We already have much interest. Plan to partner only with one team in USA. We offer exclusive USA only license. Maybe later we offer team in EU.
So why are you here drumming up business if you dont want any more.
Im lost, and out, enjoy your day.
We need more acess to FPGAs. We can currently spoof 203 TiB per watt. Very more energy efficient than hard drives.
I am going to sleep now. Best way to talk is WhatsApp.
Well I hope Chia Devs will find a way to combat plot grinding with FPGA, you seem to be kinda of smart yet ignorant about the damage you causing to the chia economy and farmers who invest in equipment and keep the price high.
It does not only do extreme harm to chia economy and liquidity but also exposing the whole ecosystem to a 51% attack vector.
If devs couldnât come up with a protective solution fast enough you will be the reason for Chia death. Congratulations
1 Like
We are not first with Chia FPGA. Our #2 competitor is 82 TiB per watt. Chia Devs know about plot grinders. They keep reducing plot filter making job harder. Only strong survival. If can do will do. Not our problem.
ÂŁ ÂŁ ÂŁ
they are securing the chain, cheaper will always win.
Ppl said the same about compression ruining chain security.
But most gobbled it up and ran with it.thinking the same as this outfit, ÂŁ ÂŁ ÂŁ.
Not looking for a debate, i realise the consequences of all this.
Enjoy uour day!
Yeah right. Did you include that 2 TB RAM + Xeon Server in the calculation?
Thatâs already ~500W right there, so that would be 500 * 203 TiB = 99 PiB.
I call bullshit on that sorry. Thatâs like plotting ~4000 k32 plots every 9 seconds, or a k32 plotting time of 0.00229 seconds.
Or maybe you mean 203 TiB per 1000 W? Thatâs something I can see happening.
2 Likes
One server controls a stack of FPGA controllers. The server is not included in the calculation of 203 TiB per watt (same as others). You can also use old Power10 Server with 64 TB of memory and gain efficiency. We only use old server. Very green than hard drive.
Still itâs hardly believable.
There doesnât even exist any FPGA with enough HBM to plot a k32. And if youâre not using HBM then it surely wont be power or cost effective.
Iâve partly developed an FPGA for GH 1.0 myself, it was a ton of work and nothing was easy. And it was for low compression, so most of the work is still done by HDDs.
The more you go towards grinding the more RAM / HBM you need. The biggest you can buy are 16 GB HBM FPGA, and they cost a fortune (around 10k USD).
203 TiB per Watt is 50 to 100x more efficient than HDDs. Full grinding is not expected to be more efficient than HDDs. So if your claims are real itâs gotta be based on some other technique.
2 Likes
FPGA boards and network very expensive. Not big advantage as think but still order of magntitude consider total cost. Problem is FPGA sourcing. Need USA partner. No cost to license but share 50%.
madMAx43v3r copy NoSSD so we careful with tell you tricks
We do not use latest FPGA to save cost. One technique I give you is shared-memory multiplexing. By time you catch up we will work on new plot format so do not try. We already offer free if you buy equipment and share 50%.
Wait, is the deal you are offering is that you will give someone the VHDL schematic, let them set up an FPGA, and not take any money until they have won blocks with their contraption?
Not going to lie, that seems like a good offer. If you are asking for money up front, then you will need to provide some sort of evidence that it works. Your claim is incredible, in the sense it will require incredible evidence to prove.
1 Like
What do you mean by shared-memory multiplexing? Spatial multiplexing or temporal multiplexing? How many FPGAs are in a stack?
What model Xilinx? I thought Xilinx is now AMD. How many are you trying to source?
To get the numbers youâre claiming you need some mathematical exploit, not hardware tricks.
Youâre basically claiming that one server with 2 TB RAM and a couple of old and cheap FPGAs can spoof 100 to 1000 PiB.
Heck, if you have trouble sourcing FPGAs, then you might already be the entire netspace. Assuming 100W per FPGA, it would only require 1240 units to spoof the entire current netspace.
Youâve got to be trolling or youâre off by 100 or 1000x.
2 Likes
Who is the other competitor? You said the #2 competitor is 82 TiB per watt. Is that only the FPGA also? What is the system level TiB per watt? Why are you not charging for the license?
Letâs talk. Iâll be in Shenzhen next weekend. Am from Europe. Energy cost is not an issue.
Are we talking Xilinx Virtex 7 series?
GrinderPro steal our code. Bad men. Please no use. They have same problem as us. No FPGA because ITAR. No trust.
Good to see your Icon!!!
1 Like