I Have Cut My Time in SL In Half

[Love Zhaoying]

1 hour ago, Wulfie Reanimator said:

"I don't understand the cloud, so I don't like it. The problem must be the cloud. I refuse to help myself because it has to be cloud."

Here lies the thread.

Darn evil programmers.

[ChinRey]

3 hours ago, Prokofy Neva said:

This is the remedy merchants give, however. I think it would be nearly impossible to disseminate through the ranks the view/information you're trying to articulate here, because it won't be absorbed or believed.

It's informantion, not view - well and truly documented, tested and explained. If it was as easy as those merchants claim, LL would simply have set the LoD factor to 4 by default and be done with the whole problem.

But yes, I know it's not a very popular fact. People want to both have their cake and eat it - that's a fundamental part of the human nature.

 

2 hours ago, Prokofy Neva said:

WiFi has generally always worked with SL. For years.

The "for years" is an important part of the explanation. SL requires a lot more bandwidth today than it did only a few years ago.

There's more mesh for a start. A mesh can easily require several thousand times more data to be donwloaded than a prim or a sculpt.

Then there are textures. There are far more of them now than there used to be and people also tend to use higher resolution ones. A 1024x1024 texture can easily take a second to download even on a high speed connection. On a modest 3 Mbps connection it'll take more than five seconds, with 1 Mps 15 or more seconds. That's for each texture. There may well be a thousand or more of them in a modern SL scene and that means even with the top rank high speed connection it takes several minutes just to download the textures. Fortunately SL does have mechanisms to soften the blow, we have caching where textures and objects are stored locally so they don't have to be donwloaded again next time they are needed and we have a rather elaborate priority system where the objects and textures that are presumed to be most noticeable are downloaded and rendered before the viewer starts working on the finer details. Those features help a lot but there are limits to how effective they are.

[Qie Niangao]

9 hours ago, ChinRey said:

Increasing the LoD factor won't cure render failures, it will actually make them worse. To use another travelling metaphor, it's like trying to flog a horse within inches of its death to speed it up when it's already running as fast as it possibly can.

With the LoD factor you set the LoD level you want your viewer to achieve. When rendering stalls at too low a LoD level it's because the viewer is overworked and can't manage to reach the target. If you try to push it even harder by increasing the LoD factor and it will only struggle even more.

I'm skeptical. At least what I tend to see is corrupted, partially completed downloads that somehow make it into cache and survive across login sessions. Sometimes sculpties, more often textures, and very rarely mesh geometry too. I don't think those arise because rendering is overworked, but rather that network delivery is interrupted and (somehow) the viewer is rendering and caching the partially downloaded content.

I guess I could concoct a just-so story that the viewer is so busy rendering inflated LoDs that network data is falling in the bitbucket, but... seems unlikely.

(EDIT: That's not to say inflating LoD render factor is a good idea, whatever merchants may say to make their cheezy content acceptable to the less informed. It's a super bad practice. But I think the two problems are pretty independent.)

Edited November 17, 2018 by Qie Niangao

[Callum Meriman]

2 hours ago, Love Zhaoying said:

Darn evil programmers.

The evil ones blockchain their cloud, or was that cloud their blockchains.

[Callum Meriman]

19 hours ago, Prokofy Neva said:

SOFTMARE. This is my new term.

I like this softmare term.

5 hours ago, Prokofy Neva said:

what are THEY doing different NOW?

Oz did comment (I think) in the townhall that they recently had had some problems with the Bake Server as they brought in Bakes on Mesh. The bake server is a prime factor in the orange cloud.

Now, while it might not be the ongoing problem you have seen for years, it is something that would have made things noticeably worse in the last week only. Now the bake server fixed, the problem will return to it's normal state of 'errors less often'. I don't think they will ever fix the softmare completely, Prok.

[ChinRey]

16 minutes ago, Qie Niangao said:

I'm skeptical. At least what I tend to see is corrupted, partially completed downloads that somehow make it into cache and survive across login sessions. Sometimes sculpties, more often textures, and very rarely mesh geometry too. I don't think those arise because rendering is overworked, but rather that network delivery is interrupted and (somehow) the viewer is rendering and caching the partially downloaded content.

