My Puget Systems Benchmark Results

Get answers to your questions about color grading, editing and finishing with DaVinci Resolve.
  • Author
  • Message
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

My Puget Systems Benchmark Results

PostTue Sep 10, 2019 1:55 pm

Puget Systems has released their benchmark test over at https://www.pugetsystems.com/labs/artic ... mark-1523/. This benchmark test exercises Davinci Resolve with 5 different levels of processing on 5 different codecs.

I downloaded the benchmark and ran it on my system. Here are my results in FPS for the 4K files, my system is not quite up to the 8K tests.
psFPS.PNG
psFPS.PNG (29.43 KiB) Viewed 5116 times


While running the benchmark, I logged the loading in percent on the CPU, GPU Core, GPU Memory Controller, and GPU Memory usage. I used Open Hardware Monitor to do this. Then I took the resulting log file and generated a chart for each codec. The order of the 5 tasks applied to each codec are given in the lower right corner of the chart, to determine when the task changed over look for discontinuities in the plots.
ps4KCinemaRaw.png
ps4KCinemaRaw.png (97.08 KiB) Viewed 5116 times

ps4Kh264.png
ps4Kh264.png (94.66 KiB) Viewed 5116 times


As the forum software limits attachments to 3, I'll attempt to add them in a separate comment on this thread.
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostTue Sep 10, 2019 2:02 pm

The other 3 charts:
ps4kPR422.png
ps4kPR422.png (99.7 KiB) Viewed 4983 times
ps4KPR444.png
ps4KPR444.png (99.19 KiB) Viewed 4983 times
ps4Kred.png
ps4Kred.png (99.73 KiB) Viewed 4983 times


From reading the results of the benchmarking that Puget Systems has done, a 2080Ti GPU would be about twice as fast as my 1070. For the tasks where my GPU was 100% loaded, you'd expect the frame rate to be twice as fast. But if the frame rate increases so does the loading on the CPU so if the CPU is now loaded more than 50% it would be the bottleneck.

Again, if you want to understand the benchmark better or run it yourself, visit Puget Systems https://www.pugetsystems.com/labs/artic ... mark-1523/
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline

MishaEngel

  • Posts: 1432
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: My Puget Systems Benchmark Results

PostTue Sep 10, 2019 4:12 pm

59.94 fps seems the standard at puget.
Offline
User avatar

Jean Claude

  • Posts: 2973
  • Joined: Sun Jun 28, 2015 4:41 pm
  • Location: France

Re: My Puget Systems Benchmark Results

PostTue Sep 10, 2019 4:22 pm

MishaEngel wrote:59.94 fps seems the standard at puget.

It is surely to really stress the system?
"Saying it is good, but doing it is better! "
Win10-1809 | Resolve Studio V16.1 | Fusion Studio V16.1 | Decklink 4K Extreme 6G | RTX 2080Ti 431.86 NSD driver! |
Offline

MishaEngel

  • Posts: 1432
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: My Puget Systems Benchmark Results

PostTue Sep 10, 2019 9:43 pm

Jean Claude wrote:
MishaEngel wrote:59.94 fps seems the standard at puget.

It is surely to really stress the system?


Their customers are youtube focussed and have never seen a cinema from the inside.
Upto and including DR15 their test where pretty good, now it's useless(a Davince Resolve test without BRAW, how incomplete can you be).
Offline

Dan Sherman

  • Posts: 1185
  • Joined: Fri Jul 01, 2016 11:07 pm

Re: My Puget Systems Benchmark Results

PostTue Sep 10, 2019 10:01 pm

MishaEngel wrote:
Their customers are youtube focussed and have never seen a cinema from the inside.
Upto and including DR15 their test where pretty good, now it's useless(a Davince Resolve test without BRAW, how incomplete can you be).


That's not even close to true. YouTubers don't buy Pudget grade hardware, and they don't use 8k red footage either.

Pudget benchmarks are lacking, but not because they are targeting YouTubers.
AMD 7950X | AMD 7900XTX (23.20.24) | DDR5-6000 CL30-40-40-96 2x32 GB | Multiple PCIe 4.0 X4 NVME | ASUS x670e HERO | Win 11 Pro 23H2 | Resolve Studio 18.6.5 B7
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostTue Sep 10, 2019 10:37 pm

Dan Sherman wrote:Pudget benchmarks are lacking, but not because they are targeting YouTubers.


