zen3 has slower L3 cache

zen2 avg L3 latency 39 cycles, zen3 avg is 46 cycles

which I suppose is unavoidable since they unified it across the 8 cores and doubled the size

<xx> how do we beat intel hyc bro

<xx> xmrig.com/benchmark/1

that's per-core? not really a useful number

since AMD has far more cores now than Intel

the multi-thread result is the only thing that matters. xmrig.com/benchmark

<xx> yes but its actually the most fun to bench

<xx> we do know that hyc

<xx> that 10900kf btw is 5.6GHz intel with 5ghz ring/cache

<xx> ddr4 4200 16-16-16-34

<xx> hyc: single thread wise, zen2 is still faster than zen3

the ryzen 5000 numbers aren't looking too impressive yet

but I'd guess an immature BIOS is part of that

<xx> that 3600 and 5600X of sech1 is both on 4.7GHz

<xx> theres this weird thing on zen3's bios

but yeah, zen3 L3 cache is slower than zen2, so this may be as good as it gets

<xx> i dont know if itll matter

<xx> 0xC86E One Of: Enhanced REP MOVSB/STOSB, VarStoreInfo (VarOffset/VarName): 0x1C3, VarStore: 0x5000, QuestionId: 0x1C, Size: 1, Min: 0x0, Max 0x1, Step: 0x0 {05 91 41 00 42 00 1C 00 00 50 C3 01 10 10 00 01 00}

that's not an instruction that randomX uses

<xx> useless then

<xx> cdn.discordapp.com/attachments/7253…88A-C267-4131-A8CD-C0BC80A24CB5.txt

<xx> common bios options for zen3

<xx> how about this

<xx> One Of: RdRand Speedup Disable, VarStoreInfo (VarOffset/VarName): 0x294, VarStore: 0x5000, QuestionId: 0x22, Size: 1, Min: 0x0, Max 0x1, Step: 0x0 {05 91 4E 00 4F 00 22 00 00 50 94 02 10 10 00 01 00}

<xx> i guess zen3 will have this working

<xx> they disabled this on zen2 shit was buggy

we don't use rdrand either, the mining nonce is our RNG seed

sech1: the 3700x - you can't get the RAM there to run well > 3200MT ?

I had that issue on asus strix B450-i

No, it's not stable

Well, my newer kit ran at 3600 MHz

and even my old kit could run (but unstable) with the same timings/voltage. Both are b-die

what board chipset?

But my old kit doesn't like high voltage, so I have to run it at 1.4V and lower speed

B450M-K

right. B-die should be happy with a good deal of voltage. But haven't tried my Falre-X over 1.4 either

Flare-x

Yes, but my old kit seems to be highly temperature sensitive. It errors out at 1.5V even if I run at 3200 MHz with stable timings that worked at 1.4V

hm maybe I should try pushing my 3600MHz bdie more

Not sure why the per-core h/s falls so much when I'm using more cores - temps are under control, clock seems stable

total miners: 44591