Comments

Kinvermark wrote on 10/4/2018, 12:12 PM

Nothing. Measuring the cpu / gpu load is like measuring a car's tachometer rather than the speedometer.

Test your performance using real-world metrics like render time, playback fps under load, etc.

zaheer-abbas wrote on 10/4/2018, 12:25 PM

this is exactly what i have done, when i do a bench mark test or a cpu stress test the cpu show 100% percent useage on all cores. When i render on any of my intel based cpus the first thing to be used is the cpu and it hits 100 percent right away which shows its rendering at max speed. So in theory i should be getting roughly half the render times on my 16 core as compared to my intel 8 core. At the moment its seems to be the same speed. Very frustrating as i though vegas 16 would utlise all cores and all threads to the max, unless sothing is bottle necking

 

fr0sty wrote on 10/4/2018, 1:36 PM

The point is, some codecs are not going to utilize multiple cores as effectively, the only way to speed them up is to make the individual cores faster and more powerful. Some codecs will use the GPU better than any CPU will, many can't use it at all. There are aspects of Vegas that your threadripper is making much faster, but for the things that do not scale well across multiple cores, you aren't going to notice a huge improvement if any. Throwing 2x the cores at something almost never results in 2x the performance, the only real exception being things like Cinema 4D CPU rendering that scale almost perfectly across multiple cores.

Think about it this way, I need to empty a pond. I have 3 big trucks that have high speed pumps on them and can fill their tanks in 20 minutes each. I also have 20 smaller trucks, that hold the same amount of water together as the 3 big trucks, but it takes 10 minutes for each to fill their tanks. Which group of trucks is going to drain that pond the fastest?

Now, I have 20 ponds that need to be drained, a 10 minute drive apart from each. Which group, the 3 big or 20 small, will drain those the fastest?

Now you see why throwing more cores can improve some situations, but sometimes fewer more powerful cores is the better option. Some tasks in Vegas do well across multi-cores, but others need stronger individual cores. For rendering, first see if GPU acceleration is able to help, as it is usually faster than any CPU. Magix AVC and HEVC support it.

Last changed by fr0sty on 10/4/2018, 1:46 PM, changed a total of 5 times.

Systems:

Desktop

AMD Ryzen 7 1800x 8 core 16 thread at stock speed

64GB 3000mhz DDR4

Geforce RTX 3090

Windows 10

Laptop:

ASUS Zenbook Pro Duo 32GB (9980HK CPU, RTX 2060 GPU, dual 4K touch screens, main one OLED HDR)

zaheer-abbas wrote on 10/4/2018, 2:07 PM

I tried something that got some interesting results, i disabled 8 cores on the 1950x and the cpu usage shot up to over 90%. So clearly Vegas doesn't suport more then 8 Cores. I would have though Vegas 15 and 16 would have used them as all my setting remaing the same, ie codec , bitrates, frame rates, resolution, etc and theres no noticable imporvement. Ill keep trying

fr0sty wrote on 10/4/2018, 2:52 PM

What that could also mean is that the type of work being calculated doesn't scale well above 8 cores, so the 16 cores all get used, but not to their fullest capability. Vegas supports a max of 32 render threads, which is all the threads your threadripper can throw at it, so that shouldn't be the issue unless there is a bug there.

Systems:

Desktop

AMD Ryzen 7 1800x 8 core 16 thread at stock speed

64GB 3000mhz DDR4

Geforce RTX 3090

Windows 10

Laptop:

ASUS Zenbook Pro Duo 32GB (9980HK CPU, RTX 2060 GPU, dual 4K touch screens, main one OLED HDR)

Kinvermark wrote on 10/4/2018, 4:30 PM

+1. Looking at CPU / GPU usage is too simplistic. The TechGage article a few weeks back suggested that Vegas did benefit from 32 threads on the new threadrippers. @BruceUSA has also claimed great benefits from threadrippers. Maybe he will comment here. Forget about the Windows % usage stats and see if you are getting better timeline performance. Media decoding is heavily influenced by the CPU, so a good test for that would be multiple streams of 4k highly compressed media - but of course, your disk subsystem may bottleneck this if it's not fast enough. Renders are influenced by GPU (especially NVENC and VCE modes) so they are not necessarily a good test of the CPU only.

PS - Personally, I think your choice of CPU is a good one. Parallelism is the only way forward now for many computing problems. Clocks aren't increasing. ASICs are fast, but by nature are specialized and thus prone to becoming obsolete.

