What is the power consumption?

i would guess virtual memory issue

As others wrote - mining at loss is the way to go

Not so long ago people speeding lies about 8 being the limit, I thing the are lying because they don't know, not intentionally.  But then there is very few real info and lot of lies. So maybe it is intentional to scare people away by putting out lot of bullshit.

And then Metroid is trolling hard for years now.

I don't think that nvidias, at least 10 series with gddr5 like the strap changes, as there is no way to see memory errors.
I stupidly did not backup my pre-strap OC, but now even my most stable, with straps, rig crashed after 122 hours 36 minutes.
I wonder if maybe older drivers or windows7 are less prone to memory crashes as someone mentioned here or if is it just bullshit.

wifi is ok, cables do go bad too, I have good snr on wifi with lot of rigs, however I can't use 2.4G logitech keyboard from more then about 1.5 from the rig, so maybe try different channels

Does someone other experience better hashrates with 391.24 then newer drivers?

I think you should always define what you mean by stable - for me it is at least week without crash due to mem OC. It seem to me that people consider mining for hour or few hours as table, while I know from experience crash can occur even after two days of continuous mining due to high OC.

Microns strap 1,100 - 25.6 on best cards (+640mem), while keeping <70w per card (by undervolting, but I guess it is now equal to PL70).  Drivers: somee recent ones, i think 430something
Did not get to samsung jet as I still have lot of stability problems.

Do you think, that maybe using PL instead of UV there is better stability with straps?

I think nvidia part of the miner is really beta ,
multiple gpus crash with strap 0 intensity 0, +0 mem while before took +700 mem with ethminer or phoenix
some gpus with straps crash after 46hours of mining - this I think might indeed be to much OC, but still I would expect it to crash sooner
but really I think it is some driver (DCH on rigs with most issues) or even just NVML issue (again DCH seems to have differences in NVML) - well thinking of it now, issues on rigs without DCH drivers are kinda manageable and be indeed cause by straps or too much OC, but on rigs with DCH it seems to just not work correctly
Of all the nvidia rigs only one just works after applying straps, while mostly nvidia gpus are same with same memory in all of the rigs.


I dunno maybe i just don't know how to do things, but my amd rigs use straps (in bios) and work 100% stable. Some autotune for nvidia or something would be nice.

Maybe a way to know memory errors like hwinfo does on AMD? But AFAIK NVML does not report it for consumer cards, so some hack possible?

Please don't get me wrong - I'm glad that you are doing actual improvements after so long (timing straps for gddr5 nvidia)! Just as it is such a new thing for general public it may need more ... support?

Is there now any way to watch mem errors on nvidia? Because straps brought all the pain from amd tuning, but at least amd reports mem errors which helps a lot to make rigs stable. Now I get to wait 20+ hours just to see one gpu in rig fail and try again.

not a typo and i have replied to him, obviously amds are crap with about 1/2 efficiency per watt

in nvidia inspector you can go as low as you wish, 650mV is not as low as it will go


However after the straps I have problem on two rigs, one card is crashing on each, I have even tried settings like this
-strap 1,1,6,1,1,1,1,1,1,1,1,1 -sintensity 100,100,2,100,100,100,100,100,100,100,100,100
-strap 1,1,0,1,1,1,1,1,1,1,1,1 -sintensity 100,100,0,100,100,100,100,100,100,100,100,100
and 0 mem OC, and even disable undervolting
then cards now gets about 20MHS and still crashes, while before with ethminer or phoenix both cards could get 23 or 22mhs undervolted overclocked

Well no rig is truly stable now, but that is expected while finding new OC values, best I got was 22h58m uptime while before the straps months

Maybe just first dead GPU, but so far these things were unkillable.

nvidiainspector, but I think there is issue wit p2 states after claymore driver hack, I think undervolting no longer applies maybe, maybe p2 disable deactivation is why i'm having issue with cards even with -sintensity 0 and -strap 0 even when these same cards worked without issue before hack

I don't know why, but on some rigs I needed to disable force p2 state, on others not, and I have issue with one gpu on 2 rigs now, so maybe this?

EDIT: hmm so force p2 state for cuda is removed from new drivers, so that is not it

Hmm, i'm still having crashes with nvidia gpus even when using strap 0, these gpus worked before (using ethminer) even with much higher clocks that what i'm trying now.

Does sintensity apply to strap 0 too?

wtf, what strap are you using to get that?

I wonder if it is possible to destroy card using strap. 2 of my nvidias now crash even without straps, while before they worked for months with no issue ...

Ah, nvidia straps are gonna really be fun
I dunno why my gpu2 can't take any OC now, same memory as all the others, but that card always took less OC IIRC
Seems like I will need to do same card GUID to OC to Strap tool

O wanted to try new AMD version, I mined XMR for the long time now
I get this crash:


I have fairly recent amd drivers 19.4.2. I don't want to upgrade drivers as from my experience updates over RDP (and other remoting) usually fails and whole rig is unusable.

BTW since I did not mine eth for so long - what is ideal core clock and voltage for rx570 and rx580? Same as XMR 1150mhz@840mV or can I go lower?

I know it is not virtual memory issue, rig has 90+gb pagefile (for dual thread xmr mining)


Update: Log ends with
start building openCL program for GPU0
Which could be intel, but it seems that it skips intel.

BTW same thing with ethminer, phoenixminer works

So I have deployed new straps thing on about 1/2 of nvidia rigs and i'm able to get about 8% more in real, on pool hashrate, but also about 5% more power consumption to get there. Rigs i'm trying this on use almost exclusively micron memory cards.
Still some improvement is better then none, however in this case it will not ever pay for the downtime for resting each rig

Much more testing needs to be done, but I will share what I got so far
Micron 1060:
675mV, strap 1,100, mem +600 (on cards that can take it) - 300MHs@850w - so about 25MH/s on a good car, while remaining really restricted in power consumption (lower then PL70!!!).


Tomorrow I will try to find some samsung cards.

So let the finding of new ideal values for each memory type begin yet again.


Yes, seriously awesome work - I mean - it was so hard to believe I thought it was just trolling, I think that speaks for itself