Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1563
Article Thumbnail

DaVinci Resolve GPU Roundup: NVIDIA SUPER vs AMD RX 5700 XT

Written on August 21, 2019 by Matt Bach
Share:

Introduction

Whenever we do GPU testing, we always look forward to benchmarking DaVinci Resolve since it has long been the best NLE at utilizing the power of your GPU. This means that more than almost any other video editing application, you typically want the best GPU (or multiple GPUs) that you can afford.

Recently, both AMD and NVIDIA have launched a number of new video card models. From AMD, we have the Radeon RX 5700 8GB and RX 5700 XT 8GB. From NVIDIA, the new models are a relatively minor update that includes the GeForce RTX 2060 SUPER 8GB, 2070 SUPER 8GB, and 2080 SUPER 8GB. Beyond a small performance bump, the most notable difference with the new NVIDIA cards is that the RTX 2060 gets an upgrade from 6GB to 8GB of VRAM.

In addition, DaVinci Resolve Studio 16 has left beta which makes this round of testing even more interesting than normal.

NVIDIA GeForce SUPER vs AMD Radeon RX 5700 XT in DaVinci Resolve

Since we are interested in how these new cards compare to the previous generation, we will be including in our testing the full line of NVIDIA GeForce RTX cards as well as the AMD Radeon Vega 64 and Radeon VII. If you would like to skip over our test setup and benchmark sections, feel free to jump right to the Conclusion.

Looking for a DaVinci Resolve Workstation?

Puget Systems offers a range of workstations that are tailor-made for your unique workflow. Our goal is to provide the most effective and reliable system possible so you can concentrate on your work and not worry about your computer.

Configure a System!

Test Setup & Methodology

Listed below are the specifications of the test system we will be using for our testing:

To test each GPU, we will be using the fastest platform currently available for DaVinci Resolve Studio - most notably the Intel Core i9 9980XE. This does mean that the AMD Radeon RX 5700 XT will not be able to utilize PCI-E 4.0, but we have an entire series of posts coming up that will be examining PCI-E 4.0 in professional apps like DaVinci Resolve. For that testing, we will be using the new Ryzen 3rd Gen platform since that is currently the only way to get PCI-E 4.0 functionality.

The benchmark we will be using is the latest version of our (soon to be) public DaVinci Resolve Studio benchmark. With each test configuration, we made sure that the GPU processing mode was manually set to either CUDA (NVIDIA) or OpenCL (AMD) as well as having the GPU selected for hardware accelerated decoding of H.264/H.265 media and RED Debayer.

One thing to note is that since our 8K media benchmark requires ~20GB of VRAM in order to run without "out of memory" errors, we will only be running the 4K media and Fusion benchmarks for this test.

Full details on the benchmark is available at:

Benchmark Results

While our benchmark presents various scores based on the performance of each test, we also wanted to provide the individual results. If there is a specific codec or type of grade that is a hindrance to your workflow, examining the raw results for that task is going to be much more applicable than our Overall scores. Feel free to skip to the next section for our analysis of these results if you rather get a wider view of how each GPU performs in DaVinci Resolve.

4K Media Benchmark Analysis

NVIDIA GeForce SUPER and AMD Radeon RX 5700 XT 8GB DaVinci Resolve 4K Performance Benchmark

With the Overall Score from our 4K Media benchmark, we get a great look at how each of these cards perform in DaVinci Resolve Studio 16. Starting with the NVIDIA SUPER cards, the RTX 2070 SUPER and 2080 SUPER are slightly faster than the "normal" versions, but it is only be a few percent. The RTX 2060 SUPER, on the other hand, saw a terrific 14% performance gain. In addition, the increase from 6GB to 8GB of VRAM makes the RTX 2060 SUPER a much more attractive card since having enough VRAM is always a concern in Resolve.

The AMD Radeon RX 5700 XT 8GB, however, did not fare nearly as well. The (sadly EOL) Radeon VII 16GB still does terrific, which initially gave us high hopes for the 5700 XT. Unfortunately, the RX 5700 XT ended up at the very bottom of the chart with the Vega 64 8GB outperforming it by 15% while the similarly priced RTX 2060 SUPER was about 24% faster.

