Topic: ASIC Testing on Scrypt? - page 5. (Read 17504 times)

Well yes. I keep noticing this.
If I had sucessfully created an ASIC for Scrypt I wouldn't share it either, so.. 

I just looked up the pricing on EK waterblocks, as I don't do water cooling and I hear they're good. Looks like I was wrong, they retail for about a third of the price of a 7950, therefore ordering them in massive quantities would get a significantly better price. You're still probably looking at increasing your price per GPU by 10%-15%, though, which is why I would probably go with air cooling anyway.

Also, WindMaster, that is pretty fucking awesome. I'll soon be running 5 GPUs out of my bedroom, lol. You're WAY outta my league.

Lol,  make sure to charge top dollar for it. LTC mining could be bigger than BTC!

really interesting ..

Well, had I thought they would pay retail price, I would have said 50%. I just guessed they'd get a pretty hefty discount for ordering in bulk.

It would have been cooler if it was. 

its a fact that its been done - people are stupid , we have learned this by observing how docile people are .

most of the population are damaged goods, unfortunately. 

There are professional ASIC fab companies known to be exploring Litecoin ASICS. No reason why they couldn't have built one by now and are keeping quiet.

30% of $750,000 is $225,000. Even if you have that kind of cash to drop on cooling, it's stupid to do because it's a fuck of a lot cheaper to get good AC.

 

I saw that farm and I love seeing all those rigs- but still, that's 250Mh/s or so of Scrypt.

Wind - One more note, I read through your history as you piqued my interest, and I think I better understand as to what you are doing by putting the pieces together. You've got a perfect mix of skills to set everything up along with the fact that the infrastructure is really there, at least in part back in May, to support the 3D Film rendering server leasing that you did. I am imagining that you run a lot more GPUs per motherboard than I was thinking, and this is an edge. Not sure how all the infrastructure costs you have wouldn't kill your ROI, but I'm going to assume with the multiple uses of the infrastructure, and tax breaks, this is factored in.

Certainly, a limited number of people have the skills, capital and cheap power, but I do understand your point that certainly the infrastructure exists and while out of the range of a "normal" or even an "advanced" mining operation, there are professional setups that can easily achieve this goal.

It's obvious you have the capability to have one of the larger Scrypt mining operations out there, however, I don't believe you're the 2 Gh/s mystery, although a quite respectable operation.

When this all is over, I'd love to hear more about your setup. We have some time left to mine though, so maybe next year.

Pay attention to the part in bold.

Just saying...

Cool, thanks for posting. You must have ramped up a bit since you were at 18,000 kh/s in May. I do wonder why you're thinking you have some sort of trade secret though, except the fact that you're running more GPUs per rig with custom engineering and using custom frames with forced cooling.

The cooling is probably a very "cool" design, but the number of GPU's per rig honestly isn't worth the cost delta to me. Those are the cheap parts of my rigs - Memory, MB and boot drives cost me $140 per rig or less, so stuffing double the cards in the rig saves $140, minus whatever the parts you're using to rig it, which I assume would be minor in cost. You may save minor power, but you pay .065, and I pay .048, which doesn't really impact too much.

I think we got off topic though, your setup is not running anywhere near the 2 g/hs we're discussing, and while your initial message that stated it would be simple to create, just because we all know in theory and can multiply our own setups out, this is still a very odd thing to pop up as it did. Regardless, I'm sure we'll see this a lot more in the near future. I admire your eluded to setup, and you were correct, I wasn't clear on what 1 MW would look like going in. It's okay to admit that anything over 100,000 kh/s using GPUs does take a lot of unique problem solving to implement successfully in order to fit in leased spaces that are readily available with cheap power and is not common. Which brings us back once again to the original point.

Water blocks for each GPU would easily raise the cost per GPU by 25%-30%. Sure, if you have unlimited funds, you can do it, but other than that, it's just stupid.


Hashrate that bounces all over the place is a pretty good indicator of a botnet. You can't go by the number of connections, obviously, but one thing they can't really hide well is bots coming online and going offline.

Windmaster,you made my day ..f*cking awesome

So long story short ..Windmaster YOU are the mystery 2600 gig hash ??


Also what are your GPS co'ords and what are your views on Al-Cadia

I envisioned a huge warehouse piled high with stacks of mobo's and GPU's with arcs of electricity flying around. 

jk but your setup looks pretty sweet.

I agree that it'll be a fun time once GPU mining becomes uneconomic, and everyone starts posting photos.  For that matter, it would be interesting to get together a group of large mining farm builders over a beer to "shoot the shit" after this is all over.  There's a lot of interesting engineering and tricks of the trade that people have developed for large mining farms.



As best I can tell, most of the people here have just never seen what a ~1MW service entrance looks like, or are imagining something massive and expensive.  So here, I ran outside with my camera just before dark in an attempt to educate people on what a service entrance of this size actually looks like.  I'm not going to post any mining rig photos for the reasons I noted a few posts back.  However, I'll post service entrance photos since there isn't exactly anything in these photos that will help anyone else build a GPU farm.  Our service entrance in the below photos has a maximum capacity of 900kW (2500A x 208VAC x 1.73):

The transformers out on the pole.  I bet no one would take a second look or think anything was at all unusual if they saw this next to any warehouse or office building:







Electrical distribution room:



The 4th panelboard contains 3 2500A fuses and the disconnect.  I can't take a photo of the disconnect and fuses, as this panelboard is sealed by the power company:



The other 3 panelboards contain an assortment of 100A, 200A, 400A, 600A and 1000A breakers that feed the other subpanels throughout the building:







Here's a 400A disconnect ahead of one of the subpanels that fanout power to each row of racks:



Note: I Photoshopped all of these (well, GIMP'd them).  I resized them down to a reasonable size, cropped them to the area of interest, and in the case of the power pole photo, removed the PUC tag # for the pole and the serial # on the transformer (I value my privacy, and any lineman in half of the western US could look up the exact location of that power pole otherwise).


And no, I will not post rig photos until after GPU mining of all cryptos becomes uneconomic and the contents of the photos will not aid anyone else in building their GPU farm.  However, it's not going to take long for someone to calculate an upper bound of how high our hash rate could be for the amount of power we have available.  It's between 0kH/sec and that upper bound, that's all you'll get for now.  

That's certainly not in "unlimited budget" territory.  I trust that the above photos look significantly different than what everyone else was envisioning when talking about a service entrance in this size range?  Were you guys envisioning a substation and switchyard or something?