Ok. I was mainly responding to the LoD factor thing. What you're talking about happens earlier in the chain of course and has nothing to do with rendering so the LoD factor won't make any difference whatsoever in those cases.

When a file is downloaded, the viewer is supposed to send a receipt back to the server with some info (typically a checksum) to confirm that it got it all. That failsafe fails depressingly often in SL and I'm not sure why. It may be overload - not in the gpu but the cpu, connection or server - but this should be a high priority task so that shouldn't be a factor.

Btw, one thing we have to be very clear about here, is that this thread covers a lto of very different lag/render/performance issues with very diffrent causes and very different (if any) solutions.

As for lag in general, the only two solutions are that people realise that you can't fit a pint into a quart pot or that LL takes some drastic measures to keep the load in check. Neither is going to happen this millenium so for now, Second Life is what it is. Take it or leave it.

 

[Theresa Tennyson]

1 hour ago, Qie Niangao said:

I'm skeptical. At least what I tend to see is corrupted, partially completed downloads that somehow make it into cache and survive across login sessions. Sometimes sculpties, more often textures, and very rarely mesh geometry too. I don't think those arise because rendering is overworked, but rather that network delivery is interrupted and (somehow) the viewer is rendering and caching the partially downloaded content.

I guess I could concoct a just-so story that the viewer is so busy rendering inflated LoDs that network data is falling in the bitbucket, but... seems unlikely.

(EDIT: That's not to say inflating LoD render factor is a good idea, whatever merchants may say to make their cheezy content acceptable to the less informed. It's a super bad practice. But I think the two problems are pretty independent.)

Do you have HTTP pipelining enabled? (It will be unless you have specifically turned it off in debug settings.) When I had it enabled I would get bad "stereogram"-like textures on objects occasionally. It causes problems for quite a few people, apparently.

[Qie Niangao]

5 hours ago, Theresa Tennyson said:

Do you have HTTP pipelining enabled?

Thanks, I'll try turning it off in the LL viewer and Firestorm. It was already FALSE in Catznip (although I don't remember turning it off) which is the viewer I use most often lately. (Now I wonder if this could be part of triggering the old weird problem with alpha mode flipping from Masked to None, which I unsystematically associated with incidents of visible texture corruption.)

On the tangent of my crashes: looking at logs, it seems (always?) to be a memory allocation error, usually images followed by vertex buffer object pool. That's probably not particularly surprising but there's still the mystery of why it triggers so readily after I've run and exited Firestorm normally, or other viewers in combination, and the weird interactions with Chrome. (fragmentation? seems awfully retro though.) Anyway, I'm still collecting crashes, trying to be a bit more systematic about it.

[Prokofy Neva]

12 hours ago, Wulfie Reanimator said:

"I don't understand the cloud, so I don't like it. The problem must be the cloud. I refuse to help myself because it has to be cloud."

Here lies the thread.

No, I understand it all too well. It is you who are deceived. It's other people's computers -- therefore, you really can't control it as fully as you would like. But it saves you money so you pretend not to care.

[Prokofy Neva]

9 hours ago, ChinRey said:

It's informantion, not view - well and truly documented, tested and explained. If it was as easy as those merchants claim, LL would simply have set the LoD factor to 4 by default and be done with the whole problem.

But yes, I know it's not a very popular fact. People want to both have their cake and eat it - that's a fundamental part of the human nature.

 

The "for years" is an important part of the explanation. SL requires a lot more bandwidth today than it did only a few years ago.

There's more mesh for a start. A mesh can easily require several thousand times more data to be donwloaded than a prim or a sculpt.

Then there are textures. There are far more of them now than there used to be and people also tend to use higher resolution ones. A 1024x1024 texture can easily take a second to download even on a high speed connection. On a modest 3 Mbps connection it'll take more than five seconds, with 1 Mps 15 or more seconds. That's for each texture. There may well be a thousand or more of them in a modern SL scene and that means even with the top rank high speed connection it takes several minutes just to download the textures. Fortunately SL does have mechanisms to soften the blow, we have caching where textures and objects are stored locally so they don't have to be donwloaded again next time they are needed and we have a rather elaborate priority system where the objects and textures that are presumed to be most noticeable are downloaded and rendered before the viewer starts working on the finer details. Those features help a lot but there are limits to how effective they are.

