Can't really blame them if they deliver more than was promised. It just seems weird that the Jalapeno only gets a ~29% increase, while the Single and Minirig both get 50%.

I've found plenty on ebay. Less than $4 is cheap enough for me.

I believe Catalyst 12.1 is the last version that's compatible with SDK 2.1. I tried 12.2, 12.3 and 12.4, and ran into the exact problem the OP described.

250MB per day would be enough for my 2GH/s rig. But do you mean that it costs £2 a day to run even if you stay within that 250MB limit?

I'm talking about the minimum memory clock that 7970 cards enforce, which is 150MHz less than current core clock. If you really found a way to get around that limitation with just cgminer, I'm sure many would be interested in how you did it. But it's more likely that you just think you are running at 300MHz memclock, while in reality the card picked something much higher when you requested it.

AFAIK, getting around this limitation is only possible with some Windows software that use the cards in undocumented ways. For example MSI Afterburner and Sapphire Trixx (both happen to be made by card manufacturers).

Attempting to downclock 7970 memory to 300 with cgminer is almost guaranteed to not work. Try setting it to 775, and see if the temps drop a little.

Fully understanding DGM is a tough task, but the main points are that it's fair, no hopping possible, and it usually has less variance than PPLNS (which is also fair, and easier to understand).

I don't think having those SatoshiDice transactions in your donation address transaction history helps getting more donations.

Problem solved! I noticed that even MSI Afterburner couldn't control the fans, even though it showed the rpm readings. That was easier to search on google. The card turned out to have a bios that blocked some sensor readings. I read about it in this thread: http://www.overclock.net/t/637836/xfx-5970-no-fan-control. The old and new bios versions were exactly same according to atiflash, but flashing the new bios solved this problem anyway.

I tried to create yet another mining rig with a new 5970 + 5870. The OS is Debian Wheezy, and I'm using Catalyst 11.12 + SDK 2.5, just like on my other Linux rigs. All the GPUs are detected in cgminer, and it's hashing as expected. The problem is that GPU0 (the first core of the 5970) doesn't have any temperature reading, and both cores of the 5970 don't have a fan speed reading. The 5870 has both readings. My other 5970 cards with the same Accelero Xtreme HSF display their temperatures and fan speeds just fine. The new 5970 is even the same XFX model as the card I bought before it, which works perfectly.

Strangely, cgminer doesn't detect the new 5970 as a dual GPU card, when checked with cgminer -n. Also, aticonfig --odgt can't display the temperature for the first adapter either, it only works for adapters 1 and 2. fglrxinfo lists all the cards, and I'll say again, all the GPUs are hashing. cgminer also states "hardware monitoring enabled" for all the GPUs.

cgminer -n output for the problematic rig: http://pastebin.com/UDUCvZKG
cgminer -n output from a rig that works fine, showing the dual GPU detection: http://pastebin.com/pSCV0mCe

I believe the "Failed to ADL_Overdrive5_FanSpeedInfo_Get" error is not connected to this issue, because it also happens on my other rigs with similar 5970 cards, while the fanspeeds are displayed anyway.

To rule my Linux setup out of the equation, I even tried to install Windows on the rig. MSI Afterburner shows the missing sensor readings, but cgminer still can't show them. Temperature for GPU0 and fanspeeds for GPU0&1 are still missing, and cgminer -n doesn't detect the 5970 as a dual GPU card.

So, all the GPUs are hashing normally on both Linux and Windows, but the only sensor reading cgminer can get out of the 5970 is the temperature of the second core. Anyone else ever heard of a "special" 5970 like this?

I already disassembled this one, but the third card was so elegantly fastened I had to keep the picture.



That card slot cover kept the top card firmly in place! I could take the rig in my hands and shake it in any orientation, and it wouldn't budge.

The ziptie on the middle card actually kept it as one piece. The fan shroud and some plastic shrapnel fell off when I took the card from another rig, but fixing that was easy!

Does anyone know if TeamViewer is faster than TightVNC on gigabit lan? I'm currently using TightVNC and it's awfully slow, no settings seem to help it much. Windows Remote Desktop was so fast on the lan that I sometimes forgot I was using a computer remotely. Too bad it doesn't work well when mining.

Just a wild guess: try disabling ULPS.

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4D36E968-E325-11CE-BFC1-08002BE10318}

There are numbered keys in this registry location, perhaps more keys than you have GPUs. Some of them might be leftovers from earlier driver installs. You can check from device manager what the registry keys for your cards are. Just double click on the GPU, select details, and then view Driver key from the list. After you've found the right registry keys, change EnableULPS under them from 1 to 0. Then reboot.

I'm getting 0.59MH/s per MHz from a 7970 that's also running the Windows desktop. Optimal yield should be 0.6, and I can get that with intensity 11, but it brings the CPU usage bug on Windows. You mentioned using intensity 6, try 9 instead. That's the max I can go without getting extra CPU usage.

I had this too, on 3 different rigs both Linux & Windows. With an invalid pool 0 it closes immediately after starting. On Linux I'm using latest git now, let's see if that helps.

What will be the difference in payout, after the fee hits, for those donating 2%, and those using PPS? Both schemes would then effectively mean a 5% fee with protection from orphan blocks. The only boon for DGM users would be merged mining, so it would be a bit more profitable, but how much?

Even now, those who donate 5% should get paid just as many Bitcoins as those who use PPS payments. Still, some people donate 5% and subject themselves to higher variance from the DGM payment scheme. Is merged mining the only reason for doing this?

After setting up all my miners, I noticed that one of them started with a very high fee balance compared to any of my other miners. The balance was over 2000 shares right after the miner showed up on the list. I'm just curious whether this is working as intended, since all the other miners (even including the obscured ones) have much lower fee balances.

I understood that the fee balance grows steadily by 1 share for every 20 shares a user has submitted to his own pool accounts. However, this miner that had a fee balance of over 2000 right away had already been mining for about a day and half at ~1400MH/s before I set up the port forward. I only restarted it afterwards to change my api settings and add the pools, but the old settings allowed read-only access too. Could Puppet Master have picked those 40k+ shares that I had submitted before actually using it, and calculated them on the fee? At least the math would add up. In that case, the initial fee would be uncalled for, and for some other users, this problem might be even worse.

Since cgminer constantly spawns new connections, I presume tor might route them through different IP addressess every time. Or does that only happen when the user requests it? I've only used tor on Windows, where you can press the "New Identity" button on the GUI. Either way, it's prone to being seen as multiple IP addressess. If you really want to mine using tor, you could use a service like Air VPN. You can connect to Air VPN using tor, and pay them in bitcoins.

https://airvpn.org/

Flash drives are very cheap, and usually mining rigs should be able to restart quickly. I doubt anyone has released something like this for bitcoin mining. If you really want to, you could do it yourself using tmpfs and switch_root / pivot_root. I tried it and it was pretty easy to hack together. There might also be some live CDs with the feature. If you find one, you can modify the contents to include a bitcoin miner.

Would there really be significantly more abuse if the limit was two addressess instead of one? I don't claim to know how the abusers are spreading work, but I know that people mine at home and at work, for example. Of course there will be people who legitimately mine at more than two locations, but the home / work locations are pretty common for obvious reasons.

I wouldn't be comfortable with running anything over 100C. I'd be surprised if any large scale miner allowed their cards to run so hot.