Forum Replies Created
February 17, 2014 at 14:10 in reply to: 4.1.28 compatible issue with Win7 widget and outlook 2013 #2361
In Server setting, Performance, disable Use acceleration for display processing. See the screenshot attached.
This DirectX acceleration is what gives the big performance boost to NX 4.1 on Windows. It was extensively tested in in the last 5 months on Vista, 7 and Windows 8, with a wide range of applications and games. Not all possible games and applications, obviously.
We would love to introduce features in Beta versions and collect useful feedback from a big number of users for a couple of months, so that, when the feature is finally released to the public we can be reasonably confident that it will work in all possible combinations of HW and SW, but unfortunately things are not as easy as we would like to be. For example we made 7 different previews available from 2010 to 2014 and still in the 4.0 we had to solve a big number of bugs that had been in the software for 4 years and that nobody had ever reported. This is because NoMachine is not Microsoft and the number of users who are available to help as beta testers is proportionally small. Add to this that NoMachine, at the present moment, is almost exclusively used by professionals that only care about getting the job done. They have no motivation to play with something new. You can understand by yourself how getting feedback and bug reports while the SW is still in Beta is everything but simple. At the same time the number of possible HW and SW versions we can test is small, compared to the hundred of thousands users that download NoMachine.
But there is another option: introduce a feature as experimental and give users the possibility to try it. Unfortunately very few users in 2014 want to try experimental features. Very few users in 2014 ever go and change the configuration of something. Ten years ago it was normal for people to install a software and then play with the configuration to get the best performance. Not so anymore. Today everybody wants the software to “just work”. This is demonstrated by the fact we need to explain how to improve the quality of the display in this forum, something that would be easy to experiment with, if only people would be willing to experiment. But people are not willing to experiment, because they (rightly so) have something more important to do in their life. So for us, poor software vendors, is really difficult to strike a balance. In fact the way it works nowadays is that vendors introduce features and get feedback that they otherwise would be never able to get, while, at the same time, being quick to fix the bugs that they were unable to find by themselves. For example at the present moment I get 2 or 3 crashes a day from an application that is absolutely essential for my work. I never provided any bug report to the vendor, regardless of the vendor kindly asking to do that at each new crash. The application I’m talking about is free, it is used by hundreds of millions in the world and has corporate backings that, in theory, should be enough to iron out all possible bugs before distributing it to the users. But me and the vendor made a tacit pact: I use it for free while the vendor improves it, possibly introducing more bugs. The advantage for me is that I get a better application from one release to the other. The advantage for the vendor is that it gets bug reports from people like you that the vendor would have never gotten, even if it had kept the application in testing for many additional months.
I’m joking, of course, and we are surely not treating our users as beta testers. I think people underestimate the complexity of a software like NoMachine. They see “remote desktop” and think about a software that, at most, can be a couple thousands lines of code. It is not like that. The lines of code are in the number of millions, the number of layers a packet or a function have to traverse, the number of regression tests, new and old functionalities, different use cases, platforms and operating systems, packages and combinations we need to test at each new release is appalling. Of course we make mistakes and there is no excuse for that.
Did you try to play with the display settings?
While the session is running, open the menu panel and go in Display settings. There are two check-boxes: “Disable network-adaptive” and the “Disable multi-pass”. Be sure both boxes are checked. This will maintain the quality constant and will disable the progressive refinement. You can also play with the display quality. Moving the slider to the right will improve the quality to practically lossless.
The improvements I’m talking about are twofold: first we need to give better quality to text since the first encoding pass. We know what’s text, but video codecs are quite rigid in the way they allow the application to modify the behavior of the encoder. Enabling a feature of the codec often means disabling some other, so to get the best results we should modify the codec, but in this way we would lose the ability to link at run time and use whatever codec is available on the machine or use the HW encoder/decoder on the GPU. Secondly the “lossless” we obtain at the maximum quality is still the result of a YUV color-space conversion. This color-space conversion includes a down-sampling making highly-contrasted foreground/background combinations visually blurry. To solve this we need an additional RGB pass. This additional pass is not ready for the prime-time and we had to disable it in the 4.1 release.
You posted to say NX 4 is lagging. I don’t think it’s lagging in any way. I challenge you to show me a remote desktop system that is less “laggy” than NX 4. I think that, for other people reading this forum and that may have the experience of other remote desktop systems that, compared to NX 4, are WAY MORE lagging in every respect it was important to put your statement in the right context.
When I upgraded to NX 4.x from 3.5 I didn’t know such a big revamp in NX, otherwise, I would just stick with NX 3.5 (though it is no longer available)….
