Amph
Legendary
Offline
Activity: 3248
Merit: 1070
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
June 30, 2016, 11:50:53 AM |
|
this assumes that you know the gpu chip structure... not sure nvidia or amd want to share that with anybody ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) unless it's themselves doing that. yeah he is putting it like they can simply look at the circuit and replicate the chip and everything perfectly, also designing a chip it's not free... it does not work like this, all i know is that is very easy for a big farm to add other gpu in their already big farm a 1000 gpu farm can easily add 100 gpu each week, which is 12GH a month, network is onyl 3.85 tera currently
|
|
|
|
thevictimofuktyranny
Legendary
Offline
Activity: 1092
Merit: 1004
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
June 30, 2016, 12:02:17 PM |
|
this assumes that you know the gpu chip structure... not sure nvidia or amd want to share that with anybody ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) unless it's themselves doing that. yeah he is putting it like they can simply look at the circuit and replicate the chip and everything perfectly, also designing a chip it's not free... it does not work like this, all i know is that is very easy for a big farm to add other gpu in their already big farm a 1000 gpu farm can easily add 100 gpu each week, which is 12GH a month, network is onyl 3.85 tera currently Well, the engineers who produced the Dash X11 ASICs have published a lot on how they did it? The engineers trying to break the Dagger Algorithm aren't going to be working any differently? Are they?
|
|
|
|
Amph
Legendary
Offline
Activity: 3248
Merit: 1070
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
June 30, 2016, 12:18:02 PM |
|
i'm not saying that it's not possible, but there is too much risk for them for something that is supposed to go pos in few months
they need to invest upfront a very big portion of their money, it is more easy for them to simply buy gpu at this point...
|
|
|
|
thevictimofuktyranny
Legendary
Offline
Activity: 1092
Merit: 1004
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
June 30, 2016, 12:38:35 PM Last edit: June 30, 2016, 12:49:10 PM by thevictimofuktyranny |
|
i'm not saying that it's not possible, but there is too much risk for them for something that is supposed to go pos in few months
they need to invest upfront a very big portion of their money, it is more easy for them to simply buy gpu at this point...
Specs for Ethereum, found this from an obscure citation on the internet, so I cannot say for certain it is correct (SO READ WITH CAUTION). 60 million Ether - Crowdfunding sale prior to August 2015. 12 million Ether - Allocated to long-term development fund. 18 million Ether - allocated to POW mining. Total supply is: 90 million Ether, before POS and POS will be targeted towards zero inflation. ---------------------------------------------------------------- Currently supply is 81.5 million. That means another 8.5 million Ether still to be mined at POW. This means POW will end in 9 months time, plenty of reasons for engineers to be working ASIC Hardware. At current prices: $106 million worth of Ether is still available on POW!
|
|
|
|
Amph
Legendary
Offline
Activity: 3248
Merit: 1070
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
June 30, 2016, 12:51:48 PM |
|
yes but the current network is filled with something like 120K gpu, assuming 30MH for each gpu
we can assume that each miner in the world even newbie, can easily add 1/3, i mean a guy with one rig can without problem add 2 gpu to his mining setup every month(actually i think he can add much more the increase is not linear is somehow exponential...)
so the whole network can grow by 1/3 each month, which is 12k gpu, so 360 GH, and this by counting the bar minimum that anyone can add to his farm
|
|
|
|
thevictimofuktyranny
Legendary
Offline
Activity: 1092
Merit: 1004
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
June 30, 2016, 01:22:34 PM |
|
yes but the current network is filled with something like 120K gpu, assuming 30MH for each gpu
we can assume that each miner in the world even newbie, can easily add 1/3, i mean a guy with one rig can without problem add 2 gpu to his mining setup every month(actually i think he can add much more the increase is not linear is somehow exponential...)
