In the last post I worked on with my friends from Meta-Luban, we discussed the third hashrate revolution and made some predictions of our own regarding Liquid-cooled miners. Since the time of that post, we’ve had quite a number of inquiries.
There are still some questions from miners in the field which we would like to address in this post. And we have some questions ourselves.
If you are a traditional miner, using air-cooled systems, will you wait for the new liquid-cooled miners to come out and capture the market before you jump on board, or will you take a step in that inevitable direction, accept the change that coming, and embrace it ahead of your competitors?
Well, if you’re a series 19 Antminer user, we’ve got your golden ticket to the liquid-cooled world.
How did we achieve this upgrade?
Today, we will be conducting a liquid cooling upgrade test with a traditional air-cooled Antminer S19 Pro (110T).
Let’s take a look and see what kind of performance we can get out of the TSMC chip and whether or not it is possible to pull the liquid-cooled world of mining closer instead of simply waiting for the market to change.
Environmental Preparation for this Testing
This is a simple, closed environment test, just to go through the logic and technically verify both our own process and also see if we can actualize the evolution from air-cooled to liquid-cooled.
This technical test is purely for experimental purposes and of course, we want to share that with the world. The test will be broken down into seven parts.
- Upgrade packages.
- Upgrade testing procedures.
- Three liquid-cooled hashrate optimizations.
- What is S19pro with the best liquid-cooled system?
- Significance of liquid-cooled hashrate optimization.
- Why can the 19 series be upgraded with liquid-cooled technology on large scales.
- The shortcomings of this liquid-cooled test and the follow-up plan.
Let’s begin upgrading!
1. Upgrade packages.
[A] The liquid-cooled power supply of 6500w (oil-cooled as a substitute for now).
[B] Customized liquid-cooled heat dissipation board.
[C] Inlet and outlet liquid pipes.
[D] Fan simulator.
[E] Hardware system upgrade.
2. Upgrade testing procedures.
The whole procedure of this test is as follows.
Step one: Prepare a liquid-cooled tower.
Step two: Remove the fans.
Step three: Remove the dissipation board.
Step four: Clean the chips.
Step five: Put silicon grease on the chips.
Step six: Install the liquid-cooled heat dissipation board.
Step seven: Install the fan simulator.
Step eight: Install the upgraded software system (confidential).
Step nine: Connect the inlet and outlet liquid pipes.
Step ten: Prepare the liquid-cooled power supply (now using 6500w oil-cooling as a substitute).
Step eleven: Start testing.
3. Three Liquid-Cooled Hashrate Optimizations.
Three optimization modes of liquid-cooled technology.
Mode One — The best optimization (a hashrate increase of 70%).
- Hashrate: 181TH.
- Power consumption: 6500.
- Chip temperature: 73 degrees.
- Testing time length: 2 hours.
- Power consumption ratio: 36WJ.
Mode Two — The normal optimization (a hashrate increase of 35%).
- Hashrate: 152TH.
- Power consumption: currently untested — please wait for our second test result.
- Chip temperature: 64 degrees.
- Testing time length: 1 hour and 37 minutes.
- Power consumption ratio: untested and unrecorded.
Mode Three — The basic optimization (a hashrate increase of 20%).
Power consumption: currently untested — please wait for our second test result.
Chip temperature: 61 degrees.
Testing time length: 1 hour.
Power consumption ratio: untested and unrecorded.
Applications of the different modes.
- Basic optimization: Gain a small increase in hashrate with a slight increase or without affecting the power consumption ratio. This model is to guarantee the lowest shutdown price for the downward movement of the coin price.
- Normal optimization: Moderately increase the power consumption ratio to obtain a partial increase in hashrate. The coin price is in a continuous sideways phase, and this model is to moderately reduce the machine position, reasonably use the power resources and increase the coin production.
- Best optimization: Double the power consumption but improved power consumption to coin production ratio significantly. Get maximum hashrate improvement, boost coin production, and gain excess revenue in the case of industry bulls. This is the recommended model for obvious reasons as it allows for the modulation of power consumption and hash rate based on the prices e.g. a higher hashrate will lead to higher returns, but also your power will go up. When the Bitcoin price is high we can use the high power mode to get higher hashrate and higher returns.
4. Why is the S19pro the best candidate for a liquid-cooled system?
On its own, with zero modifications, it is a traditional miner with strong perseverance.
Expanding on that;
- When it is liquid-cooled, we can boost it to 180TH.
- It is a fully muted, non-disturbing product which makes its hashrate remain stable all the time.
- Although the power consumption ratio grows when we liquid cool it, its additional output is similar to a Whatsminer 70T M20S of 3250w only comparatively it’s still far better. Our consumption ratio of 46WJ is better than M20S of 48WJ. The chip temperature is 20% lower than that of the air-cooled system and its theoretical usage of time is greatly extended.
