Author

Topic: High Discard Rate (Read 3885 times)

member
Activity: 98
Merit: 10
May 02, 2013, 09:19:35 PM
#8
I posted on a separate thread, but I think it fits in with your OP. So I use BFGMiner on stratum for BitMinter mining, and have two workers:

Worker A = 260 MH/s
Gets about 68% discarded work
A:3192, R:0, S:6
ST:2, DW:7059, GW:1867, LW:17310

Worker B = 1800 MH/s
Gets about 24% discarded work
A:11414, R:8, S:0
ST:2, DW:3677, GW:924, LW:22936

Any idea why my slower worker (A) seems less efficient and seems to have more discarded work? How I can reduce this to mine more effectively?
newbie
Activity: 13
Merit: 0
April 25, 2013, 09:53:39 PM
#7
I'm seeing this same issue with a 5770 mining on slush's pool.  I haven't noticed it on other pools but I haven't tested the other pools enough to be sure.  It doesn't seem to be affecting hash rate or payout, i'm getting what calculators say I should be getting.  Pulling about 200-214MH/s.

Discards: 249 197.6%
Accepts: 253 98.8%
Rejects: 3 1.2%
H/W Errors: 0
Stales: 0
Get Fails: 0
Rem Fails: 0
newbie
Activity: 21
Merit: 0
April 15, 2013, 07:46:37 PM
#6
Thanks for the replies. I think the card is getting plenty of power, although I may check that as a last resort.

I've played with the thread concurrency a bit:
8000 didn't make any difference.
4096 I get HW Errors.
Any higher than 8192 and I get black screen and the card crashes.

I did the setx stuff.

I'm trying yet another pool now to see if that will make any difference. Will report back.

Update: The other pool seems to be having server issues(wemineltc), so not good for testing right now. I'm still at a loss.
legendary
Activity: 1358
Merit: 1000
April 15, 2013, 05:26:59 PM
#5
thread-concurrency set wrong?
newbie
Activity: 12
Merit: 0
April 15, 2013, 04:31:28 PM
#4
The most likely explanation I can think of is that you aren't giving enough power to the video card. 600W is a lot but when you are running the card 100% then it draws much more than normally. Consider upgrading the PSU.

One thing you can do is take the video card out of the PC and put it in another PC to try to correlate if it's the graphics card itself that's the problem.
newbie
Activity: 21
Merit: 0
April 15, 2013, 09:56:00 AM
#3
That is not normal.

What is your OS, CPU, video card, video driver version and how much Watt does your PSU output? How hot does the GPU get and how are you cooling it?

Thanks for your reply. Here is more info:

OS: Windows 7 32bit
CPU: E8400 @ 3GHz
Vid Card: XFX 7850 DD 1GB
PSU:OCZ 600w ModXtreme Pro
Vid Driver: 13.1 catalyst
GPU Temp: 60 C at full load
Core Clock: 1045
Mem Clock: 1310
GPU Usage:99%
No Overvolting
Stock XFX Double D Cooler (2 fans, copper pipes etc)
newbie
Activity: 12
Merit: 0
April 15, 2013, 09:47:07 AM
#2
That is not normal.

What is your OS, CPU, video card, video driver version and how much Watt does your PSU output? How hot does the GPU get and how are you cooling it?
newbie
Activity: 21
Merit: 0
April 15, 2013, 09:39:07 AM
#1
I'm using CGWatcher 1.1.2 with cgminer 2.11.4 and I noticed that alot of my work is being thrown away. I have about 2x as many "Discarded" as "Accepted". I assume the discards are caused by the GPU not getting the work done fast enough before the server moves to the next block.

Here are some details:
Code:
Video Card: Radeon HD 7850 DD
Hash Rate(Litecoin) : 363 Kh/s
Accepted...............1377
Rejected...............91
Hardware Errors........0
Stale...0
Discarded...2382
worksize: 256
thread concurrency:8192
intensity:15-17 depending on desktop usage
Efficiency: 115%
Utility(shares/min):2.53

Mining Statistics:
Mining duration............... 9 hrs, 12 min, 55 sec              
Current accepted work rate.... 0.042/sec (2.530/min)
Current discarded work rate... 0.073/sec (4.359/min)
Current rejected work rate.... 0.003/sec (0.166/min)
Current stale work rate....... 0.000/sec (0.000/min)

Network statistics:
Times Sent.................... 1,495
Bytes Sent.................... 169,526
Times Received................ 2,702
Bytes Received................ 830,288
Net Bytes Sent................ 169,526
Net Bytes Received............ 830,288

I've tried a couple different pools, but still high discards.

Is this normal? Is there any way to minimize the number of discards? Would worksize make a difference? Is it a network latency issue? Anything suggestions for what I can tweak to improve this would be appreciated.

Edit Update: I switched to another pool and discard rate has dropped from 200% down to 60%, so I think a large part of the problem was the pool.

Edit Update 2: Saw over in the cgminer thread that discards are "meaningless" which is why it was removed from the cgminer screen. So I guess there isn't anything that can be done about it. I think its related to hash rate - seems the lower the hash the more discards. This would make sense because the miner can't keep up with the pool. The pool moves on to a new block while the miner is still working on the old one, so the work gets tossed, if I understand things correctly.
Jump to: