Video has been disproved, is clickbait nonsense. Ignore.
This. The video is total BS. And, I'm actively selling these boards -- that should tell you something. |
|
|
|
I know that the PDU is 16A. I never stated that it was 20A. My original question, which has yet to be answered, is can the PDU run 16 Amps continuous or do I have to use the 80% rule and only run it at 12.8 Amps. If someone here has server room experience I would like to know an answer to that question. EDIT: After more google searching this is the best I could find as an answer to my question: Some of the APC PDUs I've had had circuit breakers, some haven't. If there's a breaker on it, it will be labeled with the trip value.
https://arstechnica.com/civis/viewtopic.php?f=11&t=1158437Yes, I'm aware of what you're asking. That's what I was answering, it sounds like it's 16A rated not 20A rated. If it was 16A continuous then it would obviously have to be a 20A breaker to handle that power. No? If the breaker says 16A on it, it's probably not derated. The easiest way to tell would be to push 15A of power through it and see if it trips. If it trips, it's not derated. To reply to some of your other points... The 6-20R/P don't come loose.. They're also like $1.50 for 2 outlets and you're using the exact same length, gauge and count of wire. The only thing that changes is you've added a couple boxes to hold the outlets. Here's my old mine... https://bitcointalk.org/index.php?topic=873046.0I know a little something about what i'm talking about.. Even if my name is senseless.. We had 10MW capacity and 1MW built out before the 2015 crash shut us down. Whats more interesting to me these days are these tripp-lite units using 415/240Y. You can place a single primary voltage -> step down at multi-mw level to 415/240. From there you can use 480/277 panels to deliver insane amounts of power to your racks. These tripp-lites are hardwired. There is one caveat though, going this method increases AIC and it is recommended required you add in-line fusible disconnects to reduce AIC. The fusible disconnects can be rated above the breaker amperage so that they never blow. The fuse itself though will reduce AIC due to it's design. DO NOT DO THIS UNLESS YOU HAVE AN ELECTRICIAN WHO KNOWS WTF AIC IS AND HOW TO CALCULATE IT. THERE IS ARC FLASH POTENTIAL HERE. Most electricians haven't a clue. This a european style electrical system thats used a lot in iceland for instance. It's still 60hz (in the usa) but in EU (50hz) they have a tendency to use fusible disconnects more than they use breakers. The fusible disconnects reduce AIC and arc flash potential. By reducing the number of transformers and step downs they significantly reduce the cost of building and maintaining their electrical grid. https://www.tripplite.com/28.8kw-3-phase-switched-pdu-240-230-220v-outlets-24-c13-6-c19-hardwire-415-400-380v-input-0u-vertical~PDU3XVSRHWBI suspect that using these will pay for itself by reducing man hours during installation/setup. Plus, who wouldn't want API based remote reboot? -- In addition you'll save costs on various panels, sub panels, isolation transformers, wiring costs, electrician costs, etc, etc. I expect a 415/240 setup at multi-mw to cost about 50-75% of a traditional 480/277 with isolation step down. @sundownz let me know if this is interesting to you. Happy to talk to your electrician. I really want to see one of these setup but have no use for it myself at present. You're probably getting close to filling up your transformer by now. |
|
|
|
What kind of operation is Xilinx running? Can they produce enough volume to bring these down to sub 1k$? 300$?
Don't have too high hopes. Your expecting something that hashes 10x more than a 1070ti and uses same wattage for one card to cost the same as a 1070ti? Why would anyone sell it for that cheap with the performance to wattage it can offer. This type of equipment is not meant for kids. There are a lot of reasons. |
|
|
|
What kind of operation is Xilinx running? Can they produce enough volume to bring these down to sub 1k$? 300$?
High volumes are not their typical business model. You can consider this them putting their toes in the water. I think they're pretty happy with how things went / are going. I know I'd be if I was them. We'll have to wait to see where it goes from here  Edit: I should say, volumes high enough to get the price to what we want it are not their typical business model. |
|
|
|
It's really great to see FPGAs making a comeback, and in a way that's more accessible to individual miners. I have some technical questions, if I may; - It was mentioned in a previous post that the units will come with encryption keys burned in. I'm somewhat familiar with the IP protection mechanisms of previous generation Xilinx devices, so is my understanding correct that these VCU1525's won't load a bitstream unless it's been encrypted using your key?