The performance with the 5700 XT reminds us a lot of what we saw in our Photoshop testing. In that case, our contacts at AMD noted that the Radeon RX 5700 XT currently includes software optimizations for gaming, but not yet for office/professional workloads. While it is possible that post launch software optimizations will improve performance in applications, we will have to wait to see if that happens.

GPU performance with Temporal Noise Reduction TNR in DaVinci Resolve

If we switch over the look at just the scores for the "3x Temporal NR" tests, we get a look at what should be among the maximum difference between each of these cards. Here, the RTX SUPER 2060 is actually 26%(!) faster than the standard RTX 2060, and the (EOL) Radeon VII manages to pull ahead of even the Titan RTX by about 8%.

Unfortunately, the Radeon RX 5700 XT 8GB continues to be at the bottom of the stack with the RTX 2060 SUPER 8GB outperforming it by about 34%.

Fusion Benchmark Analysis

NVIDIA GeForce SUPER and AMD Radeon RX 5700 XT 8GB DaVinci Resolve Fusion Performance Benchmark

Fusion is interesting since it is a fairly recent addition to Resolve and so far, does not utilize the GPU to anywhere near the level of the Color, Edit, or Delivery tabs, Instead, it is much more dependent on the per-core performance of your CPU. Because of this, there is almost no difference between any of the NVIDIA cards we tested.

The AMD Radeon VII and Vega 64 also performed about the same, although once again, the Radeon RX 5700 XT lags behind. In this case, it is about 11% slower than the Vega 64 or about 20% slower than the RTX 2060 SUPER.

How well do the NVIDIA SUPER and AMD Radeon RX 5700 XT perform in DaVinci Resolve?

In DaVinci Resolve, the new RTX 2070 SUPER and 2080 SUPER are only marginally faster than the "normal" versions, but the RTX 2060 SUPER 8GB is on average about 14% faster than the normal RTX 2060 6GB. Considering the fact that these cards are typcially cheaper than their predecessors (other than the 2060 SUPER which is $50 more), this is overall a win for NVIDIA - especially on the low end.

The AMD Radeon RX 5700 XT 8GB, however, did not fare nearly as well. We can hope for driver optimizations to come in the future, but for now the previous generation Radeon RX Vega 64 8GB is about 20% faster than the new 5700 XT, while the RTX 2060 SUPER 8GB is about 24% faster on average.

What this means is that with the exception of the AMD Radeon VII 16GB (which is no longer being made), NVIDIA GPUs are the clear choice for anyone using DaVinci Resolve. While the exact performance gain over AMD changes based on the type of grade and codec used, the RTX 2060 SUPER 8GB is overall about 24% faster than the Radeon RX 5700 XT 8GB while the RTX 2070 SUPER 8GB is 11% faster than the Radeon RX Vega 64 8GB. This is a pretty significant lead for NVIDIA and one that is going to require a lot of work on AMD's part to overcome.

Keep in mind that the benchmark results in this article are strictly for DaVinci Resolve. If your workflow includes other software package (like After Effects, Photoshop, Premiere Pro, etc.), you need to consider how these GPUs will perform in those applications as well. Be sure to check our list of Hardware Articles for the latest information on how these cards perform with a variety of software packages.

Looking for a DaVinci Resolve Workstation?

Puget Systems offers a range of workstations that are tailor-made for your unique workflow. Our goal is to provide the most effective and reliable system possible so you can concentrate on your work and not worry about your computer.

Configure a System!

Tags: NVIDIA vs AMD, RTX 2060 SUPER, RTX 2070 SUPER, RTX 2080 SUPER, Radeon RX 5700 XT, DaVinci Resolve
Sicofante

Great tests, funny conclusions, as usual. Your Nvidia bias is definitely something you should try to address one day.

1. The Radeon VII is available in good enough quantities. It's not a good gaming card, so it won't fly from the shelves. Your recommendation should have been "buy it while it lasts".

