We'll see if it actually gets that. On Ampere, rather than crashing the memory will generate more and more errors. Actually it probably was somewhat true in the previous generation too, but to a much lesser degree.

I have 3 Polaris cards running. 1 card is weaker in this aspect than the rest. None of the cards get memory errors nor do they get incorrect shares. But the weaker card needs the memory clock to be backed down little bit further otherwise the effective hash rate drops compared to the reported hash rate.

I've seen other systems on Etherscan where the effective hash is consistently 5-7% lower than the reported hash rate. This is what happens when the card is pushed too hard but not so much that monitoring utilities are able to see the problems.

920GB/s bandwidth is required to achieve 115MH/s theoretically. That's 23GT/s.

In reality though, the efficiency is about 85-90%, so you'll need 25-26GT/s to get 115MH/s. So its possible on the 3090, but not on the 3080.

Compare it to Polaris.
32MH/s requires 256GB/s bandwidth theoretical. 256GB/s bandwidth needs the memory clocked at only 8GT/s. But user results show they need the memory to be 8.8, or even 9GT/s to reach that.

It's unrealistic to expect 115MH/s. They are liars.

That won't work.

25MH/s require 200GB/s. 100MH/s requires 800GB/s.

You need 4 HBM2 memory dies running at 800MHz to do this. The only cards capable of being in the ballpark are the Volta based Tesla GV100 and the Titan V costing $3000. HBM2 memory has been cited as the reason for the pricing being so high. You won't get mainstream cards getting that much anytime soon.

350MH/s from a 6 GPU rig requires close to 60MH/s. Vega 64 gets ~40MH/s. RX 470/480/570/580 cards get ~30MHz, if you crank them to the max. Again you need Titan V costing $3000, or GV100 at $10K+.

Are you trolling? Because it sounds like you are.

It's 1%. 1 out of 100 times. You criticize others about greed, but you can't see your own. He has enough coding knowledge to make mining work and he's getting paid for it, nothing wrong with that.

Lots of people don't get that most successful people that get paid a lot, went through a lot themselves to get there, and did hard work. There are greedy folks with lots of money, but there are also poor folks that are lazy and greedy as well.

For me, I came back to Claymore miner from Phoenix because PhoenixMiner was causing problems. Sure, it saves me 0.35%, which is a WHOPPING 35% of the 1%, but running it stable is worth paying that extra.

They are calculators, not Crystal Balls. They cannot predict what will happen 6 hours from now, nevermind in 30 days.

mmmKay!

I'm talking about memory overclocks.

You said 29.8MH/s. You need to overclock quite high to get that, not a "mild" overclock. Plus just a few months dual mining could have your cards start working weird.

Do you dual mine? Or you used to and they are older cards? Are they overclocked to the maximum?

Genoil performs much less than either of the miners we are discussing so the demand on the card would be naturally much less.

You sure the problem is not due to the hardware/driver? Get something like OpenCL hangs on GPU 0, moments later the application starts itself only to get lower hash rate. That's likely a hardware problem, because I have an older GPU that does that occasionally. Not always, just when I wake up the display to use the system.

I am using PhoenixMiner on one of my machines as a test. It will take few days to verify.

Same hash rate on client side, even though peak is higher in some cases. The effective hash rate looks better, but as I said it will take few days to verify. Stales seem somewhat higher than Claymore. The ability to use nearly identical settings as Claymore is a plus, because it makes migration easier.

For all the knowledge you have, you should act more civilized.

Blockchain is faster for some, even with the compute enabled on the latest drivers. It also has another advantage, that you don't need to manually switch to compute mode every single time on a driver install.

From my experience there's a noticeable hashrate drop on Eth mining on Claymore with the driver anything other than the Blockchain one. And YES I tried the compute mode.

Blockchain hash: 28MH/s
Newer driver gaming mode: 20MH/s
Newer driver compute mode: 26MH/s

If you can't use more than 8 GPUs with the blockchain driver, I think it might be better to get a second system rather than having 12 GPUs on one system but getting lower hash/s.

How do you get the new drivers to have equal performance to the blockchain one? Even with compute the performance loss is there. 28MH/s to 26.5MH/s for example.