so the whole network can grow by 1/3 each month, which is 12k gpu, so 360 GH, and this by counting the bar minimum that anyone can add to his farm
A GPU crypto coin only needs to produce $5 million worth Alt Coins per annum, before engineers start making prototype ASIC hardware for it's algorithim. Ethereum is well over that threshold.
|
|
|
|
pallas
Legendary
Offline
Activity: 2716
Merit: 1094
Black Belt Developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
June 30, 2016, 01:24:39 PM |
|
yes but the current network is filled with something like 120K gpu, assuming 30MH for each gpu
we can assume that each miner in the world even newbie, can easily add 1/3, i mean a guy with one rig can without problem add 2 gpu to his mining setup every month(actually i think he can add much more the increase is not linear is somehow exponential...)
so the whole network can grow by 1/3 each month, which is 12k gpu, so 360 GH, and this by counting the bar minimum that anyone can add to his farm
A GPU crypto coin only needs to produce $5 million worth Alt Coins, before engineers start making prototype ASIC hardware for it's algorithim. Ethereum is well over that threshold. Where did you get that information? How is it calculated? EDIT: BTW it's offtopic
|
|
|
|
thevictimofuktyranny
Legendary
Offline
Activity: 1092
Merit: 1004
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
June 30, 2016, 01:37:13 PM |
|
yes but the current network is filled with something like 120K gpu, assuming 30MH for each gpu
we can assume that each miner in the world even newbie, can easily add 1/3, i mean a guy with one rig can without problem add 2 gpu to his mining setup every month(actually i think he can add much more the increase is not linear is somehow exponential...)
so the whole network can grow by 1/3 each month, which is 12k gpu, so 360 GH, and this by counting the bar minimum that anyone can add to his farm
A GPU crypto coin only needs to produce $5 million worth Alt Coins, before engineers start making prototype ASIC hardware for it's algorithim. Ethereum is well over that threshold. Where did you get that information? How is it calculated? EDIT: BTW it's offtopic Look at Dash coin POW stats for 2015
|
|
|
|
bathrobehero
Legendary
Offline
Activity: 2002
Merit: 1051
ICO? Not even once.
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
June 30, 2016, 11:54:36 PM |
|
Considering Eth's market cap and profitability (eg. total daily mined coins in BTC) I'm preeeeetty sure someone figured out how to solve the TLB trashing issue and have a crazy high efficiency we could only ever dream of.
In fact I'd go ahead and guess that there might even be private ASICs simply because they would reach ROI way before PoS (which will very likely be late anyway).
People with deep pockets will jump the gun and spend crazy amounts of money when it comes to such high marketcap coins. And they can also manipulate the market to their will.
I mean Eth is the second biggest coin ever and look at how much money was sank into developing ASICs for BTC.
|
Not your keys, not your coins!
|
|
|
scryptr
Legendary
Offline
Activity: 1797
Merit: 1028
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 12:17:03 AM |
|
Considering Eth's market cap and profitability (eg. total daily mined coins in BTC) I'm preeeeetty sure someone figured out how to solve the TLB trashing issue and have a crazy high efficiency we could only ever dream of.
In fact I'd go ahead and guess that there might even be private ASICs simply because they would reach ROI way before PoS (which will very likely be late anyway).
People with deep pockets will jump the gun and spend crazy amounts of money when it comes to such high marketcap coins. And they can also manipulate the market to their will.
I mean Eth is the second biggest coin ever and look at how much money was sank into developing ASICs for BTC.
, JUST SUPPOSING-- There was a basic design with 3GB shared common memory, a chip to communicate with the ETH pool and load the DAG on the memory, and 16-64 ASIC chips to hash the DAG independantly, send their hash through the COMM chip, and do the work of a GPU, but specialized, stripped down for only hashing the DAG. Think it would fit in a shoe box? Run on 25% of the power needed for 16 GPUS? Deploy prior to the POS phase? --scryptr
|
|
|
|
bensam1231
Legendary
Offline
Activity: 1750
Merit: 1024
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 01:02:31 AM |
|
pure speculations :-) gtx 1070 on the way to me... let's see if I can fix neoscrypt on it.
