Back To Listings RSS Print

Adobe CS5 and nVidia:  First Impressions

or..."Breaking the First Rule of NLE, Part 3"

By Bruce A Johnson | May 19, 2010

image

I guess once you break the big rules, you get used to it. You may remember my mini-series from last January, when I replaced a several-year-old Dell Pentium-D workstation with a fire-breathing HP Z-800 eight-core Xeon box. At that time, I installed my existing Matrox RT.X2 video accelerator card and an ATI Radeon HD4870 video card, to work in collaboration with Adobe Creative Suite CS4. The system ran pretty well, but it wasn't a month later that the news started leaking out about something big on the horizon - something called Mercury and CUDA, to be included in the new version of Adobe Creative Suite - CS5.

April brought my yearly trek to Las Vegas for the NAB Convention, and one of the first places I went to was the Adobe booth. The demos of of the Mercury engine running with the nVidia CUDA cards were incredibly impressive. I knew instantly I wanted to torture-test this combo. A few phone calls by the PVC brass brought to my door (eventually) an nVidia Quadro FX4800 video card and the Adobe CS5 Master Collection. As what seems to be the lone member of PVC that edits on a Windows box, I intend to use this combo for ongoing torture tests for PVC. But first I had to see if it could even be installed in a calm and controlled manner.

image

Ahhh, memories. Here are the guts of the HP Z800 with the Radeon HD4870 video card, and the Matrox RT.X2 nestled below it. The hardest part of removing these cards was un-installing the software and drivers.

image

The backplane of the nVidia Quadro FX4800 features something this old PC guy hasn't had to deal with before - two DisplayPort slots. My current monitors don't support DisplayPort, but a web-visit to Monoprice.com yielded a DisplayPort-to-HDMI adapter and a reasonably-priced HDMI cable. Of the two DisplayPort and one DVI plugs, only two can be active at once (and don't quote me, but I understand the Mac version has a different selection of plugs than the PC one.)

Page 1 of 2 pages 1 2 Next »

Editor's Choice
PVC Exclusive
From our Sponsors

Share This

Back To Listings RSS Print

Get articles like this in your inbox: Sign Up

Comments

Jim Hines: | May, 19, 2010

I’m excited - seriously - give us all the details - particurlarly as it might apply to dynamic link.

Chris Meyer: | May, 20, 2010

Hi, Bruce -

Just confirming that the Mac version has 2 dual-link DVI connectors instead of the DisplayPort plugs. Also, one of those mysterious s-video-like connectors not mentioned in any of the online docs I’ve seen so far; need to dig a little deeper on that one…

Although nowhere near as thorough as Bruce’s review, I’ll chime in that the Mac version is plug-and-play (no mounting difficulties), using the same internal power connector as the 8800 cards that came with our MacPros. Unexpected benefit: it’s quieter, especially on power-up (no power tool/dentist drill over-rev on first boot). OpenGL drivers are in the system (running 10.6.3); CUDA drivers must be downloaded from NVIDIA.

- Chris

Bruce A Johnson: | May, 20, 2010

I hear the funny-looking mini-DIN connector has something to do with…

...wait for it…

...3D playback.  We’ll see.

BAJ

scottieb: | May, 20, 2010

I couldn’t spring for the 4800 myself, so I got a GTX 285 for Mac, but I can confirm the same things as Chris said - 2xDVI, perfect fit, uses the same power cables, and seems to be quieter (although it does seem to give off more heat from the exhaust). Did a quick DSLR edit in Premiere and it was really quite smooth! Now… think we can get the canon import tool (like for FCP) for adobe? Timecode and metadata would be nice smile

Scott Gentry: | May, 20, 2010

I installed the NVIDIA card on my Mac using two of the newer Mac Monitors.  They use the Mini Displayport connections and the NVIDIA card for mac doesn’t support them directly.  Meaning, I too had to buy two powered adapter (approx. $300 total) arrgghhh.