NX 3.5 is available, now, in the form of lightweight mode for version 4. From our benchmarks the lightweight mode is even faster than 3.5, at the protocol level, since it benefits from many additional X protocol optimizations that never went into 3, plus all the goodies and the rewritten core that went into 4.
I will see whether I can somehow get LXDE on CentOS 6 (currently not available in its official repository)….
Well, it seems the the world is moving forward. People willing to run a terminal or an IDE in Linux are becoming a shrinking minority. Probably they are insufficient, numerically and financially, to keep a company or even a small development team afloat.
Lagging compared to what? Lagging compared to some other remote desktop system? Because if you can show me a remote desktop system that is faster and less laggy than NX 4 I’ll be glad to spend some time learning how they did it.
Or do you mean lagging compared to NX 3? If it is lagging compared to NX 3 when running lightweight X applications like a terminal or an IDE, then I’d say that it’s rather normal. You can’t compare NX 3 to NX 4 in this respect. NX 3 is simply unbeatable at running such X11 applications. On the other hand, as a system designed to transport applications built on top of X11, NX 3 shares with X11 most of its limits and most of its technical deficiencies. At least all the technical deficiencies that we couldn’t work around because we had to respect the X11 protocol semantic and follow the way X11 applications throw the X11 protocol at us. So we started afresh and designed NX 4 to be a remote display system with none of these limits and technical deficiencies. We did a good job, I think, if our only competitor seems to be our own “old version”.
We tried to keep a single code base and move ALL users to the new system because keeping 2 completely different display systems in place is expensive, but since NX 3 was so good, we couldn’t make everybody happy. Some noticed a small performance decrease when running the same applications (and this was an outstanding achievement), some were extremely happy to see that now they could do things that they were never able to do with NX 3, and finally some others found the new system inferior to the old for the use cases they were interested in, that’s all that counts for them (then there were those who didn’t like the fact some things that were free went into a pay-for version, but that’s a different story). I believe in a couple of years nobody will look back to NX 3 with nostalgia, but when you have a system that works well and make people happy, why throwing the baby out with the bathwater? So we introduced this “lightweight” mode. The lightweight mode is available in the Linux Workstation and all products providing Linux virtual desktops (so no lightweight mode for the free Linux version that doesn’t have virtual desktops). It works in the same way as NX 3, that is by using the glorious X11 protocol. Fonts will never be blurry and the responsiveness will be the same as you were accustomed before, but probably you won’t be able to watch Netflix or YouTube, do videoconferencing, play a game or connect 10 users at the same desktop. But if you don’t care watching Netflix, doing videoconferencing or play a game on the machine you use at work, lightweight NX 4 sessions are what you are looking for.
In 4.0.x it was possible to have two connections with the same user. I often used this feature together with a co-worker to show some stuff on the server machine. However, currently when a user with the same username connects, the other connection is dropped. Why was this behavior introduced? What is the current version of having two connection to the same physical display simultaneously?
This change was one of the most requested “improvements” and the topmost reason users of the 3 complained about the 4. They argue that if you are connecting to the same desktop you should do by a different user, not hand over your credentials to somebody.
The old behavior can be restored by changing the server configuration. Somebody will post to tell you how to do it. The problem is obviously that the guest user connecting to your machine should not have at all a system account, but for this to work you will have to wait the NoMachine Anywhere service.
This is an area where we need to work further. People want to be able to watch a movie and play games on NX and at the same time work in a terminal and display sharp fonts, that are two different needs difficult to reconcile, especially if you are not talking about blitting glyphs over a PCI bus but there is the network involved. To respond to your question, yes, pixel-perfectness will be much improved going forward.
I have a server HP ML350G5 on which I had an installation of nomachine freenx 3.5.0
Do you mean NoMachine Free Edition 3.5.0 or freenx from freenx? I’m asking because they are very different software. I don’t know what freenx may have changed in your environment but for sure NoMachine 4 handles the upgrade from NoMachine 3 in a way that has demonstrated to be quite seamless for most users. Additionally I have really no idea of what NoMachine 4 may have changed in your environment, since NoMachine 4 doesn’t change ANYTHING in the user’s environment, except starting whatever desktop environment you may have configured, pointing it at a different X server. Bug or unintended side effects excluded, of course.
One can be this:
There is a TR for this problem:
The solution is being released today as part of the 4.1.
About the machine not starting on the physical X server I don’t have any idea. I suggest you check if the article applies to your case and then try to start the X server on the command line without loading any desktop environment.
If the X server loads, you can try some simple X clients, like xterm. Post the results here, so that we can start from there.
