6 x 1080 Ti - 4 at 720+ Sol/s but two at ~680 Sol/s

Any ideas ?

Hashrates:
GPU0: 722 Sol/s
GPU1: 727 Sol/s
GPU2: 717 Sol/s
GPU3: 686 Sol/s
GPU4: 681 Sol/s
GPU5: 725 Sol/s

Each of those uses 600w, right? Do you have atleast 3600w for the just the cards?

Mine should have been in 1 hour ago, so I’m sure I’ll be back with all sort of problems lol.

600W each? I think you’re mistaken :wink: 1080 Ti = 250W

2 Likes

600w ? what card makes that ?

each card takes about 250W, and i have 2 x 1300W (each running 3 cards) its way over what i need.

Yup, you’re right - that’s the power limitation as listed on newegg

can you tell more details how you get this hashrates? mine is 4x asus gtx 1080 Ti and they work two at 600-620 Sol/s and two at 400-500 Sol/s

Can you post the make, model and sol/s of each card along with the current voltage/gpu clock/memory overclock settings?

I also read somewhere that some manufacturers use Hynix memory on the gpu cards and it underperforms beside the manufacturers that use Samsung memory in their cards.

1 Like

There are 1080ti - 250W
There are other 1080ti - 280W. Depends on card/type/model.

For example EVGA FTWs are usually the more powerful types (1060s, 1070s, 1080s, etc) + their boost clocks go a bit higher too.
with the 1080ti spcs consume up to 280W.

ero77: I’ve found with my ASUS cards that they tend to perform better at Equihash if I run them at 100% TDP or a little bit more. If down-power to …say 60-65-70% they run at lower SOL efficiency.
May be check it out - I dont know how yours are setup.

MColeman: Yes. 1060s and 1080s usually can go with either of the 3 types of memory
samsung, micron, hynix. Latest “batches” sold seem to be a mix Micron/Hynix. Not sure if they even sell with samsung memory anymore in the past 1-2 months or may be rare cases.
But that (memory type) should not matter much for Equiash/Zcash, etc type of algos.

280W = 112%TDP
You cant get it with stock settings
I saw one on this forum with a modded 1080ti at 305W(only to the psu) its totally dumb why not use a LN2 bios it will be more funny :fire:

I remind you that 8 pin pcie power connector is designed for 150W and 6 pin pcie power connector for 75W
The extra power is pulled from the PCI Express that provides a maximum of 75 W

All of them are Gigabyte 1080 TI

How about their temp? I observed that my 1080ti are having more sol/s when in low temp.

Actually, the PCI-E power cables , either 6pin or 8pin, can handle more power each.
the 6pin can handle 150 watt without much trouble.

say they come with a 3+ wires for example. By ‘spec’ they are rated to cover 2A and yes 75watt for 6pin by spec.
but the cable length and wire gauge which comes with good PSUs that has 8pins allows for much higher amperage. Lots of reserve above the 150watt. Of course if you are not using some tier5-6 cheap PSU cabling…:slight_smile:

example. a good PCI-E power cable comes at 18AWG wire x 3. Length usually 600mm to 750 mm
250Watt is possible on those PCI-E 8pin. The issue may come if you attach 2 GPUs where each consumes 250Watt (total 500W) on a dual PCI-E 8pin (2 x 8pin connectors).

The crappy "GPU Boost 3.0"
You can try to mislead it a bit with black pencil or conductive thermal paste
Under 60°C my 1080 was already throttling in overclocking

Very true as jimmy97one61 said.
Lower temp, get a bit more stable OC and computing.
that is why people invented water cooling for pc components I think.

Please educate me on this concept as I am total newbie to this. How will the gpu be throttling when it’s under 60C? I do have a water-cooled 1080 ti which I believe will last longer if I keep it at 60C and I only OC it <+100 on both core and mem. If that’s not the case then why?

These cards have some “Performance states” 1, 2 and 3. When doing compute applications (mining included) it enters state 2 which downclocks the cards automatically.

That’s why when you close the miner you can notice that your clock increases! That’s because the card goes from P2 to P3 or P1. This variations between Pstates make overclocks less stable and decrease the potential on these cards. I don’t know how to disable them. Might only be possible through bios modding.

from what I know (may be wrong of course) ,
it its usually a curve trying to match clocks vs temperatures not a flat line… and we dont know what the Manufacturer put in the bios for settings.

Forcing State to P2 under mining, can be disabled with nvidia profiler tool (think that was the name)
but I dont see why we should mess with it.
also the P2 state downclocks the memory mainly from what I’ve seen. -200 on the memory.

So back to the OP - as others suggested, check Temps and Watts usage per card.

temps are ~60 - 65 , never over 65