HeavyHemi

beekermartin
That's what I figured.  Thanks for the link.  
 
I will test it both ways running Heaven and 3dmark11 benchmarks to monitor temps and performance.  I know Furmark is not a real world test for temps but I will give that a shot as well.  If the temps are significantly lower using slots 1 and 5 and the performance difference is really only 1-3% then that is what I will do.  When I had the Cyclone 460s the top card was over 20c hotter while running Furmark with them in slots 1 and 3.  With them in slots 1 and 5 the top card was only a few degrees warmer.  Which is normal of course because it is the primary card.  
 
The Gigabyte cooler is much better than the Cyclone's so I am curious to see the results.  They are also much more powerful but from what I read in the link it still won't effect performance much at 8x.  

I'm running TITANS in SLI with as Asus Phoebus in the middle PCIe x16 slot. Our processors really can't push the GPU hard enough that x8@2.0 really makes that much difference. My 3DMark and Heaven scores were within a percentage point or so regardless of which slot I had them in.

beekermartin

HeavyHemi

beekermartin
That's what I figured.  Thanks for the link.  
 
I will test it both ways running Heaven and 3dmark11 benchmarks to monitor temps and performance.  I know Furmark is not a real world test for temps but I will give that a shot as well.  If the temps are significantly lower using slots 1 and 5 and the performance difference is really only 1-3% then that is what I will do.  When I had the Cyclone 460s the top card was over 20c hotter while running Furmark with them in slots 1 and 3.  With them in slots 1 and 5 the top card was only a few degrees warmer.  Which is normal of course because it is the primary card.  
 
The Gigabyte cooler is much better than the Cyclone's so I am curious to see the results.  They are also much more powerful but from what I read in the link it still won't effect performance much at 8x.  

I'm running TITANS in SLI with as Asus Phoebus in the middle PCIe x16 slot. Our processors really can't push the GPU hard enough that x8@2.0 really makes that much difference. My 3DMark and Heaven scores were within a percentage point or so regardless of which slot I had them in.

Are you running two Titan cards in quad SLI?  If so and you didn't notice a difference running them in slots 1 and 5 at 16x/8x then I am sure I won't with two GTX 670s.  

beekermartin
My current card doesn't boost to 1202 at stock speeds.  I have to overclock it to get that high.  How do I overclock two cards differently?  I assumed that I would have to set my overclock to run at the lowest of the two cards.   So if my current card overclocks to 1202 and my new card can do 1250 then I would set the overclock to 1202 so I don't pass that level on either card.  If my new card will only do 1150 then I will have to lower my overclock to that level.  
 
I could be wrong of course.  

beekermartin
So I guess what I need to find out is what both my cards kepler boost is.  I imagine they will be different.  Once I know that and what the max stable clock is of both cards I will then set the GPU offset to an amount that doesn't allow either card to pass its max stable level.  

beekermartin
So the new card I received runs at slightly higher boost clocks, stable @1250 mhz versus 1202 max of current card, but the temps are @10 degrees hotter.  The problem is that if I set the GPU clock to +105 mhz, which is what my original card needs to hit 1202 mhz, it will push my new card beyond 1250 mhz and it will crash immediately.  
 
So what do I do to max both cards clocks to their highest stable overclock.  I can't just plug in +105 core, like for my original card, since my new/second card can't handle that since it pushes it past 1250mhz and is unstable.  I was able to run with the new card alone at @1240 mhz at +35mhz core and it was stable.  Once it breaks 1250 it crashes.