DiyLightAnimation
Software => Light Show Pro => Topic started by: zwiller on September 03, 2012,
-
Wanted to get some other opinions that my trusty old Dell 100L 2.4G Intel Celeron with 2G RAM is not gonna cut it with a display of maybe up to 1,500 nodes with 100 dmx channels. I am running 2.5. PC was already acting slower with just some preliminary testing of a window frame of 60 nodes and only one transition. I was hoping to squeak by with keeping the sequences basic which is fine with me since it is first year for both LSP and RGB but I don't know if I will be lucky.
Guessing it is time, so I have this one in mind. http://www.newegg.com/Product/Product.aspx?Item=N82E16834230581 (http://www.newegg.com/Product/Product.aspx?Item=N82E16834230581) Plan to give it a go with the stock 4G RAM and max RAM out to 16G if need be. I need to do some further digging as Intel specs state max RAM is 16G but ASUS says 8G...
Any thoughts? Thanks in advance.
-
Go with the max ram from ASUS, as they would be the ones that actually designed it. The Intel chipset may be capable of handling 16, but if Asus made some change, that could decrease the amount it would use.
-
A friend of mine has one of these machines and he has put in more than 8 gig and it took it but there may be some instability issues above 8 gig.
-
Are you planning just to run the show using this laptop or is this your sequencing machine too? What adaptor are you using for your show? USB or Etherdongle? Performance with an Etherdongle is much faster than USB. If you are using USB, I would recommend building an Etherdongle instead of buying a new laptop. The bottleneck is USB throughput and not CPU or memory.
-
Thanks for comments so far. Maybe there is hope the old Dell after all!
I originally planned to use the Dell to sequence and run show both with LSP this year. For some reason I only expected issues with running show, not sequencing, as I heard the scheduler was fixed with 2.5 and I have a few more options for running show: use the Conductor if it arrives this season or run show with xlights. In addition, I have the etherdongle and use it now. Sorry for not mentioning it.
As I type this I now realize my concern is primarily sequencing right now. Seems like I am seeing the Dell bog down already with just a hundred channels or so and I am just starting to use LSP. Maybe I don't have it set up best for sequencing. Right now I haven't drew any channels in and just have a SSC hooked up to make sure its blinking with a test sequence.
-
Depending on how old it is, you may want to stick another GIG of memory in te Dell and see what happens. LSP is a bit of a meory hog which may be what is slowing you down.
-
I've got the Dell maxed out at 2GB. Good idea though. I bet it would help if I could have 4GB in there.
After Mickpat’s post I started to think about how LSP works. As he pointed out sequencing and running show are two different things which I suppose I got mixed up based on years of vixen programming. Then I watched more of David's vids relating to sequencing and I noticed even on his machine (which is got to be nice) the processing of transitions and macros take time. Then it hit me: what I am noticing is only the processing time to generate sequencing effect. It will be interesting to see how long on the Dell it will take to generate a transition on a fully populated visualizer. I am going to finish mine then try and see.
-
For what's it's worth, I'm running nearly 5000 channels of LED Flexstrip on LSP on a Laptop, Win7 64 bit, i7 Quad core, 8Gig RAM and dedicated graphics cars with a solid state drive.
I did try LSP on a Core 2 Duo, 2Gig RAM, i7 laptop and the time to generate effects was pretty bad, it took a minute of so to add in a single transition across the flexstrips when sequencing, so that wasn't going to fly.
Generating a single effect (using one of the color slider macro) across all the flextrips takes about 10-15 seconds on the Quad core Laptop, so even on the higher end machine sequencing can be quite slow.
Looking at David's machine, I believe he's using an HP z820 machine, so that's an eight core, workstation and hence why his effects get generated in a few seconds.
-
I'd invest my money in an upgrade to my home machine that I'd use for sequence generation -- and take any old system (including your laptop) and use it as my show computer to run xLights.
The visualizer adds extra processing to your show, if LSP can be configured to disable the grid view and the visualizer, it might get comparable speeds... but i think xLight's file format which is optimized for playback, gives it an advantage.
I also think RAM is less important for xlights, and if you wanted to spend money on improving it further, I'd upgrade the laptop disk drive to use an SSD drive instead -- much more reliable, and better performance than the 5400rpm drives normally found in a laptop.
-
Then reality set in... :o
Pete, you talking 5,000 RGB channels (150,000) right? Or like 1,667 RGB (5,000)? I am planning upwards to 1,500 RGB channels (4,500).
