-
ErCiccione[m]
-
ErCiccione[m]
for non-twiter folks: "Full validation sync of monero 0.15.0.1 to block 2025045 on my benchmark machine took 1 day 2 hours 40 min. Caching policy is effective; only read 9GB from disk while writing 297GB. Bottleneck appears to be software rather than hardware."
-
ErCiccione[m]
from Jameson Lopp
-
sech1
Well, full validation sync is slow
-
sech1
It takes at least 12 hours on a modern CPU to verify all PoW hashes from 2+ million blocks
-
ErCiccione[m]
Yeah, what hits is the huge difference with a bitcoin core node 0.15.1 (i know our tx are way bigger and our times are slower for a reason), which he synced in 162 min last year. assuming he used the same machine for both.
-
sech1
Cryptonight hash verification takes 20 ms on an unoptimized PC (I doubt he configured large pages properly). Now compare it with Bitcoin's SHA256.
-
smooth
12 hours is assuming one core, which is unlikely for most modern PCs
-
M5M400
took about 3h on my ryzen laptop. btw the "trusted daemon" checkbox in gui should prevent verifications, right? seems it doesn't work in gui 0.15.0.2
-
selsta
prevent verifications?
-
selsta
there were problems with the checkbox in the past but not in v0.15.0.2
-
Inge-
how well does monerod node sync utilize multiple cores?
-
moneromooo
1 day isn't bad for full verification. Does he mention how much it took before ? I know he did that experiment before, and historical sync got sped up since.
-
smooth
Inge-: It seems to use the first few reasonably well, after that it might not be that great
-
smooth
Possibly some tuning would help, did not ever try
-
moneromooo
It depends what it's verifying. IIRC PoW uses everything. Other things vary. It *could* preverify PoW so there's always some PoW ready for starved CPUs.
-
hyc
M5M400: the trusted daemon option for wallets only means the wallet will do some RPCs that it otherwise would avoid
-
hyc
and yes, currently monerod will use all available cores
-
hyc
whether that's a good idea or not is debatable. Intel CPUs are cache-starved, so it's better to use fewer threads than cores.
-
hyc
AMD tends to be fine with threads == cores
-
Insight
is it technically possible to run a gui wallet on a private testnet?
-
moneromooo
Yes.
-
cohcho
Is it possible to be notified when core isn't busy without polling or patch that triggers --block-notify within cryptonote::core::on_synchronized() ?
-
asymptotically
cohcho: i was just wondering the same thing today :)
-
cohcho
I need it because of my hardware may freeze unexpectedly and i'll need to resart monerod and pool as soon as possible
-
asymptotically
cohcho: maybe we could add a new flag, like --delay-block-notifications