AMD VCE Rendering Questions

Comments

gpasko wrote on 1/18/2019, 12:59 PM

More AMD VCE fodder...

UHD to UHD

FHD to FHD

FHD-120 FPS to FHD-30 FPS

I can't wait to give this another try. Hopefully, I'll be able to get to this tonight.

BTW- How much RAM does your rx480 have? I have the smaller 4GB card.

john_dennis wrote on 1/18/2019, 1:25 PM

I went with 8 GB because my upgrade cycle is four years. 

Don't discount the work done by 6 physical cores with hyper threading.

gpasko wrote on 1/18/2019, 7:55 PM

Sorry for my missing my mouse cursor in this video. I used OBS to screen capture.

I just made a video of my CPU vs GPU w/ VCE on my RX480 4GB reference card that has a bios upgrade to a RX580. I used the standard old school 1080P Red Car video test. My CPU is a dual Xeon E5-2690 (16 full CPU's w 32GB RAM) w/ hyper threading disabled as it seems to slow down renderings ever so slightly. After uploading this video I tried rendering with VCE in 720P and I was getting pretty fast renderings (30 to 100fps+) similar to my dual GPU setup using "legacy rendering settings" using a single-side of a GTX590 and RX480 or even a R9-270X card. Is there some setting that I'm missing where anything 1080P and up slows down my FPS down to a very low 6fps?

bob-h wrote on 1/18/2019, 9:05 PM

The VCE encoding engine that your card uses is known for being slower and lower quality than Nvenc or QSV but 6fps at 1080p where you are not bottle necked by cpu/gpu surely is too slow?

 

gpasko wrote on 1/19/2019, 1:51 PM

Since it renders quickly at 720P could it be a RAM limitation of a 4GB card to render anything 1080P or higher at the slower pace? Has anyone else tested any of AMD 4GB card?... I’m considering unsinstalling all iterations of Vegas and reinstalling v15.

JN_ wrote on 1/19/2019, 3:26 PM

I don’t have a vce enabled machine, but I was able to get the use of a 4gb laptop to test. It has VCE VER. 3.1 as best I can make out ...

Laptop ...

Amd A6-9210 Radeon R4  7th. Generation.  Stoney Ridge, AMD Radeon R4 Graphics, 3rd Gen Graphics Core Next.

Video Encoding ... VCE 3.1. Video Decoding ..  UVD 6.3.

I updated the BruceUSA graph with the results, using Wwaag's app. I wanted to see what effect varying the data rate a bit would have, quite a bit actually in the pecking order.

The above may not be strictly speaking addressing you VCE query, but what I did find was that I simply could not render to a relatively high data rate (135/100) or even close to it. I was able to do so on a higher specced laptop using Nvenc and a PC for Qsv.

So either the Vce encoder is limited by a lower specced machine or its not yet integrated within VP as it should/can be.

 

This above graph has the higher data rates that I was attempting to add a VCE value to ...

 

At the end of the day I don't think it matters an awful lot, I mean if you have a VCE machine or an Nvidia, or QSV enabled one, that's what you will use if you intend using hardware encoding.  There isn't a huge difference between them anyway, certainly data rate is important..

Some updated tables ... https://www.dropbox.com/sh/xh8vkexgxif53m5/AAAH135JcFgIuqoQgFp5-o1za?dl=0

 

 

 

 

Last changed by JN_ on 5/15/2019, 5:18 AM, changed a total of 10 times.

 

Desktop and Laptop basic specs ...

Both run Win 10 ...

Running latest ver. of Vegas Pro with latest updates.

VP13 B453 also.

Vegaseur and Pluraleyes installed on both ...

Quicktime 7.79.80.95

PC ...

i9 9900K, Intel Graphics 630. Nov 2018.

Mem. 32gb DDR4 Nov. '18.

Graphics card .. Nvidia Rtx 2080 Ti

Nvidia Graphics driver .. latest Studio driver.

Latest Intel Graphics driver

 

Laptop ... (Acer Predator G9-793-77AC)

CPU .. i7-6700HQ Skylake-H

Memory ..16GB DDR4 

Graphics card .. Nvidia GTX 1070, latest Studio driver.

gpasko wrote on 1/19/2019, 7:34 PM

I uninstalled NVIDIA drivers, AMD Radeon drivers, and other non-used programs on my PC. After re-installing the AMD driver I was greeted with an unfinished install error. Ultimately I used a free program to clear out any GPU drivers installs that was not completely uninstalled. It was recommended by AMD and is called Display Driver Uninstaller. I did this in SAFE MODE as that is what was recommended by the DDU program. After doing this and removing any remnants of NVIDIA and AMD I successfully installed AMD's driver. My rendering times had improved to be almost as fast as the "legacy rendering settings" using NVidia's Fermi and AMD's OpenCL. Here's a video of the RED CAR 1080P on my system (btw- this time I did disable hyperthreading) using straight CPU and VCE renderings in Vegas Pro 15.

Thanks for everyone's responses... This confirms RX480/RX580 4GB cards working just fine. I'd even guess that any other AMD with VCE 3.0 would be close to the same speed (other than the GPU's RAM speed).