Puget Systems has come up with 5 different tasks for a 5 different codecs. It took over 10 GB of files and about 45 minutes to run the benchmark. How are Puget Systems' benchmarks lacking? And how many more GB of test files and how much more test time would it take to fulfill the deficiency?
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline

MishaEngel

  • Posts: 1432
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 12:13 am

Michael_Andreas wrote:
Dan Sherman wrote:Pudget benchmarks are lacking, but not because they are targeting YouTubers.


Puget Systems has come up with 5 different tasks for a 5 different codecs. It took over 10 GB of files and about 45 minutes to run the benchmark. How are Puget Systems' benchmarks lacking? And how many more GB of test files and how much more test time would it take to fulfill the deficiency?


Test Media (59.94 FPS) soap opera frame rate.
4K CinemaRaw Light C200 1Gbit/s, 59.94 fps means highly compressed(24..30 fps would be better)
4K H.264 150mbps 8-bit C200 AVC soap opera interview codec, possible
4K ProRes 422 possible, but most use this codec at a lower frame-rate 25 and 30 fps broadcast
4K ProRes 4444 typical 24 fps codec for low end movies and TV show.
4K RED What is the compression ratio? 3:1 or 22:1 makes a big difference
8K H.265 100mbps This is a delivery codec, nobody shoots this with a camera or edits it.
8K RED This is from Linus at 22:1, nobody but Linus shoot 22:1 8k.R3D for more than 5% of it's time(5:1 at 24 fps is commen when you value picture quality).

24,25 and 30 fps are the most used framerates (59.94/60 isn't).
BRAW is missing
ARRIRAW is missing
Sony X-OCN is missing
Offline

Dan Sherman

  • Posts: 1185
  • Joined: Fri Jul 01, 2016 11:07 pm

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 2:00 am

Michael_Andreas wrote:
Dan Sherman wrote:Pudget benchmarks are lacking, but not because they are targeting YouTubers.


Puget Systems has come up with 5 different tasks for a 5 different codecs. It took over 10 GB of files and about 45 minutes to run the benchmark. How are Puget Systems' benchmarks lacking? And how many more GB of test files and how much more test time would it take to fulfill the deficiency?


First, I'm talking about the reviews they have been publishing on their blog for years, as that is what the vast majority of people look at.

for example this one.
https://www.pugetsystems.com/labs/artic ... ethodology

All the benchmarks tell you is what fps is maintained under the following scenarios.
  • basic color correction + 4 power windows
  • basic color correction + 4 power windows + 3 OpenFx
  • basic color correction + 4 power windows + 3 OpenFx
  • basic color correction + 4 power windows + 3 OpenFx + TNR

What these benchmarks don't tell you.
  • Is the cpu & gpu bouncing off 100% utilization, or is quietly working away at 30%
  • how much ram is actually being used. They always use 128GB for all their tests.
  • How are are they stressing the drives. They always use 1tb m.2 nvme drives, because they never test with a large amount of footage.


And yes even the studio benchmark is lacking.
https://www.pugetsystems.com/labs/artic ... mark-1523/

This is all it tests.
For the 4K and 8K render tests, we benchmark 5 different levels of grade:

"Optimized Media" - No effects applied in order to simulate the performance when generating optimized media
"Basic Grade" - Simple color wheel and similar adjustments with a base grade plus 4 power windows.
"OpenFX - Lens Flare + Tilt Shift Blur + Sharpen" - Basic Grade plus four OpenFX
"Temporal Noise - Better 2 Frames" - Basic Grade plus a single TNR node
"3x Temporal Noise - Better 2 Frames" - Basic Grade plus three TNR nodes using splitter/combiner nodes
The "Optimized Media" timeline is rendered out to MXF OP1A DNxHR LB at 1920x1080 while the others are all rendered out to Quicktime DNxHR HQ at the native resolution of the timeline (UHD or 8K).



This doesn't test.
  • how Fairlight effects/plugins effects playback and rendering
  • How transitions effect playback and rendering
  • how hardware will handle multiple streams in multi-cam or sync bins scenarios
  • how various titles effect playback and rendering

Resolve is a lot more than just grading!
AMD 7950X | AMD 7900XTX (23.20.24) | DDR5-6000 CL30-40-40-96 2x32 GB | Multiple PCIe 4.0 X4 NVME | ASUS x670e HERO | Win 11 Pro 23H2 | Resolve Studio 18.6.5 B7
Offline

Dan Sherman

  • Posts: 1185
  • Joined: Fri Jul 01, 2016 11:07 pm

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 2:03 am

MishaEngel wrote:
Michael_Andreas wrote:
Dan Sherman wrote:Pudget benchmarks are lacking, but not because they are targeting YouTubers.


