Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1388
Article Thumbnail

DaVinci Resolve 15: NVIDIA Quadro RTX Performance

Written on March 18, 2019 by Matt Bach
Share:

Introduction

While the NVIDIA Quadro cards are often positioned as being the type of GPU you would want to use in a workstation, a significant number of video editors and colorists use either NVIDIA's GeForce or AMD's Radeon line of video cards since they give terrific performance for their price. Especially in the higher-end production fields, however, there is often a solid reason to use one of the NVIDIA Quadro GPUs instead. In most cases, there are typically three main reasons why you would want to use a Quadro card in an application like DaVinci Resolve:

  1. Reliability. Over the last three years, we have recorded roughly a 60% higher failure rate with GeForce cards over Quadro (1 in 52 chance of failure versus 1 in 84). This includes both DOA and failures that occurred after our customer received the system, but the difference is even more stark if we remove the DOA results (leaving only the failures that actually affected our customers). When we do that, we found that GeForce cards have a recorded 1 in 117 chance of failure while Quadro cards have yet to have a single card fail within this time period.
  2. Higher VRAM capacity. If you are working with 8K media and do a lot of noise reduction, the 11GB maximum that is currently present on GeForce cards may not be enough. Even a mid-range Quadro card can have 16GB of VRAM, however, and the high-end cards can have as much as 48GB! Note that there are some newer cards that have plenty of VRAM such as the AMD Radeon VII 16GB and the NVIDIA Titan RTX 24GB - although both of those cards have cooler designs that make them unsuitable for 3-4x GPU configurations.
  3. 10-bit display support. Unlike the GeForce or Radeon cards, Quadro cards can output 10-bit video signals in OpenGL applications like DaVinci Resolve. However, if color accuracy is a concern, you likely should invest in a video monitoring card like the Blackmagic Decklink series as they will output a more accurate signal than what GPUs are currently capable of.

One thing you will notice is that pure performance is not one of the reasons we listed. There are some applications where Quadro cards do give better performance than GeForce (primarily in the CAD/engineering software), but we are not aware of any video editing application where this is the case. If raw performance in Resolve is what you need, we recommend checking out one of our recent DaVinci Resolve GPU articles where we looked at the performance of the NVIDIA GeForce and AMD Radeon series cards.

NVIDIA Quadro RTX DaVinci Resolve

Before we get into our testing, we did want to look at the most relevant specifications for the Quadro cards we will be testing in this article. Normally we wouldn't do this since real-world performance is more important than specs, but we wanted to point out that the Quadro RTX 6000 and RTX 8000 have the same number of CUDA cores and the same theoretical FP32 (single precision) performance. This is because these two cards are actually identical except for the fact that the RTX 8000 has twice the VRAM. Because of this, we don't expect to see any higher raw performance with the RTX 8000 even though it is a more expensive card.

CUDA Cores FP32 Perf. VRAM MSRP
M6000 3,072 7 TFLOPS 24 GB $5,000
P6000 3,840 12 TFLOPS 24 GB $5,000
RTX 4000 2,304 7.1 TFLOPS 8 GB $900
RTX 5000 3,072 11.2 TFLOPS 16 GB $2,300
RTX 6000 4,608 16.3 TFLOPS 24 GB $4,000
RTX 8000 4,608 16.3 TFLOPS 48 GB $5,500

Like we mentioned, specs don't always reflect real-world performance so the only way to see exactly how these cards perform is to put them through our DaVinci Resolve benchmark process. If you would like to skip over our test setup and benchmark result sections, feel free to jump right to the Conclusion section.

Test Setup & Methodology

Listed below is the system we will be using in our testing:

Our testing primarily revolves around the minimum FPS you would see with various media and levels of grading in the Color Tab. The lowest level of grading we test is simply a basic correction using the color wheels plus 4 Power Window nodes with motion tracking. The next level up is the same adjustments but with the addition of 3 OpenFX nodes: Lens Flare, Tilt-Shift Blur, and Sharpen. The final level has all of the previous nodes plus one TNR node.

We kept our project timelines at Ultra HD (3840x2160) across all the tests, but changed the playback framerate to match the FPS of the media. For all of the difficult RAW footage we tested (CinemaDNG & RED), we not only tested with the RAW decode quality set to "Full Res" but we also tested at "Half Res" ("Half Res Good" for the RED footage). Full resolution decoding should show the largest performance delta between the different cards, but we also want to see what kind of FPS increase you might see by running at a lower decode resolution.