2. A Radeon Vega 64 is some 30% cheaper than the 2070 Super, so the obvious conclusion is to buy it, not the Nvidia part. If money is no objection the choice is neither anyway.

3. Since you won't mention it, readers should know that Vega is a great compute but poor gaming platform, while Navi is just the opposite. (The Radeon VII is basically an Instinct card with video outputs.) AMD will keep manufacturing Vega-based cards, just not for gaming, so they will get more expensive. But you should keep an eye on upcoming Radeon Pro parts. Unless your love for Nvidia forbids it, that is.

There's a very good reason to buy Nvidia, though, and that's CUDA. If you need a graphics card for any CUDA renderers out there, not just for Resolve, then AMD is a no-go. Otherwise AMD is pretty much the best option for Resolve, as your tests clearly show to anyone but apparently you.

Posted on 2019-08-22 01:18:40

I think one thing you need to keep in mind is that the purpose of our testing and articles is NOT for the general public - it is for our customers. The fact that it is useful to so many others is great and one of the reasons why we don't put them behind a registration wall or anything like that, but it definitely is going to influence our conclusions.

For example, the Radeon VII being EOL makes it immediately a non-contender for our product line, just like any other EOL product. The fact that you can find some out there doesn't mean that the supply is anywhere near consistent enough for us to offer it in our product line. If you want to use it, absolutely, go ahead!

As for pricing discussions, we always go by the MSRP - just like most other hardware reviewers. Current market price is constantly in flux with various sales going on, and publishing a static post with conclusions based on temporary pricing is never a great idea. So yes, you can find Vega 64 cards cheaper than the RTX 2070 SUPER right now, but even then that really doesn't change much since the RTX 2060 SUPER is also faster than the Vega 64.

Beyond the Radeon VII (which again, feel free to use if you want, but we can't offer it in our product line), I don't see how you can declare that "AMD is pretty much the best option for Resolve". The fact that the RTX 2060 SUPER is faster than both the Vega 64 and 5700 XT, and at worse the same price (looking at current pricing for new cards on Newegg), that makes NVIDIA a pretty clear winner at the low-end. And once you get above that (RTX 2070+), AMD simply doesn't have anything that can compete with NVIDIA beyond the Radeon VII which we've already talked about.

Testing the Radeon Pro cards is something we have in the works. Depending on how things shake out, we may wait until the next update for those cards, or we may go ahead and test the current Radeon Pro and Quadro cards, and do a followup with new cards in the future.

Posted on 2019-08-22 02:32:36

If its not for the public then it should have been in a newsletter to your customers, or make the page review private / But for your customers, i mean the public (are all your customers private?) i suggest watching some EposVox https://www.youtube.com/wat... https://www.youtube.com/wat...
Regards Alan

Posted on 2019-08-22 12:12:26
citanon

Seriously? Are you seriously suggesting they stop making public a valuable and unique resource because the results don't show AMD hardware in a favorable light?

AMD fanboyism has truly reached all time high lameness levels.

Posted on 2019-08-22 13:07:56

Firstly: I never said anything of the sort regrading STOPPING this resource being publicly available, please read again what i said. I was replying to Matts comment saying that this article is not for the public. Second: AMD actually commands a price / performance lead in Resolve and finally the only fanboi here Citanon seems to be you objectively speaking. And Randy its like you never looked at the results And regarding EOL status (denied by AMD and reported https://www.tomshardware.co... looking in retail the top 5 in UK all in stock for Radeon Vega II

*edit* added price /

Posted on 2019-08-22 14:12:15

Radeon VII is 100% EOL, we confirmed that directly with AMD before we started this round of GPU testing. Leftover supply does not mean it is still being manufactured.

Posted on 2019-08-22 15:05:49

Thanks for the confirmation Matt / I think people do tend to label others as fanboi when stating a preference / Like you & Randy mention below its about value and finding the right tool for the Job for some its AMD for others its Nvidia.

Posted on 2019-08-22 15:17:36
Sicofante

Let me insist: so what? What stops you from grabbing 10 or 50 of these? They're incredible value for the type of customers you address, aren't they?

