What is the best cost-effective PC configuration in terms of CPU, graphics card and RAM for VPX 13?
Hi.
What is the best cost-effective PC configuration in terms of CPU, graphics card and RAM for VPX 13?
In some respects that is a very loaded question. While the programs' specification page lists the minimum requirements to run VPX 13 it does add
Some advanced program features demand more of your computer's processing power. To get the most from these features, your system should at least meet the recommended requirements.
This certainly seems to apply to some of the offered free third party plugins and what resolutions and effects you may want to work with.
Also there is the question of 'Future proofing'. Needs you may have in the future. Either working at higher resolutions because your new smart phone can take higher resolution video or you want to add a specific effect not currently included in the base program or you may wish to upgrade VPX in future.
I also notice on the specifications page that support for the new Intel CPUs (11th and 12th generation) as 'coming soon' has been removed.
I have no idea if that means the newest processors are now supported or that they have removed that promise for now. There is still no notification for official support in Window 11yet.
Personally, at present I would recommend an i7 ninth or tenth generation CPU with inboard graphics and an nvidia 1650 Super, with 16GB of ram minimum on the motherboard for a basic functioning version of VPX 13 but will not be the fastest or necessarily the smoothest experience if working above HD content or using some third party effects or using a lot of multiple tracks. It may also not be the best set of options for future proofing. I would also recommend the use of NVMe or at least M2 drives where possible for the operating system and project drives. Fast SATA drives may be sufficient for video storage or archiving old projects.
Beware of buying a gaming laptop or similar as you may not get to be able to use both GPUs in such a system.
If you intend on using a PC and only intend to use one monitor you may also need to buy a 'headless ghost' to plug into the opposing GPUs port to keep it activated.
While the use of two GPUs are not strictly needed, the program runs much better with the aid of both an Internal Intel GPU with an nvidia graphics card.
While AMD alternatives may seem tempting we notice on the forums that such machines seem to need to be way above the recommended minimum specs before they experience a similar problem free experience.
Ray.
Thank You Ray for the very valuable clarification.
I'm going to buy new computer, and hesitated between Ryzen and Intel configuration, but now I can take a very well based decision.
RTX3080/GTX3080?
Hi.
If you look at my system specs in the blue + SIGNATURE at the base of my posts and each contributors forum posts, you will see who is using what on their systems. You may notice in my signature that I'm running an i9900K with an nvidia 1650 Super. The i9900K at the time was possibly overkill and the nvidia 1650 Super slightly under-performing for my needs. I say the latter about the nvidia card because I work mainly in 4K now and really a graphics card with 6GB of vram or more would have been a better choice but funds were limited. I also have 32 GB of motherboard ram and because I use some third party effect plugins the programs I use often use 24GB of that ram or more during processing or playback.
When using Boris FX titling plugin or the external ProDad Hide program the amount of ram used gets even higher and has thankfully so far topped out at around 28.9GBs of ram use, regardless of whether I am running at 4K or HD resolutions. Those programs and plugins slow down any system regardless of system specs. As does Neat Video, a noise reduction plugin.
My personal problem has always been if I upgrade anything I push it to the limits and eventually surpass the capabilities of a given system to cope. I then have to find creative ways of continuing projects as they slowly grind to a halt.
Ray.
Hi Al.
If you can actually find one of those for sale and don't mind paying up to double retail price and if going for a prebuilt machine they often don't include a CPU with an Intel GPU as they expect the nvidia card to do all the work.
I don't think Radeon cards (AMD) graphics support NVENC either which would reduce the efficiency of the hardware acceleration, hence the need for much more powerful AMD components compared to Intel and nvidia choices. Not that that choice wouldn't work but the needed increase in component capability means the AMD alternative would work out more expensive in comparison.
Ray.
@CubeAce there is no "Matrox" AMD graphics cards. They are Radeon and yes, they do not support NVENC, but VCE.
Out of curiosity, it was mentioned that someone would need a higher spec'ed PC if using AMD hardware, why?
And how would that relate to cost effectiveness?
If I were building a "new" computer, I'd consider several factors including what additional software I am using that pushes the computer limits. Is someone using photo editing or other NLE's? Rendering 3D? The Infusion Engine 3 supports Intel's iGPU, AMD and Nvidia for video acceleration and export. And the OFX plug-in interface supports plug-ins requiring the use of more than one processor core.
Hi.