Codec Resolution FPS Bitrate Clip Name Source
H.264 3840x2160 29.97 FPS 80 Mbps Transcoded from RED 4K clip
H.264 LongGOP 3840x2160 29.97 FPS 150 Mbps Provided by Neil Purcell - www.neilpurcell.com
DNxHR HQ 8-bit 3840x2160 29.97 FPS 870 Mbps Transcoded from RED 4K clip
ProRes 422 HQ 3840x2160 29.97 FPS 900 Mbps Transcoded from RED 4K clip
ProRes 4444 3840x2160 29.97 FPS 1,200 Mbps Transcoded from RED 4K clip
XAVC S 3840x2160 29.97 FPS 90 Mbps Provided by Samuel Neff - www.neffvisuals.com
XAVC Long GOP 3840x2160 29.97 FPS 190 Mbps Transcoded from RED 4K clip
Blackmagic RAW 4608x1920 24 FPS 210 Mbps A001_08122231_C008 Blackmagic RAW
RED (7:1) 4096x2304 29.97 FPS 300 Mbps A004_C186_011278_001 RED Sample R3D Files
CinemaDNG 4608x2592 24 FPS 1,900 Mbps Interior Office Blackmagic Design
[Direct Download]
RED (7:1) 6144x3077 23.976 FPS 840 Mbps S005_L001_0220LI_001 RED Sample R3D Files
RED (9:1) 8192x4320 25 FPS 1,000 Mbps B001_C096_0902AP_001 RED Sample R3D Files

With the addition of the "Fusion" tab in Resolve, we are also going to be including some basic tests for that tab as well. At the moment these are relatively easy projects that specifically test things like particles with a turbulence node, planar tracking, compositing, and 3D text with a heavy gaussian blur node. These projects are based on the following tutorials:

Benchmark Results

While our benchmark presents various scores based on the performance of each type of task, we also wanted to provide the individual results in case there is a specific codec or level of grade someone may be interested in. Feel free to skip to the next section for our analysis of these results.

NVIDIA Quadro RTX DaVinci Resolve Studio 15 Benchmark Performance Results

Benchmark Analysis

Our results are divided into several categories based on the level of grade as well as an "Overall Color Grading Score" that is a combination of each grading score and a dedicated result for Fusion (which we will ignore since it doesn't use the GPU much). For most users, the Overall Color Grading Score should be a pretty accurate breakdown of how you would expect each GPU to fare in DaVinci Resolve. However, if you tend to do very heavy grades with things like noise reduction, you may want to scroll to the fourth chart which has the results for a grade with a combination of Power Windows, OpenFX, and TNR.

Starting at the top, the Quadro RTX 6000 and RTX 8000 performed exactly the same which is to be expected since the only difference is the VRAM amount. Having enough VRAM is incredibly important, but just like normal system RAM, having an overabundance of it won't affect performance. Since our benchmark only needs about 6-8GB of VRAM, having a huge amount of video memory simply isn't a factor. Just for reference, the ~820 score the RTX 6000 and RTX 8000 achieved in our benchmark puts these cards right in-between a GeForce RTX 2080 and RTX 2080 Ti.

Moving down the chart, you may be surprised to see that the previous generation P6000 is a bit slower than the RTX 5000 even though the GPU specs say that the P6000 should be faster (12 TFLOPS vs 11.2 TFLOPS). This comes down to what we mentioned earlier about how part specifications is not a reliable way to measure real-world performance. Specs can be useful when comparing models within a single product line, but especially when comparing different generations of card, performance specs are almost meaningless.

This is further demonstrated by the Quadro M6000 results. It should only be slightly slower than the RTX 4000 in terms of raw TFLOPS, but in DaVinci Resolve it is actually about 12% slower. The higher VRAM may definitely be useful for many users, but you would have to find a M6000 for less than $900 (less than 1/5 the MSRP) in order for it to be a better buy than a Quadro RTX 4000 in terms of performance.

Which Quadro RTX GPU should you use for DaVinci Resolve?

Like any piece of hardware, what Quadro GPU you should use is going to depend on a number of factors including your budget, the amount of performance you need, and various other factors like how much VRAM your projects require. Because of this, we decided to go through each model individually and explain when and why you would want to use that specific card:

Quadro RTX 4000 8GB for DaVinci Resolve

With a MSRP of $900, this is currently the lowest cost card in the Quadro RTX lineup. While the performance is not anything terribly exciting (it is only slightly faster than a GeForce RTX 2060), it has the robustness and reliability that you can expect from a Quadro series GPU. If you are doing basic grades and simply need a highly reliable GPU (or one that can output a 10-bit video signal), this is a great choice.

Quadro RTX 5000 16GB for DaVinci Resolve

While roughly on par with the GeForce RTX 2070 8GB in terms of pure performance in Resolve, the Quadro RTX 5000 16GB has twice the VRAM, plus the higher reliability that comes with Quadro cards. The RTX 5000 is a solid choice if you are doing moderate grades or a decent amount of noise reduction while also needing the most reliable GPU possible.

