Vegas 12 Workstation Specs

[Dave Haynie]

[John Rofrano] “IMHO, this should weight heavily in your choice of cards. Today you may just be using Vegas but tomorrow you might want to branch out into Motion Graphics and so you purchase Adobe After Effects only to learn that it only supports CUDA so you’ll need an NVIDIA card.”

Adobe is moving to OpenCL, too. I say “moving”. Their “Mercury” video engine, for CS6, does both, though unlike Vegas, it’s currently doing both, not moved entirely to OpenCL yet. I’m not an Adobe person for video, so I don’t know, but I suspect this is a process, and you may still be better off with CUDA than OpenCL for Premiere.

The newer Mercury Graphics Engine, for Photoshop, is pure OpenCL. That seems to be where they’re headed.

Part of this shift is the obvious case that CUDA is proprietary. That may be fine when nVidia and AMD/ATi are the only choice, and it certainly was back when nVidia was inherently faster, and the only company really pushing for native GPU computing. But now there’s Intel doing GPUs that are actually capable of GPU computing, even if at lower performance levels than the others. But also developments like Intel’s “Phi” — that’s 60 x86(ish) cores on a single under-$2000 PCI Express card. And already supported by OpenCL.

[John Rofrano] ” didn’t benchmark with two cards like Dave did, but I did compare rendering with the CUDA option and OpenCL option and for my Quadro 4000, the CUDA option was much faster. I’m not sure how an AMD card would have done with OpenCL”

I would definitely expect the same benchmark to run a bit faster on CUDA than OpenCL on nVidia. For one, CUDA was designed based on how nVidia GPUs work. But also, OpenCL is built on top of CUDA… nVidia didn’t trash CUDA and start over with OpenCL, they just added to what they already had.

It may also be benchmark dependent. In my tests — admittedly just two hardware sample points in a very complex question — I got measurably better performance out of the AMD than the nVidia in Vegas. And yet, the nVidia has the edge in pretty much every gaming test done between the two. I was actually a bit surprised, given that AMD uses more, simpler processing elements versus nVidia (about 3:1 in this generation).. but nVidia seems to be moving more in that direction in their future chips.

I also found that the nVidia was using far more of the GPU, 70-80%, versus around 40-50% on the AMD. While that suggests the AMD was actually much faster, spec-wise, than the nVidia, it also kind of suggests it’s a harder architecture to use. That kind of suggests that, if you’re looking for performance at any price, a super-high-end nVidia might deliver more than a super-high-end AMD. I did not test this. I do have a pretty fast i7 Gen 3 at work, and a nVidia 4000-series GPU.. maybe worth a benchmark at some point.

-Dave

[Pete Schaefers]

Hi all!

So I have my 6 core, 4 channel bus CPU, 16GB RAM, and Q2000 card up and running. I did some benchmarking comparisons but I’m not seeing the results expected…

I through together a 30 second HD timeline with several various, transitions and rendered it out to 1920 for Internet MP4, using the stock (NV300) video card and RAM, then with the Q2000 and RAM upgrade.

Results:
NV300, 2GB RAM, 16 threads-
GPU-off = 1:12
GPU-on = 1:11
GPU-on, 8 threads* = 1:10

* I thought max threads was cores x 2, so I tried setting it lower. Also I got a high CPU usage warning on the first two, but not the third.

Q2000, 16GB RAM, 16 threads-
GPU-off = 1:09
GPU-on = 1:07

At this point I checked the Q2000 driver and updated it to the latest, but I didn’t see any change, I think (my notes got confused at that point).

—-

Two questions:
– Any input on thread usage selection?
– Shouldn’t I see a significant render time decrease when enabling GPU acceleration, and is there a set of benchmark files somewhere online that I could use to compare against?

Thanks!

[John Rofrano]

[Pete Schaefers] “Shouldn’t I see a significant render time decrease when enabling GPU acceleration, and is there a set of benchmark files somewhere online that I could use to compare against?”

If by Q2000 you mean a Quadro 2000 I wouldn’t expect it to be faster than a hex core. I wouldn’t use anything less that a Quadro 4000 (which is why I bought one) and in some cases (i.e., rendering to MPEG2) my hex core is still faster with the Timeline GPU Acceleration disabled.

~jr

http://www.johnrofrano.com
http://www.vasst.com

[Pete Schaefers]

So in essence, I could have just stayed with the stock NV300 and gotten nearly the same speed. (It performed well even with only 2GB of RAM.) Not very impressive…

(Side note: does anyone else have trouble with the “Quote” feature when replying? I click Reply, then up to under what I am replying to I click “Quote”, but nothing happens. Yes,I have j/s enabled.)

