or selling the coins, the market is likely to be dominated (own research, but I'm fit to research the concept) by online market (of the "secret" variety) places, not places actually accepting bitcoin... where actual currency is also accepted.
There are also so many coins now. The advantage is also momentary because of difficulty adjustments, unless one pool dominates which is bad.... urgh.
On the + side, if more than half of the network is compromised is getting harder to do, but as bitcoins are not a practical currency, but more of a hobby for "traders" and such... it's sad.
[There might be a handful of people who trade with success, maybe it's their primary income! But the majority of trading on exchanges is from those dodgy markets using the coin and hobbyists having a go - I'm not saying there are none, for those who'll choose to pick on that point]
This post has been deleted by its author
Biting the hand that feeds IT © 1998–2020