Two months ago isn't "years," though. And it worked two months ago. THAT is the problem. There aren't significantly more textures now than there were two months ago. I don't need lectures about 1024 textures. I myself require 512 textures in malls. People think, oh, it's this deck on this one sim that is poorly textured. No. It's everywhere. On all sims. Even in an empty Linden ocean. It's not about caching. Every time I return to the same sim I go every day, it's grey squares all over again and loading from scratch.

 

[Prokofy Neva]

13 hours ago, animats said:

The main grid sim servers are not in "the cloud" yet.

The asset servers have been on Amazon AWS, front-ended by Akamai's web content delivery network, for years. The asset system works just like the Web; the viewer makes an HTTP request for a URL and it (usually) gets a texture or a mesh back. That's all standard technology.

If the viewer is crashing, something is wrong in the viewer or on the machine it is running on.

If Firestorm is crashing, you can probably get help from Firestorm support, and there are tools for analyzing Firestorm crashes. If the LL viewer is crashing, you'll have to rely on Linden Labs support.

Have you run the "Valley" test yet to check out your machine?

No.

If the asset servers on Amazon AWS for years etc etc and it's crashing now, it's  not about my viewer or my machine, which has been the same for the last 18 months when it worked.

It's about the Lindens' softmare.

For the millionth time, I don't and will never use Firestorm. I'm not interested in analyzing Linden's crash reports. They can do that. That's their job.

There is absolutely no need for me to be downloading some test that is 1,276 MB because there's something wrong with the Linden's software.

Stockholm Syndrome.

 

[Love Zhaoying]

10 minutes ago, Prokofy Neva said:

I'm not interested in analyzing Linden's crash reports.

Good luck.

[Theresa Tennyson]

7 hours ago, Prokofy Neva said:

No.

If the asset servers on Amazon AWS for years etc etc and it's crashing now, it's  not about my viewer or my machine, which has been the same for the last 18 months when it worked.

It's about the Lindens' softmare.

For the millionth time, I don't and will never use Firestorm. I'm not interested in analyzing Linden's crash reports. They can do that. That's their job.

There is absolutely no need for me to be downloading some test that is 1,276 MB because there's something wrong with the Linden's software.

Stockholm Syndrome.

 

As you're running Windows 10 then your machine has made changes to its operating system several times over the last 18 months, some of them major. Microsoft sends updates frequently whenever they feel necessary or whenever they want to add a new feature. They will upload and install the changes on their schedule without your knowledge or consent. This is not a conspiracy theory but the standard update path for a Windows 10 consumer installation. That's the main reason I've never installed Windows 10 on the machine I use for Second Life.

[Whirly Fizzle]

9 hours ago, Qie Niangao said:

On the tangent of my crashes: looking at logs, it seems (always?) to be a memory allocation error, usually images followed by vertex buffer object pool. That's probably not particularly surprising but there's still the mystery of why it triggers so readily after I've run and exited Firestorm normally, or other viewers in combination, and the weird interactions with Chrome. (fragmentation? seems awfully retro though.) Anyway, I'm still collecting crashes, trying to be a bit more systematic about it.

Sound like an out of memory crash.  Are you running a 32bit or a 64bit viewer?

[Qie Niangao]

1 minute ago, Whirly Fizzle said:

Sound like an out of memory crash.  Are you running a 32bit or a 64bit viewer?

It does sound that way, but I run a selection of 64 bit viewers. At the moment, they're:

• Firestorm 5.1.7 (55786) Jul 13 2018 20:00:04 (64bit) (Firestorm-Releasex64) with Havok support,
• Catznip R12 - Dec 31 2017 12:50:12 (64bit) (Catznip Release), and
• Second Life Release 5.1.9.519298 (64bit)

The machine has 8 GB physical RAM, and Windows 10 has somehow decided to allocate about 12GB paging space on my SSD. I mean, it's an old i5, nothing great, but it doesn't seem less capable than the average SL user's PC that doesn't have this problem. Hence a puzzle.

