logo

Live Production Software Forums


Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error

Options
Go to last post Go to first unread
h2video.nl  
#1 Posted : Friday, October 10, 2014 2:00:56 AM(UTC)
h2video.nl

Rank: Advanced Member

Groups: Registered
Joined: 3/7/2012(UTC)
Posts: 503
Man
Location: The Netherlands

Thanks: 9 times
Was thanked: 45 time(s) in 39 post(s)
hi all,

pc: cpu and gpu developments are going very fast. amd and intel rapidly develop new cpu/gpu combinations.

I would welcome a thread where vmix users can compare performance based on a standard test or benchmark so performance can be compared.

i have good results based on using amd a10 chips and considering using intel i5 with intel hd graphics 4600 chip, but have no reliable info of the expected performance. for simple 2 to 4 input systems, basic recording and basic streaming an i5 system could be sufficient. for higher performance requirements p.e. streaming with three streams and recording etc more than 4 capture inputs a higher spec could be beneficial. but in the end its about price, performance and reliability of all components and this could be made more visible with a benchmark.

any ideas how to do this?

stefan
Charssay  
#2 Posted : Friday, October 10, 2014 5:39:40 AM(UTC)
Charssay

Rank: Advanced Member

Groups: Registered
Joined: 11/24/2012(UTC)
Posts: 97

Thanks: 5 times
Was thanked: 18 time(s) in 18 post(s)
Dear Stefan,

I like your Idea, and this can be give more user one answer form this request, How and with what can I Build a basic system and it will work in one shot with acceptable cost.

But the problem is, when we see a benchmark on a System, they use software that can measure it according to the work the computer is doing.

With Vmix, what can we measure to show the power of a kind of Unit, a data all can relate on ?

- The CPu charge is the one we can see rapidly!
- The graphic charge : If someone can know a software that can measure it at live
- The HDD buffer/Write use

Then We must have according to me 3 kind of basic Unit already ready for it, having 2 input (HDMi or SDi), a good Internet Connection for the stream, like

1. One Unit Base on Intel HD Grafic with a Common I5
2. One Unit With I5 and a Nvidia Card
3. One unit with an AMD Cpu with Core graphic

All these CPu must have the same core number and the average same frequency

For these one, we intend that if someone build something higher than this, He can have a good result of it., and if He build something exactly like this, he will knows that his system will work like a charm also.

All these 3 systems must capture, recording and stream in HD only at the same time when the bench is measure.

C.
h2video.nl  
#3 Posted : Friday, October 10, 2014 6:13:27 AM(UTC)
h2video.nl

Rank: Advanced Member

Groups: Registered
Joined: 3/7/2012(UTC)
Posts: 503
Man
Location: The Netherlands

Thanks: 9 times
Was thanked: 45 time(s) in 39 post(s)
well, thats what i am thinking off: have a standard test, lets say 2 hd inputs, 1080p25 record to mp4 1080p25 and stream at 2,5 mbit at 1080p25 and then run a test, record cpu and gpu performance and buffer usage etc etc. .

this would give feedback on relative performance differences between specs: intel ams, ssd vs hdd, 8 or 12 gig memory, graphics cars yes no make etc.

but we could start by a standard vmix usage definition to make performance relatively comparable: is a i7 a musthave or could an i5 perform in most cases, when do you need an extra graphics card of is intel graphics hd 4600 enough in most cases.

a simple log ficle could help: maybe Martin has soxzme trickt hiden in vmix already...
stefan
Thremir  
#4 Posted : Friday, October 10, 2014 7:39:44 AM(UTC)
Thremir

Rank: Advanced Member

Groups: Registered
Joined: 9/17/2013(UTC)
Posts: 173
Man
Location: Norway

Thanks: 10 times
Was thanked: 56 time(s) in 29 post(s)
Yes! Please!

To kickstart this we've got to do some work ourselves to crank some numbers in there.

Personally I've got quite many computerparts at home and will try to fill in some i7-4770k, i5-4670k builds with various graphics cards and I've got some four- and sixcore AMD ones too.

I've created a spreadsheet and a text document which will be the "how to benchmark" guide.

Now we need to define what the benchmark is going to consist of;

Firstly, we might want to run two configurations where one of them won't be as demanding as the 2x record and 1x encode of 1080p25 inputs, maybe a 720p25 and a SD benchmark aswell? (same 2x record and 1x encode for them aswell)

As for the benchmark itself:

What about GPU-Z for graphics?

1. Start benchmark
2. Open GPU-Z and go to the sensors tab
3. Click the "GPU Load" percentage until it reads "AVG."
4. Read percentage when benchmark is done

And then we need to find equivalent apps for CPU/RAM, which can read an average load.

Setup in vMix then? We need to get some inputs everyone has in their setup. Due to limitations of which I don't think one can record other than "camera" inputs, we might have to create some virtual ones. Any ideas on this part?


Cheers,
Håvard Njåstad
Njastad Webshop


doyousee.me  
#5 Posted : Saturday, October 11, 2014 12:45:59 PM(UTC)
doyousee.me

Rank: Advanced Member

Groups: Registered
Joined: 8/4/2012(UTC)
Posts: 129
Location: the Netherlands

Was thanked: 10 time(s) in 10 post(s)
Hi,
i have a bunch of motherboards / cpus (all i5 intel but different configs) and several low to high(ish) end gpus but little time.

i can test some combo on request though. (personal experience says the internal gpu won't cut it in my productions, but those were not 1 or 2 cams tbh)

M@ik

(Stefan : you know where to come and play with my 'studio' gear and are Always welcome to :) )

M@ik
h2video.nl  
#6 Posted : Tuesday, October 14, 2014 5:51:18 PM(UTC)
h2video.nl

Rank: Advanced Member

Groups: Registered
Joined: 3/7/2012(UTC)
Posts: 503
Man
Location: The Netherlands

Thanks: 9 times
Was thanked: 45 time(s) in 39 post(s)
the question is how to compare performance of systems used.

i would suggest two video input, one running over lay with a ticker witgh standard tekst, recording in 1080 p25 and streaming at 1080p25 and then record cpu load.

when we agree a standard then we can test and compare. maybe we can come to some indicator, with 100 as best and 1 as worst, or just a percentage cpu load would do...

how to mesure gpu load i don t know but i guess whhen gpu load is max cpu will compensate...

an other issue is what is recorded, a statick image or with a lot of movements. that why the ticker can be used as a tool and this could be standardized.

so what would be a benchmark approach? how can we come to an objective comparisin of i3i5 and i7 systems, and compare intel and amd... how to score a vmix system?
stefan
Users browsing this topic
Guest
Forum Jump  
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.