Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 860. (Read 2347659 times)

But only if the 2 extra 4 pin molex'es are connected to the psu.

You can run ccminer with the -e switch to turn it off.

I don't know why you are bringing up connector types and power pins because it has nothing to do with
what I'm saying. I'm talking about the aggregate power through the PCIe bus of the motherboard.

A quote from you in a previous post:

"The difference is that the BTC MB can provide 75W to ALL the slots together."

I agree entirely. Then I said:

"I wouldn't be concerned with drawing 75W from an x1 slot if the x16's weren't already fully loaded.
200W is the max draw I would put on the PCIe bus unless the MB is specifically designed for higher."

How does that conflict with your statement?

does anyone known what this is extranonce.subscribe and how to get rid of it

I assume you are reading it wrong. It says that the card can request 75W regardless of the type, even though it should be a graphic card; no mentions of the connector type.
Furthermore, the powered pins are all on the 1x part of the connector.

I don't have the actual spec but this seems to say otherwise...

https://en.wikipedia.org/wiki/PCI_Express#Power

I wouldn't be concerned with drawing 75W from an x1 slot if the x16's weren't already fully loaded.
200W is the max draw I would put on the PCIe bus unless the MB is specifically designed for higher.

And I add: all mobos can supply 1 or 2 pcie slots with 75W. Better one maybe 3, and that is why Pro BTC have molex on it: to give 75W to all 6 ( or7) pcie...

also Crysx... if you use both mobo mollex nothing will happen. Mobo already have its 12v enough ...

All the motherboards can provide 75W thru the 1x slots, it's by specifications. The rest of the pins do not provide any power.
The difference is that the BTC MB can provide 75W to ALL the slots together.

Powered ribbon risers or just ribbon risers? I use power ribbon risers and they don't cause any problems. I also have some USB risers and it doesn't seem to change much.

It's extremely hard to clip the 12v to the board on a USB riser as well and most feed power from the board in addition to their connector.


It's not like there is a limitation on the amount of current carried across a wire. There is a limitation where the wire will carry so much current it will start melting, but unless it's built into the card, it's entirely possible to deliver 200w over a 8pin vs 2x6pin. Just because a 8pin is rated for so many watts in spec does not mean it will necessarily be limited to that in practice.

The powered riser I have doesn't have the slot power connected so it would be of no use to have extra MB power available.
I presume most, if not all, powered risers are the same.

The main benefit of the BTC MBs is the ability to supply 75W to 1x slots which are normallly limited to 15-25W, which makes
the use of unpowered risers possible.

That's real close, even if the direction of the airflow on the cards is favorable.  I'm using 8cm gaps even between 750 Ti's which might be overkill but at least they don't throttle (65°C temp target) even in the summer at around 80% max fan speed curve.

having the powered risers AND the molex connectors on the motherboard at the same time? ...

all the cards have a 2.5cm gap between them on the aluminium rail i built for them in the unit ... however - that is changing to a more compact designed frame with rapid airflow in a closed system ( open initially for testing ) ...

#crysx

Feel free to do it, BTC mobo is working nice with 6 cards ( 960 ) and all 6 non-powered risers ( 2 molex on mobo fitted ) for me. I have 2 of them, other is with 5 cards + 1 on USB3.0 because had to be placed further for better cooling ( try to have some 8-10 cm between cards )
And any risers need maintenance - cleaning from time to time, that could save you some cards

i guess thats the same sort of point that we are both making sp ...

as long as the cards get the power they need - from which ever source - they will run stable ...

non-powered risers will not have the stability nor the durability ( in my experience with the ribbon cable risers ) unless the there is adequate power getting to the cards without drawing too much from the power pin off the pcie connector - hence the two molex connectors ON the motherboard ...

my experience with these ribbon risers have been atrocious ... killing quite a few cards ( the amd ones ) in the process due to the lack of adequate power for the amd cards through the non-powered ribbon riser ... BUT - i would have guessed this could have been due to a number of reasons - one of which was that there wasnt that extra power input ( like the two motherboard molex connectors ) ...

i decided to run the other way - to be much safer ... there is no need to supply the motherboard with extra molex connectors AS LONG AS the risers you use are powered risers ... i didnt want to take any risks with the nvidia cards due to the failures on the amd cards ... so i purchased high quality usb 3.0 powered risers - and have used them for a long time also with out issue ...

and in a farm environment - heat plays a massive part in it too ... the hotter the environment - the harder the cards have to work - the more power they use - the more they draw ... ribbon risers from what i have seen - are not up to the job ... burnouts and shorts - as well as melting - just made for a mess of thefarm when i had them in there ... replacing them with powered risers was the best decision ( though it did cost a bit - i admit ) ...

i wonder what would happen if the powered risers had power connected AND the 2 molex connections were connected at the same time? ...

theoretically - it shouldnt matter as the system will only draw what its needs ... practically - im too chicken to try ... i really dont want to find that my theory is way off and destroy 6 cards in one hit ...

#crysx

But I use the same setup with non power risers and with the two extra 4pin molex connectors connected to the mother board. Been running stable since mar-2014 (H61 BTC) windows 7. Some of my rigs have older drivers, and it seems to help in the lyra2v2 and quark algos. +100-200KHASH per 750ti card.

very nice

the power doesnt need to be ALL motherboard provided ...

ALL of the non-powered gigabyte 750ti oc cards run from the 1x pcie usb 3.0 powered risers ( 6 of them on a motherboard ) ... they are powered not only by the single pin on the 1x pcie connector from the motherboard - but also the power connector on the riser itself ... so they don't require any more power ( or draw any more power ) from the motherboard itself ...

the risers are getting their extra power directly from the psu using sata power connectors on a separate rail ...

i have this type of setup running on both the asrock h81 motherboard ( without the two extra 4pin molex connectors supplying the motherboard ) as well as 5 x pcie gigabyte motherboards ( that have NO extra motherboard connectors ) ...

and they ALL run really well - with throttling only happening when they overheat ... and im fixing that issue also ...

#crysx

Let's put it simpler.
You can draw more power if you have more pins, thus you have more chance of reaching a certain frequency.

NVIDIA cards will trottle and downlock if you don't have enough power. (If the power exceeds the TDP)

6 non powered 750 ti's on a single motherboard is not a problem as long as the motherboard is equipped with extra power. Like the h61 and h81 asrock BTC boards. Default tdp of the oldest non powered 750ti cards is 38w (in the bios) This can easily be checked with the NVIDIA inspector tool.

for once I agree with you    the power is drawn (?) from the card, meaning it doesn't depends of the number of pins (however if there are not enough pins, the card won't be able to get what it needs to work... and/or the pins will try to carry a higher intensity than the one designed for...