I owe you this along with some core dumps and logs, but just in case it's useful, here's the whole About Firestorm configuration info:

Quote

Firestorm 5.1.7 (55786) Jul 13 2018 20:00:04 (64bit) (Firestorm-Releasex64) with Havok support
Release Notes

CPU: Intel(R) Core(TM) i5-3450 CPU @ 3.10GHz (3103.05 MHz)
Memory: 8151 MB
OS Version: Microsoft Windows 10 64-bit (Build 17763)
Graphics Card Vendor: NVIDIA Corporation
Graphics Card: GeForce GTX 660/PCIe/SSE2

Windows Graphics Driver Version: 25.21.14.1694
OpenGL Version: 4.6.0 NVIDIA 416.94

RestrainedLove API: (disabled)
libcurl Version: libcurl/7.54.1 OpenSSL/1.0.2l zlib/1.2.8 nghttp2/1.25.0
J2C Decoder Version: KDU v7.10.4
Audio Driver Version: FMOD Studio 1.10.05
Dullahan: 1.1.1080 / CEF: 3.3325.1750.gaabe4c4 / Chromium: 65.0.3325.146
LibVLC Version: 2.2.8
Voice Server Version: Not Connected
Settings mode: Viewer 5
Viewer Skin: Firestorm (Grey)
Window size: 1636x1036 px
Font Used: Deja Vu (96 dpi)
Font Size Adjustment: 0 pt
UI Scaling: 1
Draw distance: 64 m
Bandwidth: 500 kbit/s
LOD factor: 2
Render quality: High-Ultra (6/7)
Advanced Lighting Model: Yes
Texture memory: 384 MB (1)
VFS (cache) creation time (UTC): 2018-11-15T8:53:16 
Built with MSVC version 1800
November 17 2018 11:24:55 SLT

 

 

[Prokofy Neva]

7 hours ago, Theresa Tennyson said:

As you're running Windows 10 then your machine has made changes to its operating system several times over the last 18 months, some of them major. Microsoft sends updates frequently whenever they feel necessary or whenever they want to add a new feature. They will upload and install the changes on their schedule without your knowledge or consent. This is not a conspiracy theory but the standard update path for a Windows 10 consumer installation. That's the main reason I've never installed Windows 10 on the machine I use for Second Life.

Windows 10 is on most normal people's machines. And tens of thousands if not more of SL residents have Windows 10. You can choose to accept the updates or not. Yes, even if you choose to opt in, they sometimes update against your will. I personally held off from installing Windows 10 itself forever, and when it tried to sneak it, batted it away. It finally barged through and installed anyway. And low and behold, nothing bad happened. SL worked. Slack worked. Etc. BTW I don't like Slack and it's gone now so that's not relevant.

But since I don't go around saying Windoz and Windblowz and all that stupid stuff and realize Microsoft is the most normal of all those Big IT companies for many reasons, I don't care if it updates or installs against my will. The whole concept of having to update softmare all the time is a built-in horror and perhaps some day geeks will learn how to make softmare right the first time without having to correct it all the time.

The Lindens, whatever you want to say about them, aren't stupid. And while they still include Unix nerds and they tend on the whole toward the Mac cult, being geeks, they are still sensible enough to test their software with Windows 10 because many customers use it and probably some of them even do.

If there was something about the latest Windows 10 installment that disrupted their software in some way -- unlikely -- they'd address it. I think they'd even put a message on the forums if they couldn't fix it, asking people to roll back to Windows 7 or the previous Windows 10 installment or whatever.

Therefore, since there's nothing like that, It's extremely unlikely.

That is, I don't think the Lindens would conceal a thing like that. They might, but it's not likely.

So it's something else.

Edited November 18, 2018 by Prokofy Neva

[Prokofy Neva]

14 hours ago, Love Zhaoying said:

Good luck.

They can analyze their own crash reports and fix their softmare.

