Author

Topic: Crash when underclocking memory (Read 1852 times)

legendary
Activity: 1344
Merit: 1004
April 26, 2012, 09:31:24 PM
#8
Hi,

I have two Asus 5850 DirectCU's. Using Afterburner and CCC 12.1.

AB can change core clock and memory only in GPU1. I can not do anything to GPU2.
With Trixx I can change the memory and core clock on both cards.

System crashes after 2minutes of mining. Without any OC'ing it is stable but MHash/W is so crappy I don't want to run it like that. Any solutions? Tried CrossFire too but same problem.

1000W PSU
GA 990GXA-UD3
GUI Miner

One of the 5850s has to be the problem. That same computer can run two 5870s without any problems.

Check ASIC quality on the one that is crapping out,  I bet it is below 65%.  

You must have a 7970 (or similar). Cant get ASIC quality of older GPUs (6xxx or lower).

To the OP, Sapphire Trixx 4.3.0 is out now, which now supports undervolting in addition to overvolting, as well as 7xxx over/underclocking. I don't see any reason to use afterburner anymore. Also there are "bad" memory frequencies where the timings to a certain frequency are too stressful to cause it not to work until it gets past a certain point them jumps to a different set of more stable timings. Try different increments of 5 mhz memory until you find one that works for you. For me, I found 220Mhz+ and 370MHz+ to be breaking points for both of my 5870's.
member
Activity: 84
Merit: 10
FPGA convert
April 19, 2012, 03:50:32 PM
#7
Oh yes, forgot that.

725/250, 725/300 with a range of different voltages from 1050mV to 1165mV.


I have 25 5850s ... best card out there.

Anyways. anything under 300 is gonna be hard for a 5850 on windows with AB.  Stop using all that other crap and just use AB.  Trixx sucks ass and CCC is stupid too.

1149 voltage, 950 core and 300 ram will get you a good 400 mhash.

If your PSU cannot handle that ... try 1112 voltage, 900 core and 300 ram.  

If you STILL have trouble ... just do 1087 voltage, 820 core and 300 ram.
zvs
legendary
Activity: 1680
Merit: 1000
https://web.archive.org/web/*/nogleg.com
April 08, 2012, 03:17:26 AM
#6
I have one 5870 that crashes instantly (when mining) if I try to set the memory below 250 or so.  My other two handle it fine (they run at 975/215).  Never figured it out.

Though the two at 975/215 require 1.12v to be 100% stable.... the one that I can't clock below 250 runs stable at 1000/300 at 1.04v... truly epic
sr. member
Activity: 466
Merit: 250
March 31, 2012, 11:49:04 AM
#5
725/280 seems to be working in GPU2 and GPU1 runs smoothly with whatever I put there. Well, I consider this solved.
full member
Activity: 210
Merit: 100
March 31, 2012, 09:17:15 AM
#4
I've seen some particularly bad core-memory clock combinations result in hardware errors and/or instability on a few of my cards.
Try some other memory clocks, doesn't have to be very far from you current setup, say 725/280 or 725/340.
sr. member
Activity: 466
Merit: 250
March 30, 2012, 05:15:06 PM
#3
Oh yes, forgot that.

725/250, 725/300 with a range of different voltages from 1050mV to 1165mV.
full member
Activity: 944
Merit: 101
PredX - AI-Powered Prediction Market
March 30, 2012, 04:31:41 PM
#2
Don't have 5850's myself but at the least, you'll have to give us what core and memory clocks and voltage settings you're trying to use before anyone can give you more help.
sr. member
Activity: 466
Merit: 250
March 30, 2012, 03:17:51 PM
#1
Hi,

I have two Asus 5850 DirectCU's. Using Afterburner and CCC 12.1.

AB can change core clock and memory only in GPU1. I can not do anything to GPU2.
With Trixx I can change the memory and core clock on both cards.

System crashes after 2minutes of mining. Without any OC'ing it is stable but MHash/W is so crappy I don't want to run it like that. Any solutions? Tried CrossFire too but same problem.

1000W PSU
GA 990GXA-UD3
GUI Miner

One of the 5850s has to be the problem. That same computer can run two 5870s without any problems.
Jump to: