Max number of pixels

[Powers_Ronald]

Just wondering what is a realistic limit to the number of pixels an LOR S5 display can run? I know that in Network Configuration you can define 999 DMX universes which would come out to 168,830 pixels, but can an LOR network handle that much traffic? I am only about 8,000 pixels and was just wondering how much larger I could increase my display before I had to worry about over loading my network.

 

Thanks, 

Ron 

[lkcubsrule]

This is a great question, and unfortunately there is no finite answer.  Part of it depends how you're running your technology.

If you're running all your pixels on an Enhanced LOR network, I'm going to guess you might hit a ceiling near the +/- 10,000 mark, depending on your speed, programming, and device layout.

If you're running your pixels on an E1.31 network, you can throw tens of thousands without a hitch, granted your computer can output that many at once.  It really comes down to CPU / Graphics output and network throughput...those will be your limiting factors at some point.

These are relative figures, and your mileage WILL vary.

[stevehoyt]

Curious too...

cub do you mean +/- 10K per LOR enhanced network running at 500k

[k6ccc]

First of all, LOR network and DMX don't mix.  A RS-485 DMX network is limited to one universe - or 512 channels or 170 pixels.  As such, anyone running LOTs of DMX is doing it by way of E1.31.  More on that in a moment.

As for LOR networking (also runs on RS-485), there is no hard and fast number, but as I understand it, at Expo 2017, the LOR show demo was running 2400 pixels on one LOR network (presumably 1000K enhanced).  How much you really can get on a single LOR network depends on how fast and how many channels are changing.  That means that the number can vary A LOT.  LOR currently supports 16 LOR networks.  Using the 2,400 pixels per 1,000K Enhanced network as a working figure, that comes out to 38,400 pixels.

Now onto E1.31.  E1.31 is DMX over Ethernet.  You can run hundreds of DMX universes over a single Ethernet connection.  Several years ago, I calculated that each DMX universe takes about a quarter Megabit per second of bandwidth.  Therefore, even on a 10Base-T connection (yes, there are still connections that slow in use today), you should be able to get a couple dozen universes.  At 100Base-T, the number climbs to several hundred universes.  Gigabit would easily carry the 999 universes that LOR is currently limited to.  Just as a reference point, Friday evening I did a test with my P10 and P5 panels overnight.  I happened to be running it from my laptop over one of my WiFi networks at home.  I was driving 108 universes and it worked just fine.  Mind you, I don't really recommend putting that much traffic on your WiFi for a production operation, but this was just a test and it worked fine.  For what it's worth, my 2018 Christmas show will be running 66,639 channels.  About 1,000 channels of that will be on LOR networks, and the rest will be E1.31.  Based on my show testing, I do not expect any issues running that.

I guess the biggest question is how many channels can a show computer process using the LOR Show Player software.  That I can't answer...  I don't know if the software or the computer hardware would be the limiting factor.  If it's the hardware, get a faster computer.  Off hand, I doubt that will be the limiting factor - as the show computer does not have to do much to play a show - particularly when using Intensity files.

 

[stevehoyt]

Thanks for the info K6ccc.

I have about 12,000 pixel channels spread over 2 enhanced 500K networks. I think the LOR protocol should be able to handle anything I should ever need in the future. At my age I am not interested in tackling DMX or e1.31

Steve

 

[SPaschall]

Don't know if this helps but when S5 was still in the Alpha stage we ran over 46,000 channels of pixels on a commercial job with no issues. Matt has done a great job with improving the performance since that time. Unless you really go all out with overloaded universes I think it will handle 99% of any residential display

SPaschall

[Powers_Ronald]

Thank you everyone for the input. Based on this information, I should be OK for now.  All my controllers are connected via Cat5e cables (no wifi) and my laptop should be fast enough. Still, I am concerned about the limitations I may face soon. 

[k6ccc]

20 minutes ago, Powers_Ronald said:

Thank you everyone for the input. Based on this information, I should be OK for now.  All my controllers are connected via Cat5e cables (no wifi) and my laptop should be fast enough. Still, I am concerned about the limitations I may face soon. 

How many channel are you anticipating?

If you are expecting to run enough channels that 16 LOR networks can't handle it, you really should be running E1.31.  If you desire to stay with LOR hardware, the PixCon 16 controller can operate from either E1.31 or LOR networks.  Using E1.31 adds a little to learn, but it's no big deal.  As I said in my earlier post, I don't recommend running a show over WiFi.

 

[MattBrown]

We have tested 750 full universes over E1.31 - the equivalent of 127,500 pixels - with the S5 Sequencer. It works. As mentioned by others above, it does depend on what equipment you are using.

