06-10-2021, 05:10 PM
I'm sorry but I saw that tweet and would still ask question about LHR and more details about the tests to verify.
If I remember correctly NVIDIA said that the limit take effect after they see a specific memory usage for some time and if that pattern is gone (for some time) they will lift the limit again. Thus I would not suspect any penalty by running hashcat benchmarks even if they are effected, too.
Well but NVIDIA also said for 3060 that the limiter will only be there for etherium mining and that it is not only the driver but a combination of driver and firmware which will make it unbreakable -- and then they break it with their own beta driver for developers.
With all this in mind (and that often people think that the limiter is there all the time) it would be nice to know how it was tested that these cards are not affected and what other tests are planed. To be honest I am not sure how I would test that, because maybe only 3 or maybe only 23 or maybe 53 algorithm of hashcat are affected and the rest is not. Maybe this was all planed and considered by the test already done, that would be awesome; it would be nice to know.
But the limit is not the only thing NVIDIA did to prevent hobby mining etherium (the big companies would use their own driver and firmware anyway). They limit the number of cards that could be used for full speed to mostly one card and they did not break that with their driver. Well but the miner could get around that to use up to 7 cards in one rig (which seems to be the normal number, I guess).
Are there any tests with multiple cards an hashcat? Up to now I only saw one card used.
Oh and because there was a question about if the above card is a LHR card or not: All 3080 TI and 3070 TI on the market should be LHR card -- all 3060, too, if you ignore the driver flop. For 3060 TI, 3070 and 3080 there should be both options LHR and non-LHR cards and it seems that NVIDIA force their vendors to mark the LHR cards as LHR in any way.
So to speak, the above card should be LHR because there are only LHR cards of 3080 TI, only the pre release test cards where partly non-LHR (and these chips should have other ids, meaning you cannot use the test driver for the market gpus).
If I remember correctly NVIDIA said that the limit take effect after they see a specific memory usage for some time and if that pattern is gone (for some time) they will lift the limit again. Thus I would not suspect any penalty by running hashcat benchmarks even if they are effected, too.
Well but NVIDIA also said for 3060 that the limiter will only be there for etherium mining and that it is not only the driver but a combination of driver and firmware which will make it unbreakable -- and then they break it with their own beta driver for developers.
With all this in mind (and that often people think that the limiter is there all the time) it would be nice to know how it was tested that these cards are not affected and what other tests are planed. To be honest I am not sure how I would test that, because maybe only 3 or maybe only 23 or maybe 53 algorithm of hashcat are affected and the rest is not. Maybe this was all planed and considered by the test already done, that would be awesome; it would be nice to know.
But the limit is not the only thing NVIDIA did to prevent hobby mining etherium (the big companies would use their own driver and firmware anyway). They limit the number of cards that could be used for full speed to mostly one card and they did not break that with their driver. Well but the miner could get around that to use up to 7 cards in one rig (which seems to be the normal number, I guess).
Are there any tests with multiple cards an hashcat? Up to now I only saw one card used.
Oh and because there was a question about if the above card is a LHR card or not: All 3080 TI and 3070 TI on the market should be LHR card -- all 3060, too, if you ignore the driver flop. For 3060 TI, 3070 and 3080 there should be both options LHR and non-LHR cards and it seems that NVIDIA force their vendors to mark the LHR cards as LHR in any way.
So to speak, the above card should be LHR because there are only LHR cards of 3080 TI, only the pre release test cards where partly non-LHR (and these chips should have other ids, meaning you cannot use the test driver for the market gpus).