Posted on 2019-08-22 22:34:06
Håkon Broder Lund

I'd take most YouTubers and other sites with a grain of salt when it comes to professional workloads. They are usually geared towards gamers. When these other sites test premiere pro and resolve they usually only test export performance. Missing playback at various decode resolutions, which is important for workflow. Often on h.264 footage and basic (LUT) if any grading. This is not reflecting the needs for professional use cases. Don't get me wrong. I'll listen to them in their field, like streaming from EpoxVox, but not for in depth pro workloads.

Try to find any other source that publish tests done on the same verity of professional formats, resolutions and varying workloads. Not just h.264/h.265 with a simple LUT on top.

The tests performed by Puget Systems are the most in depth and varied that you can find around in their field of expertise and that they publish this for free is greatly appreciated.

Posted on 2019-08-22 15:25:00
Sicofante

The tests are astounding. That's what makes me come here. The conclusions are almost always biased. That's my issue with these articles.

I understand it's much easier for a system builder to stick to a manufacturer, know it well and deep and prefer it. I built systems for a while (for content creators, precisely) a few years ago. At that time there was no question: Intel+Nvidia was all I had to think about. That's not straightforward anymore, so a system builder must be ready to accept what their tests are saying. And what these tests are saying is: go buy tons of AMD Radeon VII cards while you can. Period.

Posted on 2019-08-22 22:46:32
Scot Dixon

Of course the conclusions are biased. They are biased toward the specific needs of their customers. That's their job, and there is nothing wrong with it. They aren't getting paid off by NVidia or anything. They are simply biasing the results based on their objectives.

Posted on 2019-09-28 03:21:47
alamilla

Have you even watched his videos on the 5700?
He literally samples several codecs, Blackmagic RAW files and the Puget bench tests

Posted on 2019-08-23 09:56:58
Rakesh Malik

I sold my Radeon VII because my 1070 was a LOT faster with Sapphire, Mistika, and Scratch.

As for "not for the public" you need to read more carefully; that's not what he wrote. This newsletter is for current and prospective Puget Systems customers, so it's focused on what works for them, not for AMD fanboys.

Posted on 2019-08-26 19:02:24
Scot Dixon

Alan, you are off base here. Matt never said the information was not for the public. They published it publicly, and thank goodness for that. They are providing valuable data that cost them a lot of time/money/resources to compile. (Thank you, Matt, and the rest of the Puget Systems crew!) What Matt was trying to explain was that the REASON they compiled the data was for their customers, or more specifically, so they could learn how to built the best systems to meet their customers' individual needs. He was simply trying to explain that their CONCLUSIONS were geared toward what their customers want... and not to offer 5,000 different options for each type of general public user. If their customers are looking for the best performance available, then that is what they will conclude is best. PLEASE don't antagonize the Puget people and cause them to stop their generosity. Just say "Thank you" and move on. Wow. So many people today feel so entitled. They pay nothing at all but want free service on their terms.

Posted on 2019-09-28 03:19:34
Randy

It's as if you didn't even read his first two sentences.

Posted on 2019-08-22 13:27:50
Brandon

It's really sad that you would let your own bias blind you so much that you suggest making one the best pro benchmark sources go private.

Posted on 2019-08-22 16:29:46

Again Brandon, I only rebutted what Matt said ; its sad that you came here to make an accusation. I suggested that if the article was not meant for the public then perhaps, make it for customers. But it is public and I think its commendable, I always research my workloads and test myself. I like yourself am entitled to my opinion and conclusion. Some of the best resources are private and behind paywalls (unfortunately) these days not because of commentators, but plagiarism . I am not a fan of this at all, If Matt and the team feels like that again, that is their prerogative. But petty stabs at my initial post are trivial and quite pointless, can you actually focus on the material in question and not your "feelings" regarding my opinion.

Posted on 2019-08-22 16:46:51
AlbertS

Alan - Any thoughts as to why EposVox NAVI vs RTX 2080 results and timeline performance are more promising than Puget Benchmarks?

