-
gingeropolous
i wonder how long the road to 3 gh/s is gonna be
-
gingeropolous
.c 1+1
-
gingeropolous
no bot i guess
-
gingeropolous
what was the spike from azure agan
-
gingeropolous
2.44 gh/s
-
nioc
mebbe 600 MHz by themselves
-
nioc
anybody who actually knows is sleeping so I took a guess
-
gingeropolous
-
gingeropolous
it'd be the equivalent of 10k dual epyc miners, if those dual epycs can to 100 kh/s
-
gingeropolous
*can do
-
gingeropolous
so it could be a long road
-
nioc
senile man strikes again \o/
-
gingeropolous
well, i guess presumably at some point the infrastructure of the large pools will hit its limit
-
gingeropolous
and pool ops will either have to upgrade / add more infa, or find a way to encourage ppl to move
-
gingeropolous
because performance could be degraded for everyone
-
hv-bridge
<\x> lol
-
gingeropolous
total miners 78170
-
gingeropolous
minexmr.com 32.7
-
Inge-
unknown isn't that big either
-
midipoet
OMG I forgot about the bullet catching drones
-
midipoet
Also, wasn't there a whole company. What was it called? They wanted to mine in fridges and ovens
-
midipoet
Kristy, I think, right?
-
moneromooo
Kristy Leigh Minehan (spelling might be off a tad).
-
midipoet
was that also bat shit crazy or legit? The idea was IoT mining devices, which didn't seem too off the wall.
-
Inge-
reading scrollback midipoet ? :D
-
midipoet
maybe
-
lkcl_
scrolling back to the bit about RAM: OpenCAPI, OpenHMI and Gen-Z are all designed to be 3D stacked memory for GPU and Supercomputers
-
lkcl_
and AMD has started doing chiplets for multi-core SMP
-
lkcl_
putting 2+2 together to make 5, it's not outside the realm of possibility that every-day computers / GPUs end up with insanely-fast multi-chip NoC'd DRAM access
-
lkcl_
(translation) a large batch of DRAM ICs, which gives great throughput, backed up by a Network-on-Chip that routes through and gives general high-performance access
-
lkcl_
somewhere in 5-6 years i am planning to implement precisely and exactly that (if another fabless semi company doesn't do it before then)
-
lkcl_
and i know of at least one person looking to design an alt-currency ASIC-based-on-GPU with the same type of 3D-stacked layout
-
lkcl_
multiple 20 gigabit SERDES to multiple DRAMs *well* in excess of what is "traditionally" done in any CPU or GPU
-
lkcl_
these types of designs will literally eat RandomX for breakfast
-
Inge-
go for it
-
sech1
DRAM access is not a bottleneck in RandomX even just with well-tuned DDR4, there's no need for much faster RAM
-
sech1
RandomX is not just about memory access lkcl_ but good luck anyway :P
-
Inge-
it would probably eat CN for dinner
-
hyc
CN was toast 3 years ago tho
-
hyc
no challenge there