One miner stops randomly

[opiuman]

Hello everyone this is my first post so hopefully it's not too horrible. I'm having this problem where running dual 5870 I have one of the cards and it's always the same one sorta stops working after sometime.

Windows 7
Phoenix
SDK 2.4
5870@900/300 and tried at more default 900/900

Specifically what happens is that randomly phoenix still shows like 390 Mhash/s or whatever it was in the last good state but seems to do nothing further. (It seems to be stuck in the last state it was in before it crashes and doesnt update) and checking under GPU-Z the GPU is at 0% load doing nothing. I just kill that particular phoenix instance and run it again and it seems to work fine. It doesn't drop to 0 Mhash in the display or anything it just stays stuck.

Running with

-k phatk -a 6 DEVICE=1 VECTORS BFI_INT FASTLOOP=false AGGRESSION=11 WORKSIZE=128 -u [deepbit...] similarly for DEVICE=0 with the same phoenix instance

Furthermore, it ALWAYS DEVICE=1 (the same GPU that stops working). Temps are in range (75 at max load while running) so I'm unsure of what is going on here so any help would be appreciated. Thanks.

[1714144650]

1714144650

[1714144650]

1714144650

[1714144650]

1714144650

[1714144650]

1714144650

[1714144650]

1714144650

[badfrog]

Heyho,

I used Phoenix too, sometime ago, the Problem for me was, after a Time it said "Wirk Queue Empty, Miner is Idle" and stopped Working.
Which pissed me off, so i switched to POCLBM with phatk Kernel, which works great.
I Suggest you to try the Same, just try another miner. And see if it has the Same issue.

Greetings from Germany,
Frog

[opiuman]

Hmm I tried GUIMiner with same flags using both the OpenCL miner and Phoenix miner option and performance is just a lot worse. Using OpenCL I get almost 50 less Mhash and using Phoenix option I get about 20 less. Even using -k poclbm in the first one reduces performance by about 20 too. Not sure what's really going on here T_T

[badfrog]

Use "-k phatk AGGRESSION=11 FASTLOOP=false BFI_INT VECTORS WORKSIZE=128" for mining with a 5870.
you can use POCLBM.exe -k phatk too (this is what i use).
Parameters here : "poclbm.exe -d0 --platform=0 --host=xxx --port=xxx --user=xxx --pass=xxx -v -w 128"

Greetings,
Frog

[opiuman]

My poclbm doesnt seem to be able to take -k phatk (I might be misunderstanding what you want me to do).

Anyway running it with the flags you suggested still gives me 20-30 Mhash less than what I got before. However, I will say that the GPU that keeps dying DOES have a dummy plug in it that I made. I mean I made sure to push it on all the way and everything but that might be it.

[badfrog]

Well, just enable Crossfire to use both Cards, and you can remove the Dummy Plug

Do you have the latest version of poclbm ?
If not, i could share my poclbm with you, it has already the Improvements mentioned in the other Forum.
(Kernel optimizations)

Greetings, Frog

[opiuman]

Wait seriously? Just put the bridge in and enable it in CCC? Didn't know you could do that and I could've sworn someone suggested against it because they got a performance hit. And sure send me your poclbm through mediafire or something I downloaded the one off of https://en.bitcoin.it/wiki/Poclbm. Looks old though; I was about to check on the software board. Sorry I don't have a lot of experience with poclbm and I've been using phoenix+phatk but its just WEIRD why this happens. The miner doesn't say idle or anything but it just stops doing stuff.

EDIT: After a little more searching I found http://forum.bitcoin.org/index.php?topic=24816.0 which is similar to what happens with mine I think. I haven't seen it happen while I was there at the computer but I suspect that this is what is probably happening. Still though its only a 50 Mhz overclock from stock and the temperatures are still reasonable so I'm confused how this could happen. My crash happens maybe once every 12 hours and usually not frequently but still this thing is a pain in the ass already.

EDIT: Just looked through the Event logs and I do see warnings with code 4101 saying that essentially the display driver crashed and that it recovered. Odd that a 50 Mhz overclock is causing this.

[badfrog]

No  Problem, the Newbie Section is for things like that.

Well, i use both (My Server, and my Main PC) with Crossfire, and i cant talk about performance loss.
I have in main Main Both Cards 1000/1200 Crossfire Enabled  450mh/s per Card (1.274 On Water)
In Server , Both Cards 950/1200 Crossfire Enabled 425mh/s per Card (Stock Voltage On Air)

To Enable Crossfire, you have to attach a Screen while Starting (Or your Dummy Plug)
And then Connect Via TVNC to control it, this is what i do.
When Crossfire is enabled, you just unplug the Screen/Dummy.

I Uploaded the File here:
http://www.mediafire.com/?r16a4femd9x37ez

I also wrote a Small Installer (Installer.bat) , run it, and poclbm is at C:\poclbm,
and a batch File to start it is at your Desktop.
(You have to add your host/port/user and passwort to it first, to make it run)
But Pararmeters etc. are the best for 5870.

Some Cards are Better for Overclocking some not, i can hit my Cards unter Water over 1100mhz core.
But when i open any Site that contains Flash, it crashes

Hope i could help you a Bit.
Greetings, Froggy

[opiuman]

Thanks for all the help I just found it odd that one of the 5870s is crashing when they're both the same model/brand and everything. I'm using Hamachi/Logmein right now so that would be useful since I already have that setup. I'll probably try the CF method soon and see if it still crashes. How much does water cooling cost anyway (I'm curious).

[badfrog]

For my Main i paid:

 80 € for 280 Radiator
 30 € for 240 Radiator
120€ for 2 Aquastream XT Pumps
 80€ for EK Supreme CPU Cooler
160€ for 2 EK Fullcover 5870 GPU Blocks
~75€ for Tubes and nozzles

Makes around 550 € overall.
Which is alot Overpowered for my System so far
You can get a MidRange Watercooling solution for acounr 350-400€, When Having 2 Cards.

Greetings,
Frog

[BTCPOOLMINING]

Hi,

Most efficient way is to move the cards further apart by using a x1 PCI-E Cable to x16 PCI-E Cable, you simply can place these into the x16 slot or a X1 slot and they work great, especially for spacing the cards, they provide a 12 CM cable length. They are readily available on ebay. Usually no more than $8 USD Delivered.