Puget Systems has come up with 5 different tasks for a 5 different codecs. It took over 10 GB of files and about 45 minutes to run the benchmark. How are Puget Systems' benchmarks lacking? And how many more GB of test files and how much more test time would it take to fulfill the deficiency?


Test Media (59.94 FPS) soap opera frame rate.
4K CinemaRaw Light C200 1Gbit/s, 59.94 fps means highly compressed(24..30 fps would be better)
4K H.264 150mbps 8-bit C200 AVC soap opera interview codec, possible
4K ProRes 422 possible, but most use this codec at a lower frame-rate 25 and 30 fps broadcast
4K ProRes 4444 typical 24 fps codec for low end movies and TV show.
4K RED What is the compression ratio? 3:1 or 22:1 makes a big difference
8K H.265 100mbps This is a delivery codec, nobody shoots this with a camera or edits it.
8K RED This is from Linus at 22:1, nobody but Linus shoot 22:1 8k.R3D for more than 5% of it's time(5:1 at 24 fps is commen when you value picture quality).

24,25 and 30 fps are the most used framerates (59.94/60 isn't).
BRAW is missing
ARRIRAW is missing
Sony X-OCN is missing



Honestly, I think this is because they only test with what they can freely get their hands on, or someone gives to them.
AMD 7950X | AMD 7900XTX (23.20.24) | DDR5-6000 CL30-40-40-96 2x32 GB | Multiple PCIe 4.0 X4 NVME | ASUS x670e HERO | Win 11 Pro 23H2 | Resolve Studio 18.6.5 B7
Offline
User avatar

Marc Wielage

  • Posts: 10901
  • Joined: Fri Oct 18, 2013 2:46 am
  • Location: Hollywood, USA

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 2:34 am

Dan Sherman wrote:All the benchmarks tell you is what fps is maintained under the following scenarios.
  • basic color correction + 4 power windows
  • basic color correction + 4 power windows + 3 OpenFx
  • basic color correction + 4 power windows + 3 OpenFx
  • basic color correction + 4 power windows + 3 OpenFx + TNR


Small clarification: not all OFX plug-ins are created equal. A Dust Fix OFX plug-in won't drag as much as a Lens Flare. Lens Flare may not be as bad as Face Recognition. Face Recognition won't be as bad as a BorisFX OFX plug-in. And the BorisFX won't be as bad as the Neat Video OFX. I think they need to specify which OFX plug-ins they're using, because even the ResolveFX plugs from BMD react differently with different material and different speeds. And Performance Mode affects these results if you select an uncompressed format vs. slightly compressed vs. more lossy.

I typically don't do less than 15-16 nodes, and I might have 4 or 5 simultaneous windows in quite a few shots (including vignettes), sometimes more. Masks & Shapes are generally not the problem; TNR, SNR, blurs, and problem source material are far more challenging.
marc wielage, csi • VP/color & workflow • chroma | hollywood
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 3:41 am

Let's review the purpose of these benchmarks. They are to setup some standardize usage cases so that different hardware configurations can be compared A Vs B. Since there are a wide range of codecs and an even wider range of variance in how much processing can be applied, it is necessary to limit the number of cases. Since the universe doesn't revolve around any one person, it's unlikely that the number of usage cases selected will match any one person perfectly. If you feel that the benchmark doesn't match your usage closely enough to make it worthwhile, then ignore it. Or create your own benchmark, make all of the files available on a hosting site, and persuade lots of people to download the benchmark, run it, and publish their results.

Responding to some specific objections:
--No cameras encode to h.264 or h.265. Maybe yours don't, but lots of DSLRs, action cameras, and drones do.
--Most filming is done at less than 59.94/60 Hz. Well yes, probably. Why does that make the benchmark worthless? If Codex X recorded at 60 FPS is processed at framerate Y, do we expect Codex X recorded at 30 FPS to be processed at a framerate significantly different than Y?
--The benchmarks don't tell you which part of your system is being stressed. Absolutely agree, which is why I ran the Open Hardware Monitor and published the results. Can you point to a better Benchmark software that gives the kind of insight you think it should without using another tool?

