Topic: Liquid cooling for GPU miner (Read 4923 times)

As someone that runs pretty much the largest farm in the US, I can say 100% that liquid cooling is not something I would ever look at.

1. Inital setup costs are astronomical at scale.
2. All hardware warranties will be voided after they are soaked in oil, especially when all of the serial number stickers and everything else would just fall off.
3. Mining hardware in general is not reliable, you cant just stick it in an oil filled box and never expect to have to service it.


I would go so far as to say there is no market for this stuff in crypto datacenters. Bitfury tried to base their last chip around immersion cooling...go ask them how that worked out for them on the sales side.


Edit: I just looked at the promo video and it is even less likely anyone would ever use it. It is not universal, and in the mining game that is instantly a no go. Your video is discussing server chassis and other things that just dont exist in this market. Your reduction in power consumption you claim also have literally nothing to do with the crypto market. I think you guys dont have a grasp of this industry and might want to think about targeting traditional datacenters.

To my knowlege, there is only ONE "large farm" in the US that is working with immersion cooling.
The up-front much higher cost seems to be a major issue and the benefits are minimal compared to a well-designed air cooling setup in any area of the US that has VERY LOW COST POWER available in quantity (sub-3 cents per watt).

Hey LCS,

Looked at your website and sounds really interesting

Could you tell us if in one year your business had expend to include some mining farms ?

See U

Nico

I would say large scale miners are a rarity in the US, and most of the time even they optimize for cost over stability.

Here is what one might look like: https://www.youtube.com/watch?v=LR8cpsc66I8&t=172s

Mining is different from server applications because 100% uptime (and often fire safety and proper cooling) isn't a primary concern for many if it means cost savings. This is because miners need to ROI as fast as possible, not only due to the volatility of crypto prices, but also new hardware making old hardware obsolete or unreasonably inefficient (rising difficulty).

Generally, people who mine at a large scale do it in cold regions in China, where electricity is cheap and exotic cooling isn't necessary. Put bluntly, I don't think it is feasible to get miners to pay a premium for liquid cooling.

Really appreciate your advice, Xeon. You've offered some very valid points.

One of the unique aspects of our product that we are hoping to leverage is its extreme portability and the ease of recycling waste heat. For example, they could be used as heating units. A hobby miner could buy one for the bedroom, one for the living room, or perhaps retrofit their central heating system. For farms, we could see the heat being reused to heat an adjacent office building, etc. Any thoughts?

We can accommodate pretty much any GPU, but the technology is better suited to high-power NVIDIA GPUs that generate a lot of heat. However, I understand that it's preferable to have many mid-range GPUs. Is there something inherent about this approach that makes it more profitable, or is it a software/firmware thing? Is there any way to leverage higher-performing GPUs (say Tesla K80s) to yield a higher hashrate?

Great question!

Yes, there are several other liquid cooling vendors. What makes us unique is our patent portfolio; LCS is the only company licensed to produce liquid cooled systems in which all components are fully immersed in a completely sealed chassis. All other vendors will have at least one component outside the system, meaning LCS will always have the competitive edge on efficiency.

In the case of the link you sent, GRC's systems are not sealed, which is messy and cumbersome in our humble opinion. Also, GRC's solution is rather inflexible - you have to fill the tank with equipment. LCS's sealed chassis makes it much easier; you can buy them one by one. Plus, there's no need to drain the tank when changing out components. With ours, you just drain each individual unit, swap out your parts, refill, and get on with your life!

Here are our US patents (note we have several international and pending US patents as well):

PAT. NO.      Title
1   9,451,726   Full-Text   Method and apparatus to manage coolant pressure and flow for an array of liquid submerged electronic devices
2   9,426,927   Full-Text   Scalable liquid submersion cooling system
3   9,223,360   Full-Text   Rack mounted liquid submersion cooled electronic system
4   9,176,547   Full-Text   Liquid submersion cooled data storage or memory system
5   9,128,681   Full-Text   Liquid submersion cooled power supply system
6   9,086,859   Full-Text   Liquid submersion cooled electronic system
7   9,051,502   Full-Text   Nanofluids for use in cooling electronics
8   8,654,529   Full-Text   Liquid submersion cooled network electronics
9   8,467,189   Full-Text   Case and rack system for liquid submersion cooling of electronic devices connected in an array


Yes, farms would probably be our target segment. Any recommendations for penetrating that market? Is this where the farmers hang out, or are there other more specialized forums?

