Since migrating into the Second Life wiki, the Second Life viewer directory has been resorted rather interestingly. By crash rate.
Second Life viewers (both first party and third-party) send crash statistics to the Lab (along with other statistics). This most recent incarnation of the viewer directory lists viewers from the least-crash-prone to the most.
Interestingly, Linden Lab places all of the official viewers above all of the third-party viewers, with the assertion that they crash less, which doesn’t really match up with my experience, but does list the first-party 1.23 viewer right at the top (which very nearly does).
I don’t know how often the ordering of this list might be updated.
Found this interesting? Give it a bump!
I could believe it. 2.5+ hasn’t really crashed for me. Imprudence doesn’t much either.
The same question occurred to me also, since you’d expect variation in different versions.
Sorry I edited that question out after reading: “The rate used for each viewer is that of the highest version number with at least 200,000 minutes of use in the sample period (generally, a one week sample from Monday to Sunday).”
Originally asked: How’re they deciding which version(s) of TPVs to sample from and then sort?
Still worth questioning considering nightlys, experimentals, betas and so on.
And we’re supposed to believe LL’s word without any (detailed) evidence. Um, Right.
@Net Antwerp
I’m not quite sure how (for example) you or I would benefit from seeing the raw data. I’m open to suggestions, though.
As much as I hate it, the current official Linden viewer 2 is the most stable and best performing of all viewers, at least for me. However, I find myself using Phoenix 90% of the time simply because I am addicted to the very long range radar and command line utilities. Why can’t someone put my chocolate and peanut butter together?
Firestorm could one day be that Reeses Cup, but it has a long way to go.
Ascent works for me… has the chocolate and P nut butter thing going… texture uploads seem as fast as any.. and faster than some. The only crashes I experience are from doing dumb things like attempting to upload a texture at a Welcome Area full of avatars.
Ascent is likely not rated cause so few of us use it.
@Tateru Nino:
Does it hurt Linden Research in any way by publishing anonymous crash statistics?
Revealing crash statistics as an optional download/link benefits existing and potential customers, as they can verify the data if they want to, and even possibly give Linden Research some feedback about the collected data.
Releasing crash statistics is a win-win for existing/potential customers and Linden Research.
The crash rate became really good (lower
) during the past 3 years.
I remember when we had to be careful – just pressing a wrong button could crash the viewer.
Nowdays the only ‘unexpected’ exits I see are the disconnects on TP (does it also count as crash ?).
LL’s assessment is not reflective of my and my husband’s experiences. I use Phoenix (the version before 908); my husband uses SL viewer 2.5. He crashes far more frequently than I do. And his crashes often cause his computer to blue screen.
The Cool VL Viewer is probably the stablest viewer around: it probably contains the highest number of *actual* crash fixes (including fixes to crashes that still affect v2 viewers). The reason is simple: I use it every day, and should I find any crash in it (which becomes rarer and rarer over time), I *immediately* fix it…
By “actual” fixes, I mean fixes that when not applied, *do* lead to crashes (there are also other *potential* crash fixes, which all TPVs address as well).
Since the Cool VL Viewer is not in the directory (for privacy concerns, since LL requires your snail mail address when you want to register a viewer !!!), it is of course not listed. You could use Dolphin v1 as an approximation (since Dolphin v1 is nothing else than the Cool VL Viewer v1.25 rebranded and with a few minor patches added), but apparently (and while Dolphin was classed first in the beginning of this reliability list), LL decided that they didn’t have “enough” data to class it any more… Is LL afraid of teh fact “old UI” viewers are best ?…
A few things that annoy me:
Why isn’t Dolphin in the first tier TPV list yet? It’s a solid viewer.
Why is Emergence still taking up space in the TPV list? LGG said it was a one-shot, part of the Emerald political strife. It will never be updated or anything.
As far as the official LL 2.x viewers being the most stable, I think they might have reported the least crashes so far because they are the newest. Note how they are counting 2.6.1 and 2.5.2 seperately, but they aren’t counting each different version of the TPVs. That’s why. 2.6.1 has, what, a half dozen users, who only reported 20 or 30 crashes last week, so overall, of course 2.6.1 has less crashes than the rest of these viewers.
In the end, this is just another example of LL cooking the books. They don’t care if we believe them or not when they lie, and they don’t care about our good will either. They want to go bankrupt and god damn it, come hell or high water they’ll do it!
Is this rate calculated PER 1000 users/hours or, is it total crashes….. if twice as many people use Phoenix…etc.
Also…no difference between Mac and PC?
I use Imprudence and 1.23 and really never crash except when SL is playing up.
I find my personal crash rate has more to do with other variables than what viewer I’m using. Just a guess, but I would suspect that users who are socializing and shopping crash less often than those rezzing out a coalesced object of thousands of prims or having a gunfight or flying a plane at high speed. And I also suspect that those using more-advanced third-party viewers are more likely to be doing these crashier things.
Forgive the negativity of this post; viewers are a sore point with me.