And a dark knight appears. There is a guy in the Eth miner thread and also post here stating that the bug with NeoS was due to a WDDM issue, which will be fixed in this coming month. He has details on it as well. Good news everyone - the driver for both Win 78/Win 10 insider preview should be available next week - on this driver all Pascal GPUs are able to do the same as on Linux on ether in Win 10 Insider preview without any change But with Win 7/8 and other algos for win 10 insider preview devs need to take action If any devs are interested in details PM me - Genoil is informed, it would be nice to fix neoscrypt - if anyone is up to the task - welcome
He explained it differently in Genoil's thread. Something about block size. With difficulty of Ethereum jumping by 35% in 3/4 weeks ![Shocked](https://bitcointalk.org/Smileys/default/shocked.gif) Do you reckon ASIC mining hardware is shipping amongst insiders? From Dash coin, engineers where making small batches of ASIC X11 miners, long before commercial products became available to the community in April this year! No... Dagger is designed in a way that makes it extremely ASIC resistant (in addition to being better 'gpu' resistant). This is just megafarms scaling up in addition to all the Eth babies. It's more like someone ( i.e. Genesis etc.) bought batch of 1070s or 1080s or 480s... and paid someone to make /adapt miner... Trend starts before all 3 cards began shipping. 12Ths increase in difficulty every 4 weeks, looks more like a small batch private ASIC hardware is shipping out each month. I just mentioned it: people might be rushing out to buy more GPUs for Dagger, when ROI is looking unlikely! chinese can add gpu like you add cheerios to your cup of milk 30 gpu = 1 tera, so that s 360 gpu, which is nothing for some folks i doubt there are asic for this, too hard and too costly to do They may be costly to produce, but they are out there! When Ethereum crashed below 0.014BTC ![Shocked](https://bitcointalk.org/Smileys/default/shocked.gif) 10THS of difficulty disappeared ![Shocked](https://bitcointalk.org/Smileys/default/shocked.gif) There was no increase in hashpower on Lyra2v2 or neoscrypt Alts ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) Locked hashpower on Dagger algorithm is summarised as: ASIC Hardware ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) Big investors have dumped all their positions in EXPANSE, because they know there are ASICs out there. just lag i think , it would be stupid to spend million for an asic chip that can mine ethereum at current rate which is not so profitable like in the good old days pure speculations :-) gtx 1070 on the way to me... let's see if I can fix neoscrypt on it.
good to hear, i hope you can fix it and increase to 1200-1300 khs ASICs can hash exponentially faster then GPUs... BTC for instance. A 1070 would hash at a fraction of a percent of a current ASIC. Depending on how fast they can speed up their ASICs, they could potentially outrun the difficulty increase, since it's not a reduction in block reward... Although I don't know how long they could do that for. OK, I can tell you don't understand the R&D process. Essentially, engineers look at GPU, with best performance and put together a cheaper (ASIC) prototype, but these have to be produced in batches of 500-1,000. These prototype batches, strip the extra GPU functions not necessary to hashing the algorithm. The engineers keep repeating this process with each batch of prototypes. This process is done in secret and is self-financing so long as news never gets out that this is being done, because the prototype batches produce Alt Coins, which can be sold for BTC and converted into cash to pay for the next generation of prototypes. What the engineers are hoping for is some kind of breakthrough, in this process, which produces chip design that makes vastly more hash output per a unit of electricity. On Dash, X11 prototype batches where being produced 6 months, before the engineers made the breakthrough. The situation on Ethereum is slightly different, the prototype batches have become highly lucrative with Ethereum prices about 0.014BTC to 0.028BTC. So, those engineers who are working on Dagger algorithm are simply chucking out bigger and bigger batches of the prototypes Not sure how any of what I said would constitute that I don't know how ASICs are funded or made. That aside, I believe you're thinking of FPGAs... ASICs are generally built from the ground up, which is why they're financially much more intensive then FPGAs and also take much longer to construct and bring to market (or to mine with, whatever you choose), but also yield much better results. Also once again aside, there are plenty of ASIC makers that were crowd funded off of BCT, check the forums. In other news, DO NOT buy the Zotac 1XXX models. The amp edition cooler is garbage. The shroud and backplate wraps the card so tightly there is no room for airflow compared to other manufacturers cards, which is really quite funny because the coolers and cards themselves feel quite solid. I'm currently in the process of removing them.