[John Rofrano]

[Pete Schaefers] “So in essence, I could have just stayed with the stock NV300 and gotten nearly the same speed. (It performed well even with only 2GB of RAM.) Not very impressive…”

I get some improvements when using the GPU for rendering. I don’t always get improvements when using the GPU for timeline playback. I’m still experimenting so it’s hard to say.

The key is to buy a GPU that is more powerful that your CPU. Having a Quadro 2000 with a Dual Core is probably beneficial, having it with a Hex Core probably requires a Quadro 4000 or 5000 to see a benefit over 6 core. So you have to match your components. In other words, you have an underpowered GPU for your CPU.

[Pete Schaefers] “(Side note: does anyone else have trouble with the “Quote” feature when replying? I click Reply, then up to under what I am replying to I click “Quote”, but nothing happens. Yes,I have j/s enabled.)”

You highlight/select the text that you want to quite and then press the “Q” key or the [Quote] link.

This does NOT work for me on my iPad (and it use to) so it doesn’t always work but it works in Chrome and Firefox for me.

~jr

http://www.johnrofrano.com
http://www.vasst.com

[Dave Haynie]

[John Rofrano] “The key is to buy a GPU that is more powerful that your CPU. Having a Quadro 2000 with a Dual Core is probably beneficial, having it with a Hex Core probably requires a Quadro 4000 or 5000 to see a benefit over 6 core. So you have to match your components. In other words, you have an underpowered GPU for your CPU.”

That’s pretty much it. Here’s basically what happens.

For CPU-only, you have a traditional multiprocessor, which can handle pretty heavy lifting. Other than waiting on I/O, the CPU can be kept pretty well fed in rendering, this is all pretty well understood by programmers. A few things, like tapping SSE “vector” instructions or efficiently breaking up a render across multiple CPUs requires some talent to do well, but it’s not new, and it’s “immediate”… the whole program runs on that CPU.

The GPU is a strange ranger. It’s got somewhere between 200 and 2000 simple but fast floating point processors. These aren’t necessarily ideal for general purpose computing, but they have been evolving that way. The job of actually scheduling code to the GPU is complex enough that you never do that — you write code in abstract under the OpenCL (or CUDA) computing model, which is compiled and sent to the GPU of choice as needed.

Those last pieces are critical: the CPU does a job other than rendering… compiling that OpenCL code. Then it has to copy stuff to the GPU. Only then the GPU starts working, and it signals the CPU when done, which then does a bit of cleanup and unpacking of results. It looks like of like this:

CPU-Only Activity:
CPU ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
GPU ________________________________________________

CPU + GPU Activity
CPU ^^^^^^^XXX_________XX^^^^^^^^^^^XXX_________XX^^
GPU __________^^^^^^^^^________________^^^^^^^^^____

So, in solving just one problem set, the GPU is taking over for the CPU, and there’s the relatively small bit of overhead, represented by XXX, compilation and communications. It’s easy to see here that, unless the GPU is at least a bit faster than the CPU, you might be better off just running the CPU.

Now, this is vastly oversimplified. For one, you may have multiple CPU threads, all doing things with the GPU. Ideally, there’s enough work to pipeline everything, so that no CPU is actually idling (the _____ in my graph). In practice, not yet… I get a nice boost in rendering and even editing with the GPU (my six core is AMD, so it’s maybe half the performance of the Intel i7 six core), but I do see CPU performance drop from nearly 100% to 75-80% during a GPU render.

Another thing… it’s pretty complex to go from regular CPU coding to GPU coding. So there’s a certain learning curve, and there’s no reason to expect it’s even close to optimal yet.

-Dave

[John Rofrano]

[Dave Haynie] “…but I do see CPU performance drop from nearly 100% to 75-80% during a GPU render.”

Yea, and I see an even larger drop. With CPU rendering my 12 threads are blazing at 98%. With GPU rendering my CPU’s drop all the way down to 15% – 25% and the GPU is cranking at 85%. So as Dave said, unless the CPU can keep the GPU busy with things to do and unless the GPU can do them faster than the CPU, you will see no benefit from using the GPU and may even see a slow down.

Bottom line: The Quadro 2000 is probably not going to boost anything for a hex core computer. If you were using Maya, or 3D Studio Max, or some other 3D modeling program you may see better visualization of your 3D model, but for raw video rendering, it’s just not powerful enough to make a difference against a hex core.

~jr

http://www.johnrofrano.com
http://www.vasst.com