Adding CUDA cores from Multiple Different Quadro Cards

I had Quadro FX 1800 … and I decided to go for an upgrade to a Quadro 4000 (2 GB Windows PC Version).
I didnt see much difference.
The new card was installed on a I7-990X , 24 GB Ram, asus Rampage III extreme, OCZ SSD 120 GB HDD, running windows 7 Ultimate 64 Bit
My question is :

I’ve heard that all (or most) Quadros use the same DRIVERS … so … what if… Could I Add my old Quadro FX1800 into another PCI slot next to Quadro 4000 ?
could that card exist in the same system/windows installation ? would the Cuda cores of both cards utilised as a global CUDA tank ?
would that increase performance by any bit ?

Thank you in advance.

Yes you can have two cards, but no CUDA won’t automagically use both devices as some sort of fused device. CUDA multi-gpu must be programmed explicitly at the application level, at the moment.

I CANT BELIEVE IT!! My Quadro FX1800 is FASTER (Clocks) than my Quadro 4000 (2.0 GB PC version)

!!! WHY ? … I added both cards in my machine. the nvidia control panel automatically chose FX1800 for PhysX calculations (dedicated) …

I also when I choose Hardware Mercury Engine renderer in Premiere Pro … the red line , wont turn to yellow !!! not even with 2 cards inside.

why is that? I’ve seen youtube videos , I’ve read reviews … and none of them had such problems… I have also installed the latest drivers. Could it be hardware problem ?

I paid so much more money to upgrade my quadro only to notice that the new quadro is actually Slower than my old one but just more memory. That’s all.

Any ideas ? Its not stellar… its not even High-End ! its not even mid-range like my FX1800… whats wrong with it ?

Its the exact opposite of what the reviews describe it to be.

What benchmark demonstrates that the new card is slower? A Quadro FX1800 has 64 CUDA cores clocked at 1.3 GHz with 38 GB/sec memory bandwidth, whereas a Quadro 4000 has 256 CUDA cores clocked at 0.95 GHz and 90 GB/sec memory bandwidth. The shader clock is lower for the Quadro 4000, but the number of CUDA cores is 4x larger. (And that is not even counting the difference in CUDA architecture between the two.)

CPUID (Graphics Card info) gave me huge difference between core clock, shader clock (the memory clock was faster on ddr5 quadro 4000)

Does Quadro 4000 with the new drivers supports now SLI Frame Rendering and SLI Antialiasing or I would better not repeat the error and not buy a second one for SLI configuration schemes ?

viewPerf 11.x Shows better results in most of the tests except the first two ones… And the thing is the card’s core temp is at 85 C !!! What is the Temp Limit ? ?? It is really silent though.

55 C when idle in windows.

I have two Quadro devices. I want to execute matrix multiplication program on the both devices simultaneously to improve the performance. How can I do this ?