Today is the day for the v8 fork – and given that hash rate on Dwarfpool just took a precipitous drop I assume that it just happened. As such: Make sure to check your miners, and update to 12.1 as soon as v8 is active!
For those using the Phi 7220/7240 PCI cards – make sure to use 0.12.1, not the 0.12.0 I posted earlier this week – 0.12.0 works on socketed phis, but had a bug in the MPSS offload code, which got fixed in 0.12.1.
With that – happy mining!
PS: And of course, also change “-a xmrv7” to “-a xmrv8” in your mining scripts!
My PHI 31S1P Cards waiting for luk-knc-native update,expect it.
LikeLike
Some as for the answer I just gave to DDG: I couldn’t get _both_ KNLs and KNCs fixed in time, and the KNLs had (much) higher priority … so it’ll have to wait ’til I find a few days… 😦
LikeLike
The hash rate drops a lot for phis after v8 fork. Any potential solution to improve the hash rate back?
LikeLike
True that …. but not all that much that can be done about that: The v8 fork adds another (third) step to the inner loop, _plus_ bumps several of the 16-byte read/writes up to 64-byte read/writes. Even assuming the “four times the bandwidth” gets absorbed by the fact that those 64 byte are still the same cache lines as the 16-byte reads were (and thus, not _really_ more bandwidth), there’s still that additional third stage, which is quite costly. So back of the envelope 50% to 100% more work shold translate into a 33-50% drop in hash rate. There’s _some_ room for improvement, but probably not too much (and no, nothing you can do as a user). On the upside, that algorithm becomes more costly for other sorts of CPU and GPUs, too, so world hashrate – and thus difficulty – should at some point drop, too, so we’re not _really_ losing anything in terms of relative hashrate.
LikeLike
The problem is Monero mining is no longer profitable on phi, while Vega doesn’t lose hashrate on V8. (It does require a bit overclocking and consumes more power on Vega). Network hashrate is back to 500m already.
LikeLike
I’m at it … see latest post. I’m still slower than before the fork, and _some_ performance penalty will likely remain, but with the right changed we should be way better than what we currently are. As said – I’m working on it …
LikeLike
Hi, is there a reason you didn’t make the knc-native version this time?
Which version do I need to use on my 7120s?
LikeLike
Simple reason is that I haven’t gotten around to doing it, yet – the problem with the KNCs are _very_ different from CPUs and KNLs, with a totally different vector instuction set, special compiler, and even a special machine for developing and testing (that I can’t even access remotely from my starbucks :-/) …. So in short, fixing that KNC back-end is a _ton_ of work, and with order 3000 KNL phis that were working on monero yesterday the phis just had higher priority …
Once things cool down a bit I’ll port that code over, too (as I did for v7), but at least for now it’ll have to wait a bit…
LikeLike
The problem with the KNCs is that they use a _totally_ different instruction set than CPU and KNLs do – the latter two obviously had _some_ differences, but could share a lot of code; but the KNC codebase had to be completely different, which basically means “a lot of work” updating to new coins. On the upside, many of the changes I’m just doing for improving v8 performance will make the KNL version _much_ closer to the KNC one, so there’s now “some” reasonable hope that the latest rewrite will eventually also support KNCs again.
LikeLike
I didn’t even know people still used KNCs at this point.
LikeLike
My AsRock 7210 dropped from 2800 to 1550 H/s…Anything we can do to improve?
LikeLike
Hi,
Why didn’t you add CPU host mining for other algos for Phi card version? For example xheavy. Why does support only xmr?
LikeLike
The new phis do support all those algorithms; only the old ones don’t.
LikeLike
When miner for KNC ???
LikeLike
Working on it. v8 performance improvments for KNL have higher priority, but I’ve heard loud and clear that poeple _are_ still interested in KNCs (which a month ago I wasn’t so sure of, BTW).
LikeLike
well hashrate dropped from 13500 h\s (4xphi 7240 + 2 xeon 2690) to 8100 h\s. :(. that’s a lot..
but i don’t see such a dramatic hashrate drop on a videocards. why so? is there anything to do with this?
LikeLike
There’s some performance improvements coming up soon, which should get at least part of that performance back. I’ll write a blog post as soon as the release will be available.
LikeLike