What we see here is yet further evidence that Linden Lab at this point has a zero “trust” rate among users. We don’t trust what comes out of corporate, certainly don’t trust their claims, PR or promises, and propaganda is obvious.
It interests me that Imprudence is at the top of the list. While one of my favorite viewers design-wise, my experience with that viewer was continual crashing, 4-5 times a day or more. I tried it on a fresh install of XP Pro, tried totally uninstalling and re-installing, even contacted Imp themselves. They asked for crash logs, I sent crash logs. I’m not at all anti-Imp; I kinda like the people and concept. It’s just crashy.
In contrast, I do not crash with SL 1.23 or Hippo; I can work all day on Hippo 6.3 and not crash a single time. That indicates it is not my system, nor experience unique to me; evidence points to severe instability in the Imprudence viewer.
I’m not sure there is a “good” viewer out here at this time. All of them were based on Linden Lab viewers to start with, which themselves are such a collection of extremely poor and mangled code it’s like trying to build a house on top of a rotten foundation. I don’t have a solution for that– other than I use the viewers that work best for me.
And that’s the trick; people should use what works best in their case… because all of these viewers seem inconsistent and flaky in performance. Experience seems to differ greatly from one user to another. I am very cautious of security issues– especially from people who have breached faith and trust before. It takes little or no effort for a 3rd party viewer company to hack the password sequence. So I am extremely cautious whose viewer I use. As I said, right now I’m using Hippo. It’s stable *for me*. What other folks use depends on who they trust… and what works best in their experience.
One thing I do know for a certainty: these days I trust no claims that come out of LL corporate. No PR, no promises, no charts… NONE. Consequences for actions. Trust is not warranted. Not being snide; just a statement of fact. When someone punches me in the face several times, I get the hint he’s not exactly working in my best interest. ; )
So that viewer page means nothing. I’ll trust independent posts like Tat’s… but nothing coming out of corporate.
I’d love to use Kirsten’s client, but it crashes on me in a way that I’ve seen on other viewers, but which Kirsten’s now does for me VERY regularly: after a short interval, it allocates all available RAM, the frame rate plummets to around 1-1.5 fps, and then the client crashes. (I should mention that I use Ubuntu 10.10, 64-bit version.)
Once Imprudence moves to a 2.x base, I may very well switch back, because it alone has a 64-bit version, and thus it’s the only SL client that lets me hear streaming audio. (Come on, everyone else, bite the bullet–do OEMs even make 32-bit x86 flavored systems any more save for netbooks you wouldn’t even consider running SL on anyway?)
Right now, I’m very happy with the Firestorm preview #2.
The big problem with 64-bit Viewers is that Linden Lab aren’t doing anything. All the third-party viewers are based on the code that Linden Lab has released. There are tools which allow 32-bit programs to benefit from 64-bit systems, such things as more memory available, but there are so many complications.
We’re getting that sort of lightweight dusting of 64-bitness, and there’s stuff that’s only a compiler-switch away, but a true 64-bit program is a really big job.
The page has been updated with some more information about how the crash-rates are calculated.
The crash-rate standings are completely bogus. The official 2.5 viewer will always crash on me within an hour so I rarely use it. Imprudence was crashy too so I dont use that one anymore. I use Singularity — the rez and upload/download rates blows away anything else I’ve used and I haven’t crashed a single time since I’ve been using it — about a month now.
I think crash rate with any viewer is hardware specific and unique to the user. The fact that I crash constantly using Phoenix V1 doesn’t have a lot of bearing on someone else crashing with it, or not crashing … unless we have identical machines. Also, it doesn’t surprise me that LL viewers are the most stable for the broadest segment of the SL population since LL has by far the most people working to bug-fix their viewer across the broadest array of hardware. Personally, I saw the writing on the wall months ago and switched to V2 exclusively because as XMPP chat, mesh, avatar physics and other features roll out to the platform a V1 viewer just won’t cut it. For now I use Snowstorm (latest build) and KL’s viewer. Almost, but not completely, crash free.
Many people I know use Phoenix or Firestorm for their support of RLV. These people use tons of huds and scripts and animation overrides to enhance their kinky roleplays. I could imagine that it really pushes the viewer to its limits often. What I mean is that we must take into account what users do with their viewers; this could bias the statistics, making Phoenix appear to crash more often than Viewer2. Dunno, just an idea ! And I’m not blaming the kinky roleplayers :p.
Actually, HUDs and scripts are pretty much server-side entities. They don’t really affect the viewer very much. As for RLV/RLVa, that’s common to most third-party viewers.
Sorry to necro an older article but I’d like to point something out: The users here that have stated that crash rates are machine specific are quite correct. The various crashes I have had while testing out other clients (in addition to the official client) do not seem to appear nearly as much with my client of choice – Firestorm.
Each of them shares a single, broad band crash: Certain textures and load types silent crash the client. Mouse look overloads the GPU buffer after a while as well, across the board.