Benefit of using a Pro (non-gaming-optimized) video card

Dumb question for the gallery…

Most will be using a consumer (gaming) optimized video card almost by default.

But has anyone seen/experienced the benefit of using a Pro style graphics card, such as Nvidia Quadro P4000 and up for this type of work? Somewhat overkill probably, unless one also does Video, since in photo processing there is only one “frame” to be processed, but still.

I would imagine that even the absolute worst imagined rendering of very large image Impression processing should be almost instantaneous?

Anyone out there in Topaz land using a “grown-up” graphics card?

Was looking towards an upgrade and I am still overflowing with drool from looking at Nvidia Quadro P6000 or even a Quadro GP100. (Although I can’t imagine spending $5000 - $7000 on a graphics card.

Me neither.

@Kathy_9, No not for this kind of manipulation (single photo thing). P6000 and GP100 are mostly for Pros with a need for real-time video editing and manipulation. Think of 4K Video as say 60 highres images to be transformed per second of video.
And we complain about transformation time on a single image file. Hah… Multiply that by 300, just for 5 minutes of video.

If a company pays people to do fancy animation or Video (think Adobe After Effects or Adobe Premiere Pro type uses), then they cannot have their employees sitting around just waiting for video to render. Using apps like those, you can spend your time waiting and waiting, if you do not one or more real GPU cards. I have all that Adobe Software and it is a serious hassle unless you have fast GPUs.

Also, the Pro cards are optimized for accuracy and detail work, whereas the average consumer/gamer rig sacrifices accuracy for just blasting faster frame-rates, which is why switching to a Pro type card might be better for high-resolution images and Video.

But the Quadro P4000 is much more reasonable in price. $899 MSRP (on Amazon for around $700 and up). Very comparable to a higher end gaming card in the GeForce series (like 1080 Ti, but optimized for real work; not just for high-frame-rate gaming.) Nevermind even the Geforce Titan gaming GPU at $1200.

And I have NO interest in computer games. Hence my question of whether anyone have tried a Pro type card, like the Quadro Series.

Another reason is that when using software generating math-based image structures, there is an ENORMOUS difference between using the CPU (one fairly simple image in low-res: 2-5 minutes), switch to calculate using my current GPU (4-6 seconds)… But with an upgraded Pro card, adding to the TFLOPS available, maybe just fractions of seconds.

Despite the complaint discussions I see here in the forum, there is a reason Topaz Studio folks have chosen to switch to push everything into the GPU, which causes some people to complain about Topaz’s fairly simple and minimalistic “Minimum Requirements” for a good GPU card. Because the calculations and rendering needed to run complex and highres Impression and Simplify are massive. Also why these discussions comparing Impression manipulations with other simplistic “photo Development” plugins (Nik and similar) is like comparing Apples and Oranges.

1 Like

I believe that there are a couple of options you need to look at with NVIDIA cards …

  • Pro Series of you are looking for accurate color grading and using multiple monitors (more than 2)
  • Notebook use the 10 series cards.

It is always the best idea to get the highest quality card available, especially as if you have a desktop you can take it with you when you upgrade.

I have a 5 year old notebook that is still up to date but I choose the one with the top of the line card with 4GB vRAM.

Yeah. Best quality of course.

But the difference between Pro cards and what is in most by default (consumer gaming cards) is memory, accuracy and speed (especially when texture) or heavy calculative manipulations (think Impressions).

My current card is “consumer” level, but still supports 4 x 4K monitors (I have only 3). But it only has 4 GB of memory and could need more Ooomph for other purposes. And all their driver updates to support games I would never use annoy me. Pro cards are by nature more stable and do not change software or download gaming updates every day.

(I am not that patient, you see. :slight_smile:
I hope to see instantaneous transformations in Topaz Impressions on large highres images. Right now I have to wait sometimes a second or two between checking presets. :slight_smile: )

I have a notebook i7 with 16GB RAM and a 1050 card with 4GB vRAM and it is pretty much instantaneous. I do have 2 SSD’s as well though.

I have the system (OS and app installs) on one SSD, and all caching, tempfiles, and such on two other SSD drives.
I7-4770K (8 logical cores) running at 3.89 GHz.
32 GB system RAM (for CPU) plus the 4GB GPU dedicated VRAM.