For an example of how to use the benchmark, I could compare my usage case to one of the codecs and a similar amount of correction processing and evaluate whether it would be worthwhile upgrading to a 2080Ti given that their benchmark runs give about a 2x increase in performance. I could put the benchmark files on a spinning hard drive to see how that effected it. YMMV.
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 4:51 am

Coincidentally, there has been some discussion over at the Puget System website on this issue. https://www.pugetsystems.com/labs/artic ... mark-1523/
Scroll down to the comments and read that Puget System will add more codecs if they can get the rights to use the footage in their benchmark. If you have a camera with a unique codec, offer to donate 14 seconds of footage to them.
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline

Andrew Kolakowski

  • Posts: 9207
  • Joined: Tue Sep 11, 2012 10:20 am
  • Location: Poland

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 8:22 am

Those tests take quite a lot of time.
In the same time it should not be that difficult to add more formats, eg. ARRI has plenty of footage on their website which can be used for such s tests. They can also ask user for some footage. It's just quite time consuming.
Offline

Dan Sherman

  • Posts: 1185
  • Joined: Fri Jul 01, 2016 11:07 pm

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 1:19 pm

Michael_Andreas wrote:Since there are a wide range of codecs and an even wider range of variance in how much processing can be applied, it is necessary to limit the number of cases. Since the universe doesn't revolve around any one person, it's unlikely that the number of usage cases selected will match any one person perfectly.


No one said the benchmarks need to match any one persons workflow/use cases. However, when they negate entire sections of the application it's an issue. As I mentioned Fairlight can have a large impact on playback performance if you are running something like NR or de-reverb.


Michael_Andreas wrote:--The benchmarks don't tell you which part of your system is being stressed. Absolutely agree, which is why I ran the Open Hardware Monitor and published the results.


Michael_Andreas wrote:For an example of how to use the benchmark, I could compare my usage case to one of the codecs and a similar amount of correction processing and evaluate whether it would be worthwhile upgrading to a 2080Ti given that their benchmark runs give about a 2x increase in performance. I could put the benchmark files on a spinning hard drive to see how that effected it. YMMV.


I think you missed my point. You could upgrade to a 2080ti and find that you only see a 10% to 20% performance increase, even though your cpu and gpu are both well under 100%. The reason could be something as simple as the memory controller in the cpu or the gpu can't keep up.
AMD 7950X | AMD 7900XTX (23.20.24) | DDR5-6000 CL30-40-40-96 2x32 GB | Multiple PCIe 4.0 X4 NVME | ASUS x670e HERO | Win 11 Pro 23H2 | Resolve Studio 18.6.5 B7
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 1:24 pm

The Open Hardware Monitor software I used to track loading during my test has dozens and dozens of parameters it monitors. I selected 4 of them so the charts wouldn't look like plates of spaghetti. One of the plots happens to be the GPU memory controller.
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline

MishaEngel

  • Posts: 1432
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 2:08 pm

Michael_Andreas wrote:Responding to some specific objections:
--No cameras encode to h.264 or h.265. Maybe yours don't, but lots of DSLRs, action cameras, and drones do.

Don't be a politician
"8K H.265 100mbps This is a delivery codec, nobody shoots this with a camera or edits it."
"4K H.264 150mbps 8-bit C200 AVC soap opera interview codec, possible"

Michael_Andreas wrote:--Most filming is done at less than 59.94/60 Hz. Well yes, probably. Why does that make the benchmark worthless? If Codex X recorded at 60 FPS is processed at framerate Y, do we expect Codex X recorded at 30 FPS to be processed at a framerate significantly different than Y?


59.94/60 fps 8k.R3D 22:1 runs on a laptop, 23.976/24 fps 8k.R3D 5:1 cripples a $10k workstation with an i9-9980XE and a RTX Quadro 8000.

Michael_Andreas wrote:--The benchmarks don't tell you which part of your system is being stressed. Absolutely agree, which is why I ran the Open Hardware Monitor and published the results. Can you point to a better Benchmark software that gives the kind of insight you think it should without using another tool?


When they only way to go from the US to Europe is swimming, doesn't make at a good solution.

