leliel Mirihi

It was a driver issue. The viewer asked the driver how many TIUs it had and the driver said "I got 32 man" and the viewer said "Cool dude, I believe you 'cause like you know what's going on, so I'm gonna use all 32 now" and the driver said "Oh noes! I don't actually have 32. Arg crash!". So the fix was to make the viewer not believe everything the pot smoking hippy driver said.

SSDs have a constant seek time, it doesn't matter where the data is on the drive. Defragmenting a drive is a legacy from spinning platters which have a variable seek time depending on where on the platter the data is and where the head is at that moment.

Ann Otoole wrote: I have been unable to get an answer on this. LL was working to fix depracated openGL calls. So exactly what driver version is supposed to work with SLv3.2? Not sure I want to upgrade and disable my use of TPVs to use this beta version. The 'bug' has been fixed in the shining-fixes branch. The problem turned out to be that nvidia's driver was lying about how many texture image units the chip had, saying it had more than it really did. But I'm shure the myth of depreciated opengl calls will stick with us for years to come.

Mannie Clowes wrote: its not difficult to compare SL to modern games at all, even if you have 10 people all with different variations of those necklaces standing next to each other you wont get anywhere near the graphics load that any of the new games will have, compared to them SL is poorly coded thats why you need a machine with a little umph to run sl nothing more The reason you don't get as high of a graphics load is pipeline stalls. GPUs can only draw a limited number of textures at a time, most modern chips can do 16 or so max. So the more textures you have the more you have to split up the work in to batches, the more batches you have the more time you have to spend changing state as you switch between them. The ridiculously high textures counts in sl is one of the main reasons for the low frame rates we get.

Dresden Ceriano wrote: Six Igaly wrote: also @ Dresden ..about this alpha glitch bug thingy. There is no longer such thing if you are on one of the 2 latest beta viewers (yes the offical). I was working on a project that supposed to get advantage of the glitch and after an update it was gone. And now latest beta release it still is gone. The viewer handles transparancy perfectly well. Unless of course if it turns out in next update the bug was just bugged. :smileyhappy: If this is, indeed, the case, it's a major accomplishment. I was under the impression that it would take so much processing power to calculate that it would be completely impractical to achieve. Makes me what to check out the latest viewer so I can see it with my own two eyes. ...Dres Your impression was correct, the 'new' feature is something that was added in 2.2 for deferred rendering which will now be enabled all the time, and it's really just a smarter version of a dirty hack that's been in the viewer for many years. It minimizes the sorting problem but does not eliminate it. We are still a few years away from having GPUs fast enough to use the perfect fix in sl, altho current high end cards are fast enough to use it in professionally made games that don't have a million alpha textures.

Well that looks like a new one. Create an issue in jira for it.

Verena Vuckovic wrote: You can't create a 1/200 scale model of Titanic and then claim that it's just a matter of 'personal opinion' and an 'internal model reference' if you make the Captain taller than the masts and the iceberg the size of a cocktail ice cube ! Yet people do exactly that every day in sl. Have you ever tried mixing furniture from different stores? You'll be too big for some and too small for others. That's because apparently most creators in sl have no idea what the word scale means, they probably think it's just an alternate spelling of size.

MrsSissyMorrison wrote: to get this how yu want- it is simple, make yur texture however dark you want it and set to full bright on the prim. when set to ful bright the testures are not impacte by the sun changing. They aren't effected by any other lights either, it looks just as bad in the opposite way. Especially if you make light colored prims fullbright, then at night time avatars will look super dark in a bright room, very weird.

Cerise wrote: The old non-deferred antialiasing still works here in davep_shining-fixes 243593, so it will be good to check all the setup first. I see you're right, I spoke too soon. Looks like it still has the FBO based AA, I was thinking of the sampler2DMS one.

Cerise wrote: leliel Mirihi wrote: The shining-fixes branch removes the MSAA code and replaces it with FXAA so that's not really a problem anymore. The FXAA stuff is used in deferred rendering ("lighting and shadows") mode. It is a good addition because hardware FSAA did not like to work at all with deferred rendering, but it really does not work as well as the hardware version that is still used when deferred is off. The MSAA code was completely removed, I just tested now and AA doesn't work unless deferred is enabled. Should file an issue in jira for that, there's no technical reason why the viewer can't use FXAA or any other shader based AA method with the forward renderer.

Cerise wrote: FSAA still works in SL on Linux, but the fancy effects that recent desktops use can interfere with it. I'm looking at you, compiz. I think recent xfce can interfere too. This is not only an SL problem, it breaks other things like games that use FSAA too. On Ubuntu, you can choose the "classic desktop with no effects" from the login screen menu, and that will free up the resources to allow FSAA. The shining-fixes branch removes the MSAA code and replaces it with FXAA so that's not really a problem anymore.

WanLi wrote: TP-ing to busy sims causes the avatar to (1) freeze - not move at all; or (2) turn but will NOT walk straight ahead or back, sideways or fly ; or (3) will walk for awhile in quiet sims but even then, after walking short distances, the avie will keep on walking (walking motion in place) until the program eventually crashes and offers the IM-Chat only option The viewer isn't crashing, it's getting disconnected due to high packet loss and/or high latency. Double check your networking hardware, any software you have monitoring or using your network (firewalls, anti-virus, bittorrent), and test your network connection (don't trust what your ISP says).