Steve, do you think the machine I posted will work or do I need a Ferrari like yours? ;D
-
Then reality set in... :o
Pete, you talking 5,000 RGB channels (150,000) right? Or like 1,667 RGB (5,000)? I am planning upwards to 1,500 RGB channels (4,500).
Steve, do you think the machine I posted will work or do I need a Ferrari like yours? ;D
I'm talking about 5000 channels of RGB (or in other terms 5000/3 = 1666 nodes) so similar to what you are planning.
I've tested the LSP scheduler and on my most complex sequence, it uses about 400-700MBytes of memory when running the sequences in the scheduler and consumes about 15% - 20% of the CPU.
Now I have also tested the same sequence in XLights by getting FrankR to convert it to the XLights format
see thread here: http://diylightanimation.com/index.php?topic=9266.0
The same sequence run in XLights the CPU utilization is so low, it doesn’t even register a % (couldn’t believe it at first, had to check a couple of times to make sure)
Memory utilization is ridiculously low at 49MBytes, again barely registering
So XLights is super efficient running the converted sequences.
-
For the playback/show computer, most of my upgrade advice is not needed.
For sequencing however, I think anything you can do to upgrade will lower frustration, and make your sequencing experience more enjoyable. When I do sequencing or programming, having a fast computer allows me more time to try something, throw it away, and try something else. With a fast computer you spend less time frustrated by the tools.
Is 2.5 better than 2.0? ...every indication leads me to that conclusion, but I haven't had the time to play with 2.5 yet, and I can't say if the software improvements go far enough.
I've heard no one say that having too much memory hurts LSP. :)
So, even if you don't max out your computer now, having the ability to add more memory is something that should help later.
myself... I'd spend money to optimize the system in the following order:
- memory
- cpu cores
- disk
- cpu speed
-
To add to what the others are saying you want your sequencing machine to have a minimum of 8 gigs ram (16 would be better) a 3.0 MHz quad core processor (or better) and a good mid level video card (to take the video processing load of the CPU) . You can find some good deals on bare-bones systems at Tiger Direct but you might be better off just piecing together a system from NewEgg. Do your home work and remember that a AMD processor will be cheaper that and equivalent Intel processor.
-
So I did some more testing last night. It did not go well. In fact, my Halloween sequence which is quite simple, just 4 aethers with random long fades and lightning lasting 2 minutes long completely maxes out the CPU. The virtualizer is jittery. I confirmed no malware, etc and I don’t have anything else running. Does this seem right? Sure, I expect issues with some high channel counts but just 12? At this point it’s clear I need to upgrade if I plan to use LSP unless I have something wrong. Hoping its a newbie mistake.
-
That sounds familiar, when I started earlier this year, I loaded LSP onto an Intel Dual core set up and also on an Intel Core 2 Duo, neither ran LSP adequately. That's when I bit the bullet and got a new Laptop i7 with Quad Cores.
I know that the LSP sequencer is pretty heavy on memory and also the CPU
Did you optimize the sequence in the sequencer to see if that helped?
-
You can go to the dell outlet here
http://www.dell.com/us/dfh/p/?cs=22&c=us&l=en&s=dfh&~ck=bt&redirect=1
I picked up a nice Alienware laptop last year (M17x R3 with a 3d screen). Now if only I had time to watch a 3d movie or play a 3d game ;) With 12 gigs of ram and fast quad core processor it has no problems with LSP.
-
Thanks Keith. The 17R is a contender. i7 quad, 2.7Ghz, 8G RAM, 17†screen for around $600. No room to add more RAM though.
Pete, only thing I could find relating to optimizing was a user preference which I turned on. Re-saved sequence. No change. I looked up optimizing in the help areas and did not find anything.
-
To optimize the sequence- Right click in the audio area and you will see options like whole sequence or selected area.
-
Thanks Keith. The 17R is a contender. i7 quad, 2.7Ghz, 8G RAM, 17†screen for around $600. No room to add more RAM though.
Pete, only thing I could find relating to optimizing was a user preference which I turned on. Re-saved sequence. No change. I looked up optimizing in the help areas and did not find anything.
Check on the max RAM configuration. Sometimes you can upgrade by swapping out the old and replacing with higher density DIMMs. Memory has gotten cheap enough that this might be an option you'd consider.
-
I'm planning on swapping out 2 x 4Gig modules for 2 x 8Gig modules in my Laptop, but I wonder where I would see a difference in LSP with more memory.