Quadro RTX 6000 24GB for DaVinci Resolve

For those that work with 8K media and do a lot of noise reduction, the Quadro RTX 6000 24GB is one of the few cards available with both the performance and VRAM capacity you likely require. The Titan RTX 24GB is a bit faster at a lower price point, but due to the design of the Titan RTX's cooler, you can only use one or maybe two of those cards in a system. With the Quadro RTX 6000, however, you can have up to four cards in a suitable workstation, or even up to 8 cards in an 8-GPU rackmount.

Quadro RTX 8000 48GB for DaVinci Resolve

Very few DaVinci Resolve users are going to need the Quadro RTX 8000 48GB since the only benefit of this card over the RTX 6000 24GB is the larger VRAM capacity. In most cases, if you are a user that will benefit from the 48GB of VRAM, you likely already know that you need it. If you aren't sure if you need 48GB of VRAM or not, you likely don't.

To be completely transparent, most of the workstations we sell for DaVinci Resolve use the NVIDIA GeForce line of GPUs. A large part of this is due to the fact that Resolve is becoming much more accessible to the masses, which is where pure price-to-performance is typically a primary concern. If you want to simply get the best performance for your money, a GeForce (or AMD Radeon) GPU is simply going to be the best option. If you are in production environment where maximum reliability is key or where you need a lot of VRAM for 8K or heavy noise reduction, however, there is no replacement for workstation-class cards like those in the NVIDIA Quadro line.

It is also worth pointing out that while Quadro cards do not prioritize raw performance as heavily as GeForce does, the new Quadro RTX cards are significantly faster than previous generations. Compared to the Quadro P6000, the new RTX 6000 is about 15% faster while having a MSRP that is $1,000 cheaper. Or, if you go back another generation to the Quadro M6000, even the $900 Quadro RTX 4000 is able to beat it in terms of pure performance.

Overall, these cards are not going to be the right choice for all DaVinci Resolve users simply due to their cost. But if the higher reliability or VRAM capacity is something you need, these new Quadro RTX cards are excellent.

Looking for a
Content Creation Workstation?

Puget Systems offers a range of workstations designed specifically for video and image editing applications including Premiere Pro, After Effects, Photoshop, DaVinci Resolve, and more.

Tags: DaVinci Resolve, Quadro, RTX 4000, RTX 5000, RTX 6000, RTX 8000, P6000, M6000
Misha Engel

It's the memory bandwidth that makes the difference. Since driver 419.17 NVidia doesn't make use of their agressive branch prediction which could lead memory attacks (looks a bit like spectre with intel CPU's) . Now it's just raw memory bandwidth(same as for AMD).
Compute wise they are (RTX6000/8000 16.3 Tflops fp32) a lot faster than Radeon VII(13.4 Tflops fp32), they loose it on memory bandwidth 672 vs. 1024 GB/s (might be one of the reasons they drop their prices so drastically).

Posted on 2019-03-19 00:57:07
Turing

https://www.nvidia.com/down...

https://www.nvidia.com/down...

"[DaVinci Resolve]: Performance drop occurs. [2532482]"

Nvidia has fixed the performance drop bug in the latest GeForce 419.67 WHQL & Quadro 419.67 WHQL driver.

Posted on 2019-03-21 07:56:37
Misha Engel

We will see if that is true when Matt updates this article https://www.pugetsystems.com/labs/support-software/DaVinci-Resolve-15---Performance-loss-with-NVIDIA-418-419-driver-1376/

Posted on 2019-03-21 15:26:44

Yep, I'm confirming that it is completely fixed right now actually. Just a clarification for anyone else coming across this comment - we only saw the performance regression with the GeForce drivers. Quadro was unaffected.

Posted on 2019-03-21 15:54:03
Misha Engel

Then NVidia has a lot of work to do. From the cheapest RTX 4000 ($900) to the most expensive RTX 8000 ($5500), all are slower than the Radeon VII ($ 700).

Benchmarking also becomes more difficult when RTX2080ti's with more headroom in the turbo frequencies are faster than top of the line quadro's with more cores and memory bandwidth.

Posted on 2019-03-21 18:54:23

To be fair, the Quadro cards are intended for a very different use-case than Radeon. GeForce vs Radeon is a much more accurate comparison, although even there the Radeon VII is really, really good compared to GeForce in Resolve. About on par with the more expensive 2080 Ti, but it also has more VRAM (important for 8K media). So, yes, I agree that AMD is definitely on top right now for DaVinci Resolve in particular.

Quadro cards are made for reliability and stability first (both in terms of hardware and drivers), performance second. They absolutely have a large price premium, but they are for customers that cannot tolerate part failures or the downtime that comes with those failures. For example, a major Hollywood studio isn't going to care about spending $10,000 on video cards if there is even a tiny bit of reliability improvement. To them, if there is ever a single issue, that is going to cost them way more than the cost of those cards.

Quadro really isn't for average Resolve user, it is for whose budget for workstations is just a sliver of their total operating costs.

