soundtrack for this post: https://www.youtube.com/watch?v=m89BQJoBQ90&t=155s

So now that the dGPU market prices have stopped being entirely crazy, as the first wave of GPU demand shock has hit, I'm going to explain why this is just the beginning. Where March 2021 through March 2022 was an AWFUL time to buy a GPU. November 2022 through June 2023 will be a GREAT time to buy.

First, a very brief recap:

• Higher than average demand for consumer computing products due to the pandemic hit in 2020.
• leading node FAB capacity constraints hit at the end of 2020.
• GPU and CPU availability is problematic but prices are only somewhat above MSRP.
• Crypto boom launches at the start of 2021.
• GPU prices go nuts, up to 2.5x MSRP in many cases
• CPU prices and availability normalize -- GPU prices driven by something that doesn't affect CPUs. HINT HINT.
• Etherium implements a change that reduces miner payout significantly in May 2021 -- GPU prices almost immediately respond with a drop. HINT HINT.
• Crypto crash in early 2022.
• GPU prices subside

And here we are today. Good raw data https://www.reddit.com/r/hardware/comments/xbhrsk/the_graphics_card_market_in_the_cryptomining_hype/

​

So what comes next? This story is not done yet!

A few bit things are still yet to affect dGPU pricing.

1. GeForce 4000 series announcement and launch.
2. RDNA 3 announcement and launch.
3. General macro-economic problems globally
4. Etherium Merge is happening in a few days. If you deny this fact and want to make a fool of yourself with a snarky quip, I've got this soundtrack for you: https://www.youtube.com/watch?v=26SvLTb5qM4&t=83s There is more than enough information about what is actually happening for anyone not lazy enough to look for it. I'll have some interesting links at the end of this post for those who want to know more.
5. Intel releasing GPUs. J/K! Whatever they do it will be insignificant IMO, except maybe in the sub - $125 space or maybe if they magically release a working Battlemage generation next Summer that has ok performance.

​

Combined, the first four things above will be HUGE in keeping the GPU costs low. Not as low as people want, but still very good, especially considering the last few years. Of course, some markets in the world will not be as low prices as others, as any location with a healthy used dGPU market and extra old mining GPUs will probably have lower prices than elsewhere.

​

-------

Lets take the new GPU announcements first. These are the simplest.

Even if NVidia and AMD delay their releases somewhat, and push the highest of high end first, this will depress prices on the whole stack. Imagine that we low-ball the next gen rumored performance enhancements, and the 4090 and top end RDNA3 are 'only' 50% faster than a 3090 ti.

And they set the price to $1200 for the new GPUs... This puts a price buyers will be willing to pay for a prior gen significantly slower card at perhaps $650 to $750. And that puts a cap on the next GPU down the stack, compressing everything to lower prices.

But we are fairly sure, that the performance increase will be bigger than 50% at the top. And we are rather sure that we'll be getting something 50% faster at the next tier (e.g. a 4080) at a price significantly lower than $1000 ($700 or $800?). And that will significantly squeeze the whole price stack down, even before the 7700XT or 4070 are announced or launched.

---------

Next, what about the ETH Merge?

This one is HUGE. There are roughly 15 million 3070 equivalent GPUs worth of GPU hashing power mining Etherium today.

After the merge, yes, there will be other coins. BUT that is largely irrelevant. Those other GPU mineable coins produce approximately 5% of the global GPU mining revenues. Since mining revenues are _shared_ amongst all miners mining, the vast majority of hash power will be taken offline, or else it will just lose money.

There are many, many estimates of how many GPUs can remain profitable after the merge. But for our purposes here, it doesn't really matter. On one extreme some calculate that over 90% of GPUs have to be offlined. On the other, some calculate 50%. At either extreme, there will be a FLOOD of used GPUs, bigger than we saw in 2019.

I can go into much greater depth of just how many GPUs need to go offline, but to keep this post short, note that the following GPUs are the most power efficient for mining (depending on the coin mined, different ones win), and power efficiency will be king after the merge:
3060ti, 3070, 6700XT, 6800, 6600, 5700XT 5700.

Note, 3080Ti and especially 3090s are not on that list. They burn too much power. Likewise, 2080ti is not very efficient, and higher end AMD cards as well (A 6900XT mines as fast as a 6800, but just uses more power doing it and is far more attractive for gamers to buy).

Expect a lot of GPUs of all types on the used market. Expect 3070 FE and 3060ti to strangely hold a bit more value than the others, as they are the most flexible (across coins) and efficient currently available and miners will hold on to those longer.

----------------------

My prediction is that $300 sued 3070s will be common enough. 2080ti's less than that. Where does that leave mid range new cards?

The 6600 is already getting close to $200 used, I expect the 6600XT to be the one hitting near $200 used and $225 new. A 3060 is not a fantastic card at $350, but for $225 many will happily upgrade.

To put it simply: I think we have another 25% to 35% price drop to go on the low end, and another 30% to 65% to go on the high end of existing cards, over the next 6 to 8 months.

​

-----------------

Informational links:

Don't believe the merge is happening? type in "eth merge" in google.

There is a lot of (scattered) info out there, but a moderately tech heavy info dump is in these two links:

https://www.galaxy.com/research/insights/what-to-expect-ethereum-merge-upgrade/

https://www.galaxy.com/research/insights/how-to-watch-the-merge-ethereums-biggest-upgrade-ever/

​

​

RE: ASICs mining ETH and the impact on GPU prices. There are ASICs that mine ETH, these aren't 'real' ASICs, in the traditional sense, but are effectively purpose built hardware that is somewhat more efficient at mining ETH than a GPU. Estimates for what percentage of all ETH hash rate is ASIC are all over the place. But no matter what estimate you believe, the largest GPU POW Coin that is not ETH, is ETC, and a large percentage of the ASICs will work on that coin.

What this means is that less than 2% of the total mining rewards will be available to GPU miners after the merge. ETC is 60% of all the rest, and will be devoured by ASICs. The remaining few percent is what all the GPU power remaining has to fight over, in a race to the bottom of a combination of efficiency and power costs.

The most efficient GPU miners with the lowest power (and other) costs will remain. These will likely be at below $0.05 per kWh electric cost with more efficient GPUs. Everyone else will either sell their GPU, or hold it in storage hoping the next big mining boom happens before their GPUs are worthless.

​

Personally, I don't think there will be a 'next GPU mining boom'. This is because two things must happen for it to be real:

1. The overall crypto market has to boom again. Sure, one day this will happen.

2. One of these leftover, mostly shitcoins, have to become the next big thing -- at least a top 5 market cap coin. Doubtful. Very doubtful. Most new coins are PoS, most remaining GPU coins are nothing special that aren't going to attract much attention.

​

So, lets all hope that we never again have to deal with 2.5x MSRP GPU prices due to this, and GPU mining becomes a thing like CPU mining -- a fringe thing that never really affects the cost of the commodity.