5. Significance of liquid-cooled hashrate optimization.
In order to obtain a higher hashrate, we have to first understand what exactly is preventing us from doing so.
The answer to that, as many miners will already know is that usually, this means higher temperatures are created. These higher temperatures will affect the chip directly unless cooled. This in turn means even higher power consumption is required to cool the chip. And so it’s a ‘one step forward, two steps backward situation’.
The problem is therefore not caused by the increase in the hashrate of a single chip, but by the heat generated from the increase in the hashrate. This means you can increase hashrate, as long as you don’t increase heat, or better put, have a way of efficiently decreasing this heat.
Is there a better way to dissipate heat?
The air-cooled method is limited when it comes to energy efficiency. In the traditional air-cooled way, we simply cannot keep the chip temperature at 70 degrees while maintaining high hashrate performance. But with the emergence of liquid cooling, we have a solution to this problem. It not only improves the miners hashrate but also reduces the chip temperature at the same time, which will greatly extend the life of the chip.
Of course, in order to obtain a higher hashrate, the chip calculation needs to be faster, so it needs a larger power supply as a guarantee. However, the air-cooled dissipation efficiency of the power supply is also limited. That’s why a liqiud-cooled power supply appeared. And because of the higher power consumption, and stronger power output, the use of an aviation head is for a three-phase power supply, significantly reducing the current of the traditional single-phase power supply, making the power increase, but the thickness of the wire does not need to increase. In the same thick line, the power of three-phase is more than 2.5 times the single-phase, so that the power supply will not result in high temperature, burning power, or fried power supply situation.
Fan cooling is a product of the air-cooled era. Sure, air-cooled systems can increase the fan speed to 8000 rpm, 10,000 rpm, or 20,000 rpm. However, physical damage occurs with such high-speed rpm. Maintenance problems occur, and parts requirements will also be higher, more frequent, and more expensive. And you also don’t want a mining machine that sounds like a helicopter.
Many things eventually reach a physical limit. In the field of traditional air cooling, a big investment only brings you a small improvement, so fan cooling is already a poor economic indicator when it comes to cooling in the high-hashrate era.
6. Can the 19 series be upgraded with liquid-cooled technology on large scales?
The short answer is yes.
Due to the usage of TSMC chips.
With the 19 series liquid-cooling upgrade, the biggest replacement part is the liquid-cooling heat sink. Carrying out the 19 series liquid cooling upgrade is very simple. We only need to remove the heatsink on the chip side. Unlike the heatsink of the Antminer 17 series which needs to be blown down with a soldering gun (if the temperature goes up to 500 degrees during the operation, you will need assistance from a professional engineer BTW).
Theoretically speaking, operations and maintenance can be independently done by anyone, especially for those miners who are already doing things on their own. Of course, if the number of miners is large, you will need more manpower. The only caveat is that in addition to completing the air-cooled heatsink, the original hashboard needs to be deep-cleaned. If too many impurities remain on the hashboard, there will be a risk of scaling, causing chip damage after the liquid-cooled machine starts running.
In theory, one person can complete 20 liquid-cooled units per day, which is a simple, and repeatable thing.
7. The shortcomings of this liquid-cooled test and the follow-up plan.
In such rudimentary and poor conditions, we verified that the 19 series can achieve the evolution from air-cooled to liquid-cooled products.
However, because the liquid-cooled power supply was not in place, this product test can only be considered half-finished. In addition, because the power consumption meter was not prepared, it was not possible to accurately examine the power consumption.
These shortcomings will be improved in the next liquid-cooled test.
We plan to do a comprehensive liquid-cooling test on the main models of the 19 series in mid-June including S19\S19j\S19A\S19i. We aim to verify the performance of the 19 series liquid-cooled products in depth again. We could also use help from any miner who has different S19 series models. By helping us you are helping the entire mining community who are eager to understand the liquid-cooled systems. If you are interested and want to help, contact us.
By late June or early July, a standardized liquid-cooling retrofit will be achieved, so that we can clearly see how liquid-cooled products perform in concrete terms on the premise of scaled-up liquid-cooled products. At that time, we will see anywhere from dozens to potentially hundreds of units from the 19 series, that will have completed their transformation into liquid-cooled systems.
Let us all look forward to it.
Notes: The testing of our S19pro upgrade liquid-cooled products as of the time of this posting is 18 hours of stable testing, hashrate remains at 181T, power consumption 6500W, and chip temperature around 64°C.
After the S19 series achieves liquid cooling at scale, the chip temperature will be around 70 degrees, the liquid-cooled power supply will be in place, and the power consumption will be optimized for the best mode and will be below 6500W. The bad rate of the machine will be significantly reduced, the overall risk resistance will be significantly improved, and with that, the S19 series will become a symbol of change in the era, and a long-term presence will be established in the new era of high hashrate systems.