The system pulls another 17 large drives (spinner type), but they are not involved when Topaz Studio is up and running with image initially loaded.

Most things are also (virtually) instantaneous, even with images running around 140 MB or larger, but with large images, you can see the progress bar.

Mostly it might have to do with loading up (unused) presets. If I hit an Impression preset that has not been used before, the progress bar runs around 70-80% across, then stops for a second or so before completing. Or since it is at around 80%, it might be on reloading/repainting image at the end of processing?

On subsequent clicks on the same presets, maybe 1/4 to 1/2 second visible total on the progress bar.

Slow progression when testing presets is not really a speed issue. But when it “repaints” as you change parameters in the slider settings for a chosen preset, that is where you really need the speed so you can move the sliders without delay. Also in my case near-instantaneous, as in you can move the sliders and it repaints the 140 MB image in near real time.

But speed is relative to the perceiver. :slight_smile: It can always get faster, if throwing the right hardware at it.

But as mentioned, my main speed issue is not with Topaz, but will heavy math based software creating images using the GPU… Before they might reach Topaz. Where currently finishing up a final image in 140MB + high-res can run for 15 minutes or more. :slight_smile:

Knowing how a pro-type card would handle Topaz is merely knowing whether it would make the progress bar closer to invisible on setting changes. :slight_smile:

Sorry… I just thought someone at Topaz Labs might have tested Studio on a real powerful card to see what the effect would be.

THIS. Several of our technologies render heavily in the GPU, using the OpenGL language. That’s the important part, rather than the overall speed. A workstation/pro graphics card, like the Quadro or FirePro series are going to lag behind consumer-level GPUs, with our software, primarily because we process in OpenGL. As you correctly noted, @Rover, most consumer-level GPUs have a bunch of optimizations for gaming. This allows these specific GPUs to run much more quickly when given very common OpenGL instructions (light point, shadow location and depth, etc.), than a workstation/pro card. Why? Power/speed/efficiency is definitely important, but all that power will do nothing without the right “transmission” (car analogy, in case I lost anyone!). In the case of a pro GPU, you’re looking at a card that has been optimized for use primarily in OpenCL applications. What’s the difference? OpenCL is another language that enables an extremely large amount of simultaneous computations that would otherwise bog down a consumer GPU. This is referred to as parallel computing – the more operations you can do in parallel, the more sections of your video you can render at once, or the faster you can get that skyscraper (shear, load, etc.) designed and off to the client for review.

In the simplest of terms: not as great as you’d expect. Workstation GPUs are much more expensive, because they have more computation throughput than a consumer GPU, and they’re used accordingly, in the design of infrastructure, high resolution video processing (8k and up), and processing large data models. That’s not to say that a workstation card isn’t going to work well, but you may find a comparable consumer GPU will outperform, in OpenGL applications. That’s why our software will run better with a consumer GPU. Here’s a bit of information on the differences between OpenCL and OpenGL:

https://www.quora.com/What-are-the-differences-between-OpenCL-and-OpenGL

1 Like

Very useful… Thanks. I did not remember the OpenGL/OpenCL difference.
So essentially, without using the card for heavy or specialized CAD work or VR rendering, a Pro card MIGHT not be worth the money. A 1080 Ti or a Titan it must be then. :slight_smile:

This short benchmark might show some of the same. Quadro P5000 and GeForce 1080 just about equal in all benchmarks. Except for certain OpenGL functions, like two-sided Lighting render, the P5000 blew the 1080 out of the water by a factor of 7 on low-res (1280×720) and a factor of 20 on hig-res (3840×2160) images.

http://www.geeks3d.com/20170515/test-nvidia-quadro-p5000-vs-geforce-gtx-1080/2/

Anyway… Thanks.

If you don’t have any use for the OpenCL computation power, then the 1080/1080 Ti is the better bet than the Quadro card. Generally speaking, you’d have to go WAY up the GPU scale on Workstation cards to get comparable performance. Considering you could get a 1080 Ti build for ~$1,800, and a Quadro P5000 build can cost you upwards of $7,000, it’s all a matter of need.

