[vecio]

Centralization of Filecoin is an issue, but not the point of these tweets. The tweets focus on rekt Filecoin miners, they have spent more than $200M on hardware, but seems impossible to get back that with FIL mining, and since most miners are from China, no west media report this thing yet.

[roywiggins]

My question is, why did the miners not anticipate this- aren't the rules more or less agreed to in advance? Or were the rules changed out from under them?

Or is this something nobody anticipated would be an issue?

[vecio]

Most miners thought they bought 10TB storage, they will mine $FIL with that 10TB, but according to the rules, they can only start from like 1TB, most of the storage they bought are useless for mining.

[vecio]

Like in the tweets, many miners don't understand this, nobody tell them the risks in the community, not even the Protocol Labs.

I think mining equipments are over sold, maybe 10 times, due to the lack of enough warnings.

[ashtonkem]

It’s highly unlikely that they would get their money back if they lived in the west either. Most victims of crypto currency fraud are not made whole.

Genuinely not sure why people keep falling for this though.

[noxer]

Regardless of the actual reason why they cant use the 200M hardware, isnt int kinda funny that they even assumed that they could make money from the full capacity at day 1. Who the heck would be the customer of cloud storage hardware worth 200M at day 1. Clearly demand grows over time, that cant be skipped. Reward "miners" for whatever they provide even if no one uses it - sound unsustainable.

[wmf]

Rewarding miners for reserved but unused capacity is an explicit feature of Filecoin that I guess is intended to build up capacity in advance of customer need. These miners just did what Protocol Labs told them to do.

[znpy]

> These miners just did what Protocol Labs told them to do.

Without a contract? I'm not talking about smart contracts, I'm talking about real world, legally-binding contract.

If they just went and bought $200M USD of hardware basing on marketing... I guess that's the problem, right there.

[noxer]

Reserves sure, but at the start you dont know how much is needed so to just assume reward for whatever show up is comically naive. It could take months for real demand actually using a significant portion of the storage. Pay everyone else anyway is literally trowing money away.

Just imagine if one of the large cloud storage player would have set this up and simply offered all their reserves which they could do as there is no way anyone would use it. If they would get paid for that It would be very lucrative.

[etaioinshrdlu]

Well, they can always sell their hardware as it's generic server equipment right? They can also sell their FIL holdings.

[mariopt]

Brand new hard drives drop a lot in value once used. Buying 256 gigs of RAM (specially ECC memory) and 2080 TI is not exactly cheap either.

[mariopt]

Oh, I forgot a super important bit. Using a 2080 TI/consumer grade GPU for cloud purposes is against Nvidia terms.

To avoid being sued you need a Tesla GPU, the cheapest one is $7,374

[WrtCdEvrydy]

Datacenter license only counts if you're leasing out the hardware directly. It's fine to use consumer grade but you can't subdivide it or sell it to customers (for ML as an example)

[mariopt]

Bit of a grey area: https://www.datacenterdynamics.com/en/news/nvidia-updates-ge...

Quoting: “To clarify this, we recently added a provision to our GeForce-specific EULA to discourage potential misuse of our GeForce and Titan products in demanding, large-scale enterprise environments.”

According to this article, they want to discourage but I would expect legal retaliation if FileCoin proves to be profitable. There is one area where AWS never launched: gaming servers and I suspect this could be the reason.

[kalium-xyz]

Its almost always specialized cause crypto mining software for proof of work extensively uses GPU to increase hashes per second rate.

[rsa25519]

This is filecoin. Afaik it's storage-based, not compute-based.

[jdmichal]

I looked at the requirements earlier and was blown away. Their official recommendation is three servers with some very expensive hardware. Three Epyc processors, one machine with 128GB of RAM (and a lot more swap space), and one with 2x GPUs. And all that to "seal" like one TB an hour. But it's OK, because this is totally not a proof of work system.

https://filecoin.io/vintage/mining-hardware-config-testnet-v...

[Dylan16807]

How often (on average?) do you need to seal each TB you have?

For example if it's four months and you can do a terabyte per hour then that's about 3 petabytes and the cost of the sealing servers is a small fraction of the cost of the storage server. Is it much less?

But also the tweet says something about one terabyte per day?

[iamnotnicola]

You only seal (the expensive operation) when you add power.

Once you stop sealing you don't need to seal again for ~1 year (depending on your sector lifetime).

Keeping storage power is a cheap computation in comparison.