Posted on 2019-03-21 19:11:09
Sebastian PĂ©rez

I've been reading every article you guys post about pc builts performance in photography and video editing for the past 5 months. I really appreciate all the effort put in giving professionals a reliable source of information to make the most well informed decision.
Right now, I'm kind of thorn apart with the graphics card, I've been looking to migrate to Resolve 15 for editing and color grading, 4k material XAVC-I 4:2:2 10-Bit, and multiple Blackmagic formats ProRes 4444XQ 4:4:4 10-Bit and 12 Bit BM Raw in a diverse type of frame rates.
Because my work is 50/50 divided in architecture/advertisement photography, and editing/grading all sorts of video for web, the quadro rtx 4000 seemed like the best option to obtain the 10 bit output i need in both Photoshop, DaVinci an Premiere and a reliable performance, so I've been waiting for this card patiently. But I keep reading that the 2080 ti performance on Resolve will just crushed this low tier Quadro. For the type of material I'm looking to handle, do you think the Quadro can manage a fluid workflow and playback (mainly color grading on Resolve) or will I regret this immediately?
I'm currently running an Intel 10 core i9 9900x with 64gigs of Ram.
I really hate losing that 10 bit output because already own a BenQ Sw320 monitor, I know that I could obtain that output via BlackMagic's declink. But for Photoshop It wouldn't help regarding that matter, at least in real time while editing a photo.

Thanks in advance for the help guys!

Posted on 2019-03-24 10:13:15
Misha Engel

AMD Radeon Vega Frontier Edition Liquid Retail $900 at Jeff Bezos.
16 GB, 13.1 Tflops FP32, support for OpenGL overlay at 10-bit color depth.

Posted on 2019-03-24 12:52:35

You are definitely right that the 2080 Ti will be way faster in Resolve than the RTX 4000. Premiere and Photoshop probably won't be too much slower, but Resolve definitely could be - it just depends on the level of grading you are doing. Basic stuff shouldn't be a problem for live playback, but if you want to use OpenFX or noise reduction you definitely will notice the hit (about 60% lower FPS). Exporting will definitely be a bit slower no matter what, however.

Photoshop is really what throws a wrench in things since a Blackmagic Decklink card won't let you see what you are working on live. But... mixing GeForce and Quadro (or any other line) usually results in a sub-par experience overall so you really don't want to do something like a Quadro primary plus GeForce 2080 Ti for compute in Resolve. Most apps will work fine, but Windows updates especially likes to get confused and mess up the drivers at random times which can be annoying to have to fix.

I really don't think there is an easy answer to be honest. You could look at the AMD Radeon Pro cards since you tend to get more performance per dollar with AMD and you can keep the 10-bit output. But in our experience, driver stability is not as good as NVIDIA and you are still giving up performance compared to a 2080 Ti. Alternatively, you can just make the call that 10-bit isn't worth the performance trade-off. That is really only a call you can make. Your BenQ monitor is still really good for color accuracy (assuming you calibrate it regularly of course), and it's always worth keeping in mind what people are viewing your work on. If it is just going online, it isn't like anyone is actually going to see it in 10-bit anyway.

Or, just take the chance with the RTX 4000. Buy from somewhere with a good return policy, load up some of your recent projects, and see if the performance is good enough for what you are doing.

Posted on 2019-03-25 16:50:44
DE

Any comps coming with the 2080 Ti soon?

Posted on 2019-03-26 08:32:39

You mean a performance comparison with the RTX 2080 Ti in Resolve? We have several articles with that card, but this is the most recent: https://www.pugetsystems.co...

Posted on 2019-03-26 17:22:30

I would recommend having two geforce cards because of one fails you can still work on the other until you replace the dead one. You could run three to be extra safe, but it's worth noting resolve does not benefit from more than two cards as much as it used to so you probably only really need two. Then there is no real need for quadro at all. Original deck link cards allowed use as a desktop. It was great for Photoshop. I was so upset when they changed. I have never found a way to build a macro or action to use bmd preview quickly and it forgets settings each when you close the app. Just do your Photoshop work in after effects! Then everything is non destructive! Thanks also for the recommendation about using nvidia new creator drivers. Sounds closer to the vendor testing the quadro drivers get?

Posted on 2019-03-31 09:16:36
Turing

https://www.blackmagicdesig...

"The revolution in editing starts at 9AM PDT on April 8th 2019."

Matt, can you please test DaVinci Resolve 16 when it's out of beta? Thanks.

Posted on 2019-04-07 11:14:24

Yep, I'm at NAB right now so I'll get to talk with Blackmagic and check it out. I probably won't be able to until late May, however, simply due to my current schedule. April is always crazy since everyone (Adobe, Blackmagic, etc.) wants to launch new versions during or right before NAB.

Posted on 2019-04-07 15:44:52