Posted on 2019-08-26 08:47:39
SpenReyn

It's not really comparable results due to EposVox using an AMD R7 / X570 platform with much faster memory and a PCIE4 NVME drive vs Puget's testing which was on an x399 Platform with relatively slow memory and no PCIE 4 support. People say PCIE4 shouldn't make a difference, but looking at the differences in the numbers here shows there ARE differences and as the platform matures those differences could get even bigger.

Posted on 2019-11-05 18:06:00
Sicofante

You're right about the low end. I was thinking about some other tests you made before where the Vega 64 won the GTX 1000 series easily, being much cheaper. I have to admit I didn't pay too much attention at the 2060 Super winning here. The new RTX 2000 Super series is the right decision in the low end.

But on the high end I think you are completely missing the point. Being EOL should not affect your decisions on system building and of course you can offer it in your product line, exactly the same as Amazon is offering them to me as I write. If I were building systems for creators today (I did this for a while in the past) I'd be grabbing as many Radeon VII cards as I could. That would be real value for my customers, not a RTX 2080 Ti. Driver support is not going anywhere and you know that, so I don't get your logic.

EDIT: Don't get what's the real point regarding these articles being written for your customers. We, your readers, come here to see tests for GPUs and CPUs in our regular environment: video production and content creation. Who else do you expect to read this anyway? Aren't we precisely potential customers of yours? Your conclusions are wrong no matter if I plan to buy from you or not (arguably, making such mistakes makes you less interesting as a provider...)

And EDIT2: I ***LOVE*** your articles. Never make them private. No matter my tone (I'm old and grumpy) this is the only source I trust when it comes to Resolve, which is my main source of income these days. Your data is gold. If I don't agree with your conclusions, I'll make my own ones.

TO EVERYONE CALLING FAN BOY: I have never used an AMD card. I'd love to, because everyone loves the underdog. I just can't because besides Resolve, I like playing with AI, and AMD is too far behind there.

Posted on 2019-08-22 22:32:51
TheGreatPascalian666

Matt. Meet the toxicity that is called AMD fanboys. Always "muh amd" and nothing else.

Posted on 2019-08-23 07:40:29

Is EOL for Radeon VII ever confirmed or announced by AMD, the only original source for the EOL rumor mill is Matt Bach from Puget Systems, who says an AMD representative told the custom PC builder that the Radeon VII is indeed now end of life (EOL). Matt Bach words, not AMD's.
Is anyone knowing what they actually talked about? It seems there talk was in line with the launch of the Rx5700, so the AMD representative may have told Matt Bach the Rx5700 is a better bet considering there customers. Maybe... Who knows...

It's like I as an end user should say "An Puget System representative told me that all active development for DaVinci Resolve has now reached an end!".
I have seen several perfectly good products disappeared from the market only because of the industry rumor mill.
Killed by the industry, not the mfg.

Posted on 2019-09-24 10:28:44
Randy

As a consumer, I recently started researching components for a modest video editing rig running Resolve. Only when I began researching did I discover this inane brand rivalry. I don't care whose brand I buy, I just need a card that meets my needs without blowing my budget. I was very excited to see AMD announce their new cards a while back (along with their new processors) since the 2060 wasn't quite beefy enough but the 2070 was a little pricier at that time. The 5700 models potentially dropped right where I needed them. While AMD clearly wins my choice for CPU (and Puget was perfectly happy to sing their praises), as more info becomes available it's clear that the 2060 Super is the best performance for the price. No other card makes sense since they are either more expensive or give less performance at the same price. I'd be happy to buy AMD, but right now they just don't make sense for me. Brand has nothing to do with it.

Posted on 2019-08-22 14:19:39

I agree - the 2060 Super in particular is a great value. Honestly, however, almost every single system we sell for Resolve ends up with either a 2080Ti, Titan RTX (for 8k) or multiples of those cards. For our customers, it is generally about solving a specific problem and for what they tend to be doing, the extra cost of the higher end cards pays for themselves.

It has been shifting a bit recently, however, as more hobbiest/students/YouTubers are starting to use Resolve. For those kinds of people, there typically isn't anywhere near the return on hardware investment, so price to performance is a much larger factor.