Quite correct. Senility is creeping in as they once produced graphics cards almost exclusively using AMD components. No more. Matrox still exists but have now specialized in other markets. I will amend my comment above accordingly. Thank you for pointing that out.
I'm not sure if it is a Magix thing or more a Windows thing as Microsoft has much closer ties with Intel than AMD.
Microsoft depend more on Intel based machines than AMD ones. (AMD still only account for 11% of the market share of machines that use Windows.) Win 11 did not exist until Intel announced their own 'chiplet' processor designs.It was only after they had Intel processors working reasonably well they switched their attention to AMD. If you look through the posts on the forum we get more complaints of MEP / VPX working from AMD users that have lower specified CPUs than we do Intel ones that one paper at least should perform roughly the same. That seems to go away on the higher performing AMD CPUs.
The Radeon cards not only do not support NVENC which is really only a concern if using H265 files or wish to export in VPX but they also seem to have on average less pixel shaders which seems to impact how well they cope with 2D and 3D rendering. Also MEP and VPX do take advantage of the inboard GPU where present and working as separate entities, both during exporting and playback.
You say OFX plugins have the use of more than one core which is true but they also can take advantage of both GPUs.
Compare the CPU usage against the Intel GPU and even the nvidia graphics card at this point of the render while it renders a Boris OFX based plugin. That is for just fifteen seconds of video taking over 34 minutes to render.
I would actually like to see an AMD/Radeon comparison for something like that. While it doesn't seem to use the nvidia card it certainly makes use of the cards Vram as well as the motherboard ram which as the processing continues can creep to above 28GBs.
Then there is the disconcerting referencing from Magix who have always to date based their published results on testing Intel nvidia based machines. Make of that as you will but I suspect they would go with the system that gave the best results and not seem too esoteric. Call me cynical, that would be fair.
So far I have only seen one AMD 64 core processor with an nvidia RTX 3090 that that has beaten any Intel based system on the forums.
Please remember that this is a recommendation based on my opinions and personal observations.
You are free to disagree and put forward your own case.
Ray.
What is needed is a project that everyone can test on their particular systems and the results can be compared. Then users with the same plugins can compare with a "base" project. 34 minutes to render... wow....
Suggestions?
I have Boris Continuum Complete 2022 and some Sapphire plugins on my computer.
Yeah, Gid's computer is a beast....
Tthreadripper and Nvidia....
Hi.
What is needed is a project that everyone can test on their particular systems and the results can be compared. Then users with the same plugins can compare with a "base" project. 34 minutes to render... wow....
I did that some time back. See here. That project did not include things such as 3D titles or too many effects and was mainly concerned with differences in export times across different specified machines.
The main problems will be finding specific effects we all have in common and the our video resolutions and frame rates. How would we decide which were to be included? How about how many drives are in use for importing workload and exporting?
The problem if there is one is the PC platform itself as it gets more components from more manufacturers over more generations of development get thrown into the mix.
Magix or any other software company are about getting users. They will be careful in their publicity blurb. They will be as illusive as possible about the problems with expectations and their 'get out clause' of 'Some advanced program features demand more of your computer's processing power. To get the most from these features, your system should at least meet the recommended requirements.'
They are of course telling the truth but not the specifics. How can they? It would run into pages of information most people would not even bother to read.
I often wonder what it is that tempts people to upgrade if the last version was still working. Buying new for the first time and not realising the possible consequences of just how much processing power could be needed I can understand. We all go through that learning curve.
For me personally I upgrade when I think it should still work and have a beneficial outcome. Or, if I have bought a new piece of equipment the old program can no longer handle. I have a long list of unsupported hardware components from over the years. This is why I have hung on to VPX 11. It works. MEP has become my test program. For the most part MEP still does the majority of work I throw at it but when it doesn't or I need a bit more refinement from one of the basic plugins I'll then open the project in VPX which surprisingly works most of the time.
If anyone is up for making a new project to test I'll be happy to try it out.
Ray.
Hi.
Finding out how things are implemented is the key problem.
After looking at MEP's manual I forgot another part of the program that can take advantage of Intel's' Quick Sync Video' hardware-accelerated encoder for MPEG-4 AVC/H.264 video codecs if you can access the Intel GPU currently from the HD 520 and upward variant although I think John Baker has said that the HD 460 may have limited use. Again that would be missing from any AMD CPU. Again, at the mid to lower end of the CPU spectrum, this in my view would be a good cost vs, performance choice.
Ray.
@CubeAce if you only have an Intel CPU with iGPU, yes, it is more cost effective. But if you have a recent video card, then that is something different.