Ann Otoole wrote: I can tell you this. 6 months ago the LL viewer was vastly superior to what it is now. I'll certainly agree with you that the viewer was performing much better 6 months ago. Oh and you are full of it where PSUs are concerned lol. You should read some forums on the topic sometime. I had to buy a 750 to run a 240 and the rest of my computer properly. The GT 240 doesn't have a PCIe power plug which means it gets all its power from the PCIe slot and per the spec it can only draw 75w max. Anyway back on topic AA works fine for me on linux. This is on Ubuntu 11.10 with a GTX 560 TI and the 280 driver.

SappaDallagio wrote: What I haven't seen in this discussion is whether or not the iPad supports OpenGL. Does it? I don't have one and I'm not inclined to spend time searching for that, simply because I have no desire to own an iPad. But, to all you iPad owners who believe this device can render SL, does it have the OpenGL implementation within it to do so? All most all mobile devices support OpenGL ES 2.0. The only exception I can think of would be Windows Phone 7, but nobody buys that trash.

Several of the bugs have been fixed, they just haven't been pushed out to a release yet. Unlike many TPVs LL runs the viewer through a lengthy QA process before releasing. Laugh at that all you want, and I'll even agree with you to some extent, but keep in mind that the TPVs aren't developed independently, they're working off of LL's code. They only fix bugs that LL missed[*] so of course they will be "better". * Except for bugs they introduced with their own features obviously.

Min Fairweather wrote: Hi Iellel, The mac was bought in 2007 but I guess that doesn't mean it's not a 2006 model. More info... It's a late 2006 20", that machine should not be struggling on low. What exactly are the problems you're having?

Something doesn't sound right, a 2007 iMac should have no problems running on mid with the latest viewer. Unless you have the late 2006 17" iMac with a Core 2 Dou and a GMA 950.

Void Singer wrote: My thoughts on land valuation are driven by general behavior of users (in any industry), people spend what they can afford to, and when something comes along that improves that, the majority take that as a bonus and stick with paying just as much and enjoy the newfound value for their dollar. some will downgrade, but others will move up seeing an increased value to a product they didn't see as worth the cost before, and the increased exposure is instant marketing gold, because it gives people a taste of more. That's only true with pay as you go pricing structures, with segmented pricing structures it's often the opposite. The best example is probably cable tv. People are forced to buy large channel packages at a hefty premium just to get the few channels they actually want. The cable providers have steadfastly refused to offer a la carte subscriptions because the know the majority of people would only get the few channels they want causing the company's revenue to drop like a rock. The same could very well be true in sl, I've seen plenty of parcels where the owner only used a portion of the land, the rest was purchased solely for the extra prims or to fill out tier.

It was just added a few days ago. It will work from the web site as well so you don't need to be in world to send messages.

Innula Zenovka wrote: It's two related problems, though, isn't it? There's the general problem with the TDR error firing too readily and the specific one that SL's using the deprecated OpenGL calls is pretty much guaranteed to make that happen. Every single person who has this problem has said they can log in with basic shaders disable. With basic shaders disabled the viewer exclusively uses the deprecated OpenGL calls, basically the viewer uses OpenGL 1.5 with shaders disabled, and OpenGL 1.2 with VBO disabled. Furthermore most, but not all, people report being able to put the graphics back on high/ultra once they have logged in. In other words it's something the viewer is doing at startup that's causing the driver to die. Here's another data point, the reason why the fixed function api is deprecated is because modern GPUs don't have that hardware anymore, the driver emulates it with shaders. But almost everyone who has this problem is using a GTX 460, if the problem was caused exclusively by the fixed function api then everything from the GeForce 8 and up should be crashing. Yet another data point, I've seen people talking on gamer forums about having the same problem with the GTX 460 when running DirectX games. Nvidia changed something in the 280 driver that makes it much more picky about things than the previous version. Matters are futher complicated by the fact that, under certain circumstances -- and, in my case, this is certainly one of them -- the error, or the circumstances that caused it, seems to damage the Nvidia driver to the extent that not even doing a clean reinstallation using the Nvidia installer fixes things, since that apparently leaves some damaged dlls and registry settings untouched. That happened to me, and after that I just couldn't in with anything. That just sounds like the driver is barfing all over itself.

I'm using Ubuntu 11.04 with an evga GTX 560 TI that I've had for around 6 months and the fan speed & temps are the same now as the day I bought it. The temperature can get as high as 80c under load and the fan will slowly speed up as it gets hotter, but rarely get too loud, when idle it's nice and quiet at around 38c. The only time the fan has ever really been that loud was on hot days during the summer, can't really blame that on the card tho.

Actually the issue has been around for several years and isn't specific to OpenGL or even Nvidia. http://forums.nvidia.com/index.php?showtopic=65161

And 1.15 was still mostly useable 3 months after 1.16 was released with support for sculpties, but what about a year later? Or right now for that matter, I believe you can still log on with 1.15, don't see too many people doing it tho. Yes I know there are v1 viewers that can see meshes, but since they ripped out the whole render pipeline (among other things) and replaced it with the one from v3 I don't think you can really call them v1 anymore, more like a v1 & 1/2.

It was just added in 3.1 I believe. STORM-1612

The official documentation refers to is as an alpha mask. But nobody reads the manual so...