-
Quadro FX baseline performance
Has anyone found a baseline performance test of the Quadro FX 3800, 4800, and 5800 with PP CS5’s Mercury engine? All things being equal,(workstation, video sequence) it would be great to know how much less rendering time you buy back by investing in the 4800 over the 3800 or the 5800 over the 4800.
-
5 Replies
-
This does depend somewhat on your CPU…
Since the video decoding is done on the CPU, a lesser CPU will decode fewer streams of video. The GPUs on the Quadro card kick in to preview effects. Big GFX card with a smaller CPU equals less layers of video with a ton of effects… Lesser CPU and the lower-end Quadro or the GTX285 card and you’re doing less of each… Massive CPU with no Mercury-compliant GFX card will get lots of decoded video streams…and will have to trade unaltered streams for effects to share that bandwidth on the CPU.
Mercury is still impressive in software mode on a big-iron CPU.
I suspect the leverage achieved by a massive CPU (Nehalem at this point), generous RAM allocation (16 GB for instance) and the beefiest Mercury-capable GFX card will be more than proportional to an alternate system where you start to diminish any of those individual components.
Once again, we’re at the point in the post industry where we can spend as much money as we have…it’s nice to know the technology keeps advancing.
🙂
TimK,
Director, Consultant
Kolb Productions, -
Thanks Tim
I appreciate the feedback. I have two XEON quad-cores on a Tyan 5396 mobo, Nvidia Quadro FX 1700 and 16Gb of ram.
I was looking for a dollar to minute equation for the Quadro FX lineup if I upgraded to the one of the cards compatible with Mercury, but maybe its not so simple? I was hoping to determine if, given a sequence that now renders out in 100 minutes. would only take 50 minutes with a Quadro FX3800 ($1,000), 25 minutes on a 4800($1700), or say 7 minutes with the FX5800 ($3100).
Buying back the time is what makes the investment a consideration; its just trying to determine what one gets (less time) for the money.
But again, maybe its not so simple
Thanks
Jim -
Yes…it’s not easy to measure it without actually running a trial with each card in your system on an identical timeline.
The format matters to the CPU…AVCHD is tougher than uncompressed and RED 4K is a real load.
More layers of more complex video will start to max out your CPU…which will at some point reach it’s limit on what it can even spool out to the GPU, regardless of the number of effects you’re using.
Drive speed becomes an issue at some point…
What effects are on the timeline? Just CC? Do you have a bunch of Magic Bullet frames on clips? etc, etc.
Also keep in mind, this is preview. It’s being able to hit the spacebar and just play whatever you just created in RT…rendering isn’t really the task involved.
It sounds like you’ve got the rest of it in hand. I would think you’ll see benefits. For me, when there is a client sitting there, it doesn’t take a lot of feedback loop acceleration to make it worthwhile…
TimK,
Director, Consultant
Kolb Productions, -
Thanks again Tim –
I now understand the primary purpose of the Mercury Playback Engine is for RT playback, I had just latched onto the line at Adobe about how it also accelerates rendering for delivery of content to clients. That’s where the primary bog is in my work flow; rendering out a sequence.
I’ll let clients and budget dictate the choice when I do get around to the upgrade.
Thanks for your input.
Jim -
Well…it does that also.
I dunno if I’ve seen the “holy bit bazooka Batman!” speed increase in that way like I’ve seen it in timeline preview. That end of things is also a bit format-dependent I believe…
All these qualifiers can really rain on a guy’s parade, I realize, but the fact that it IS a beneficial system add-on remains. the nice thing is that one of these boards is not required and the software will run very respectably relative to your configuration in software mode as well.
I think the idea may be to try the software and decide if you think you need the graphics card. I’ve not used CS5 without a Q4800, so I don’t know what the comparison really is…but I have seen some pretty experienced users pipe up and mention that they are surprised by what they’re getting even in software mode.
So, it’s always about the ROI…for me, it’s great as I like the immediate feedback while editing. The encoding advantage is good, but not as important to me personally.
Everyone’s situation is different of course…
So…just a little more gray for you to sort through. 🙂
TimK,
Director, Consultant
Kolb Productions,
Reply to this Discussion! Login or Sign Up