When I've been sequencing, I've never seen LSP use more than about 2Gigs of memory. Has anyone seen LSP perform better with more memory?
Will it speed up the sequencing when for example I'm placing a macro effect on to grid (currently it takes about 2-5 seconds for the effects to be written to the grid). Will it also speed up optimizing a whole sequence. My most complex sequence MUST be optimized before I can select the output and play it in the sequencer. This optimization takes about 10 minutes to complete before I can hit play.
If more memory will help with the above, I'd be happy to drop $90 for the memory upgrade.
-
Ok I found the optimization option. In maintenance. No change, still goes 100% cpu. I assume optimizing is a clean up function that doesn't need to be performed each time you output like deleting "virtual" things such as undo, re-dos, etc to lower file size. Some of my audio production software has this too.
-
Ok I found the optimization option. In maintenance. No change, still goes 100% cpu. I assume optimizing is a clean up function that doesn't need to be performed each time you output like deleting "virtual" things such as undo, re-dos, etc to lower file size. Some of my audio production software has this too.
Glad you found it.
Each time you make a change, it will "WANT" to re-optimize the sequence if you are trying to output to the lights. You can turn the function off so it doesn't always try to re-optimize on each change, but it will still "WANT" to optimize when you run the whole sequence back and check the output bulb.
I believe, if you make small changes, you can playback the small section with live output without optimizing the whole sequence, if you highlight the small section you want to playback (it will just optimize the small highlighted section you want to output to the lights). But if you hit play while the output bulb is checked, it will need to optimize the whole sequence before the playback starts.
On my more complex sequences it can take 5-10 mins to optimize before the whole sequence can be played back, and LSP wants to optimize the file everytime the sequence is opened and I check the output bulb. That's another reason why you need a fast machine.
-
If you have lots of RAM, consider installing a ram disk to do your sequencing. it might speed things up if you have disk I/O causing your slowdown.
look at your task manager's charts to see if you are having high memory utilization, lots of thrashing/swapping to disk, lots of disk I/O activity, and/or high CPU load.
sequencing I expect to utilize whatever system you use -- but the higher-end system will take less time to do the operations.
running your show should not take much of any resources. the worst case that I'd expect is a slow disk and a LOOONG show with LOTS of channels.
when you do a playback with output turned off, do you see similar performance problems?
-
So managed to get my 2012 layout done with the old Dell. Just under 5,000 channels and the test sequence of 1 minute actually runs, albeit jittery. LSP is finally starting to sink in.
FYI I hold each of you participating in this thread personally responsible for the purchase of my new computer. ;D
-
So I bit the bullet at the weekend and upgraded my Laptops memory from 8 Giga Bytes to 16 Giga Bytes to see if that helps with LSP 2.5
(I'm running a 64 Bit Win7 O/S, i7 Quad Core, 60Gig Solid state drive and dedicated Graphics card)
I then ran some tests to see the effect of additional memory with LSP 2.5, here's what I found
1. Macro's
The is no change in the speed of adding Macros to a sequence, if it takes 10 seconds to add a Macro with 8G of RAM, it will still take 10 seconds to add a Macro with 16G of RAM
2. Optimizing a Sequence
When a sequence is opened in the Sequencer, LSP 2.5 will automatically begin to optimize the sequence (unless you turn it off in LSP 2.5). The speed of optimization will be no faster in LSP 2.5 with 16G RAM compared to 8G RAM
3. Optimizing a large Sequence
If you attempt to play a large sequence (in my case a 38Mega Byte file) while LSP 2.5 is currently optimizing that sequence, LSP 2.5 will Throw a Memory Exception and the visualizer will stop and a Big Red X will appear in the Visualizer
4. What is the optimizer doing in LSP
I'd no code Guru, so don't quote me, but as LSP 2.5 is optimizing a sequence, it SEEMs to be creating a number of little threads that run in the background.
If you start Task Manager and look at the I/O and Disk Operations and tasks, you will see multiple tasks that are named controller 1, controller 2 etc for each of the controllers you have in the sequence, what it seems to do it to go through the effects assigned to each controller and write files out to the disk and then recompiles them into one final file when it finishes.
