Compare FPGA and 1080TI 9 BCU 1525 cards 0xtoken and Skunkhash algorithms Cooling problem can be solved by When mining on FPGA you should consider a few aspects Hello everyone Today we are going to talk about FPGA mining, its pros and cons which miners may have and discuss methods of cooling these cards

For our tests, FPGA cards and bitstreams were provided by our partners – AET Technologies While ASIC and GPU mining still remains relevant, high power consumption makes mining a pointless practice in regions, where electricity rates exceed 10 cents per kilowatt-hour in the current market situation That is why miners rediscover FPGA over ASIC and GPU devices today FPGA excels video cards in terms of performance and energy efficiency and beats ASIC when it comes to flexibility of choosing a mining algorithm As long as the majority of mineable coins use the Proof-of-Work (PoW), these cards will allow to mine any coin at any time, which is as relevant as ever in the current market situation

Before mining on FPGA you need to consider the following aspects: First, each card has limitation of its load capacity by default, but this can be changed via the programming tool for greater performance Secondly, while changing algorithm on GPU is quite simple procedure, in the case of FPGA it requires a prepared bitstream – a file that contains the programming information for FPGA, the creation of which is rather a time and money consuming process The third thing is a high heat transfer coefficient of FPGA in mining and its inability to operate in high performance mode This primarily concerns stock radiators, which cannot properly cool the card that operates at high capacities, thus it is impossible to achieve maximum performance with a standard air cooling Here, we removed the case from the card and now can see a small radiator, which we mentioned earlier

When discussing the subject of FPGA cooling, many people focus their attention on chips cooling, ignoring other components that also heat up when the card is running This is the main mistake, because chips can heat up and operate even at 110 degrees, while other components of a card will simply melt at such temperatures, for instance – the power regulator Those miners who are aware of this problem have a separate cooling fan installed specially for this regulator The problem with cooling cards can be also solved by water blocks or large radiators for air cooling However, this leads to additional costs, as in the case of our Xilinx BCU-1525 cards – one water block costs about 200 dollars

Besides money, such modifications require additional manual modifications You will need to replace the original card radiator with a water block or a huge radiator for air cooling For cooling in immersion liquid, you don’t need to apply such modifications It should also be noted that all these solutions are focused mainly on cooling chips, instead of other components which can overheat faster Therefore, those approaches are less efficient compared to immersion cooling

We are now in our lab where the prototype is located, which is our cell with 9x BCU-1525 submerged When submerging cards into liquid, all their components are evenly cooled With this cooling method, there is no need to refine the standard cooling system Liquid easily removes heat from all the spots, reducing the risk of cards overheating during operation in high performance mode Now, on the example of 2 cards, we will demonstrate FPGA performance on 0xtoken and Skunkhash algorithms

on the 0xtoken algorithm, the hash rate is currently equals 159 Gh/s on the Skunkhash algorithm, it is about 612 Mh/s In the miner you can see the frequency, hash, temperature, voltage and the amount of solutions It should be noted that the hashrate is dynamic, it varies in a certain range The maximum hashrate value we’ve managed to capture during our tests on the 0xtoken algorithm is *17,2 Gh/s, and on the Skunk algorithm it is *680 Mh/s Now we will compare these results with Geforce 1080 TI cards performance So, the comparison between FPGA and 1080TI card

You can see the consumption and performance in the table, as such FPGA is 12 times more productive than 1080TI, while remaining more energy efficient Though, at first glance, FPGA is an expensive solution for mining, still it is necessary to take into account that this trend has been actively studied right now and the developers have been constantly worked on the new bitstreams development, making them more productive and profitable, despite the high costs of the cards For instance, bitstream for x16r, aimed to make FPGA bring a $15 daily profit, according to the developer We will show you the working rack based on FPGA in the next video If you enjoyed this one, please hit the like button and subscribe to our channel

If you have any questions – write them in the comments See ya