Quote from: dogie on December 04, 2013, 05:57:28 AM

Quote from: hephaist0s on December 04, 2013, 05:14:23 AM

Quote from: hephaist0s on December 04, 2013, 12:37:54 AM

Quote from: dogie on December 03, 2013, 10:59:46 PM

Quote from: hephaist0s on December 03, 2013, 09:15:29 PM

I have 4 cubes running on 3 different power supplies, an Antec 1200, a Corsair CX750, and one on a Rocketfish 500W I had lying around. Right now, switching the clock to high makes no difference in the speed -- they all hover around 29.5GH/s, whether in high or low. When I set up the first one, it did overclock up into the mid 30's... then I got distracted setting up the rest, and never really noticed when everything changed. Any ideas at all? I'm not sure where to start on this one.

Potentially might be limited by proxy or network topology limiting hash rate.

Ok! That's kind of the direction I was going, too. I did just add a switch in order to increase ports. But I'm lousy with networking, any further thoughts? Is having a switch bad? But I needed more ports.. ugh.

While I still don't know the exact cause of the network-wide slowdown, I moved the proxy server to another machine on the network and that fixed the problem. The cubes are running in the high 30s again.

Yep, mysterious network topology Cheesy

. Sometimes its the switches playing bad and not liking the frequency they get spammed with getwork packets, regardless of how small.

afaik the network interface on blades and cubes is 10 MB. some network equipment may not deal with this properly. if it's a managed switch, try setting the port config to 10MB. however, a good switch should autosense to 10 MB. I do not know if they are half or full duplex.