|
I buy private Nvidia miners. Send information and/or inquiries to my PM box.
|
|
|
pallas
Legendary
Offline
Activity: 2716
Merit: 1094
Black Belt Developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 07:36:00 AM |
|
so amp edition is thrash... I ordered an amp extreme which should be better, let's hope it doesn't have the same flaws :-)
|
|
|
|
thevictimofuktyranny
Legendary
Offline
Activity: 1092
Merit: 1004
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 11:26:24 AM |
|
Quite easy to explain, really from the silicon wafer conundrum? When you are designing a new chip design, you cannot go to Fab and just ask for 1 ASIC chip. Instead, you get an entire silicon wafer worth, 1000's of that ASIC chip design. So, that is why you get Prototype Batches. I'll skip, the entire discussion about yield per silicon wafer, which is another reason for continuing with Prototype Batches. Example: In November 2015, Innosilicon announced it was beginning development on the A4 Dominator Scrypt miner. Basically the A2 Doniminator, but twice as many resistors by going from 28nm to 14nm. Basic die shrink improvement is 2X For 7 months, they have been ordering prototype batches ASIC of chips and then sticking them on Litecoin network and mining LTC. There engineers eked out another 0.6X performance and decided to announce a new A4 Doniminator would be coming to market in Sept 2016 2.6X better than A2. Before, the announcement a huge amount Litecoins where dumped on every major exchange - knocking $1.25 of each Litecoin (25% drop in price). That's way, engineers operate
|
|
|
|
bensam1231
Legendary
Offline
Activity: 1750
Merit: 1024
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 12:22:27 PM |
|
Quite easy to explain, really from the silicon wafer conundrum? When you are designing a new chip design, you cannot go to Fab and just ask for 1 ASIC chip. Instead, you get an entire silicon wafer worth, 1000's of that ASIC chip design. So, that is why you get Prototype Batches. I'll skip, the entire discussion about yield per silicon wafer, which is another reason for continuing with Prototype Batches. Example: In November 2015, Innosilicon announced it was beginning development on the A4 Dominator Scrypt miner. Basically the A2 Doniminator, but twice as many resistors by going from 28nm to 14nm. Basic die shrink improvement is 2X For 7 months, they have been ordering prototype batches ASIC of chips and then sticking them on Litecoin network and mining LTC. There engineers eked out another 0.6X performance and decided to announce a new A4 Doniminator would be coming to market in Sept 2016 2.6X better than A2. Before, the announcement a huge amount Litecoins where dumped on every major exchange - knocking $1.25 of each Litecoin (25% drop in price). That's way, engineers operate ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) I'm not entirely certain who you're even replying to here. I never said anything contradictory to this, nor did anyone here. However, ASICs are still, once again, much more resource intensive then prototyping on FPGAs or mining with GPUs and generally require quite a bit of funding to start out, which was mentioned earlier.
|
I buy private Nvidia miners. Send information and/or inquiries to my PM box.
