Your entire argument here seems to be along the lines of “I have a lot of low end GPUs and its not fair that I cannot mine Grin with them profitably”.
@antioch Wrong.
It should say: I and A LOT OF OTHER PEOPLE have low end GPUs.
Or let me put it another way: why didn’t you choose a 32 GB VRAM requirement (https://www.amazon.com/AMD-FirePro-W9100-Graphics-Card/dp/B01GHS2D08)?
The goal of the current VRAM requirement - as told by Grin devs - was that it should keep ASICs out and keep the apparently vast (centralized) quantities of 4GB out.
With both arguments now gone, what’s left to justify keeping lower end GPUs out? You know they will come online too, but at first by big farms only that pay for secret optimizations, thus you’re passively advocating for centralization.
If your only argument is “but you have a lot of low end GPUs” then you should stop right there, think what the argument is about and then start typing.
1 Like
What kind of cards you have exactly? 4GB or 2GB?
If that helps, I fundamentally disagree that the stealth ASIC argument is “gone”.
I should have 4GB version soon and I’m not selling it secretly to any farm. If better public 4GB miners emerge, all the better.
I mostly have 2GB RX 550 GPUs. Also have a smaller Vega farm and some 4GB GPUs.
If that helps, I fundamentally disagree that the stealth ASIC argument is “gone”.
Of course it’s never gone but a potential ASIC having to have 4GB vs. 1GB doesn’t increase its efficiency, it’s external memory either way.
I’m betting large 4GB farms have been on Grin since day 1. My guess the current focus is on unlocking 3GB 1060 Nvidia farms if that also not done already.
But it’s good that your miner comes out, will it be based on C++?
I don’t see any other way to reduce SRAM…
300MB is possible for ASIC, take the ZCASH lessons!
they are using DDR4, not really fast.
Hello,
Thank you for this information. Can you please further explain:
A) You mentioned these algos will be in place for 2 years. Does that mean there will be one algo after 24 months that everyone will mine?
B) You mentioned C29, can you also describe how C31 plays into that?
C) What is the goal with these efforts? To put GPU miners with an advantage in the first 24 months, the allow asics to take over? Just want to understand what the fundamental belief and what you are trying to accomplish.
Thanks!
A) You mentioned these algos will be in place for 2 years. Does that mean there will be one algo after 24 months that everyone will mine?
When they say 2 years they mean 2 years; there may be drift from the way difficulty rebalences off by one errors and the like, but it really should be 2 years unless you need to be an expert on the topic and you need to know the weird details pow systems can produce and we are coming close to the deadline.
B) You mentioned C29, can you also describe how C31 plays into that?
2^29 bits fits into common gpus that exist; 2^31 does not
Its mostly the same algo, the difference is the memory requirements to do things the fast way
C) What is the goal with these efforts? To put GPU miners with an advantage in the first 24 months, the allow asics to take over? Just want to understand what the fundamental belief and what you are trying to accomplish.
Asics racing is really really horrible, the asic market in bitcoin was fraudulent, rapidly outdated as new miners doubled in speed every 3 months with awful tech waste, until it started to slow down.
But asics are ultimately good for the coin health; you want the means of making new blocks to be in the hands of greedy people who cant afford to have 2ndary motives; otherwise china can declare that they need to rent some super computers to attack a coin they find inconvenient or the hashrate attacks of bch on btc, the cpu to gpu to asic jumps repersent roughly 1000:1 increases in power, so attacks simply become 1000 times harder, whatever they may be and there are allot of possible attacks.
tromps plan is trying to have the best of both worlds; an end state with asics and hopefully the asic races are slow enough that it isn’t just disgusting
Thanks for the reply. In A, I still don’t understand. Sorry, I was asking if there will be one algo after two years, versus multiple?
There should be only c31 after 2 years*
*pows are complicated organic systems and I would check the details of like 5 topics before writing miner my own code on it