Although I can't benefit as my nVidia card doesn't have the required RAM, something very odd happens with CS5. AVCHD plays without stuttering, but parts of the picture jump. Really weird. One clip today was a lighthouse, on a blue sky with white clouds. The lighthouse was stable, yet the clouds jigged up and down - downloading the latest drivers didn't help. Works fine on SD material. The jiggling is also evident when coming out to a DV device linked to a monitor. Very strange.

I was looking at the gts 250 as its cheap as chips but am wondering, given MPE uses the CUDA cores, is the fact that the 250 only has 128 a limiting factor? Cuda core count goes something like:

9800 GTX: 128

220: 48
240: 96
250: 128
260: 192
275: 240
280: 240
285: 240
295: 480 (240 per GPU as its a dual GPU card so I would assume adobe wont us both)

470: 448
480: 480

No offense to Randy but his system probably isnt the best measure to dismiss the 250, given the weird readings he is getting trying to find his card.

Would be interested if someone could throw some red footage on the time line and see how the different cards handle that with a few effects?

Otherwise it would seem that the 260 is the best bang for the buck as its half the price of the 285 and from all reports handles most anything you throw at it.

EDIT:

Found this link and a few others on google and it appears the 250 with 1gb memory can indeed work very well once you modify the list of cards

http://forums.adobe.com/message/2802886?tstart=0

So given the 250 is almost half the price again of the 260 - might be the way to go for now?

OK, just wondering... has anyone tried this "hack" on a mobile workstaton (laptop) with a cuda-capable mobile card? I'm looking at acquiring a mobile system with a Quadro FX 1800M video card. It has 1GB of memory and according to nVidia it has 72 CUDA cores. Would love to see a CS5 mobile platform with the MPE!

First of all, thanks so much for finding that out. Kudos to you! Unfortunately, I don´t really get what exactly the purpose is for step 6. Regarding step 5 => PP CS3 exists already in that menu and it doesn´t change when I pick the CS5.exe. It sticks to CS3. Why we have to change the settings of the Nvidia graphic card from ""multiple display performance mode" to ""compatibilty performance mode"? I had a system crash as I pushed the export button (not the queue button) and now I´m wondering if this problem has possibly something to do with these changes? Furthermore, after a while of working all of a sudden the audio transitions disappeared. Now I´ve set it back to ""multi-display/mixed-GPU acceleration" and until now I haven´t got any further problems. Could there be any relationship?

So, I would appreciate it if someone can explain these two different settings a bit more in detail...

Vista 64bit
GTX275

You may need to uninstall CS3 in order to enable CS5. I had the same problem with CS4 and CS5 and could not remove the existing profile and adding CS5 did not work, because it has the same file name.

An alternative is to get nHancer - Advanced Control Panel for nVidia Cards which allows you to identify the path to a file, so you can have both CS3 and CS5 in your profiles.

Thanks Harm. I gave it a go. Hopefully, I did everything right. But this question remains to be answered:
"Why we have to change the settings of the Nvidia graphic card from ""multiple display performance mode" to ""compatibilty performance mode"? " Thanks in advance...

Actually I don't know, but keep in mind that the multi display performance mode only applies to OpenGl, not to Direct3D. BTW, I have this on globally and have not yet encountered any problems.

The GTS 250 works fine - I just followed the steps in this topic and it's running fine.

Gt 240
 

The GT 240 (Zotac zone silent + Q8400 - 8gig ram) works oké.
Using my AVCHD files: no more rendering of cross-dissolves, titles etc.
With an extra AVCHD layer on top my cpu's go to 100%.
But the normal routine edits 60-80%. Using 5D2 testfiles was to much.

No more rendering is more creative freedom.

Let's try this....no it takes to long.

Now it's ...give it a shot


Rob

Has anyone confirmed that there is an advantage to having a Fermi 480 card versus a GTX 260-285 card? I know there are more cores so it "should" be more capable. However, if the Mercury PLayback Engine is programmed to only utilize certain features of a non Quadro card, is there any benefit to the 480?

I need to order a new card in the next 24 hours and I am on the fence. I do not want to order a 480 if a 260 will do the same job. I am only interested in how it speeds up MPE in CS5.....don't care a lot about games and such.

Thanks.

I can now confirm that the 470 works great with the MPE. At $350, this card is worth every penny!

Ben

I am sure it works great....but do you have any idea if it works better than a 285? I am trying to decide if the 470/480 is worth the extra money. Like I already said, it appears on the surface that double the cores would mean double the perfomance.....any chance this is accurate? I have had a good experience so far with the 260 and I need to upgrade a second computer and I just want to know if the 480 if far superior to the 260....otherwise I'll just get another 260 for now.

I think the GTX 285 is being discontinued, Not many left in stock any where. Sorry I can't campare the 470 to the 260 as I came from an 8800 card. Thats a tough call though with the $150 price difference. The 260 may work just as well. Maybe someone else can chime in that has used both cards.

Harm and I have run a beta PPBM CS5 and myH264 test was faster than his and my MPEG2 test was only a little slower than his. Harm has an i7 overclocked to 3.8GHz and a GTX 480. I have an i7 920 at stock clock and a 9800GT 1GB (112 cores).

Prior to the 9800GT, I tried a GTX 275, and it was definitely faster in encoding times. I ran my own test of 3 avchd layers and the 9800GT encoded it in 13mins and 275 was 8mins 40s.

EDIT: my 9800GT encode actually took 15mins