I think S.Herrmann refers to the fact the nxserver command is not put in the PATH environment, so that the user has to specify the full path on the command line. This was a design choice at the beginning of the 4 development to avoid making any change to the host environment that may interfere with the system (even if version 4 is a completely self-contained environment and there is little or no overlap with libraries and programs that may be installed on the host). I think we should re-evaluate this choice. While it makes NoMachine easier to manage and more secure for the advanced user that may set the path on a per-user basis, it also makes life for the average user harder.
In any case these are the paths on different operating systems:
Very good suggestion.
This is an extremely annoying problem and we are working at it, trying to solve it before the 4.1 release. In fact this is one of the bugs that we definitely wanted to get resolved at the cost of holding the release back. Since we are going to release on Wednesday, there is not much time remained.
Unfortunately the problem is not trivial. If you are interested in the challenges we have to face trying to make the clipboard work in NX, I suggest you give a look at this document, especially the part that says “It’s nearly impossible to make the clipboard shared between different desktop computers. In fact it is possible, but such an implementation would be needlessly difficult and complex”. Making the clipboard shared between different desktop computers is exactly what NX needs to do.
Add to this that 4 different X applications are likely to use the clipboard is 4 different ways and you get a picture of the situation. The reason the clipboard worked “better” in the 3 is that it was subject to many limitations. We worked hard to lift most of these limitations, but this obviously introduced new problems.
Desktop shortcuts will be possible in the upcoming 4.1
Just to clarify: desktop shortcuts work already in the 4.0. That is: you can double-click on a connection file (normally created in $HOME/Documents/NoMachine) and this will start the connection. You can also create a shortcut by hand, pointing the shortcut at a connection file, and place the shortcut on the desktop. What the 4.1 will add is the ability to create a shortcut automatically, when a new connection is created.December 24, 2013 at 15:15 in reply to: NX 4 free edition – Cannot connect to Windows host #1447
There are a number of fixes in the pipeline, many of them aimed at Windows, but, as you can read in another post, we were unable to finish testing in time to release. What I suggest is that you send the client and server logs so the support people can check if it is a known problem or something new.
Please follow the instructions here:December 24, 2013 at 13:46 in reply to: Private: Blurry fonts with latest NoMachine viewer #1438
We are working on that, believe me. We’ll give you text that is pixel-perfect, because we know that everything that is less than pixel-perfect is not good enough for text. In the 3 text was pixel-perfect because text was rendered by encoding and caching the X commands. This is not feasible anymore for a number of reasons that is too long to explain, the most important being that even X clients that still use the X protocol for rendering (many basically don’t any more and just push pixels down the link), do it in a way that is so bad and insane that the only reason the 3 worked was because there was a HUGE amount of work by the NoMachine developers to get around the inefficiencies and blindness of these clients. Even with these optimizations, the 3 could be brought to its knees by any multimedia load or any demanding application, simply because X is stateful (we made it way LESS stateful, but we could not get around the stateful nature completely) and there is a limit to the amount of compression you can do. You can compress gigabytes to megabytes, but if your clients generate gigabytes per second you have to surrender. You can’t drop content to recover it later, since the amount of state you have to store is so large that it basically becomes more convenient to use a different protocol. It is like using TCP rather than UDP. You better use UDP, when appropriate.
There is no indication that future applications are going to be less graphically demanding or that application developers are going to put more care about the networked case (basically they can’t, because making application is difficult already and the networked case is only interesting for a small minority, at least at the moment), so we designed a protocol that gives all the advantages of being stateless with all the advantages of being aware of what’s on screen, so that we can render the screen in the best way based on the network, the available hardware, the use case and content. The system is not perfect yet, and there is a long way to go, but I can say that at the present moment it already gives you the best speed and quality among all the similar systems we tested by a wide margin. It’s a good start for something that is basically a 1.0.December 23, 2013 at 18:31 in reply to: NX 4 free edition – Cannot connect to Windows host #1416
Not a license problem, as far as I can understand. The fact you are logged on the physical machine doesn’t account for a connection. It looks more like a server malfunction. What exact version are you running on the server? Be sure you upgraded to the 4.0.389.December 23, 2013 at 18:24 in reply to: NoMachine freezing intermittently when connected to Mavericks PC #1415
It doesn’t look the same problem. This is more likely a problem with the UDP traffic overwhelming your router. Quite uncommon indeed. There is congestion control in the NX protocol aimed at preventing this. The only cases where we encountered a similar problem was during development, due to some bug that prevented the congestion control to work correctly, and only with some routers. What the router should do, in these cases, is just throwing away packets, not certainly drop the Internet connection. Try disabling UDP in Edit -> Advanced settings.