Linux performance still a let down on desktop.

[BulletDust]

Except it isn't like that at all. The cause of tearing is obviously a software issue, that is quite widespread.

Yeah, it's widespread, Linux sucks, you're the man...

...Move on.

 

[Snowdog]

Yeah, it's widespread, Linux sucks, you're the man...

There are literally thousands of posts like this:
https://forum.kde.org/viewtopic.php?f=111&t=131912

This guy spent a year trying to fix tearing on his GTX 770. Much newer card than mine.

Claiming it's an isolated issue based on old broken HW is just completely false.

It's widespread on NVidia GPUs of just about every vintage.

 

[BulletDust]

I can find literally thousands of posts regarding everything from updater issues under Windows 10 to Windows 7 installer issues also, and I have and I've provided links.

Doesn't mean the issue's widespread, doesn't mean it affects everyone. I just installed five distro's without an issue. Scrolling smooth as silk, no tearing.

But hey! Believe what you want and piss Manjaro off, it's obviously not suitable.

 

[Snowdog]

I just installed five distro's without an issue. Scrolling smooth as silk, no tearing.

Sure, after I posted the solution. You yourself agreed that "forceCompositionPipeline" had to be set.

 

[MrCrispy]

Sadly my distro hopping and Linux experiments have come to a crashing halt.

My Dell laptop seems to have died while trying to update the bios. I'm very mad.

Nothing to do with Linux at all. But I don't have any other PC to try this on and I was really getting somewhere. I will post another thread to get some help but it doesn't look good.

 

[SmokeRngs]

Except it isn't like that at all. The cause of tearing is obviously a software issue, that is quite widespread.

The behavior is strictly based on proprietary driver availability which is hit and Miss on Linux. I have tried 8 distros. All the Ubuntu Distro were eventually workable, because they were all using the same driver base.

OpenSuse, and CentOS were just a complete shit show to install Proprietary drivers so I never got them installed. Manjaro for some reason doesn't have the February udpate(340.102), and is stuck on the Dec verion (340.101). I speculate this is why the ForceCompositionPipeline was rejected on Manjaro.

This is all direct cause and effect of software and settings.

Set ForceCompositionPipeline and it works properly, don't set it and it doesn't.

It really doesn't get more obvious than that.

And yet my server has very similar hardware to your system and has an 8600GT running the latest OpenSUSE Leap with the open source drivers and I'm not having any tearing or whatever issues. I will repeat, I don't even have the closed source drivers from nVidia even installed. It's the same generation card with the same base architecture as yours. You were having problems other than video on all the distros you tried. The common thread here is your system. Having stability issues with all of the distros you tried with most of them being quite popular and I'm doubting there are widespread problems like you're having points to something wrong on your end. I've been working with, troubleshooting and fixing computer issues for more than two decades and everything I know is screaming hardware issue.

 

[BulletDust]

Sure, after I posted the solution. You yourself agreed that "forceCompositionPipeline" had to be set.

Snowdog, while I can't be arsed going back through this ridiculous thread, I'm fairly sure I posted screenshots highlighting force composition pipeline before you mentioned it - I've been using force composition pipeline with Nvidia drivers for years now, it's not rocket science by any stretch of the imagination.

But I tell you what, if you want to take credit for force composition pipeline, be my guest.

 

[BulletDust]

And yet my server has very similar hardware to your system and has an 8600GT running the latest OpenSUSE Leap with the open source drivers and I'm not having any tearing or whatever issues. I will repeat, I don't even have the closed source drivers from nVidia even installed. It's the same generation card with the same base architecture as yours. You were having problems other than video on all the distros you tried. The common thread here is your system. Having stability issues with all of the distros you tried with most of them being quite popular and I'm doubting there are widespread problems like you're having points to something wrong on your end. I've been working with, troubleshooting and fixing computer issues for more than two decades and everything I know is screaming hardware issue.

Agreed, I've highlighted distro's where there was no need to enable force composition pipeline at all under Nvidia drivers and the Nouveau drivers work perfectly with no tearing at all.

Sadly my distro hopping and Linux experiments have come to a crashing halt.

My Dell laptop seems to have died while trying to update the bios. I'm very mad.

Nothing to do with Linux at all. But I don't have any other PC to try this on and I was really getting somewhere. I will post another thread to get some help but it doesn't look good.

Ouch! That does suck. Sorry to hear.

 

[ChadD]

There are literally thousands of posts like this:
https://forum.kde.org/viewtopic.php?f=111&t=131912

This guy spent a year trying to fix tearing on his GTX 770. Much newer card than mine.

Claiming it's an isolated issue based on old broken HW is just completely false.

It's widespread on NVidia GPUs of just about every vintage.

And yet mine and the 3 cards previous all worked fine. My clients are busy editing 4k video on NV hardware under Linux.... clearly you are correct. Every single Linux user in the world with NV hardware must be blind.