I hear from NVIDIA that the next board will offer options to connect directly to the newer Mac monitors.

One additional problem I had, on my tower (newer Power Mac) the extra power cable needed to run the NVIDIA board is pretty well hidden.  You need a cable, but where it connects isn’t well documented and it took a while to find.  In the end, all works pretty well.

Frustrating experience however as nothing was clear in instructions, and online data was hard to come by.  Ultimately I found the plug by pulling my computer apart and searching…

scottieb: | May, 20, 2010

Ha yeah - I admit the power connectors on the MB were easy for me to find only because I already had my 8800GT plugged in the same place! The second one (needed for the beefier cards) was still a little tricky to get at - even though I knew exactly where it was (early 2008 Mac Pro).

David Williams: | May, 21, 2010

Nvidia must have paid a pretty penny to lock CS5 to CUDA instead of using OpenCL which would allow the use of both ATI and NV professional and consumer cards.

Is it also locked to NV Quadro cards only? They use the same chips as NV Geforce series, with a few bios changes, and four times the price.

I hope NV paid an awful lot to make up for locking out a huge potential user base.

wsmith: | May, 25, 2010

Thanks Bruce! I look forward to your further testing.

I’m very interested in knowing how much of the performance is owing to the Xeon CPUs as opposed to the Cuda GPU. I’ve heard that one really does need Xeon CPUs, as opposed to i7 CPUs to make Mercury truly perform as advertised - GPU horsepower notwithstanding.   

I recently learned that, at NAB Matrox has begun touting the virtues of its MXO2 devices to accelerate Mercury, thus obviating the need for Cuda.

I was watching the various videos from NAB from various quarters such as Fresh DV etc. The Matrox re would mention this re it’s MXO2 products and it seemed to be completely over the heads of the interviewers and was just acknowledged with a “wow!” or something and the subject was quickly changed.

I was on the phone to Matrox shortly after NAB to get info re compatibility of MXO2 devices with HP’s Dreamcolor monitor, which as readers of Allan Tepper on this site know, requires a true RGB signal for the Dreamcolor engine to work. Seems like, based on my own research, the Matrox MXO2 devices are the best solution for that.

So MXO2 accelerates Mercury without Cuda and allows monitoring on a very nice and cost efficient monitor.

Benefits of MXO2 over Nivida cards: Get a complete selection of efficient Matrox codecs. Get good monitoring and scaling even if not using a Dreamcolor. Faster than realtime h.264 encoding if you get the Maxx version. 

A problem as I see it: Matrox seems to have missed the fact that acquisition has gone tapeless. To get the benefit of those Matrox codecs, you must ingest a video stream through the device.

The devices do allow monitoring of anything can that be played from timeline that is based on a Matrox project preset. Thus if your tapeless source was not ingested as a stream, you can still monitor it by dropping it onto a Matrox preset project timeline. But you don’t get the benefit of the efficiency of a Matrox codec for improved responsiveness while editing.

Matrox would do well to release a transcoder to process tapeless, file-based, video sources.

I may be wrong but something tells me that a MXO2 device, accelerating Mercury in conjunction with an efficient Matrox codec may do as well or better than an i7/Cuda solution (not sure about outperforming Xeons though).

Add the monitoring the MXO2 affords and that solution is more cost efficient and compelling, in theory.

I wonder if you might pursue some testing along this line of reasoning. I’m especially interested in getting realistic performance on Adobe’s multicam function which Mercury also accelerates.

Thanks!

Video Production Phoenix: | January, 16, 2013

We’ve been using a 2.66ghz Quad Core Apple Mac Pro for our video production editing using CS5 with an NVIDIA graphics card for some time now and the finished products look amazing. We originally were using Final Cut Pro 7, and decided not to upgrade to Final Cut Pro X because of the interface change and the taking away of many good features. Nothing but Abode products in our studio now.

Please login or register to comment