BruceUSA wrote on 10/4/2018, 5:02 PM

All of my previous posts included screen shot and videos show that my TR working great and awesome. I never seen CPU usage during rendering drop below 50%. . CPU usage during rendering had been always in the 65-90% along with gpu acceleration. For those interested to see sample videos. You can search for posts here. Videos and screen shots speaks for its self. I am happy with TR system.

Last changed by BruceUSA on 10/4/2018, 5:03 PM, changed a total of 1 times.

CPU:  i9 Core Ultra 285K OCed @5.6Ghz  
MBO: MSI Z890 MEG ACE Gaming Wifi 7 10G Super Lan, thunderbolt 4
RAM: 48GB RGB DDR5 8200mhz
GPU: NVidia RTX 5080 16GB Triple fan OCed 3100mhz, Bandwidth 1152 GB/s     
NVMe: 2TB T705 Gen5 OS, 4TB Gen4 storage
MSI PSU 1250W. OS: Windows 11 Pro. Custom built hard tube watercooling

 

                                   

                 

               

 

Kinvermark wrote on 10/4/2018, 7:09 PM

From the OP:

So clearly Vegas doesn't suport more then 8 Cores.

The last post contradicts that statement. I suggest you study BruceUSA's posts and try to figure what is different about your setup. Good luck!

zaheer-abbas wrote on 10/5/2018, 9:11 AM

Thank you for all your posts. Ok after some intense testing iv come to some interesting conclusions. When i was rendering 1080p and dvd footage i saw no real speed increase. I usually measure the rendering speed by the amount of frames that are being rendered ( 1080p approx 100 frames per sec and dvd apporx 175 per second). this was roughly the same on both my other intel 5960x oc 4.3 and my TR 1950x oc to 4ghz). Then i tried something different, i tried rendering some 4k footage, and the TR came to life, All cores at 100%. the amount of frames being rendered was less then 25 per second. I did a head to head with my intel system on 2 mins of footage. the TR did it in 5 mins, the intel in 8 mins, so around a 60% improvement. What do you guys make of this. Im starting to think there is a bottle neck somewhere even though im rendering to and from ssd drives. But it seems to me that The TR is a 4K rendering beast.

BruceUSA wrote on 10/5/2018, 10:48 AM

My TR system rendering 1080P footage I get 160 frames. That flyin speed. At 1080P my TR still load the CPU as I mention above. I am not sure why you are only seeing heavy load on 4k footage only.

CPU:  i9 Core Ultra 285K OCed @5.6Ghz  
MBO: MSI Z890 MEG ACE Gaming Wifi 7 10G Super Lan, thunderbolt 4
RAM: 48GB RGB DDR5 8200mhz
GPU: NVidia RTX 5080 16GB Triple fan OCed 3100mhz, Bandwidth 1152 GB/s     
NVMe: 2TB T705 Gen5 OS, 4TB Gen4 storage
MSI PSU 1250W. OS: Windows 11 Pro. Custom built hard tube watercooling

 

                                   

                 

               

 

zaheer-abbas wrote on 10/5/2018, 10:57 AM

do

My TR system rendering 1080P footage I get 160 frames. That flyin speed. At 1080P my TR still load the CPU as I mention above. I am not sure why you are only seeing heavy load on 4k footage only.

Do you get 100% or close on all core when your rendering 1080P

BruceUSA wrote on 10/5/2018, 12:02 PM

do

My TR system rendering 1080P footage I get 160 frames. That flyin speed. At 1080P my TR still load the CPU as I mention above. I am not sure why you are only seeing heavy load on 4k footage only.

Do you get 100% or close on all core when your rendering 1080P

not 100% but 65-80% along with GPU usage. I don't disable anything. 16 cores at default setting. My system I used NVMe. M2 for OS, Rendering and Storage. You got a high end CPU and you want high end gpu and everything else to get the best performance out of the system. Standard SSD and mid level gpu is not the best as a over all system ultimate performance.