OldSmoke wrote on 1/20/2019, 7:55 AM

@gpasko I think your CPU or the disabled Hyperthreading is holding you back, the same project renders on my system (see signature) in 20sec.

Last changed by OldSmoke on 1/20/2019, 7:57 AM, changed a total of 1 times.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

BruceUSA wrote on 1/20/2019, 10:44 AM

I am getting between 14s and 15s at over 200 frames.

gpasko wrote on 1/20/2019, 12:41 PM

@gpasko I think your CPU or the disabled Hyperthreading is holding you back, the same project renders on my system (see signature) in 20sec.

I’m sure it’s my CPU. Xeons limited clock speed is its downfall. No overclocking and relatively lower turbo speed vs I5’s or I7 and the newer clocks on AMDs. Plus the Xeon is limited on the number of CPUs that actually get to the highest turbo. On Vegas 11 Pro when I had a standard clocked GTX580 and a R9 270X using the “legacy” rendering I got 26 sec on the RED CAR 1080p. W a GTX590 and the same AMD I think I got 27 sec. My guess has been my Xeons, but I’m still ok w my setup.

gpasko wrote on 1/20/2019, 2:03 PM

@gpasko I think your CPU or the disabled Hyperthreading is holding you back, the same project renders on my system (see signature) in 20sec.

I have a AMD R9-270 installed back into my PC and re-ran a few iterations of Red Car 1080P

With Hyper-Threading ON (32 CPU threads) here are my results:

36 to 38 sec; Magix AVC - VCE (multiple runs)

1 min 04 sec; Magix AVC - CPU

1 min 30 sec; Sony AVC - CPU

18 to 19 sec; XDCAM EX (multiple runs)

With Hyper-Threading OFF (16 CPU threads)

33 to 35sec; Magix AVC - VCE (multiple runs)

58 sec; Magix AVC - CPU

1 min 20 sec; Sony AVC - CPU

18 sec; XDCAM EX

I'd say over-all hyperthreading off is better on my system. It is interesting to see that there isn't much of a speed difference between VCE 1.0 and VCE 3.0 on the RX480/RX580 mentioned previously. One difference between the two card comparison is that I cannot render Magix AVC - VCE at higher than 1080P w/ VCE 1.0. I'm considering upgrading my gaming PC's system to a R9 Nano and will run some tests w/ that in my Video Editing Workstation. Any feedback would be great!

OldSmoke wrote on 1/20/2019, 4:23 PM

I would try and leave hyperthreading ON and reduce the amount of threads inside Vegas under preferences.

Last changed by OldSmoke on 1/20/2019, 7:21 PM, changed a total of 1 times.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

bob-h wrote on 1/20/2019, 6:13 PM

 

I'd say over-all hyperthreading off is better on my system. It is interesting to see that there isn't much of a speed difference between VCE 1.0 and VCE 3.0 on the RX480/RX580 mentioned previously.

Your RX480/580 is VCE3.4 which is slower than VCE3.1, which is slower than VCE3.0 for h.264 encoding. So AMD did improve the power of encoding engine, but by the time of your card they were going backwards.

gpasko wrote on 1/20/2019, 6:59 PM

Inqould tryband leave hyperthreading ON and reduce the amount of threads inside Vegas under preferences.

Surprising results IMO w/ little variance between allowable rendering threads:

Magix VCE still using R9 270

Hyper Threading ON

39 sec (8) threads 200MB RAM Preview (on all)

41 sec (12) threads

38 sec (16) threads

39 sec (20) threads

38 sec (24) threads

39 sec (28) threads

Hyper Threading OFF

34 sec (8) threads 200MB RAM Preview (on all)

35 sec (12) threads

35 sec (16) threads

35 sec (20) threads

34 sec (24) threads

35 sec (28) threads

 

gpasko wrote on 1/20/2019, 7:02 PM

 

I'd say over-all hyperthreading off is better on my system. It is interesting to see that there isn't much of a speed difference between VCE 1.0 and VCE 3.0 on the RX480/RX580 mentioned previously.

Your RX480/580 is VCE3.4 which is slower than VCE3.1, which is slower than VCE3.0 for h.264 encoding. So AMD did improve the power of encoding engine, but by the time of your card they were going backwards.

Which cards have VCE3.1? What is the order of VCE speed? Wiki page lists RX480/RX580 as VCE 3.0. My R9 270 is listed as VCE 1.0 which seems to be about the same speed as the RX480 that I swapped in.

bob-h wrote on 1/20/2019, 7:51 PM

Looking at that wiki the fastest VCE h.264 encoding (vce3.0) was found on these cards:

Radeon R9 285, 380, 380X; Mobile Radeon R9 M390X, M395, M395X, M485X, FirePro W7100, S7100X, S7150, S7150 X2

3.1 slower and 3.4 slower again. From memory VCE2.0 is faster than VCE3.4

john_dennis wrote on 1/20/2019, 8:10 PM

@gpasko I think your CPU or the disabled Hyperthreading is holding you back, the same project renders on my system (see signature) in 20sec.