If there were a common project (we have that in the Vegas forum for everyone to test, and there is a spread sheet with the results), then we would have a better metric.
Hi.
I'm not sure about that. There are times within a project when my nvidia card is doing very little while the Intel GPU is almost maxing out and vice versa and other times neither GPUs do anything and the CPU maxes out or both GPUs work hard and the CPU is on tick-over. Those variables are not consistent between playback and exporting where it changes again and the nvidia card seems to be the more deciding factor in the smoothness of playback until effects are added. Then the Intel GPU can come into play again. It's best observed with the Task Manager open while in use.
Then there is the problem of which AMD processor is the equivalent of its Intel counterpart. I for one don't really know the answer to that one. Tech reviews online tend to only talk about gaming performance of blender tests. Neither of which really gives a user of Magix products a clue.
I think any test should only include the sets of effects and templates that comes with the base program as that still gives a wide scope for performance issues to arise.
However this is now going very off topic as it is going beyond just suggestions and maybe a separate topic is called for.
Ray.
I'm getting good results with my 2 custom PCs (specs are several years old now):
- i9 9900k with AMD RX580 8GB
- i7 9700k with AMD RX540 (4 or 8GB).
It all depends on how many tracks, effects, etc. But for my simple to moderate purposes, these work fine. 3 years ago, these would have been relatively high end. Now rather modest, I realize. But Magix' improved features have made them work well.
Hi
The ideal situation with respect to GPU's is having an Intel IGPU preferably a UHD 630 or higher and a Nvidia GTX 1050, preferably better, or an AMD GPU RX470 or better.
. . . . NVENC which is really only a concern if using H265 . . . .
NVENC only works on Nvidia GPU's, AMD GPU's will use any combination of DirectX 12, OpenGL or Open CL for hardware acceleration.
NVENC is not limited to h.265, it also accelerates the following formats:
Decoding:
Encoding:
John EB
I'm getting good results with my 2 custom PCs
Out of interest and I know it's very subjective, what encoding times are you getting? 2 times, 5 times better than the movie time assuming a basic movie?
Hi
. . . . Out of interest and I know it's very subjective, what encoding times are you getting? 2 times, 5 times better than the movie time assuming a basic movie . . .
It is as you say very subjective and subject to defining the conditions of export
As I mentioned in @Reyfox topic in a template project for testing, the Demo project (104 secs long) - on my PC it takes 11s for H.264 at the original video resolution (512 x 288) and framerate (25fps), and 25s upscaling to 1080p 25fps both using the RTX 2060, these times give ratios of 9.45x and 4.16x faster respectively.
For 4K UHD source videos and a couple of titles, crossovers and cuts for transitions, and final sharpening I can get up to 2.5 x faster exporting as 4K UHD.
More complex projects with same parameters and with collages more complicated transitions, picture in picture etc real time to 2x real time.
I prefer to reduce the speed units as the number of frames encoded per second for x * y resolution at n frames per second for as you can see in the last 2 paragraphs the encoding speeds are a bit meaningless when you say x times faster.
The above times for the Demo project of 104 secs at 25 fps = 2600 frames, so the speeds are 236 fps and 104 fps, real time would be 1 fps for a 25 fps export.
John EB
I'm getting good results with my 2 custom PCs
Out of interest and I know it's very subjective, what encoding times are you getting? 2 times, 5 times better than the movie time assuming a basic movie?
So, a typical scenario for me encodes/exports at a 1:1 ratio; sometimes a little faster. So, a 6-minute video in 5-6 minutes.
Typical details for me:
- Video project usually includes a mix of 1 to several video tracks, 1 to several audio tracks, some video processing, some audio processing, some intro and outro videos. Usually working with .mp4 video and export (i,e. h.264, avc). Sometimes with AVCHD footage. All in 1920x1080 (i.e. full HD). Usually 30fps or 29.97 fps. Usually audio is being reduced from original to 160 or 128. Often video is being reduced from 6-18MP to 4.5MP.
- However, if I have a single straight video track, in and out, with little to no processing (i.e. just cuts and fades), then the encode/export time is 2x, i.e. 2:1 ratio. (So, a 6-minute video in 3-minutes or less.)
Again, I was offering my several year-old systems (formerly high end) as a value computer now.
Thanks Craig, very helpful.
See the details in my signature..
Even that does not give me smooth playback, something that I will raise in another thread.
John CB
Thanks John, I didn't realise your long post was addressed to me.