We have clearly accepted that you have said issue... we have admitted that more then a few people have reported the same issues. (almost always with cards on the older unified driver). I have yet to see wide spread reports of people having this issue with the newer set of unified drivers for cards from Fermi onward (Released March 2010). For people reporting issues with cards from their on... I would tend to look at other factors to explain their issues. For the older cards yes some have reported the same issues with tearing that can almost always be corrected by changing a few conf settings. Not sure how that is a show stopper as the same types of hardware issues pop up in EVERY operating system.

 

[BulletDust]

Watching my daughter playing Roblox on her 6i5SYH Intel Iris Windows 10 based NUC before, there was definite tearing let me tell you! Might have to run the tearing test on that one!

 

[B00nie]

Watching my daughter playing Roblox on her 6i5SYH Intel Iris Windows 10 based NUC before, there was definite tearing let me tell you! Might have to run the tearing test on that one!

My kid loved roblox too. I never could figure why.

 

[Snowdog]

I have yet to see wide spread reports of people having this issue with the newer set of unified drivers for cards from Fermi onward (Released March 2010). .

Yes, here it is claimed that it is rare issue, confined to old broken hardware seldom seen on new cards. But on actual Linux forums it seems to be a readily acknowledged issue, often reported by long time Linux users.

From the Manjaro Wiki:
Remove screen tearing with dcompmgr - nVidia

Note: There are a LOT of different 'solutions' or
part solutions floating around. So be aware that
it seems to come down to the luck of the draw as
to whether a particular solution works for you.

From Linux Stack Exchange:
Why is video tearing such a problem in Linux?

I've used many variants of Linux(mostly Debian derivatives) for over a decade now. One problem that I haven't seen solved satisfactorily is the issue of horizontal tearing, or Vsync not being properly implemented.

I say this because I use used 5 different distros on 4 different computers with various monitors and Nvidia/AMD/ATI/Intel graphics cards; every time, there has been an issue with video tearing with even slight motion.

The Top Answer:

This is all due to the fact that the X server is out-dated, ill-suitable for today's graphics hardware and basically all the direct video card communication is done as an extension ("patch") over the ancient bloated core. The X server provides no builtin means of synchronization between user rendering the window and the screen displaying a window, so the content changes in the middle of rendering. This is one of the well-known issues of the X server (it has many, the entire model of what the server does and is outdated - event handling in subwindows, metadata about windows, graphical primitives for direct drawing...). Widget toolkits mostly want to gloss over all this, but tearing is still a problem because there is no mechanism to handle that. Additional problems arise when you have multiple cards that require different drivers, and on top of all this, opengl library has a hard-wired dependency on xlib, so you can't really use it independently without going through X.

I have pointed out before, it is NOTHING to do with the age of the card, this guy also on GTX 970 (From Archlinux forum):
Nvidia Geforce GTX 970 Screen Tearing

I have had screen tearing when watching videos, gaming or sometimes scrolling and have tried everything I could find to fix the problem. I am using the proprietary driver and KDE 4 but I have the issue on plasma 5 and GNOME too and I am so tired of this issue i'm considering getting a new video card. I have a GeForce GTX 970. I have tried reinstalling arch and am all out of ideas.
What i've tried: ...

It gets interesting when he discusses the eventual solution, which only works if he first installs the
Nouveau drivers, then the Nvidia drivers. Note it won't work if only installs the NVidia drivers, he has to first install the free drivers and switch.

Which could be another reason Arch based Manjaro that started with NVidia drivers didn't work for me.

Also that is not a sign he has bad HW, it is a sign of cobbled together software though.

I am done with this issue. I have a suitable fix, though it does confine me to Ubuntu based Distros. I am moving on to try and fix other issues. Poor handling of my dual display setup is next.

But if you want to keep claiming this is an isolated incident based on old HW, I will continue to provide counter examples, and those counter examples are for all practical purposes endless.

 

[ChadD]

Whats your point ? That the manjaro wiki says there are LOTS of solutions... there are. I can think of 20 different ways to solve lots of problems that hardly means those problem are experienced by LOTS of people. I know English can be hard but geez.

We have agreed with you yes GPUs in general can be touchy... that is true of WIndows and Linux. I can find you 20 articles describing how to solve black screens of death in windws as well... HOWEVER I will not state that everyone running NV hardware in windows is having those issues. Cause I know better. Those issues under Windows are yes shockingly often caused by MFG designs that differ greatly from the expected reference design the drivers assume.

Pointing to wiki pages intended to help folks that do run into issues... isn't some admission that 100% of people with said hardware are going to need those solutions.

https://forums.geforce.com/default/...10-creator-update-and-nvidia-drivers-issues-/
Clearly based on this thread... everyone with brand new 1000 class NV hardware should stick to windows 7 cause 10 is clearly a no go. I mean geez MS brand new hardware and your sticking people in bios reboot loops.... just not ready for prime time.

