Graphics performance tests

terrypin ha scritto il 13.03.2020 ore 21:21

Graphics performance has been much discussed here by those with in depth skills on it. I'd appreciate insights into the following aspect of it please.

I recently tested my PC's performance with Passmark's free PerfomanceTEST tool available at
https://www.passmark.com/products/performancetest/download.php

I was satisfied with most of the results but very disappointed with the graphics, especially the 2D graphics rating.

On researching I found scores of other users with similar results:
https://www.google.co.uk/search?q=passmark+low+2d+graphics+scores&ie=utf-8&oe=utf-8

Somewhere I'd read that this might have been caused by successive MS updates addressing the Spectre and Meltdown issues. I may be oversimplifying but I read this as meaning that the patches remove the security risks, but have negative performance consequences. And for solo home user the risk is very small.

As I had a WU pending (from Win 10 version 1903 to 1909) I thought I'd first use the opportunity to test the suggestion that each update negatively affected performance. Sure enough there was a further drop in 2D Graphics after the update, about 10% as you see:

To test the Spectre/Meltdown hypothesis I then downloaded another handy free tool, InSpectre.exe from
https://www.grc.com/inspectre.htm
which allows you to quickly toggle such patches on and off. I was hoping for a dramatic improvement. Maybe as much as a factor of x5 between my rating and the 'normal' level I'd seen in others' results. But sadly I didn't get it:

So my conclusion is that there must be some other cause in my case. I'd appreciate suggestions on where to look please.

Terry, UK

Commenti

CubeAce ha scritto il 13.03.2020 ore 22:30

@terrypin

Interesting Terry.

As a result I looked at some of the forum questions there.

Did you see this reply to one of them from the administrator?

"Unfortunately in modern computers the 2D rendering you see on the screen is not accelerated by the video card. (Direct2D and some video processing being the exception). There was actually more acceleration back in the days of Windows XP.

Direct2D and DirectWrite were suppose to change this, but unfortunately these new programming methods are far from optimized. So in recent tests we did the classic 2D programming methods (called GDI and GDI+) were much much faster.

The video card vendors also have almost no focus on 2D performance. Nearly all their time is spend worrying about 3D performance in AAA games. (performance in smaller games is also ignored).

We hope this will change at some point. For example the Win10 Edge browser uses Direct2D and it's performance is really really appalling. You can barely scroll the page on some web sites.

So much of the 2D performance is determined by the O/S you are using, the CPU power and the graphic's settings you have active (not the video card)."

So I wonder just how important the 2D performance rating is?

Ray.

terrypin ha scritto il 13.03.2020 ore 23:21

@CubeAce

Thanks Ray. You’re right, and I have no practical reason to worry about it as I’ve seen no obvious visible evidence of poor ‘2D graphics’. But learning that my rating is so much lower than ‘normal’ has me keen to find out why!

Terry

CubeAce ha scritto il 14.03.2020 ore 09:21

@terrypin

Just a quick couple of question please Terry. What does the 2D graphics section do as applied to me looking at a screen?

What is it responsible for? How do you think it impacts the video editing experience?

I've been doing benchmark tests on my CPU and ram and got a bit frightened as the temperatures climbed rapidly during the tests and the CPU was pushed way past what MEP seems to induce during use where more normally it just seems to tick over.

From my limited experience so far, any form of bench marking doesn't seem to translate into any performance I get from my programs. There seems too many variables in the way. Even swapping and adjusting how audio is set up in MEP seems to effect how smoothly it handles video clips (with audio) and editing in general but as I'm currently reliant on the internal motherboards audio I have no way of even checking if a performance gain could be made by swapping to a dedicated audio card.

Puzzled as always.

Ray.

terrypin ha scritto il 14.03.2020 ore 10:46

Morning Ray,

The whole test took about five minutes and during the 2D and 3D sections some pretty fancy graphics occurred. And temperature briefly rose to 55-60 C, maybe more. I was naturally trying to relate this, and the numerical result, to regular work in MEP and other apps. But I cannot, apart from assuming that if my 2D rating was say 1500 instead of 300 I would perceive some visual improvement?

It’s quick to install and run, so give it a try! I’ll brace myself for your inevitably high results.

I’d also like to hear anyone’s results. Not only 2D graphics ratings but also the Meltdown/Spectre toggling test, as I’m still very puzzled why it didn’t make the big improvement I’d anticipated. I’m going to try it again.

Terry

CubeAce ha scritto il 15.03.2020 ore 13:08

Hi Terry.

I'm loath to put additional programs onto my PC and put even more into my registry than I have to but instead I looked up more information on the role of 2D graphics.

If your observations regarding patches and subsequent slowdowns are accurate (and I've no reason to doubt your results) then it could explain why rendering of frame rates during playback and exporting slows when adding text or some graphical effects.

It would seem if I'm reading the information correctly that 2D is completely run within the CPU but can be shared with 3D rendering operations.

I'm not sure if this will be of benefit to you or just take you further down the rabbit hole but it's Microsoft's own pages on the purpose of the use of 2D.

https://docs.microsoft.com/en-us/windows/win32/direct2d/direct2d-overview

There is an awful lot of information there that could give hints as to what is happening in various parts of video editing and exporting.

If that is all true I can see no way of regaining any performance loss due to continued operating system patches.

Ray.

terrypin ha scritto il 15.03.2020 ore 16:44

Hi Ray,

Thanks, an interesting article, albeit largely above my technical skill level. From it, I suspect that my poor 2D result has little impact on my MEP work. But I'll report back if I find out anything relevant.

CubeAce ha scritto il 15.03.2020 ore 18:04

@terrypin

Hi Terry.

What I found interesting from a lot of reading is (If I'm reading it correctly) that anything 2D is being processed by the CPU which would mean possibly frequency is king for speed in that department and that any text or other process that relies on 2D can't be accelerated beyond the programing limitations. If your observation that patches to Windows are responsible for this gradual slowdown there is nothing we can do to circumnavigate that and as processor speeds seem not to be being increased but instead is reliant on more cores being added there may be no workaround unless multiple cores can be pressed into service. I can't see that happening at my end. Less than half my cores seem to do any work at any one time.

Also I have no way I can find to see how much 2D processing is going on during editing or exporting as opposed to 3D which is easily monitored. Even that does not seem that intensive. Less than 5% on my system.