I have a AMD R9-270 installed back into my PC and re-ran a few iterations of Red Car 1080P

With Hyper-Threading ON (32 CPU threads) here are my results:

36 to 38 sec; Magix AVC - VCE (multiple runs)

56 sec; Magix AVC - CPU

1 min 30 sec; Sony AVC - CPU

18 to 19 sec; XDCAM EX (multiple runs)

With Hyper-Threading OFF (16 CPU threads)

33 to 35sec; Magix AVC - VCE (multiple runs)

58 sec; Magix AVC - CPU

1 min 20 sec; Sony AVC - CPU

18 sec; XDCAM EX

I'd say over-all hyperthreading off is better on my system. It is interesting to see that there isn't much of a speed difference between VCE 1.0 and VCE 3.0 on the RX480/RX580 mentioned previously. One difference between the two card comparison is that I cannot render Magix AVC - VCE at higher than 1080P w/ VCE 1.0. I'm considering upgrading my gaming PC's system to a R9 Nano and will run some tests w/ that in my Video Editing Workstation. Any feedback would be great!


I suspect that many of your references to "Magix AVC - VCE" should actually be Magix AMD - VCE.

gpasko wrote on 1/20/2019, 8:36 PM


I suspect that many of your references to "Magix AVC - VCE" should actually be Magix AMD - VCE.

Either way but it is under Magix AVC/ACC MP4

 

gpasko wrote on 1/20/2019, 9:02 PM

In deed it is.

It would be nice to test all the AMD's rendering VCE chipsets 1.0, 2.0, 3.0, 3.1, 3.4, etc on one base system to develop a base point in speed and then ultimately a visual comparison. I think it's really great progress from Magix's development team!

john_dennis wrote on 1/20/2019, 9:30 PM

"I think it's really great progress from Magix's development team!"

I think so too. Here are a few quality measurements that I've done recently. AMD VCE holds its own for any close Bits/Pixel*Frame measurement in speed and quality.

MSE (Lower is better)    PSNR (Higher is Better)

Date: 2019/01/15  13:35:12 
Description: x264vfw CRF20 Zero Latency
Frames Processed: 659
Mean Squared Error: 15.892
Peak Signal to Noise Ratio: 36.527

Date: 2019/01/15  13:44:19 
Description: Happy Otter Scripts CRF20 Zero Latency 1080-59-94p 11 Seconds
Frames Processed: 659
Mean Squared Error: 14.181
Peak Signal to Noise Ratio: 39.159

Date: 2019/01/15  14:18:07 
Description: Vegas 15 AMD VCE 31 Mbps
Frames Processed: 659
Mean Squared Error: 12.982
Peak Signal to Noise Ratio: 39.598

Date: 2019/01/16  21:49:44 
Description: Magix Intermediate 422 HQ UHD
Frames Processed: 899
Mean Squared Error: 0.492
Peak Signal to Noise Ratio: 55.822

Date: 2019/01/16  22:49:55 
Description: Happy Otter Scripts CRF18 GOP30
Frames Processed: 899
Mean Squared Error: 4.123
Peak Signal to Noise Ratio: 47.066

Date: 2019/01/17  07:23:22 
Description: Magix HEVC Internet 4K 2160p 29-97 fps AMD VCE 41 Seconds
Frames Processed: 899
Mean Squared Error: 7.46
Peak Signal to Noise Ratio: 44.158

I'm currently running a pair of 10,000 frame tests comparing AMD VCE to another encoding method at twice the bit rate. We'll see how that turns out in a while

gpasko wrote on 1/21/2019, 8:05 AM

Looking at that wiki the fastest VCE h.264 encoding (vce3.0) was found on these cards:

Radeon R9 285, 380, 380X; Mobile Radeon R9 M390X, M395, M395X, M485X, FirePro W7100, S7100X, S7150, S7150 X2

3.1 slower and 3.4 slower again. From memory VCE2.0 is faster than VCE3.4

Very interesting! Would these cards w/ VCE 3.0 be faster at rendering vs a Vega card? A used R9-285 card can be had for under $100 bucks. For Vegas video editing what would be the downside using an old card like these vs a Fury X or a Vega 56?

OldSmoke wrote on 1/21/2019, 9:07 AM

@gpasko You also have to realize that eventually the GPU will have to wait for the CPU to feed it the frame and that is why higher clocked CPUs are doing better in Vegas compared to lower clocked high core count CPUs. Also, the Fury X is VCE 3.0.

Last changed by OldSmoke on 1/21/2019, 9:10 AM, changed a total of 2 times.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

gpasko wrote on 1/21/2019, 9:46 AM

@gpasko You also have to realize that eventually the GPU will have to wait for the CPU to feed it the frame and that is why higher clocked CPUs are doing better in Vegas compared to lower clocked high core count CPUs. Also, the Fury X is VCE 3.0.

I understand that there's a balance between components. That's part of the fun w/ new technologies. What would your best guess be between running a VCE rendering comparison on the same PC between a R9-380 (4GB) vs a Fury X (4GB)? Would we get the same performance results? I suppose it might be slightly in favor to the Fury X because of its RAM speed. I'm just not sure how much a difference the GPU clock speed makes.