Of course, the cards we’re talking about right now are well outside of the “average” person’s budget or use case requirements. Bottom line is, if you are not a professional or heavy hobbyist designer/video producer, you just don’t need the high end workstation card, and the price is excessive for the result. That said, the GTX 1080 Ti is also outside of the “average” person’s budget, but fits most consumer-level use cases with gobs of performance to spare. Our own, @HeathRobinson, has a GTX 1080 FE, and has trouble getting it to even spin up the fans, it’s so powerful and efficient. I think the 1000 series from NVIDIA really set the bar at a new high, and the competition coming from Intel is about to push a whole new level of hardware performance in both the consumer and professional grade GPUs available to us.

This is what we’re hedging our bets on. As GPUs, on average, get faster, so too will our tools. Our competitors, however, will be saddled with a design that couples CPU and GPU performance. We believe the coming processing revolution will take place in graphics rendering processors, with the vast load of computational processing taking place in non-local configurations (server farms, etc.).

I agree with you on all that. As people’s systems (GPUs) naturally get upgraded, Topaz as a product will get a significant boost along the way. Also, with all the new GPU cards coming out (and BitCoin miners switching to specialized hardware), the prices are dropping significantly on the older, but still very capable ones.
That said, you might want to consider some articles on your site educating people in the differences between various types of plugins, std photo stuff versus heavier Topaz filters, and why good graphics card (OpenGL) support should be important for any type of image work. Especially until the legacy of your older CPU based plugins kind of sputters away.
(Or at least that I what I interpret from the forum discussions around performance, or perceived lack of such when compared to simplistic “change exposure” type plugins. I could be wrong of course.)
Apples and oranges, or apples and melons comparisons.

For Photo/single image type stuff, GeForce 1080 Ti would be way overpowered, almost meaningless.
BUT… I like to play with Fractals as well, where currently my 970 coughs a little. In OpenCL for hours on high-res it spins the fans up like jet engines, and the temperatures go up… Even worse if I switch to CPU instead. I can literally smell the dust-bunnies burning when the cores start reporting around 220 F degrees and the bigger fans spin up high.

Not really thinking of upgrading right this second. But in my experience, buying cheap (and buying stock builds which by definition are cheap(er)) will cost you dearly in the long run. Home built is better and if overbuilt for it’s time will last much longer.

Soo… I built my current main system in 2013, so it has lasted soon 5 years and is still at the upper to high end of what most people use in the prebuilt category… No standard consumer rig would have lived that long for the $ investment and still be efficient to use. Which people frequently forget.

But… With only 4 cores/8 logical and limited to it’s 32 GB of memory it does not quite do it anymore. To stop Windows from yelling about being out of memory when I run too many simultaneous Adobe apps, I have had to up paging space to 90+ GB paging off SSD. The GeForce 970 is coughing some on heavy OpenCL (but is near-instantaneous on such things as single-image Topaz), and I am really longing for some M.2/M.3 based SSD speed.

So… I am starting to think/dream of my next build. Hence my dumb questions.
With a current case big enough to push around 12 drives internally (I have 10 spinning externally as well), 4 big fans for pushing away the hot air, and a 1200 W powersupply, all I have to do is to swap in an i9 based mother-board (supporting up to 128 GB new RAM), and probably an upgraded graphics card (hence this discussion). Then I am good for probably another 5-6 years. (I can hope, can’t I)

The Consumer/Pro GPU discussion is because I also run OpenCL Fractals (2D/3D) and Adobe Premiere Pro for various purposes, in addition to all the usual Photoshop/Illustrator/Lightroom/Topaz/… So I was wondering what would happen to Topaz speeds on very large images, if moving to a different type of card. Into the Pro range.
But I see that this might not matter or benefit.

My current 5 year old hardware (motherboard/RAM/graphics) can be slapped into a cheaper smaller case, and still blow most current sold pre-built stuff away on pure photo and simple-user work. So it is not wasted.

The worst of all worlds are the people that choose to buy Apple/Macs. Buying black-boxes like Apple makes them, there is little upgrade path other than buying whole new systems and start over. (Just saying. )