Michael_Andreas wrote:For an example of how to use the benchmark, I could compare my usage case to one of the codecs and a similar amount of correction processing and evaluate whether it would be worthwhile upgrading to a 2080Ti given that their benchmark runs give about a 2x increase in performance. I could put the benchmark files on a spinning hard drive to see how that effected it. YMMV.


For a GPU 3 things are important: fp32 performance, amount of memory and the memory speed.
A RTX2080ti is fast enough to de-bayer 8k.R3D 5:1 24 fps in realtime, it lacks the amount of memory to do (heavy) TNR and the memory speed is not enough to make full use of the fp32 perfomance.
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 2:53 pm

Dan Sherman wrote:No one said the benchmarks need to match any one persons workflow/use cases. However, when they negate entire sections of the application it's an issue. As I mentioned Fairlight can have a large impact on playback performance if you are running something like NR or de-reverb.

Since the benchmark test is a test of the hardware while running software, does Fairlight audio processing stress the hardware in significantly different ways than the video processing does? If, so then maybe a Fairlight benchmark is needed just as they already have a Fusion benchmark. Maybe something unchallenging on the video side such as a color bar test pattern at 1080p in an optimized codec, together with some audio files and a range of processing tasks from light loading to heavy loading. Puget Systems has indicated willingness to add other scenarios to their benchmark, you could offer something to them and see if they're interested.
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline
User avatar

Jean Claude

  • Posts: 2973
  • Joined: Sun Jun 28, 2015 4:41 pm
  • Location: France

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 4:48 pm

(maybe I read it wrong, English is not my native language)
Anyway :
what is their reference result => not found :?:
"Saying it is good, but doing it is better! "
Win10-1809 | Resolve Studio V16.1 | Fusion Studio V16.1 | Decklink 4K Extreme 6G | RTX 2080Ti 431.86 NSD driver! |
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 5:02 pm

Jean Claude wrote:(maybe I read it wrong, English is not my native language)
Anyway :
what is their reference result => not found :?:

I did not see a "not found" message. Nor could I find one when inspecting the "statuslog.txt" file produced by the benchmark script. Where did you see one?
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline
User avatar

Jean Claude

  • Posts: 2973
  • Joined: Sun Jun 28, 2015 4:41 pm
  • Location: France

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 5:09 pm

Michael_Andreas wrote:
Jean Claude wrote:(maybe I read it wrong, English is not my native language)
Anyway :
what is their reference result => not found :?:

I did not see a "not found" message. Nor could I find one when inspecting the "statuslog.txt" file produced by the benchmark script. Where did you see one?


Mickael -> not from you :
What score or result has Puget with their CPU + GPU from Pudget :?:
The scoring system used in our benchmark is based on Intel Core i9 9900K and NVIDIA Titan
RTX 24GB.
"Saying it is good, but doing it is better! "
Win10-1809 | Resolve Studio V16.1 | Fusion Studio V16.1 | Decklink 4K Extreme 6G | RTX 2080Ti 431.86 NSD driver! |
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 5:14 pm

From the Puget System article:
"When the benchmark is complete, it will give you an "Overall Score" for the type of test you ran. A log file is generated in the benchmark folder that includes these scores as well as the FPS for each individual task. You can compare your scores to those in our DaVinci Resolve hardware articles - just make sure they used the same benchmark version!"

So you have to take your results and compare them to Puget Systems' results for testing different hardware configurations in one of their other articles: https://www.pugetsystems.com/all_articles.php?filter[]=DaVinci%20Resolve
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline
User avatar

Jean Claude

  • Posts: 2973
  • Joined: Sun Jun 28, 2015 4:41 pm
  • Location: France

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 5:22 pm

Michael_Andreas wrote:From the Puget System article:
"When the benchmark is complete, it will give you an "Overall Score" for the type of test you ran. A log file is generated in the benchmark folder that includes these scores as well as the FPS for each individual task. You can compare your scores to those in our DaVinci Resolve hardware articles - just make sure they used the same benchmark version!"

So you have to take your results and compare them to Puget Systems' results for testing different hardware configurations in one of their other articles: https://www.pugetsystems.com/all_articles.php?filter[]=DaVinci%20Resolve


Thank you.
"Saying it is good, but doing it is better! "
Win10-1809 | Resolve Studio V16.1 | Fusion Studio V16.1 | Decklink 4K Extreme 6G | RTX 2080Ti 431.86 NSD driver! |
Offline

Dan Sherman

  • Posts: 1185
  • Joined: Fri Jul 01, 2016 11:07 pm

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 5:30 pm

Michael_Andreas wrote:The Open Hardware Monitor software I used to track loading during my test has dozens and dozens of parameters it monitors. I selected 4 of them so the charts wouldn't look like plates of spaghetti. One of the plots happens to be the GPU memory controller.