What's the patent #?


Total submersion has been around awhile.
https://phys.org/news/2012-09-intel-math-oil-dunk-cooler-servers.html

if you can bring cost down to just about nothing, then you would get miners interested... smaller miners that is.  You might be able to get farms interested the system if it was flexible enough where it was system independent.

bottom line, if you want to sell these to miners, the system would need to be flexible where it didn't matter which gpu card i put in the system and that i could swap cards out with minimal downtime.  basically the system would need to live through at least 2 different generations both amd and nvidia of gpu to be feasible for small miners.

at one time i was looking at a submersion tank with oil... back before the great crash of winter of 2013..  

Very interesting web site, looks expensive to do that.

Miners will do anything to save a few dollars, just look at the typical rig.

A mining rig in an aquarium filled with mineral oil and a couple of fountain pumps going into a radiator immersed into a 100 gal ice chest continually filled with cold tap water is a lot cheaper and would cool just as well.

Your products are very nice looking but they look pricey.

And that money would go a long way towards a few GPU's.

 

Oh I checked again looks its fine now. Before when I tried to basically read more about any product it would show 404 page not wound. Idk what problem was but looks like you fixed it.

Thanks for the feedback eckmar. The website is a work in progress and we are in the process of overhauling it. Maybe it's on my end but all the product pages seem to be loading fine. Which page isn't working for you?

Sorry if this came off as a sales pitch, really we are just looking for some initial guidance at this point.

Well your website is kinda fucked up. I looked to read more about your products but got 404 ... Maybe you should fix your website first before you decide to offer your services

Hi all,

My company has patented a unique method for liquid cooling electronic components. We currently design servers for data centers, but are looking to expand into other niches. Cryptocurrency mining is one of the areas we are looking into, so we are posting here in search of advice from the community as to whether this would be a good application for our technology and how to go about commercializing it.

LiquidCool Solutions employs total liquid immersion cooling, a cost-efficient technology that eliminates fans and substantially reduces the amount of energy needed to cool IT components. Liquid immersion cooling exposes components directly to a dielectric liquid coolant.

Our technology reduces cooling costs by up to 98% and has on average a 75% smaller footprint compared than an air-cooled data center. LCS coolant has 1400 times more heat capacity than an equal volume of air, making it a far more effective at transporting heat. Using our patented Directed Flow technology to circulate fluid through a sealed server chassis, coolant is constantly flowing through the system, maintaining optimal stability and reducing power-to-cool by 98%.

Unlike water, the coolant LCS uses does not carry an electric charge and is completely safe to use with any electronic component. Since all components are immersed in fluid, they are isolated from the outside environment, so they can be operated anywhere and component lifetime is extended substantially. Using liquid instead of air, the heat is easy to transport and recycle, in order to heat a home or building for example.

The system is fanless and there are no moving parts in the chassis. In fact, the only moving parts are central pumps, typically located in a mechanical room. Liquid coolant is pumped through the server chassis, removing heat from all of the components and carrying the heat out of the unit. The warm liquid then flows to a heat exchanger, where the temperature is reduced and the cycle repeats, similar to the central heating system in a home. The “cool” liquid can be as hot as 45°C, so the heat exchanger typically is a simple fan-coil system.

Here is a diagram of how coolant flows through the system: https://i.imgur.com/9JIWNEF.png
And a video aimed at the data center market: https://youtu.be/YAT_ZqWt2Og
Our website: www.liquidcoolsolutions.com

A few questions:

1. Our servers accommodate up to 4 NVIDIA GPUs at the moment. Is there any way we can make this effective for mining, or is re-engineering it the only way I can expect a decent hashrate?
2. Is there likely to be high future demand for GPU mining rigs? With Ethereum's upcoming switch to PoS, will another cryptocurrency fill the gap or will the market shift more toward ASICs?

Thanks for your time. Any thoughts or advice would be welcome!