Posted on 2019-08-22 15:12:54
TheGreatPascalian666

"The Radeon VII is available in good enough quantities"
its officially EOLed by amd lmao. Once the stock is gone, R7 is gone.

Posted on 2019-08-22 18:54:08
Sicofante

What exactly is making you laugh so loud? Honest question. R7 is on the shelves as we speak.

Posted on 2019-08-22 22:35:37
TheGreatPascalian666

Until the stock dries up. Then thats it. Amd doesnt make Radeon 7 GPUs anymore. So your whole point doesnt make sense. Sooner or later there wont be Radeon 7 to recommend anymore lol

Posted on 2019-08-23 07:36:05
Misha Engel

No.

Posted on 2019-08-25 17:18:55
TheGreatPascalian666

Well, yes. The card is officially dead.

Posted on 2019-08-25 19:56:27
Misha Engel

It's dead for PugetSystems, not for AMD and potential custormers.

https://www.tomshardware.com/news/amd-radeon-vii-end-of-life-status,39861.html

Radeon VII GPU's come from rejected MI50, which are rejected MI60. It's a waste product with expensive super fast HBM2 memory.

Posted on 2019-08-26 14:30:23
opckieran

1. The Radeon VII is available in good enough quantities. It's not a good gaming card, so it won't fly from the shelves. Your recommendation should have been "buy it while it lasts".

...But its target audience, gamers, aren't interested in it since Navi launched. So who knows how long it'll last.

2. A Radeon Vega 64 is some 30% cheaper than the 2070 Super, so the obvious conclusion is to buy it, not the Nvidia part. If money is no objection the choice is neither anyway.

That may be true, but the purpose of this article is to explore 5700XT and Super, you can't fault the author for staying within those boundaries.

3. Since you won't mention it, readers should know that Vega is a great compute but poor gaming platform, while Navi is just the opposite. (The Radeon VII is basically an Instinct card with video outputs.) AMD will keep manufacturing Vega-based cards, just not for gaming, so they will get more expensive. But you should keep an eye on upcoming Radeon Pro parts.

Again, this is either obvious through the results posted in this article, or not really the purpose of this article.

Unless your love for Nvidia forbids it, that is.

Sensationalism at its finest.

Posted on 2019-08-23 14:51:27
Daniel

Thank you Matt for another great review, fair review and write up. I and many of your potential customers greatly appreciate your work, your transparency, and your findings. It's very unfortunate that AMD fan bois or AMD themselves are ruining it for everyone.

Posted on 2019-08-22 14:29:33

Thanks for the kind words! For what its worth, AMD as a company has always been one of the nicest partners we work with. More than almost anyone else, they are always extremely receptive to feedback and very quick to respond to any questions we have.

They have been making terrific strides on the CPU side lately, so I'm really interested to see what they do with the next Radeon Pro GPUs. It seems like with the Radeon series they are strongly focusing on gaming right now, but I'm hopeful for their Pro-oriented cards.

Posted on 2019-08-22 15:18:10
Dog Has The Wheel

Hi Matt. Thank you for the fair article. I was really bummed to see that the RX 5700 XT didn't do that well in the test because I've been waiting for the 3rd party mods to come out before buying one. You said "We can hope for driver optimizations to come in the future..." What are the historical trends on new drivers bumping performance? Can we use past driver upgrades data to predict what the RX 5700 XT will maybe look like later down the road for Davinci Resolve? Thanks!

Posted on 2019-08-22 16:56:11

Really hard to say about what or how much driver improvements will make. My suspicion is that we won't see anything for pro applications like Resolve until this architecture is used on AMD's Radeon Pro line. In that case, it may trickle down to the Radeon cards as well, but if they tweak the Radeon Pro cards enough, it may not be applicable at all. There is really no way to know for sure unless AMD makes some sort of announcement about it.

Posted on 2019-08-26 19:00:01
Christopher Flannigan

Hi Matt, thanks for the great review and for your advice last week - saved me from buying the the RX 5700 XT (I think i would have cried when i read your review). I have just ordered the blackmagic pocket cinema camera 4k and plan to work in blackmagic raw - I noted this format was in your previous Radeon VII review but was not covered in this one - did you do any testing using it - I'm interested to know if the issues with the Radeon VII have been resolved.