- If the above is true, if for whatever reason you had to cease your operation, or are unable to provide devs with continued access to development hardware, would you release these encryption key(s) and any other secrets to prevent future development work on bitstreams for fielded units grinding to a halt?
Thank you both for all your efforts in this. Kind regards, David Was it answered? It's very important question to my mind... Yes, the encryption key will not prevent unencrypted designs from being loaded but will prevent other encrypted designs from being loaded. If you ever did want to run some other encrypted design you would be required to blow the fuse and the same restrictions would be in place. But, for instance, whitefire's unencrypted designs will work just fine. |
|
|
|
Any update as to the group buy?
Units are for sale now but currently only for bank transfer and crypto payments as we're working through issues with paypal. https://fpga.land/ |
|
|
|
I still don't understand what is the issue for Paypal... You are selling a legitimate product, your customer are purchasing it, it's legal...
What is their legal reasoning for refusing your business ?
PS : will be wiring fund on monday
Their biggest issue is that they're pre-orders and aren't shipping immediately. I completely understand where they're coming from. Don't you? But, It should be a simple matter to find a situation that works for them, works for us, and works for everyone else in the community. My biggest issue was, it's so hard to get the right person on the phone. |
|
|
|
|
Update:
Good news everyone!
A senior analyst at paypal just called me. Had a good long talk with them. We discussed the reasons they're unhappy with our business and possible resolutions to the issues that would allow us to continue. I think we may have some agreeable terms that we can pursue. They're currently reviewing the decision to close the account and may reopen it and allow us to continue. They said they should have a decision for me on tuesday.
So, if we're not sold out by then! Paypal may again be possible.
Thank you to everyone for your understanding about the issues we've been having and for the support we've received. <3
|
|
|
|
As far as I know, no one has made a X16R bitstream |
|
|
|
Power utilization is determined by the bitstream you're using. You can have a bitstream that uses 5 watts. 225W is maximum power availability, you should build based on 250W-300W to be safe.
Do you or will you support Linux ? Yes, our software will be linux only. We'll have distribution repos for hiveos, smos, and debian. Probably a cent one too, because I'm a cent fan.. |
|
|
|
@senseless,
if a delivery address is in US or Canada, but i pay with wire transfer from a European bank account, I do not have to pay a European VAT do I?
I'd recommend you pay by crypto. VAT is due in the country of delivery, what has to be taken into account is the country of delivery (actual shipping adress) and nothing else. I was recommending he pay by crypto because when I verify his account it's not going to look good. i just happen to have an account in Germany, whereas my current country of residence, as well as country of tax residence, is Canada. Transfering funds to Canada first will mean I loose on convertation several times. In Canada you can't have an account in foreign currency that you can use for payments. Thus first it will be converted from EUR into CAD then again from CAD into USD when I pay from my Canadian account. Makes no sense. I understand, and I've recommended you pay by crypto. We still need to make sure these wires are legitimate and not people making phone calls to old ladies and fleecing them of their savings. |
|
|
|
We are only able to talk about the work we've done on lyra2z and cnv1
I noticed on your site it says "Power: 75W PCI-E + 150W 8-Pin AUX", so would that mean it'll actually require a max of 225w combined for each card depending on the algo? Or the 150w for lyra2z and cnv1 were already measured as the overall baseline? Trying to figure out how many can fit on my PSU safely without overloading. Power utilization is determined by the bitstream you're using. You can have a bitstream that uses 5 watts. 225W is maximum power availability, you should build based on 250W-300W to be safe. Do if that's the case, does that mean you're likely able to increase the cryptonight hashrate further by using a bit more power or is it limited by other aspects of the card? Our cryptonight hashrates are conservative numbers -- If you'd like to talk about performance aspects a great place is in the discord chat room. There are a lot of developers there not just myself and gpuhoarder. This is a community effort and there are a lot of other developers releasing bitstreams. However, we're not going to use their performance numbers to try to move product. https://discord.gg/M6CyRh |
|
|
|
I'm just asking how are fpga any different than asic they are expensive which excludes people from poorer countries getting them plus how is it going to save on power at first maybe but as the hashrate goes up so does difficulty so you need to add more miners also how are you going to get a company selling a 3000-4000 dollar piece of hardware to 700-800 and that's a high priced gpu.
