No problem, I'll see what I can see set up. Do you have a preference in Linux distro? ;-)
Thanks, ubuntu or debian are the likely culprits. |
|
|
|
Interested in remote access to a 6990 rig?
A quick prod would not hurt, but I expect I'll find the same thing. Of course I can only work with linux :s |
|
|
|
|
I've just committed the first patch to the git tree that should enable 79xx cards to work properly under cgminer (delete any dodgy .bins created before this attempt). I have no idea if it performs even remotely well, but in theory at least, it won't behave stupidly as the current version does. If someone does run it, could you please dump the output of -D -T the first time you run it? Also it may be worth starting it with -v 1, and/or experiment with various combinations of -v and -w (default being -v 2 -w 128 if you set nothing).
|
|
|
|
7970 definitely needs >= 11.12 and sdk 2.6
Oh yes and the card is still as yet unsupported so xorg needs to be manually configured to allow an unsupported device. |
|
|
|
ckolivas , You have been busy. will be sending btc's your way also:
as for the 5970 fan issue, what I have done is set a minimum for fans ie: 50-85 that way if the first gpu is too cool, the fan will at least go 50 to keep the other cool.
as for the 79xx. do you want me to give you remote access ? only problem is that most of my rigs are linuxcoin, or windows 7
Thanks. Yes there are definitely ways to keep the 5970 safe but it relies on user input. Remote access would be great, provided it's some linux distribution that I can still install stuff into so that I may build software on it... |
|
|
|
|
Just an update to tell you where the next version is up to. It's currently awaiting proper BFL support on windows, which Luke-jr is fighting with at the moment. While I'm waiting, I've been given access to a 5970 remotely and trying to find some way to massage the fan to work for both GPUs. So far it's been a miserable failure since everything simply shows the devices as discrete devices. Those willing to try a git tarball or build from git, the current tree is quite stable. 79xx GCN support is non-existent as of yet. Here is the changelog so far:
- Add API commands: config, switchpool, gpu settings, save
- Implement socks4 proxy support.
- Fix send() for JSON strings
- Introduce a --net-delay option which guarantees at least 250ms between any
networking requests to not overload slow routers.
- Generalise locking init code.
- Allow invalid values to be in the configuration file, just skipping over them
provided the rest of the file is valid JSON. This will allow older configurat
- Allow CPU mining explicitly enable only if other mining support is built in.
- BitForce FPGA support
- Configure out building and support of all CPU mining code unless
--enable-cpumining is enabled.
- Allow parsed values to be zero which will allow 0 values in the config file to
work.
- Advertise that we can make our own midstate, so the pool can skip generating
it for us
- Refactor the CPU scanhash_* functions to use a common API. Fixes bugs.
- Don't consider a pool lagging if a request has only just been filed. This
should decrease the false positives for "pool not providing work fast enough".
- Invalidating work after longpoll made hash_pop return no work giving a false
positive for dead pool. Rework hash_pop to retry while finds no staged work u
- Remove TCP_NODELAY from curl options as many small packets may be contributing
to network overload, when --net-delay is enabled.
- Refactor miner_thread to be common code for any kind of device
- Simplify submit_nonce loop and avoid potentially missing FOUND - 1 entry.
Reported by Luke-Jr.
- Micro-optimisation in sha256_sse2 code courtesy of Guido Ascioti
- Refactor to abstract device-specific code
Meanwhile, I still don't have my own rig back awaiting the mobo replacement... sigh.
|
|
|
|
|
Heh yes it is. Great case for when it all must be in a case. The cards dropped a good 15 degrees or 25% fanspeed with those half-pegs.
|
|
|
|
Donating is tough to do. [2012-01-23 14:41:11] LONGPOLL detected new block on network, waiting on fresh work
[2012-01-23 14:41:12] Rejected 00000000.6f296009.749ead13 GPU 3 thread 6 pool 0
[2012-01-23 14:41:14] Rejected 00000000.be045aa1.b8a81aee GPU 4 thread 8 donate
cgminer 2.1.2 compiled from github version as of about 3.5 hours before this post. Heh well thanks for enabling donate. This run was extraordinarily unlucky because you hit the donate threshhold just as the longpoll hit and block changed. I had to make the donate affect your mining as little as possible which means I also had to tolerate more rejects by not dirtying the "block database" in cgminer with those from the donation pool. It won't affect your own mining reject rate, but I guess doesn't look as nice in your local summary. In summary, I'm willing to wear the missed good shares if you're happy to ignore the rejects  |
|
|
|
[EDIT]
After some trial and error (mostly ERROR...read: LOCKUPS...lol)... I was able to get my 6950's running at 900 Mhz GPU & 900Mhz Mem (stock speed being 1250Mhz Mem).