Posted on 2019-08-22 17:21:11
jay mahony

The 2060 Super is not the same price as the 2060.The only reason the Radeon vii would be cancelled is because they are coming out with something even better. No it's not a gaming card. Yes it's still for sale. No is not normal to quote MSRP prices.

Posted on 2019-08-22 22:49:17
Misha Engel



What is tested?


Our DaVinci Resolve Studio benchmark looks at rendering with a wide
range of codecs at 4K and 8K resolutions as well as performance in
Fusion.
Our test media consists of the following resolutions and codecs:


Test Media (59.94 FPS) Youtube/soap opera effect


4K CinemaRaw Light Indie/small studio good choice, most of the timeC200 users shoot 24p, 25 or 30 fps but 60 fps is not uncommen.

4K H.264 150mbps 8-bit youtube selfie vlogger

4K ProRes 422 16-bitdoes not exist 422 10 bit does and is often required when you want to deliver to broadcasters.

4K ProRes 4444 16-bit does not exist 4444 12 bits, often used on Apple as an intermediate codec for higher end productions

4K RED What is the compression ratio, can be between 3:1 ...22:1 at 60 fps, makes a hugh difference on CPU and GPU load


8K H.265 100mbps To my knowledge, no camera can output this and it's a long GOP rgb delivery codec

8K RED What is the compression ratio, can be between 12:1 and 22:1 at 60 fps, makes a hugh difference on CPU and GPU load

Puget Systems Davinci Resolve test up to and including Davinci Resolve 15 was a better mix of codecs (24..60 fps, BRAW, cDNG, long GOP and RED code including compression ratios and frame rates people actually use on RED camera's (95% of the time 24 fps)) for a wider audience.

When Puget Systems is mainly targeting youtube/vloggers, this is a great test....

Posted on 2019-08-22 23:04:37
Ronald

LOL you created a test where Nvidia dominates, talk about bias.

Posted on 2019-08-23 01:49:44
Misha Engel

Radeon VII Allegedly Reaches End of Life Status, AMD Neither Confirms Nor Denies.
https://www.tomshardware.com/news/amd-radeon-vii-end-of-life-status,39861.html

AMD's comment.

We expect Radeon VII availability will continue to meet demand for the foreseeable future, delivering exceptional high-end 4K
gaming and content creation experiences. You can find Radeon VII graphics cards on AMD.com.
Posted on 2019-08-25 12:34:28
AlbertS

Hi Misha - Might imply they have sufficient R VII stock to last for several months and that AMD won't have a replacement until next year. Until then, R VII is the best bang for the buck for those, like me, looking at a new workstation for Resolve 16.

Posted on 2019-08-26 07:48:55
Misha Engel

Even when you leave the "buck" out it's often the best bang.

We have 3 of them, as a replacement for the GTX970 (GTX970 couldn't handle uhd/4k).

Posted on 2019-08-26 14:25:27
Nick Clemens

In your previous davinci resolve tests, the vega 64 was performing at or above the RTX 2070. Its performing a lot worse here (excluding with red footage).

Do you think this is this due to a change in your testing, or is it better support for the new Turing cores by blackmagic?

Also will you be adding BRAW to you testing suite?

Thanks so much for all your work Matt. It helps a lot!

Posted on 2019-08-26 19:33:47

It is probably a combination of a lot of different factors. Support for the new RTX features (Tensor cores in particular) shouldn't affect the things we test, but our testing has changed quite a bit in order to get it ready for public release. We are now exclusively testing render/export performance, while before we only tested live playback in the Color tab. In the future I would love to do both (maybe even add performance in the Edit and Cut tabs), but unfortunately the only way we could find to automate benchmarking for live playback is too clunky to work in a publicly available benchmark.

There are a number of other things that could explain part of the discrepancy including Windows updates, driver updates, and just general improvements in Resolve that for whatever reason affected CUDA more than OpenCL.