CPU:  i9 Core Ultra 285K OCed @5.6Ghz  
MBO: MSI Z890 MEG ACE Gaming Wifi 7 10G Super Lan, thunderbolt 4
RAM: 48GB RGB DDR5 8200mhz
GPU: NVidia RTX 5080 16GB Triple fan OCed 3100mhz, Bandwidth 1152 GB/s     
NVMe: 2TB T705 Gen5 OS, 4TB Gen4 storage
MSI PSU 1250W. OS: Windows 11 Pro. Custom built hard tube watercooling

 

                                   

                 

               

 

Former user wrote on 10/5/2018, 2:49 PM

Then i tried something different, i tried rendering some 4k footage, and the TR came to life, All cores at 100%. the amount of frames being rendered was less then 25 per second. I did a head to head with my intel system on 2 mins of footage. the TR did it in 5 mins, the intel in 8 mins, so around a 60% improvement. What do you guys make of this. Im starting to think there is a bottle neck somewhere even though im rendering to and from ssd drives. But it seems to me that The TR is a 4K rendering beast.

Maybe it's the video processing that's causing the bottleneck, as you're not using GPU.. Encoding 4k is enough to utalise all your cpu & inefficient processing isn't noticable but with 1080p with higher fps you're getting more video processing states per second that doesn't scale well compared to encoding, resulting in lower cpu use.

In Bruce's examples, he also isn't getting 100% cpu use, but he is getting much better performance due to GPU processing. Even the worst SSD is capable of a sustained write speed of 100,000 mbit/s after it runs out of cache, but rendering is not sustained, it's burst writes which take advantage of SSD cache, so writes are much faster.

 

zaheer-abbas wrote on 10/6/2018, 6:10 PM

I have typically never used gpu for rendering m2t or mpeg2 dvd as not only is it slower then pure cpu but far more prone to crashes particularly if there is heavy use of plugins and effects. Gpu without adoubt improves timeline performance and preview play back, but i always disable before rendering. Its a very intresting point you make about cache as have experienced this on my TR 1950x system, the render will start of incredibly fast at around 170 frames per second for a about 20 seconds and cpu usuage is in the mid 90s , then it slips back to 100 fps and cpu usuage drops to around 60 -70 percent. It maybe worth investing in M.2 drives as rendering drives, albeitit this is a very expensive option. But so far there have been some positives, 4k rendering is vastly imporved. avc encoded is vastly imporved almost 80% with is outstanding, and in general time line performance is smoother, the jerky timeline effect on vegas 16 is almost completly gone albeit wiht project media window closed

zaheer-abbas wrote on 10/8/2018, 2:23 PM

Hi Bruce just wandering have you ever experimented with the Dynamic Ram preview setting to acheive better rendering times. I adjusted it from 200mb to 20gb and i got at render boost of about 80% on my 1950x system

My TR system rendering 1080P footage I get 160 frames. That flyin speed. At 1080P my TR still load the CPU as I mention above. I am not sure why you are only seeing heavy load on 4k footage only.

Do you get 100% or close on all core when your rendering 1080P

not 100% but 65-80% along with GPU usage. I don't disable anything. 16 cores at default setting. My system I used NVMe. M2 for OS, Rendering and Storage. You got a high end CPU and you want high end gpu and everything else to get the best performance out of the system. Standard SSD and mid level gpu is not the best as a over all system ultimate performance.

 

BruceUSA wrote on 10/8/2018, 5:19 PM

Hi Bruce just wandering have you ever experimented with the Dynamic Ram preview setting to acheive better rendering times. I adjusted it from 200mb to 20gb and i got at render boost of about 80% on my 1950x system

My TR system rendering 1080P footage I get 160 frames. That flyin speed. At 1080P my TR still load the CPU as I mention above. I am not sure why you are only seeing heavy load on 4k footage only.

Do you get 100% or close on all core when your rendering 1080P

not 100% but 65-80% along with GPU usage. I don't disable anything. 16 cores at default setting. My system I used NVMe. M2 for OS, Rendering and Storage. You got a high end CPU and you want high end gpu and everything else to get the best performance out of the system. Standard SSD and mid level gpu is not the best as a over all system ultimate performance.

 

 

I never mess with dynamic preview on my two system, Intel and AMD. Both of my system performed equally well. So I had no need to mess with that. I guess I have no complaint.

CPU:  i9 Core Ultra 285K OCed @5.6Ghz  
MBO: MSI Z890 MEG ACE Gaming Wifi 7 10G Super Lan, thunderbolt 4
RAM: 48GB RGB DDR5 8200mhz
GPU: NVidia RTX 5080 16GB Triple fan OCed 3100mhz, Bandwidth 1152 GB/s     
NVMe: 2TB T705 Gen5 OS, 4TB Gen4 storage
MSI PSU 1250W. OS: Windows 11 Pro. Custom built hard tube watercooling