Quote from: jstefanop on February 19, 2019, 03:43:43 PM

Quote from: Metroid on February 19, 2019, 12:46:05 PM

Quote from: wudafuxup on February 19, 2019, 12:16:59 PM

Quote from: dragonmike on February 19, 2019, 12:05:48 PM

Quote from: Metroid on February 18, 2019, 08:53:56 PM

if it was at least a trade off between them, vega56 1800mhs, rvii 2800mhs on monero, rvii is 50% shorter on computing performance than it was supposed to be and yet it costs 50% more, not worth, brainwashed trolls will make it to be worthy, trust me.

Give it time, Mr Doom & Gloom.
Unoptimised drivers, unoptimised miners - that's what we're mainly dealing with currently. Vegas were allegedly crap until somebody found out they weren't, remember?

Well the Radeon VII isn't running on a new architecture. It's essentially a Vega 56 on a smaller node with higher bandwidth and memory bus. The smaller node allows it to clock at the 2-2.1Ghz range giving it that big boost in performance. I doubt drivers will do much more than we are seeing already. The performance numbers make sense based on what we know from Vega 10 (which is essentially Vega 20 at 14nm).

This, nothing new about it, actually i think vega 14nm is way more powerful given is 14nm and i dont think drivers will help it either. If it was a pure die shrink then monero was supposed to have hashrates around 4000mhs which is not the case. I'm a lot more interested in navi.

There is no reason why Radeon VII cant hit 4k monero. Its doing 90MH eth. Something in the cryptonight kernel is not properly optimized to take advantage of the full bandwidth. Only other thing I can think of is that the algorithm is way more compute bound, so reduced number of cores is having a significant bottleneck.

I think it's the 2nd. Coz there is marginal improvement when HBM is ocd over 1150 or so. Only increasing core freq boosts the hashrate. So it's more likely core limited. cnv8 is more compute heavier than cnv7.