Posted on 2019-08-26 19:38:31
Nick Clemens

Good to know. Thanks so much for the quick response.

Posted on 2019-08-26 19:58:55
Sid

If be curious to see whether PCIe Gen 4 would make a difference for workloads that go back and forth between CPU and GPU. A comparison between a 2060 Super and 5700XT on the X570 platform would be interesting.

Thank you once again for the invaluable data you share!

Posted on 2019-08-26 21:22:29

Working on that right now! We'll be testing with the Gigabyte X570 AORUS ULTRA with a 3900X using the 5700 XT in Gen3 and Gen4, as well as a RTX 2060 SUPER in Gen3 as a comparison. Shouldn't take too much longer, we'll likely have a post up tomorrow or Wednesday unless something weird comes up that we want to get AMD's opinion on.

Posted on 2019-08-26 21:35:18
Dryden

What exact 5700 XT card was used? Was it the OEM AMD model or an AIB card? Would love to know, thanks!

Posted on 2019-08-30 02:27:42

It is one of the reference cards - gigabyte brand, although that shouldn't matter for the reference models.

Posted on 2019-08-30 02:32:21
Dryden

That would contribute quite a bit to the bad scores from the 5700 XT. The reference cards have severe cooling issues which hold back peak performance compared to AIB cards which have 2 or even 3 fans and open-air cooling.

On that note, it was stated at the start that because you tested the 5700 XT with an i9, it wouldn’t be able to take advantage of the extra boost in performance from PCIe 4.0 (because Intel chips do not yet support it) but that those tests would be coming soon. I look forward to the results, I’m sure that was the reason for the bad performance from Navi.

Posted on 2019-08-31 02:14:26

The PCIe 4.0 testing is up already - Resolve did see the biggest performance gains out of any app we tested, but it still isn't enough to catch up to the RTX 2060 Super. https://www.pugetsystems.co...

We do our testing on open air test beds, so it is really the ideal situation for cooling. It is possible the beefed up versions can maintain Turbo clocks better, but I doubt it will be enough to cover the performance gap to be honest. We kept an eye on the clock speeds and didn't see anything concerning, but we might upgrade for the next tests just to be safe.

Posted on 2019-08-31 02:59:15
Christopher Flannigan

Hi Matt, did you test Blackmagic raw? I'm interested to know if the previously reported issues with the Radeon Vii have been resolved.

Posted on 2019-09-03 22:55:01

Just checked since you asked nicely, and it looks like with Resolve Studio 16.0.0.060 and the latest AMD Radeon driver, the issue with BRAW footage we saw previously is gone. I'm not sure which exact software/driver version fixed it, but as long as you are using the latest release versions it looks like you should be good.

Posted on 2019-09-04 18:22:46
Christopher Flannigan

Thanks, that's very helpful.

Posted on 2019-09-05 07:46:33
junzhi2002 wong

I wanted to build a new pc for Davinci, CPU wise will be AMD 3700X/3800X but I do plan to game on it besides video editing and some light machine learning(some weird AI that was not going to be useful, I think so). What GPU should I get? I wanted a 2070 SUPER or higher before this article comes out...... is 5700XT better?
I am not really a pro in pc space and I am working hard to acquire more knowledge in this sector, can anyone here help me?

Posted on 2019-09-07 16:04:16
David Varela

I have the same question. If AMD fixes the drivers it will be an awesome GPU vs RTX2070

Posted on 2019-10-10 20:00:35
Marcus Sarmiento

Have the amd gpus improved since the publication of this article? Curious since I'm planning to build a pc for editing in a month or so.

Posted on 2019-10-09 07:12:22

Not that I know of. In fact, the latest AMD drivers have been causing problems in the current version of Resolve for some of our customers. Downgrading to Resolve 15 or upgrading to the latest BETA version of Resolve 16 appears to fix those issues though.

Posted on 2019-10-09 16:46:03
Turing

https://twitter.com/Blackma...

https://forum.blackmagicdes...

Matt, can you please test DaVinci Resolve 16.1 final release? Thanks.

Posted on 2019-10-18 21:23:44