Listen I'm not trying to bust balls here I'm all for power reduction it's the cost and we all know there's going to be greedy devs hiding in 10 or 15% Dev fees it was going on with zec, you look at things different because you have access to fpgas at a cost lower than most and know what your doing with them
an ASIC can only do one thing, what it was designed to do. A FPGA is like a GPU, it can do whatever you program it to do. It's even possible to program it in the same language used to program the GPU (OpenCL). There will be competition in the space, we're providing a platform that will force developers to compete on performance and fee to stay relevant. Yes, cost is still high. This is my focus right now. Dropping the cost of these devices. Our current offerings are just the first step in this process; considering that we've achieved 1/2 the retail price that others are selling them for, I think we did very well. But again, this is just the start. edit: removed the bit about claymore because it was rumor and probably not true to start with. But, even if it's not, the general idea stands. |
|
|
|
We are only able to talk about the work we've done on lyra2z and cnv1
I noticed on your site it says "Power: 75W PCI-E + 150W 8-Pin AUX", so would that mean it'll actually require a max of 225w combined for each card depending on the algo? Or the 150w for lyra2z and cnv1 were already measured as the overall baseline? Trying to figure out how many can fit on my PSU safely without overloading. Power utilization is determined by the bitstream you're using. You can have a bitstream that uses 5 watts. 225W is maximum power availability, you should build based on 250W-300W to be safe. |
|
|
|
Did you even read what I said, I don`t believe in open source for FPGA expecialy with xx big farms that will pay any developer out there, and we know there are just few ppl that will work on fpga, my personal opinion that just crumbs will be let out, profitable bitsream will be bought and only released when some over the developer head say he can.......
I also would like to be wrong and I hope... time will tell
Do you have anything to substantiate that claim? I can show you a chat room with 30 fpga devs in it right now feverishly working away to provide bitstreams for the community. What can you show me? https://discord.gg/M6CyRhI WILL get these FPGA to GPU pricing; or Xilinx (and Intel) is going to need to get a restraining order against me. I don't know how to give up, ask paypal. |
|
|
|
This is all pure FUD and nice talk....... we know as soon as HW stock is sold best profit algos will be done just for big farms....... ppl will start to curse and be mad... I hope I am wrong but that is how its done the released bitstream will get you ROI of +2 years or so , while the big one milk the coins and play stupid, and GPU miners cry someone is harvest their shit and no one has FPGA for it  PS: I still hope I be wrong, but I would not bet on that Why is it, I have to substantiate my claims but no one else does? it's ridiculous... WHERES THE CODE?!!? here... WHERES THE PROOF?!?! here... WHERES THE MORE PROOF?!?!? here... WHERES THE EXTRA PROOF!?!!? here.. I STILL DONT BELIEVE IT, GIVE ME MORE PROOF?!!? here.. THIS IS ILLOGICAL, EXPLAIN IT TO ME?!?! here... then..... "Uh ya, I don't have any proof this is a scam but I'm going to say it's a scam anyway because I'm a troll and I want to bring my FUD parade to town".. There are too many independent individuals working on this and with intimate knowledge of it for it to be a scam. And, as we've publicly stated MANY MANY times. There will only be 5000 of these things to hit the markets. We'd like as many as possible to be in the hands of independent miners. But if the independent / small miners don't want them -- there's no doubt that the farms will. I'm not going to make any claims as to profitability. I'm not a master of the mystic arts or even a financial advisor. What I will say is, we've provided very conservative numbers and we expect everyone to be very happy. However, we cannot control other players in the market, at least, not as much as we'd like to be able to. Hopefully I won't get trolled for quoting Donald Rumsfeld... But..... "There are known knowns. These are things we know that we know. There are known unknowns. That is to say, there are things that we know we don't know. But there are also unknown unknowns. There are things we don't know we don't know." |
|
|
|
|
My goal in this venture is to reduce the RETAIL cost of FPGA to under $1000, to utilize 50% of the power of a gpu, and provide 10x the performance of a gpu ... ON ANY ALGORITHM..... That's pointed directly at you ethash+progpow.