|
|
|
scryptr
Legendary
Offline
Activity: 1797
Merit: 1028
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 02:48:28 PM Last edit: July 01, 2016, 03:31:48 PM by scryptr |
|
so amp edition is thrash... I ordered an amp extreme which should be better, let's hope it doesn't have the same flaws :-)
TRIPLE SLOT CARD-- The AMP Extreme is Zotacc's top of the line. It is factory clocked higher than any other GTX of the same "rank" (e.g. 1080, 980ti). The GTX 1080 AMP Extreme is spec'd at "2.5 slots" and has a boost clock (core) of 1900Hz + . The triple-fan cooling system has a massive heat block that accounts for why the card is so wide. The GTX 980ti AMP Extreme that I bought won't fit in the case, period. I am going to build a multi-card open rig around it, instead of replacing my GTX 960 as originally intended. A "2.5 slot" card means that the two adjacent slots on the cooler side are blocked, effectively making the Zotac AMP Extreme card a triple slot card. --scryptr
|
|
|
|
thevictimofuktyranny
Legendary
Offline
Activity: 1092
Merit: 1004
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 02:57:00 PM |
|
Did find this interesting item - out of stock! But, still, here they are: http://asicminermarket.com/?product=am440-eth-minerMain advantage on this offering is electricity works out to be 40% lower than running 14 R9 390's. However, at this price you'd rather stick with GPU's
|
|
|
|
pallas
Legendary
Offline
Activity: 2716
Merit: 1094
Black Belt Developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 03:47:31 PM |
|
Did find this interesting item - out of stock! But, still, here they are: http://asicminermarket.com/?product=am440-eth-minerMain advantage on this offering is electricity works out to be 40% lower than running 14 R9 390's. However, at this price you'd rather stick with GPU's Remains to be seen if a rendering becomes a real and working product :-)
|
|
|
|
pallas
Legendary
Offline
Activity: 2716
Merit: 1094
Black Belt Developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 03:49:16 PM |
|
so amp edition is thrash... I ordered an amp extreme which should be better, let's hope it doesn't have the same flaws :-)
TRIPLE SLOT CARD-- The AMP Extreme is Zotacc's top of the line. It is factory clocked higher than any other GTX of the same "rank" (e.g. 1080, 980ti). The GTX 1080 AMP Extreme is spec'd at "2.5 slots" and has a boost clock (core) of 1900Hz + . The triple-fan cooling system has a massive heat block that accounts for why the card is so wide. The GTX 980ti AMP Extreme that I bought won't fit in the case, period. I am going to build a multi-card open rig around it, instead of replacing my GTX 960 as originally intended. A "2.5 slot" card means that the two adjacent slots on the cooler side are blocked, effectively making the Zotac AMP Extreme card a triple slot card. --scryptr Thanks for the information. I didn't know of the slot occupation. Luckily not a problem for me as I will run it in an open "case".
|
|
|
|
Amph
Legendary
Offline
Activity: 3248
Merit: 1070
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 03:56:19 PM |
|
Did find this interesting item - out of stock! But, still, here they are: http://asicminermarket.com/?product=am440-eth-minerMain advantage on this offering is electricity works out to be 40% lower than running 14 R9 390's. However, at this price you'd rather stick with GPU's 12 x 1070, can do 360MH(windows and 440MH linux but consumption is probably higher) and consume 1200w, so lol at those specs, not even worth it at all but they cost more i must admit
|
|
|
|
thevictimofuktyranny
Legendary
Offline
Activity: 1092
Merit: 1004
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 01, 2016, 04:10:56 PM |
|
Did find this interesting item - out of stock! But, still, here they are: http://asicminermarket.com/?product=am440-eth-minerMain advantage on this offering is electricity works out to be 40% lower than running 14 R9 390's. However, at this price you'd rather stick with GPU's 12 x 1070, can do 360MH(windows and 440MH linux but consumption is probably higher) and consume 1200w, so lol at those specs, not even worth it at all but they cost more i must admit Yeah, I agree, but that is the opening salvo. Now, you know why people where opposed to Hard Fork on Ethereum Network! ASIC hardware could be toast, if the code was changed in ASIC unfriendly manner.
|
|
|
|
|