Hi, what I'm doing wrong? Use the command line I put in the op post. Java on Windows requires you to use \ for path instead of /, and ; instead of : for path separators. No other OS is designed in such a braindead way to require that.
|
|
|
Vector width. I have a 4850, -w 64 -v 2 is the most optimum settings for GPUs like ours. m0mchil's doesn't seem to have a comparable option.
That makes it even slower. Desktop responsiveness stays the same. Hmm, strange. Should work on yours like it does mine.
|
|
|
Try running with -f 30; this will lag your desktop, but will reduce the overhead caused by OpenCL itself. You can go down to -f 2 safely if needed, it just slows down X.
58 Mhash/s at -f 2 after 15 minutes. May I suggest that you calculate the hash rate over an interval of a few seconds, rather than the entire running time? This would make testing much more convenient. Also, I am interested in a little justification for the "should be faster" part, especially if the OpenCL kernel remains unchanged I don't wanna do it like m0mchil does because it bounces around too much and gives a false impression of the speed. I probably should reset it after every 30 or 60 seconds or something.
|
|
|
Hrm, try my newest build. I wonder how badly the concurrency issues screwed the mhash meter.
The newest build does not give any noticeable improvement. I probably won't be testing this one very much, since the other implementation works well on my system, and I only have this one GPU for mining. After much thought, I'm not sure why you're having this issue, especially if m0's works well. Try running with -f 30; this will lag your desktop, but will reduce the overhead caused by OpenCL itself. You can go down to -f 2 safely if needed, it just slows down X.
|
|
|
Vector width. I have a 4850, -w 64 -v 2 is the most optimum settings for GPUs like ours. m0mchil's doesn't seem to have a comparable option.
|
|
|
Hmm, Windows 7 64-Bit, Radeon 4870 here.
m0mchil's poclbm is ~1000khashes/s faster and my desktop stays more responsive than with yours. I guess I'll stick to poclbm ;P
Try running my miner on your 4870 with -v 2 -w 64 and see who wins. Also, if you have responsiveness issues, try using -f 120... you shouldn't need it, but Windows sucks dick badly.
|
|
|
Hrm, try my newest build. I wonder how badly the concurrency issues screwed the mhash meter.
|
|
|
I've added the command line arguments for host and ip, and it now eats the exception when it can't connect to bitcoin so you can shut bitcoin off and the miner doesn't die. I've also fixed potential concurrency issues; it seems to have improved the mhash meter display.
|
|
|
Where can I enter the IP from RPC host? Thanks
Hmm, I seem to have forgotten something. I probably should add that.
|
|
|
On my Radeon HD 5570, the hashing rate is about 35 Mhash/s after 10 minutes. It seems to have increased slowly and ever slower, apparently it is calculated over the total running time. m0mchil's version gives 52 to 61 Mhash/s depending on settings.
What OS? And yes, my hash meter is an average of the whole run time, m0's isnt. Its not accurate until it stops going up, which might take an hour on some hardware, maybe more.
|
|
|
Does this miner have the 100% CPU usage problems that are present when using SDK 2.2 on linux?
With m0's on 2.2, I get around 50% usage for X, 50% usage for python (out of 200%) and the desktop is unusably slow. For mine, I get around 15% X usage, 75% usage for Java, and the desktop is smooth as silk.
|
|
|
Until Nvidia ships non-beta drivers that aren't ancient, it seems you can find them here.
|
|
|
|