The current pricing we've achieved is a big step in the right direction. But there's still a lot of work to be done.
I want GPUs OUT of this market. They DO NOT belong here.
|
|
|
|
If you replace a network hash for hash you end up with the same hashrate and 1/10 the ENTIRE NETWORK POWER CONSUMPTION. Disregarding the normal fluctuation in value of coins being mined. I don't see how this conclusion is arrived at. Nobody interested in these devices is planning on continuing along at the same hash rate they are now with a GPU rig or farm. They will want and need more to keep up. Nobody is going to buy one and go "OK, I got the same hash as my 15 1080ti at 10% of the electricity so now I'm good this is it." Once they re-equip en masse then network hash skyrockets. Just like it does with an ASIC machine. They hit the network and difficulty goes through the roof. People then are under pressure to buy more to keep up. And so on. Perhaps this is a better way to illustrate: I have one gpu. I hash 1000 h/s @ 1000w. I buy one FPGA. I hash 10000 h/s @ 100w. Am I going to buy just one FPGA then? No. I'm going to buy 10 of the things so I can run 100x the hash at the same power. Just like everyone else will. While we are all hashing more for the same power. We are all hashing more. And difficulty goes sky high to compensate for this. Would like to know what I'm not understanding here. Maths is hard... Let's say a network has 1000 gpu miners each with 1Mh/s. Each GPU is like 300 watts. A total of 1000Mh/s and 300kW. Now let's say we have a network of fpga using the same algo. 100 FPGA with 10Mh/s. Each FPGA is around 150 watts. A total of 1000Mh/s and 15kW. So, by switching this network over to FPGA and replacing the hashrate hash for hash we have decreased the total electrical power consumption by 285kW (95% reduction in power consumption). .. But, what we really want is ASIC resistance, right? Ok... Let's say we have a network of 1000 gpu miners each with 1Mh/s. Each GPU is like 300 watts. A total of 1000Mh/s and 300kW. Someone goes and spends 2 months and $300,000 developing a 90nm asic that gets a 4x performance improvement over GPU plus significant cost reduction at scale. Now we have a device that will have 4Mh/s and consume let's say 100 watts... This is bad right? We have a secret miner on the network who paid very little cost to rapidly have an ASIC developed and they could quickly launch a 51% attack with as little as 250 devices! Let's make the assumption that I've achieved significant cost reduction and FPGA can be bought for the price of a gpu and are available in the same quantities...... We've now replaced our 1000 gpu miners board for board with FPGA. So, we have 1000 fpga each with 10Mh/s. Each FPGA is around 150 watts. A total of 150kW (still significant power reduction over gpus...) and 10,000Mh/s. Someone goes and spends 2 months and $300,000 developing a 90nm asic that gets a 4x performance improvement over a GPU plus significant cost reduction at scale. Now we have a device that will have 4Mh/s and consume let's say 100 watts... This is bad right? NOPE! This device would not be as profitable as the FPGA and is zero threat to the coin. The device would more than likely operate AT A LOSS depending on the cost of electricity for those running it. The total number of devices required to launch a 51% attack (and costs associated with it) would be increased by 10x. By replacing these GPU board for board we've made high level cheap asics unprofitable, we've reduced the power consumption for the entire network by 50%, and significantly increased the security of the network. ... So tell me, which is the better general purpose solution? GPU or FPGA? |
|
|
|
You realize, I HAVE POSTED FUNCTIONAL SOURCE CODE SO ANYONE CAN MINE ON AMAZON F1 FPGA. Correct?
You realize, I HAVE POSTED FUNCTIONAL SOURCE CODE SO ANYONE CAN COMPILE ON THEIR VCU1525. Correct?
I have searched and I cannot find the download links, any help in locating them would be appreciated. Search for "aws-keccak.zip" on discord https://discord.gg/M6CyRhThere's like 30+ developers on discord who are going to be releasing software and bitstream support. I can't believe it's still possible for anyone to question the legitimacy of the hardware or that it exists. |
|
|
|
I've not done any work on X16R. You're asking about a design another community developer has created. You'll need to talk to them about it. These devices are like GPUs. You're going to have a LOT of people providing software support for them. We are only able to talk about the work we've done on lyra2z and cnv1 |
|
|
|
|
Show Posts