[Coby Foden]

13 minutes ago, Prokofy Neva said:

They can analyze their own crash reports and fix their softmare.

You could also analyze your computer, like it has already been suggested in this thread. But you just don't do it. Why not?
Sometimes something can go wrong in our computers without us doing anything. They are just machines, and machines can have sudden faults.

[Prokofy Neva]

1 minute ago, Coby Foden said:

You could also analyze your computer, like it has already been suggested in this thread. But you just don't do it. Why not?
Sometimes something can go wrong in our computers without us doing anything. They are just machines, and machines can have sudden faults.

I have virus programs, clean-ups, optimizers, and I run them regularly. That's not the issue.

Something changed in the Lindens' softmare. 

There's an awful lot of denial about this, and that leads me to suspect the denial spreads into the Linden camp. Or rather, originates there, and spreads out to the fanboyz base.

Pity.

[Dillon Levenque]

On 11/16/2018 at 12:04 AM, Resi Pfeffer said:

You can have 527 TB of blue fermented cranberries, if you like.  I just told, what fixed my crashes.
 

Where do you get the blue ones? I'd love some!

As for the thread...wowee.

[Tari Landar]

13 minutes ago, Prokofy Neva said:

I have virus programs, clean-ups, optimizers, and I run them regularly. That's not the issue.

Something changed in the Lindens' softmare. 

There's an awful lot of denial about this, and that leads me to suspect the denial spreads into the Linden camp. Or rather, originates there, and spreads out to the fanboyz base.

Pity.

Like you said..it works fine for most others. So look at the variables that are different between your machine and all the others it's working fine for. 

Like you said, if it was something on LL's end, and widespread, they'd say something about it.

You're negating your own arguments here, and you don't even realize it, lol. 

 

[Coby Foden]

17 minutes ago, Prokofy Neva said:

I have virus programs, clean-ups, optimizers, and I run them regularly. That's not the issue.

Those cannot repair your computer if there is some problem in it. Like in the hardware, they don't last good for ever. They cannot repair errors in the programs, not in the operating system, not in the drivers.

Edited November 18, 2018 by Coby Foden

[Prokofy Neva]

35 minutes ago, Tari Landar said:

Like you said..it works fine for most others. So look at the variables that are different between your machine and all the others it's working fine for. 

Like you said, if it was something on LL's end, and widespread, they'd say something about it.

You're negating your own arguments here, and you don't even realize it, lol. 

 

No, I didn't say that. I said *it worked fine for me* until just recently. So it's not my computer, my ISP, my settings, Windows 10, etc. They haven't changed.

And I also said Windows 10 works fine, and SL works fine on it, for many people, so the problem is NOT LIKELY Windows 10. Different.

What has changed? LL's softmare.

Most people don't complain because they think it's just their problem. So they just quit.

And that's because they're in a dreadful climate of stark, severe geek conformity and incitement of hatred of anyone who questions whether it could be something more than just their machine.

It's a rigid, totalitarian cult where you have to make enormous sacrifices like a Russian or Chinese dissident to try to get the truth admitted.

Perhaps the worst aspect is the disinformation, misquotations, deliberate fake understandings, etc.

 

[Prokofy Neva]

35 minutes ago, Coby Foden said:

Those cannot repair your computer if there is some problem in it. Like in the hardware, they don't last good for ever. They cannot repair errors in the programs, not in the operating system, not in the drivers.

It's not the drivers. It's their softmare. I'm not the only one experiencing this. The forums represents 5 percent of the user population in readership, and 2 percent of the user population in posters. Look at this thread, and count how many people say they are experiencing the same thing. The end.

[Scylla Rhiadra]

11 minutes ago, Prokofy Neva said:

It's a rigid, totalitarian cult where you have to make enormous sacrifices like a Russian or Chinese dissident to try to get the truth admitted.

Fear not! We've alerted Amnesty International and the UNHRC. We anticipate that with proper pressure, applied through international sanctions and boycotts, we should have your computer up and running SL properly in a few short years.

Hang in there, Prok! The eyes of the free world are upon you!