Yes, and take note of the fact that at times you are well above 90% in each test.

a 1070 only has 8 mcs while a 2080 ti has 11, so not a substantial increase. Nvidia muddied the waters some because they made substantial changes to how L1 and L2 cache are handled between Pascal and Turing.
AMD 7950X | AMD 7900XTX (23.20.24) | DDR5-6000 CL30-40-40-96 2x32 GB | Multiple PCIe 4.0 X4 NVME | ASUS x670e HERO | Win 11 Pro 23H2 | Resolve Studio 18.6.5 B7
Offline

Dan Sherman

  • Posts: 1185
  • Joined: Fri Jul 01, 2016 11:07 pm

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 5:39 pm

Michael_Andreas wrote:
Dan Sherman wrote:No one said the benchmarks need to match any one persons workflow/use cases. However, when they negate entire sections of the application it's an issue. As I mentioned Fairlight can have a large impact on playback performance if you are running something like NR or de-reverb.

Since the benchmark test is a test of the hardware while running software, does Fairlight audio processing stress the hardware in significantly different ways than the video processing does? If, so then maybe a Fairlight benchmark is needed just as they already have a Fusion benchmark.


Audio work in general is is single threaded and thus heavily reliant on the processor frequency, so its at odds with what's best for video.
AMD 7950X | AMD 7900XTX (23.20.24) | DDR5-6000 CL30-40-40-96 2x32 GB | Multiple PCIe 4.0 X4 NVME | ASUS x670e HERO | Win 11 Pro 23H2 | Resolve Studio 18.6.5 B7
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 5:43 pm

I'm well aware that I have loading above 90%. In fact, that's what I look for to see which component is the bottleneck. Generally, the CPU is loaded above 90% for minimal processing switching over to the GPU being loaded above 90% as more correction and TNR is added. If my results have both the CPU and GPU below 90% loading, then I'll look for another component such as the storage or memory as the bottleneck. (Which would mean adding some more "pieces of spaghetti" to the "plate".)

My 2x estimate for 2080Ti performance is based on Puget System benchmark tests. Actually it will be some time before I upgrade, so I'll look at it more closely then. Looking into the nuts and bolts at this point gives me better insight on how to use my equipment in the mean time.
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline

Michael_Andreas

  • Posts: 1672
  • Joined: Sat Jan 05, 2019 9:40 pm
  • Real Name: Michael Andreas

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 5:47 pm

Dan Sherman wrote:Audio work in general is is single threaded and thus heavily reliant on the processor frequency, so its at odds with what's best for video.


As I understand it, Fusion also is single threaded. In which case, hardware choices for Fusion could also improve Fairlight performance.
Last edited by Michael_Andreas on Wed Sep 11, 2019 5:48 pm, edited 1 time in total.
_________________________________________________
DR Studio 17.4.1 Win10Pro 21H1/19043.1320 - i7-6700K@4GHz, 32GB RAM
RTX 2070 8GB, "Studio" driver 472.39
OS,Library: 1TB SSD - Project: 1TB SSD - Cache: 1TB NVMe
Offline
User avatar

Jean Claude

  • Posts: 2973
  • Joined: Sun Jun 28, 2015 4:41 pm
  • Location: France

Re: My Puget Systems Benchmark Results

PostWed Sep 11, 2019 5:48 pm

Dan Sherman wrote:....
Audio work in general is is single threaded and thus heavily reliant on the processor frequency, so its at odds with what's best for video.


Hello Dan,
Yet, some say that if you use certain VST (Izotope or other):
Between the video and the audio of Delivery, the sound is no longer synchronous :?: :?:
"Saying it is good, but doing it is better! "
Win10-1809 | Resolve Studio V16.1 | Fusion Studio V16.1 | Decklink 4K Extreme 6G | RTX 2080Ti 431.86 NSD driver! |

Return to DaVinci Resolve

Who is online

Users browsing this forum: Bing [Bot], Charles Bennett, jakobustrop, loepark, LuciaDGE, panos_mts, Steve Alexander and 168 guests