The biggest limitation currently is the Motion Effect Generator. It calculates those effects in real time and large props can slow it down. So if you have a big tree or matrix with 10,000 pixels or more, you will want a high-end computer to speed up the calculations. Note that this is a constraint on the number of pixels in a single prop; it does not constrain the total number of pixels in your display.

Matt

[k6ccc]

1 minute ago, MattBrown said:

The biggest limitation currently is the Motion Effect Generator. It calculates those effects in real time and large props can slow it down. So if you have a big tree or matrix with 10,000 pixels or more, you will want a high-end computer to speed up the calculations. Note that this is a constraint on the number of pixels in a single prop; it does not constrain the total number of pixels in your display.

Matt

Matt,  Can I assume that affects sequence creation and playback from within the Sequencer - and NOT playing a show from intensity files via the Show Player?

 

[MattBrown]

3 minutes ago, k6ccc said:

Matt,  Can I assume that affects sequence creation and playback from within the Sequencer - and NOT playing a show from intensity files via the Show Player?

Actually, it only affects the responsiveness of the Motion Effect Generator. Nothing else in the Sequencer and certainly not the Show Player.

Matt

[k6ccc]

Intensity files for playing the show are wonderful!

I have noticed that even with a very fast gaming computer, when creating or editing motion effects on my P5 panel (12,288 pixels) and my P10 panel (6,144 pixels), it will have to "stop and think" for a while from time to time...

 

[Powers_Ronald]

Matt, May I ask to define "a high-end computer" please? My whole display is E1.31, including the CTB16PCs which I run through a E1.31 bridge.  I am planning on using a laptop with a 2.5 GHz I7 Processor, 12GB Ram and a 512GB SSD to run the show. 

Edited September 12, 2018 by Powers_Ronald

[k6ccc]

I can tell you what I'm running.  For sequencing I'm using a high end Dell gaming computer (not an Alienware).  It has an i7-8700 with 32GB of RAM, 250GB SSD and 2TB spinning disk and a NVidia 1070ti video card driving two (will be three eventually) 24inch displays.  For the show computer, I have an i5 based Dell.  Over the weekend I ran an almost 37K channel show (all E1.31 - 72 full universes) on an i3 laptop without issue.  My 2018 Christmas show is sequenced as almost 67K channels (although about 100 may not actually get built).

 

[MattBrown]

9 minutes ago, Powers_Ronald said:

Matt, May I ask to define "a high-end computer" please? My whole display is E1.31, including the CTB16PCs which I run through a E1.31 bridge.  I am planning on using a laptop with a 2.5 GHz I7 Processor, 12GB Ram and a 512GB SSD to run the show. 

The demands on the computer running the show aren't as high as the one that you sequence on. The specs for the computer you list should work for either purpose.

For a show computer, it probably doesn't matter much which generation of i7 processor you have (or even that it is an i7 - an i3 or i5 would probably work fine). The SSD will get the sequences started faster.

But for sequencing, the newer the processor, the better the built-in graphics are. So if you don't have a graphics card, the graphics in a 3rd generation Intel core processor is a bare minimum, with a 7th or 8th generation preferred. Also the number of cpu cores doubled between the 7th and 8th generation, so 8th generation is preferred for that reason also.

For others reading this, we are talking about displays with more than 10,000 pixels. If your display is smaller, you can get by with less. If you don't have any pixels at all, you can get by with much less.

Matt

[Vince4xmas]

Matt, what are the limits running pixels on a G3-MP3?

 

[LOR Staff]

When talking about max numbers, there are 4 factors that come into play:

1 - What is the network medium?  IE, is this Ethernet or RS-485?

2 - What is the speed of the medium? 

3 - If this is RS-485, are you using LOR protocol, ELOR protocol, or DMX protocol?

4 - How much is the data changing?

 

LOR USB adapters and MP3 directors ONLY output RS-485, so we can treat them the same:

For AVERAGE sequences running 500K ELOR we recommend no more than 2000-2400 pixels per LOR network.   For a G3MP3, that would be 4000-4800 total, 2000-2400 for each Network. 

Using a slower speed (19.2/57.6/115.2) will reduce than number.

Using regular LOR networking (as opposed to ELOR) will also reduce the number.

When running native DMX-512 (IE, the on the wire RS485 DMX, NOT E1.31) the limit is simple:  512 channels (it's right there in the name).