See how silly that sounds. Yet I have proof that clearly "lots" of people are having major issues. I think most of us are smart enough to understand what is happening to that group of "lots" of people isn't an across the board issue.

 

[Snowdog]

Whats your point ? That the manjaro wiki says there are LOTS of solutions... there are. I can think of 20 different ways to solve lots of problems that hardly means those problem are experienced by LOTS of people. I know English can be hard but geez.

It says "solutions" and suggests you need luck to find a working one.

You are also ignore the issue on Linux stack exchange, this problem was on 4 different computers, using every kind of video card.

Sound pretty damn widespread unless this guy was only using a series of old broken computers with old broken video cards.

And the answer that most people agree with is that X server is a wretched old piece of code. Which is an indicator that this is well known problem in the Linux community, not some edge case.

 

[ChadD]

It says "solutions" and suggests you need luck to find a working one.

You are also ignore the issue on Linux stack exchange, this problem was on 4 different computers, using every kind of video card.

Sound pretty damn widespread unless this guy was only using a series of old broken computers with old broken video cards.

And the answer that most people agree with is that X server is a wretched old piece of code. Which is an indicator that this is well known problem in the Linux community, not some edge case.

The point is the issue is SHITTY hardware manufacturing. Which is why one solution doesn't fit all. That is pretty clear to anyone willing to think about it just a little.

Yes your issue is related to a video card that I am almost 100% sure is not even close to a reference design. The fact that some solutions work for one card but not another is completely in keeping with that. The issue is Nvidias and the nature of closed source drivers... and allowing manufacturers to futz with the card designs all the time. Most don't deviate very far from what NV designs... some however make all manner of changes which can lead to issues with future driver updates. This is a known fact that effects EVERY operating system... its not just a Linux issue, nor is it an issue with every NV card sold.

So as I have said before go buy a new card if you really don't want to be tweaking things all the time.... or accept that your 10 year old NV card is going to require some tweaking to setup in Linux. That it installed and worked no issues for you in Windows is great, glad you didn't have to change any settings or deal with Bios reboot loops as 1000s of others have had to deal with in Win10 on NV hardware.

X server btw isn't an issue for a great number of people.... X is a standard framework that allows the community to build a lot of interesting software. That doesn't mean its going to be around forever. The community is in the process of replacing it with Wayland which works very much the same way with newer slimmer faster code, and of course over the years better ways of handling things like scaling have been developed. NVs closed source drivers also don't play very well with Wayland right now. To be honest ya lots of Linux developers are more then a little annoyed with NV right now. AMD is starting to get a lot of Linux development love the last year or so as they have been a lot more helpful ensuring the open source graphics stacks and drivers have the information required to ensure these types of issues don't effect ATI hardware. Still not sure what Open source politics has to do with the fact you have an ancient non reference Nvidia card that seems to be giving you grief.

As for why we are pointing at your card its simple....
http://www.fudzilla.com/news/graphics/9022-palit-changes-nvidia-8800-gt-reference-design
https://www.dvhardware.net/article22900.html

You are 100% running a card that is not using a reference design. Palit played around with the power design of their 8800 cards and made them dual slot. I'm not saying every reference card in the world is perfect and never has scaling issues.... however for people reporting issues they do seem in most cases to be using non reference designs using different power designs, different cooling designs... and in some cases such as yours completely different PCB designs as well.

My serious suggestion is when the day comes for you to buy another GPU... don't have your head turned by cards from Manufacturers that claim to have made a ton of "improvements" to NV or AMD designs. Stick with the MFGs that simply mass produce the reference cards. They tend to cause fewer issues in any operating system. (I also noticed that a large % of the Windows bios reboot loop reports where from people with non reference cards) Screwing with the suggested power solutions for a GPU seems to be bad mojo.

 

[Snowdog]

The point is the issue is SHITTY hardware manufacturing. Which is why one solution doesn't fit all. That is pretty clear to anyone willing to think about it just a little.

WTF? Do you think a better power delivery system and better cooler would negatively driver functions of a Video Cards, like disabling the default composition pipeline.

That is beyond absurd.

If my card were actually functionally different from a software perspective, I wouldn't have been been running NVidia Reference drivers on it for the last 8 years.

You are even further out on a limb, most people today look for non reference designs for better power/cooling/over-clocking. Which is exactly what Palit did. A more robust power system, and improved cooling.

 

[BulletDust]

Running Nvidia hardware under Windows using Nvidia Surround I had all sorts of issues, the drivers actually worked better on release than they did after 4k support was introduced. I used to have the taskbar swapping monitors at random, widows maximising behind the taskbar and icons moving all over the desktop - On the Nvidia forums us surround users were kicking up a storm. I waited about twelve months, if not more for a solution - In the end I was forced to ditch the surround setup altogether for a single 4k monitor as there was literally no viable solution whatsoever.

