Miner- CUDA Equihash (with devfee)

1080 have gddr5r memory and there is latency being slower, ethereum speed is also 30% less then 1070

Apparently both 1070 and 1080 use the same DDR5 VRAM and processing core. Is this wrong?

"Both GPUs have 8GB of DDR5 and although they share the same processing core (GP104), the 1070 has 25% of its resources disabled. Comparing performance between the 1070 and legendary GTX 970 shows that the newer 1070 wins by a whopping 50%. "

So there must be another explanation…

@reb0rn21
Oh, you’re right. I forgot the GTX1080 has GDDR5x memory with higher bandwidth, but also higher latency. And latency is the limiting factor for my miner right now…

@zcminer-dev, that makes sense! Hoping you can have a 1080 version or flag that tolerates higher latency, maybe at the expense of higher processing overheads/VRAM usage… I show only 37% VRAM used, 1607MHz clock, 781mV and temp 64c no matter the threads, so GPU is not working very hard.

zcminer-dev
there is a small problem…
…often goes rejected

@zatosh
Sorry, latency is currently one of the limiting factors for all cards, so if I can improve on this, all cards should benefit… well, maybe the GTX1080 more than others :wink: I will try to optimize performance in future versions.

@Hurtmanriff
You could try using less threads, this will decrease the chance of outdated shares. On the other hand it might decrease performance, but not for everyone… Try ‘-t 2’ and look if it changes Sol/s for you. If so, increase the thread number again.

GTX 1060 ~80 sol/s
But 6 cards give to total 390-405 sol/s because Celeron G1840 processor is loaded to 99%.
I tried using -t2 and -t1, but it does not help.
Could you reduce the load on the processor?

use -s3 flag, it would give you a little less hashrate but sure it’ll help with your celeron bottleneck

@Lero
Yes, switch to blocking sync (append ‘-s3’ to the command line).
Though this might cause a little performance loss.

If that’s what it takes. Right now we’re all mining Ethereum, so the hashrate is waiting there. SilentArmy is adding Nvidia support as a after thought. Not that I mind getting some of the scraps, but they’re just that. We need a legitimate miner that can compete with Claymore. No one has picked up the gauntlet yet.

1 Like

The titan X pascal also uses gddr5x and that maybe why its not performing like i think it really should.
Thats good to know

Nice, 105 sol/s on windows with gtx980 oc, why min is cc3.5 but not cc3.0? my old mobile gpu could work…

I’m another 1080 doing less than 100.
I can get it up to right at 100 going to 2200mhz with 700 mhz ram offset but it isn’t the most stable :slight_smile:

Just for reference:
GTX 1050ti is doing 37sols
GTX580 is faster with silent army (19 sols) - apparently not compatible with this miner.

@topgeek try -t2 -g0 -s1 or -t2 -g0 -s2
With my titan x pascal im seeing 133 with -s1 and 127 with -s2

1 Like

with my current card overclock settings I was at 96 sols.
I added -t2 -g0 -s2 and now I’m getting 92-95 :frowning:

have you tried the -s1 or is that where you started?

-s1 is doing 93-96 sols - sorry forgot to add that.

If I remove all parameters (i.e. stop after my userid on suprnova) then I jump up to 97-100 and then it slowly slows down to about 94.
It is using 8 threads on its own at that point.
Current overclock 2037mhz, +100mhz gpu clock and +700mhz memory clock.

the card is an MSI GTX 1080. I have the fan at 1675 rpm which has it about 55C
I am mining 4 cores of the i7-6700k at the same time.

i seem to be most efficient at -t4 -s1 getting around 140sol/s and higher i drop to 133 again
i can’t speak much to fan speeds an such because my titan is watercooled at 2025mhz +200 and +700mhz to memory aswell

Can some1 pit an example .bat my miner wont start

zcminer -l eu1-zcash.flypool.org:3333 -u t1bSsj8xWFJx6n2J1LDvjzuc8dEJ9r1DxaR.test -s3 -t4 , but you do not have to add that -s3 -t4 options (or change -u address :wink: )

1 Like