20231224 X5 Release Summary by Gingeropolous

(my html skills rule. If you don’t like the width of the text… resize your window. Or copy the text and make a PR)

It had been almost 4 years since the RandomX fork when August of 2023 rolled around. During those years, the #monero-pow channel was mostly empty. Honestly most of the log space is me asking random questions, and I don’t have the logs for before 2023 because I migrated to a new laptop and didn’t transfer over the logs. Needless to say, the monero PoW space was quiet, and those that had been on the front lines of the 2018-2019 ASIC battle were relatively confident that RandomX was fulfilling its role in promoting and maintaining the decentralization of the monero blockchain.

On August 9th 2023,  sech1 posts this: (and you can read the logs here)

Now, the hashrate of those miners isn’t available as of writing this document, but you can still see the payouts.

 

The payouts start in 2020, and notice how frequent they are! Every 1-3 hours, almost 2 xmr. This is some serious hashrate. (Though honestly, during my review of this document and the logs, I don’t see any indication that these addresses were the X5s in testing).

About a week later…

.

And a day later (August 15th), this:

Almost immediately, there is talk of developing a fork to brick these potential ASICs. It turns out there are some optimizations for RandomX that would have been released at some point, but the advent of potential ASICs provides some more momentum to move them along. One such development:

https://github.com/monero-project/monero/issues/8827

As noted, this improvement isn’t really related to ASIC resistance (it was posted in April 2023, 4 months prior). It just turns out that if there’s an ASIC, due to the nature of how ASICs work (hardwired programming essentially), then any changes to the PoW would brick them.

In addition to these changes, there’s talk of reducing CFROUND operations or something.

On August 27th, Bitmain tweets the coming of the X5. The advertisement states that its RISC-V, so we already know that the device is not an ASIC.

The 212 kh/s stands out to sech1, who is able to infer that these devices are probably responsible for the nonce patterns that have been on the network since 2021.

https://api.hashvault.pro/v3/monero/network/nonce/distribution/png?hashrate=true&nonceColor=%23ff6600

Now that the computing architecture (RISC-V) of the device is know, there is conversation about whether they are worth bricking or can be bricked. It is a CPU after all. For the following 3 days, there is a lot of conversation on the channel (which has been mostly quiet since RandomX was released). Mostly about tweaks to RandomX that would make consumer CPUs more efficient (that CFROUND thing), or potential instructions to add that RISC-V chips don’t have (which would ultimately brick these Bitmain devices).

sech1 discusses a potential tweak that’s not really related to these new devices, and highlights the need for a RandomX fork.

On September 19th, the X5 firmware is released on the Bitmain site. The text from this session was awesome to watch unfold – just by diving into this software the people in the channel were able to figure out the actual CPU in the device.

The logs from this day are thick with a lot of technical this and that, but a preliminary analysis indicates that these devices were probably not profitable, at all, for bitmain.

For the rest of September 19th, there is more technical discussion about the X5 and improvements to RandomX that have been baking for a while. Over the next couple of days, tevador publishes some code for RISC-V mining.

And finally, on September 28th, polar9669 posts some pictures of the X5.

So, there it is. How the release of the X5 unfolded on the monero IRC channel dedicated to PoW. The wizards behind these handles were able to deduce the CPU based on the firmware released, and then there was confirmation of this hardware when pictures of the device were published by an end user. Ultimately, the general consensus is that the market price of the components in the X5 is greater than the price that Bitmain is selling them for, and that in standard Bitmain operating procedure, these boards have been used by Bitmain for about 2 years. Furthermore, these devices are not ASICs, they are instead a high-density CPU unit, so any attempt to brick them would ultimately fail. Finally, there was a flurry of activity regarding improving RandomX so that it matches the developments in consumer-grade hardware, which would have the side effect of making the X5 less efficient.

If you read elsewhere that the X5 is the “ASIC” that will break Monero’s ASIC resistance, well that is simply not true. The X5 is a bunch of CPUs, and RandomX was designed to run on CPUs. There is nothing to see here except Bitmain trying their best to extract as much money from the cryptocurrency sector, and the incredible sight of watching random people on this globe work together to figure things out. Based on the economics of these devices, the PoW group doubts that Bitmain will attempt an X6.

Though, of course, that depends on how valuable Monero is in the future.