Not once did I post on forums jumping up and down blaming cobbled together software. There isn't an OS on the planet that's immune to the sorts of issues you're describing, you need to accept this fact, stop with the pointless negativity and enjoy the challenge of something new.

It's like you're sulking in front of the machine every time you try Linux and it's completely unfounded. My experience with multi monitor under Linux is that it actually works better than it does under Windows, just to tell it how it is, don't shoot the messenger.

 

[BulletDust]

WTF? Do you think a better power delivery system and better cooler would negatively driver functions of a Video Cards, like disabling the default composition pipeline.

That is beyond absurd.

If my card were actually functionally different from a software perspective, I wouldn't have been been running NVidia Reference drivers on it for the last 8 years.

You are even further out on a limb, most people today look for non reference designs for better power/cooling/over-clocking. Which is exactly what Palit did. A more robust power system, and improved cooling.

I've had non reference designs with hardware/firmware > driver compatibilities in the past. One card was a Geforce 2 GTS and the other was a 6600GT, both non reference, both had issues with certain Nvidia drivers - In fact the 6600GT wouldn't even run on all Nvidia drivers. From that point on it was reference only for me.

 

[ChadD]

You are even further out on a limb, most people today look for non reference designs for better power/cooling/over-clocking. Which is exactly what Palit did. A more robust power system, and improved cooling.

Yes I know exactly what they did to their cards. They took a reference power design, designed by a company 10x larger then them with engineers paid a whole heck of a lot more.... and threw it out the window. Designed their own "better" power scheme to feed a GPU with millions of transistors that they knew nothing about. (unlike the engineers that designed the reference setup) They did this because it gave them an "edge" over the 100s of other 8800GT vendors with cards on the market.

Should that effect scaling ? No it shouldn't. Yet your Palit designed card seems to be doing a few things that are atypical.

There is very good reason why you will never find a Quadro or Firepro card with a non reference design... least of all a redesigning of the systems delivering power to the GPU.

Gaming card MFGs have been doing it for ages as a selling point... cause sure it sounds good. More power more overclocking... sounds great. Accept its not exactly a great idea when your dealing with cards that will be getting driver updates for 10+ years... drivers that are unified and likely hardly tested on hardware when it gets that old.

 

[Snowdog]

Should that effect scaling ? No it shouldn't. Yet your Palit designed card seems to be doing a few things that are atypical.

There is no should about it. It's completely irrelevant to the issue at hand.

Think about changes they made, 3 phase instead of 2 phase power. I'll ignore the bigger fan/Heatsink assuming that you don't think that would detrimentally affect drivers.

Outcome: Less ripple/less voltage sag.

Do you think the driver checks for ripple/sad and disables the composition pipeline (only in Linux) because the values are too good.

These are software issues, plain and simple. They are widespread especially among NVidia users.

The guy's issue on the Arch forum were particularly indicative of software issue you expect with a complex system assembled from various pieces from unrelated groups.

He could get Tearing fixed with Proprietary Drivers, Only if he first loaded Open Source drivers, and then switched to the NVidia drivers.

That is a classic telltale of something not be provisioned properly by the proprietary drivers, so it only gets set up correctly if he first installs the Free drivers.

This is also exactly the kind of issue you expect with complex system assembled in from different optional parts in a large variety of ways.

If you can't understand that, it really isn't worth discussing. Start your own thread to discuss Power electronics that change the settings on your GPU.

 

[BulletDust]

There is no should about it. It's completely irrelevant to the issue at hand.

Think about changes they made, 3 phase instead of 2 phase power. I'll ignore the bigger fan/Heatsink assuming that you don't think that would detrimentally affect drivers.

Outcome: Less ripple/less voltage sag.

Do you think the driver checks for ripple/sad and disables the composition pipeline (only in Linux) because the values are too good.

These are software issues, plain and simple. They are widespread especially among NVidia users.

The guy's issue on the Arch forum were particularly indicative of software issue you expect with a complex system assembled from various pieces from unrelated groups.

He could get Tearing fixed with Proprietary Drivers, Only if he first loaded Open Source drivers, and then switched to the NVidia drivers.

That is a classic telltale of something not be provisioned properly by the proprietary drivers, so it only gets set up correctly if he first installs the Free drivers.

This is also exactly the kind of issue you expect with complex system assembled in from different optional parts in a large variety of ways.

If you can't understand that, it really isn't worth discussing. Start your own thread to discuss Power electronics that change the settings on your GPU.

Arch is a high end distro designed for the tinkerer, start tinkering!

Otherwise, use one of the more common distro's.

 

[Snowdog]

done