Topic: Why is my 2nd card slower than my first card? (Read 1497 times)

Thanks - I know I've been doing this in a pretty casual way, and need to get more serious about the data. That is, if it doesn't take too much time. For my modest rig, there's definitely a point of greatly diminished return on the amount of time I spend tweaking my cards.

Just an fyi - plotting the data and using a linear model rather than just getting an average leads to more reliable results. You can do this using deepbit and microsoft excel if they're the only tools you have to hand.

I take it you have the agression level set very high on both cards (11+)? You might try lowering this on one or both cards and see if that helps. I currently have my levels at 7 for GPU 0 and 9 for GPU 1. I've noticed that when I up aggression to 9 on GPU 0, this just lowers the hash rate on GPU 1, and my overall hash rate goes down a little. This is with fastloop left on, which may be a factor (but I wasn't seeing stale shares). I've tried going to 11 (lol, Spinal Tap) on GPU 1 (with fastloop off), and while this increases my hash rate for this worker, it seems to completely choke the share submission rate for the worker. I'm not sure what's going on, maybe I'm just a newb who hasn't yet figured out how to properyly tweak my cards, but I've notice that a higher hash rate (that appears stable) doesn't necessarily lead to a higher share rate. Rather than focusing on hash rate, I've been watching the number of accepted shares produced by each GPU (say over the course of an hour) on different settings, and trying to tweak that way to maximize my share rate. Due to variance, though, an hour may not be enough for reliable data, so eventually I need to start looking at longer time periods. Again, stale shares on not an issue here, I'm just seeing a difference in the rate of submitted shares that doesn't correlate perfectly with hash rate. Also, I don't think temp is a factor, since both cards are reasonably cool (below 80), and running at similar temps. Even if something like this helps, though, it sounds like you have some other issue as well. 

Sounds like your cards are simply to close together.

You need atleast 1/4 inch or so between cards, otherwise one can't breathe.

What kind of power supply you have ?

I did try running 2 cards with a 550W. I was not able to push both of them to 99%. One of the card was going crazy from 60% to 90%. I changed the power supply for a good 850W and I can run both cards at 99% now.

So you don't have control of fan speed with your other, hotter 6970? I haven't looked at the other post.

91C is hot enough to throttle the 6970 (done that)

I say airflow/lack of fans or crossfire is enabled/plugged in.

i'd prolly just say airflow !

Motherboard? Case? Airflow?

I thank you for trying to help; however, I don't think that's the case, unfortunately. In another thread of mine, I had asked for help on controlling both cards' temperature (http://forum.bitcoin.org/index.php?topic=28359.0).

The fact is my 1st card (DEVICE 0) is at 77C (with the control of fan speed) and my 2nd card (DEVICE 1) is 91C.

Device 0 is the first card, most likely the one on top. I'm guessing its sucking in hot air from the bottom card and is now overheating and throttling.

what does DISPLAY=:0.1 do?

you're right. The copy and paste is when the miner is not working. I turned it off so I can copy and paste without lag. But when it's running, both GPU Loads are at 99%.

Also make sure you are starting your second miner on DISPLAY=:0.1 or whatnot.

It doesn't look like you're mining at all.

Both cards are 6970s.

My DEVICE 0 is running at 360MH/s and my DEVICE 1 is running at 200MH/s.

aticonfig --odgt shows both are the same.

Adapter 0 - AMD Radeon HD 6900 Series
                            Core (MHz)    Memory (MHz)
           Current Clocks :    250           150
             Current Peak :    880           1375
  Configurable Peak Range : [500-950]     [1375-1450]
                 GPU load :    0%

Adapter 1 - AMD Radeon HD 6900 Series
                            Core (MHz)    Memory (MHz)
           Current Clocks :    250           150
             Current Peak :    880           1375
  Configurable Peak Range : [500-950]     [1375-1450]
                 GPU load :    0%


PS. Using phoenix 1.50