Re: [ANN] cudaMiner - a new litecoin mining application [Windows/Linux]
So far, thats the only explanation i can come up with. i tried both the ribbon and usb 1x to 16x powered risers, both ~30cm in length. the board im using (biostar h81S2) has one 16x slot and 5x 1x slots. i used the 16x slot when testing the single card. Has anyone else noticed reduced performance with risers tested on a SINGLE card?? Currently adding one card at a time, and seeing how they scale. just finished adding the 4th, and so far, they are scaling well. 255-258khs on bone stock settings. not sure if i'll get through the 6 card tonight. gotta work early in the am..
Yup, just tested two different cards individually on risers to a 4x PCI-E. I lose about 6.5% performance for whichever card is on the riser. I also tested and saw that using power to the risers is pointless and is actually a waste of electricity. The cards don't need additional power and unpowered risers will do just fine. Plugging the riser's molex in actually increases power consumption by 10-15w for no hashing gains.got all six going, all at stock settings. they scaled pretty much perfectly, WHEN compared to a single card that is also on a riser cable. I am convinced the reduced performance is due to the riser, given the card is completely dependent on it for power (no pci-e). i wonder what the maximum wattage these risers are rated at.
will test some more tomorrow.. i have the gigabyte cards with the pci-e plugs too. maybe after replicating this with those cards, can determine if the risers are the culprit. heres a pic of the testing setup right now, and a screenshot of em running.. http://imgur.com/61i1zO4
http://imgur.com/nTC8UWv