Topic: GTX 1060 discussion thread - page 45. (Read 289435 times)

Yes, Hynix has bad timings. Hynix usually also do not go beyond +600. Samsung can go up to +1000. If you have set your cards to +1050, this could also explain bad performance, if the clock is too high, ECC correction will take place which then drops performance. I have my 8 Samsung running stable at +800, did not test any further...

Thanks for explanation.
On windows, it is not very stable even with 5 GPUs.

Just to be clear about Samsung vs. Hynix: even when running at same clock frequency, they do different hashrate? I am able to overclock to around +1050 MHz with reasonable stability.
I will check what memory I have and report back.

I love these Nvidias 1060s

PL 70%
Memorys +950
Core - 160

Runs fine 24/7 without issues

thanks for the explanation...
some of my rigs is nv cards.
i hope i invest the right cards.


rgds,

Nvidia BIOS is encrypted, no chance for any tuning besides raising memory clock. Getting other cards or living with the below average hash rate is the only option for Hynix cards!

#2 is caused by cards with shitty Hynix memory. No way to solve it. Get cards with Samsung.

#1 and #3: not every board can handle 8 GPUs. Try with 7 and check whether the issues are gone. If yes, try tweaking BIOS to get it stable with 8 cards or get another board. I can report that ASUS Prime Z270-P supports 8x GTX 1060, 6x PCIe and 2x m.2

PCIe Extenders 1x -> 4x is already a problem. Why do people get a MicroATX board with 3 PCIe Slots and then connect 8 cards to it?! Get a proper Full ATX board with 6 or 7 PCIe slots!

Hello,

I'm sure this has been discussed already, but thread is too long to read and search did not come up with anything useful. So here it goes.

Have a rig with:
1x MBO MSI B250M PRO VD
1x DDR4 4GB
1x Celeron
8x Gigabyte GTX 1060 OC 3GB Windforce
1x 256GB SSD on USB (also tried 16GB USB pendrive)
3x PCIe expanders 1->4x from aliexpress
1x 1.5kW server PSU.
OS: Ubuntu 16.04, Ubuntu 17.04, Kubuntu 17.04, Windows 10 x64 - same results
Linux drivers tested: v375, v384. Windows - latest driver.
Mining ETH with Claymore's miner.


Problem #1
The OS lags a lot, even when no miner is running. But when I start mining, the OS becomes extremely unresponsive and crashes anytime within 10 minutes to 2 hours. Especially harsh on windows, I don't even have to start mining, it's enough to open USB drive or do any other task and CPU usage jumps to 100% on one core, saying it's being used by system interrupts. And then it gradually mitigates to second core as well, and then it freezes up.

Problem #2
I only get 15.7 MH/s per card stock, and max 18.9 MH/s OC to +100/+1000 MHz.

Problem #3
On linux, unable to overclock if IGD is enabled, some kind of problem with nvidia settings - xorg.conf file. Have to disable the IGD and connect the monitor to one of GTX cards. This is very inconsistent, sometimes I get different results, as it looks like BIOS is behaving weirdly - sometimes IGD sharing still works, even if it is disabled, but then it gets stuck on boot screen, while GTX proceeds in to displaying normal desktop GUI.

---

Has anyone tried this motherboard? It has a B250 chipset and should support up to 12 PCIe devices. Problem #1 and #3 does seem like a hardware problem, since i get same results on both Linux and Windows. I have adjusted PCI latency from 128 to 224 clocks with minor improvement.

Problem #2 is probably a driver problem? I have tried v375 and v384 on linux - same results. Should I try v372, like suggested by Claymore's miner?

No it is not true.

At the risk of using an analogy that is like saying "I have two cars which can go 200km/h. But one has a V12 engine. Does it go faster?" Of course it doesn't because you have already stated they both go 200km/h!

The number of compute units does matter for the speed but it is not something you then 'add on afterwards'. It is part of the card's architecture (which also includes memory speed, bandwidth and other components) which taken together is what gives the measured 25mh/s speed.

In the end the only thing that counts is how many shares are actually accepted by the pool.

hi guys,
some people said that compute unit does matter.
lets say my nvidia and rx run at the same speed, about 25mh/s.
the rx card have more compute unit, it will have more share thus create more profit.
is this true?


rgds,

Asus GTX 1060 3GB - TDP 24Mh/s eth

try setting, -dcri 46. i had good results w that. also, eth drops less w sia

Buy a p106-100 mining edition they are cold and they have 25mh/s without so much loading

On dual mining Eth/Dcr hashrate drops 148MH/s Eth and 1480 Dcr.

I am currently looking at making my first rig and the plan is to use 1060s. I just can't decide what cards to get. Do I get more cards for my money by getting 3gb or get less 6gb cards so I don't have to worry about dag files. Any help would be appreciated.

Have you ever try dual mining with ETH, maybe you can get more profit.

Hello.
I have 7 MSI GTX 1060 3G on Biostar TB250 and all 7 cards works fine (+800 MHZ SAMSUNG memory overclock) 170 MH/S (>24.2MH/S) at 620W power using Claymore 10 on Win 10 PRO/1703.

those of you who are mining zencash, i'm curious how you have that set up if you are not just running all defaults. thank you!

i dont think there is a way to know which mem your going to get. i'm running msi 1060s and was checking around about a week ago and saw that i could get zotac 1070s for around the same price or a little cheaper, so i'm considering how i would go about that upgrade if i chose to do it.

No, looks like you should read the last pages of this thread...

My Asus can push 24.5 but they run way too hot. For 1mh/s less I can run them at 70% power and temps in the 50's. My Zotacs get into the 25.2 range single mining. Again though way too hot. At 70%power they can hash around 23.8-24.0 single mining.