 

E1.31 is a completely different beast:

When running E1.31 (DMX over Ethernet) bandwidth is determined by your max Ethernet speed (IE 10/100/1000/etc), switching fabric speed of your network, routing, how fast your E1.31 controller can process the data and get it out to the pixels (which themselves have a limited bandwidth) and a host of other things that require several semesters of college level classes to explain properly.  Jim will correct me (as he always does for which I am grateful, since I can never remember):  an E1.31 DMX universe will consume .25Mb of LAN bandwidth if I remember correctly. 

None of our MP3 directors will output E1.31.

 

The Wildcard - How much data changes per unit of time (IE the speed of your sequence)

As others have said, there is no absolute ceiling on the number of pixels that can run on a single LOR/ELOR network - A slow sequence can run many thousands of pixels per network while a super fast data intensive sequence can saturate the network with only a few hundred.  

Even E1.31 universes can be impacted by how fast a sequence is, however because of the large amount of bandwidth available, this is rarely noticed.

 

What should NEVER come into play - what is sending the data for a SHOW:

Both the Show Player running on a computer and an MP3 director can fully saturate a network without a sweat.  That means that a sequence run from a computer should look just about identical to one running from an MP3 director.  If you are seeing lag with a sequence you can be assured that the issue is not processing power of the computer or the MP3 director - it is bandwidth.  

There is an exception to this rule, and that is if you are running from a computer, AND are playing a sequence from the Sequencer/Sequence Editor/or (possibly) Superstar.  There are a LOT of things going on when you are sitting in front of the software running a test.  The two biggest being the format of the data, and all the other stuff we have to update while running.

First, your sequence is not in an 'optimized for playback' state.  What I mean by that is the data that makes up your sequence is being kept in memory in such a way that makes it easy to be edited, not played.  Because of that, we need to compute a lot of things in real time on the fly.  That is one of the reasons why things like the Motion Effect Generator (as explained by Matt) lags a bit when you are sequencing.  Once you are happy with the sequence however, all of that data is processed down into a way that is optimized for playback.

Second there is a LOT more going on with your computer when running a sequence from the Sequencer/SE.  We are sending data around to multiple places.  We are doing all that computation from above.  We are watching for additional inputs from you.  We are updating your screen.  All of that takes time.

These are the reasons why we say that a 'Show Computer' doesn't need to be anything special, but for 'Sequencing' a better computer may perform better.  You can absolutely use the software  on slower older computers, but you may have to wait longer for things to happen.  Some cars go.  Some cars go FAST.  Both get you to the grocery store

 

[Richard Hamilton]

DevMike, wow.  That is a GREAT amount of detail and information.  Thanks for taking your valuable time to put this together.  Much of it I already knew, but I learned some new important facts !!!

Oh, and a mention about use of ELLs would be nice.  I've found that I can't use ELLS in my CCR networks as there is too much lag at that speed.   I run cat5 instead.

Edited September 13, 2018 by Richard Hamilton

[k6ccc]

Great writeup Mike!

One more thing I will say about a show computer playing a show that has a lot of DMX or E1.31 and is NOT using Intensity files.  The DMX protocol is very dumb compared to the LOR protocol.  What that means is that when running DMX or E1.31, instead of sending a single command that tells a channel to for example, change from 100% to 10% over the next 1.5 seconds (essentially what LOR protocol does), with DMX, the computer must calculate what the intensity is for each channel about 40 times per second and send that instantaneous level to each DMX channel - again, about 40 times per second.    If you are running a lot of E1.31, that DOES put quite a load on the show computer.  The solution is Intensity files.  When using Intensity files, all that calculation is done when the Intensity files is created, so all that has to happen when the show computer is playing the show is to read the Intensity file, and spit the data out the appropriate computer interface.  For you computer programming people, think a compiled program vs a interpreted program.

As a practical example of that, until last year I was using a REALLY OLD AND SLOW show computer.  A couple years ago after getting most of my show set up in early November, I was playing Brian Bruderer's version of Wizards In Winter.  That piece moves VERY fast, and was new to my show that year..  I was seeing a LOT of lag and skips on my pixel tree.  The pixel tree and star are on E1.31, so I knew it was not a LOR network congestion issue, and it was a dedicated 100 mb/s Ethernet all the way, so I did not expect a LAN issue.  I brought up the Windows Task Manager on the show computer and the CPU was maxxed out.  The solution was to convert the show to the new Intensity files that were added in S4.  After that the CPU was running abut 18% and the show played perfectly.

 

[Bumble]

Hi

Can anyone explain how to enable a second network in s5 if one network is not enough ?  I have a second red sub adapter and enough cat 5. I have 6 pixie controllers that need to be on separate networks according the pixel limits above.  Thanks ahead of time.