In my 38Mega Byte sequence, I have about 28 LED RGB controllers, most of which have somewhere in the region of 400,000 effects assigned to them, so I'm assuming that LSP is processing at least half of these (e.g. 400,000 x 14 = 5.6 Million effects) that seems to be the reason it takes my machine about 20 Mins to optimize the sequence before you can play back the sequencer in LSP 2.5
5. LSP Memory limits
LSP 2.5 never uses more than 2Giga Bytes of memory, it would seem that LSP cannot utilize more memory beyond 2Giga Bytes. You also need to take into account that other programs maybe using some of the available memory, so even if you have 2G of RAM installed, not all of the memory will be available to LSP 2.5
So in summary:
1. Faster CPUs should help with LSP 2.5 performance and wrting effects
2. Increasing memory in the PC beyond 2Giga Bytes does not increase LSP 2.5 performance either in writing effects to the sequence or when LSP 2.5 is optimizing a sequence, but increasing the total system memory may allow more memory to be available to LSP 2.5
3. Faster Hard drives should assist in optimizing sequences, since they seem to be very I/O disk intensive and should assist in writing effects/macros in the sequencer and help with the speed of optimizing a sequence
4. Very large sequences (at 38Mega Bytes and above) can cause memory exceptions when LSP optimizes the large sequence if you try to playback the sequence while LSP is still optimizing the sequence. This is far more likely as the channel counts increase, the song length increases and more complex Macros are used in a sequence
-
You might want want to revise the memory statement as other programs use memory so if you only have 2 gig the LSP won't get 2 gig. I would say 4gig is plenty for it right now but a future version "might" take advantage of more memory.
Good catch, I just edited my post above to reflect this
-
2. Increasing memory in the PC beyond 2Giga Bytes does not increase LSP 2.5 performance either in writing effects to the sequence or when LSP 2.5 is optimizing a sequence, but increasing the total system memory may allow more memory to be available to LSP 2.5
3. Faster Hard drives should assist in optimizing sequences, since they seem to be very I/O disk intensive and should assist in writing effects/macros in the sequencer and help with the speed of optimizing a sequence
Sorry to hear that more memory did not help.
SSD is a good thing, but you might consider using the extra RAM for a ram disk.
My SSD(s) are in a RAID configuration to squeeze out even more performance using striping. (RAID1) but I intend to use a RAMDISK for this year. I have 24GB in the system, and making use of that RAM makes sense to me.
I've researched a few products (and freebies) and one free one came out well in performance but was limited in max size (640MB). I found another that performed equally well, and doesn't have a max size.
http://ramdisk.tekcities.com/RAMDisk/ramdriv.htm
cost is $12 for 64-bit.
here is one place where I saw performance numbers...
http://www.raymond.cc/blog/12-ram-disk-software-benchmarked-for-fastest-read-and-write-speed/
i left my ssd performance numbers at home, but I recall I was getting RAMDISK over RAID1 SSD benefits of 10x or more.
-
Thanks for posting the real life info. I plan to do the same when I use the new box. I don't think I am unrealistic but I do expect much better performance with the new box but I am not expecting miracles either. There is a point of diminishing returns. I think we're pushing it and we are the few on the bleeding edge of this thing. I think the whole RGB think has really evolved rather exponential. Seems like yesterday folks were crying for express coops but now wondering why Ray can't ship pixels quick enough...
I think Dave is well aware that LSP is pushing hardware limitations and I expect future updates will address this. I recall a thread mentioning that he already plans to make the optimization multi-threaded. Also, I think RJ deserves kudos having the foresight to harness this technology and create our gear that outperforms commercially available alternatives. Doesn't hurt its less expensive either. ;D
-
Used the new box for the first time yesterday. Installed LSP and updates, imported sequences, etc and off I went. Definitely a night and day difference for me. Transitions and large cut and pastes that took 5-10 minutes took 5-10 seconds. More than happy I upgraded. Thanks for all the info and help guys, you rock! 8)
-
This may be a silly question, but does our video card impact our design-time performance? The biggest issue that I seem to have is delays and stuttering when using live osc/midi inputs from my korg controller and having the effects not match up with the music on playback.
-
Very timely post!
My video card is shared memory and I am not having issues with virtualizer. I just got my nanopad yesterday and gave it a real workout today. I don't think the delays are related to your video card. It's LSP. Although setting it up went fairly straightforward I don't like the results with the nanopad at all! Very laggy and sluggish. I've played music most of my life so it's not me. Something's not right.
I has stellar results using vixen's keyboard mapping music method. This was 90% done with that method: http://vimeo.com/34557018 (http://vimeo.com/34557018)
Now using mouse tap mode with good results. Debating on returning the nanopad...
-
Shoot, I thought it would be an easy fix and an excuse to get a new video card :)