Modded 8 units with the single bridge trick today. All runniung stable @ 950mhz and nearly zero HW errors after a couple hours. Pool (ghash.io) reports exactly the same 400khash as cgminer.
Tried 1000mhz on one of them and although CGminer said zero HW errors and 425khash, ghash.io reported 30(!!)khash over a 1hr average. Pretty clear that 1000mhz does not work. At least not the way i'm doing it.
Looking forward to getting ahold of that modified fork of cgminer so I can have a little more granular control over freq setting and trying 950mhz and 983mhz.
Cool, but.....
how often is 'nearly' zero hw errors? I saw red nonce's for the first time 'since I look often', in days and days of running at 850MHz. Flooky!
Do you mean that kind of nearly zero? If so, I'd be impressed with that. If not....ehhhhh.......
Zero hw errors at 1000MHz does interest me though.
I've been contemplating this phenomenon for a few hours now. Besides the obvious, It occurred to me that this too could be due to comm port overflow errors. Taken care of or at least greatly buffered by increasing the number of port calls made during full flow traffic i.e. upstream/downstream.
The overflow errors can build up sufficiently to only allow a trickle of completed work to be reported hence the lower results pool side.
Try lowering the port FIFO's a bit. See what happens. Might handle it, might not. I've had success doing this mod for similar reasons at lower clock speeds.
I have all of my comm ports FIFO buffers set at 'Recieve buffer (
' - 'Transmit buffer (14)'. No loss in number of reported shares pool side too
.
Give it a whirl! See what happens.
Let me know what happens. K? Cool!
Keep your fingers and toes crossed! 1000MHz stable here we come? Let's see!
Wolfey2014