Anything lower than that seemed to lockup the machine while starting CGMiner. My 4 6950's are non reference (2x Asus DCUII 1GB @ 810/1250 stock...& 2x XfX 1GB 6950's @ 800/1250Mhz stock). The immediate drop in temps is a welcomed change.
If your engine is stable at 900, I recommend setting the memory 125 below that as well for even cooler performance, so 775. It should still be stable. Even further, try the auto-gpu option with gpu-memdiff set to -125 so that if the engine clock speed is ever lowered, the memory clock speed is lowered with it. |
|
|
|
So.. you're saying a newer version of atiflash is all I need now?
I believe so. Took care of it for me. Thanks... whenever my freaking motherboard is replaced I'll see about it |
|
|
|
Having done this sort of drudgery in the past I appreciate what you went through to get it working in Windows. It shows true devotion to your project.
Not at all the response I was expecting, thanks! |
|
|
|
|
No doubt building on windows is nothing short of a pain. I can't even remember any more what I had to install manually but I'm pretty sure what's listed in the README covers all the essentials. Both binaries and dev libraries need to be installed and I had to manually shift some header files into the include directory to get everything recognised. I think installing pkg-config was the most troublesome, but even that I'm not sure of any more since I just kept randomly installing shit till the mother worked.
In short, windows is very much the 2nd class citizen for cgminer. The fact I can provide binaries for it at all is a miracle and I'd rather do without it myself.
|
|
|
|
Hashrate of all my cards (5830, 5870, 5970, 6870) and rigs are exacly 10% slower after version 2.0.8. So, is the hashrate really lower or hashrate calculation has been fixed?
Cgminer's hashrate is unchanged. You upgraded your drivers. It is sheer coincidence. See all the discussions about SDK 2.6... |
|
|
|
So when you build for Windows do you build under Linux or Windows?
In a windows virtual machine running XP. |
|
|
|
|
When everything is installed properly, you shouldn't need to be hacking configure or makefile scripts for it to build on mingw32. On the other hand, I never build it from git or autoreconf. I always grab a tarball (make dist-bzip2) and use that to build the windows version from.
|
|
|
|
Sapphire reference cards. They most definitely need a hardware modification to allow setting memory clock lower. There were some links showing how to do so with some wire and solder and shit and I decided against it <_<
Yeah, the link you're referring to about modification was to override the ROM write protection. It's actually not enabled on any of the ROMs I've encountered (including the recent Sapphire reference 6970 boards), but the ATIFlash utility couldn't send the correct commands to unlock the ROM to erase it if it was flashed in a particular way. For example, I had 6 identical XFX 6970s that I had purchased new & sealed from Newegg. I was able to flash 5 of them successfully but the 6th appeared write locked (which required disassembly and a 1 wire solder). To make things even more odd was that I could no longer flash any of the 5 cards after they had been flashed once. I understand why it does it now, but I was originally baffled. Now that ATIFlash has been fixed, I can flash any of the cards without disassembly. Think of it this way -- the only way you wouldn't need a hardware modification to set memory clock lower is if they stripped the variable clock generator off of the board. If they did this though, none of ATI's power management features would work properly (the cards scale the clock back when the card is idle or under low load). All of the cards would be sucking down significant power at idle. So.. you're saying a newer version of atiflash is all I need now? |
|
|
|
|
Sapphire reference cards. They most definitely need a hardware modification to allow setting memory clock lower. There were some links showing how to do so with some wire and solder and shit and I decided against it <_<
You're most welcome and thanks for donations =)
|
|
|
|
Avast fail: it report cgminer as a virus. And i don't find a way to tell it to ignore it. Avast double fail.
Just about every virus checker now considers cgminer as a virus. One of the advantages of me removing CPU mining from cgminer is that the false positives should decrease as most trojans for botnets that package cgminer as the "helper" for their evil activities do so because of the CPU mining. Anyway I keep saying any serious miner should be using linux, but I do understand why people are addicted to McDonalds, KFC and Windows. |
|
|
|
I'm also setting up a 6 x 6970 watercooled rig and hope to get some good hash out of it. Can you down clock the memory as easily as the 5 series card in something like cgminer?
What kind of overclocks and hash rates can I expect? What kind of power consumption?
thanks.
My 6970s ran stable between 965 and 1015 engine clock speed depending on the card. Unfortunately, they had that factory limitation of only downclocking the memory 125 Mhz lower than the engine speed. I tried flashing the BIOS but the last few generations are hardware limited, not software so it had no effect (this is why I put the gpu-memdiff option in cgminer). With +20% powertune and engine speeds 15 below mentioned above (I wanted stability), I was getting 400-430 Mh/s. It has been estimated each card run like that would be 280-290 Watts. I ran 4 of them with a (very good) 1250 Watt PSU and no problems. |
|
|
|
|
Show Posts
