Getting less than 1 fps in crowded areas

My system is a p4 2000 mhz, 1024 pc 2700 ram, and a geforce 6600gt, and it's running slower than oblivion at the HIGHEST settings at highly crowded areas.. I was trying to win a few dollars at a casino and the lag somehow managed to make me apply the 520$ bet 2-3 times consequitively, loosing what was left of my 1400$ balance. But there's nothing I can do about that now..

I thought i had a low-end-of-the-high-end machine, is the game supposed to run like this ??

First... Sl is not a game.

Yes with avatars around it can be painstaking slow arsed, SL download/upload alot , rendering mostly cpu based, why its so slow.

Texture sizes / amount far beat oblivion's , i snooped in their textures and it barly had any 1024x1024 sizes, wich is aaallll over SL -skins usually.

Your pc vs SL is > low , this thing need to use more gpu to do more stuff as its right now it does most on cpu wich can stress some pc's to death.

Games on the other hand do not render what they cannot see, SL tries to do that but its pretty crud at it.

(and yes something like this definatly should be stickied cause this question get around alot in forum)

Avoid casino's they are a bunch of ripoffs imo , theres no controll over who makes it or if they actually pay out anything.

So you telling me that even with a 2 GHZ machine you can expect a completely unplayable 1 FPS expieriance? And updating the graphics card won't help? I just got a brand new machine and it runs SL like horrible slow. I'm really dissapointed, I mean if the NEWEST computers can't even run SL then what's the point?

Could be other things too, what is your draw distance at? Do you have local lighting on? SL at the moment is more CPU bound than GPU, that is a week point at the moment and your settings can drastically affect your performance. Intel P4 2ghz is a 5-year-old chip, came out in mid 2001.

The best you can do is reduce your draw distance (12 and drop your detail levels a little (a notch above the middle is more than fine), and make sure you have local lighting turned off (at least till 1.9.1 is released).

SL is CPU intensive and as such doesn't really tax your graphics card (lots of texture memory is the most import aspect), but that should change starting with 1.9.1 - hardware lighting and occlusion culling - check out the release notes ( scroll to bottom)

Yes.

There is no stated intention (that I know of) to deal with that issue at present.

Instead, LL is hedging all its bets on the next major version based on .NET or Mono (IIRC), or so we were told almost a year ago. Allegedly the new engine was showing speedup of 100 X on scripting and rendering in the lab, so I guess that they are hoping that this will translate into large improvements in the real world too.

So, your frustration is probably best channeled into asking whether that redesign is anywhere on the horizon.

Sadly, all the second lifers in my local circle have channeled their frustration into leaving SL for other games/worlds/systems that don't run like dogs. Either they have no stamina, or we're just suckers for punishment.

That is an utterly shortsighted statement.

First of all, the performance that different people get on different computers varies wildly, and often does not seem to be even remotely related to their computer's speed, power, amount of RAM, video card, or what have you.

I personally have an AMD Athlon XP 2700+ runnung at 2.1GHz, 1.5 GB of RAM, and an ATI Radeon 9600 video card. ATI cards, as we have all heard, are not supposed to play nice with SL. Yet I seem to get much better performance than some people report with Pentium 4's running at 3 GHz or more, and NVidia GEForce 6600's or even higher.

Why is this? The only thing I can guess is that there is some sort of bottleneck (probably more than one) that occurs frequently, which brings even the most powerful computers to their knees. These might be in the SL preference settings (Local Lighting is one obvious, likely suspect), or in the Windows configuration, or in third party software running in the background (anti-virus utilities and such). Who knows?

The only thing I can say is not to simply say, "Well, obviously Second Life is crap," because if that were true, nobody would be getting decent performance, and that's just not the case.

P2

That's fair enough Phoenix, but the problem is that nobody seems to go beyond their "It works fine" / "It works terribly" pronouncements to discover something concrete about why so many people find it very slow compared to other 3D systems. An "I'm alright Jack" response is even less helpful than a "There may be good news in the future" one.

What we need to do is to get you and Joel to stand at the same spot in the world, set your clip distances and all other performance-affecting settings identically, and then screenshot your FPS display screens. Discussions could then be slightly more objective. Perhaps Joel's "unacceptable 1 fps" and your relative satisfaction are not comparing like with like.

At least we know that LL themselves consider current performance to be slow, since that's a key driving force behind their next generation work. And one can't argue with their own perception.

Actually, that is an EXCELLENT idea. A test group, made up of members with computers of various levels of "fastness" (for lack of a more comprehensive term), reporting on their results at a given location in SL at the same time would be a great idea.

P2

I would be interested in this also, send me an invite to a group and ill be there!
7800GTX 256MB OC
2 gig 2-3-2-5 1T ram
74GB 10k RPM WD Raptor
Soundblaster X-Fi XtremeMusic
AMD 4400+ X2 OC to 4800+

Ditto. I could assess 3 radically different systems:

1) Dell Desktop / 2.8GHz P4 + 512MB / nVidia GeForce FX5200 PCI + 256MB / WinXP

2) Toshiba Tecra M4 Laptop / 1.73GHz Pentium M + 512MB / nVidia GeForce Go 6200 PCIe + 64MB / WinXP

3) 64-bit Linux Workstation / 2.0GHz Athlon 64 3200+ + 2GB / nVidia GeForce 6600GT PCIe + 128MB / Gentoo amd64

Funnily enough, there's not a huge difference between them in terms of SL frame rate, running over 8Mbps ADSL or 10Mbps cable links, with all gigabit Ethernet internally.

An accurate test would be useful.

You know, this starement just struck me with an idea that I am not sure has ever been addressed in any of the hundreds of threads about poor performance.

There are various types of ethernet cards; 10BASE-T, 100BASE-T (so-called "Fast Ethernet", Gigabit, and 10-Gigabit being the ones that I know of. These vary in speed from 10 Mbps for the 10BASE-T to 10 billion bps for the 10-Gigabit protocol.

Could part of the variation in performance be tied, at least in part, to people having slower network cards? For example, my Netgear card is 100BASE-T. Could it be that at least some of the people having issues of this nature may still be using older 10BASE-T cards?

Just a thought.

P2

You would also need to run some other benchmarks for an independant guide to the relative performance of each of the test machines, this in itself may help to shed light on links between specific areas of system performance and SL performance.

Otherwise you could have a little get together (which I am up for) and come away more confused than when you started

Not really, not in any significant way.

Latency and bandwidth are two separate issues, and each of them has LAN and WAN component parts, and client and server program parts, and client and server hardware parts. The LAN card is just a small piece of the puzzle.

The bandwidth restriction that an old 10Mbps card or hub would impose on traffic would not be noticeable at all unless the WAN link to your ISP itself runs faster than 10Mbs, which is still rather rare except on corporate or university links. 99% of Ethernet equipment uses 100-Tx these days, and the higher end is rapidly moving to gigabit, so no modern Ethernet card or hub or switch or the Ethernet interface of a router is ever likely to throttle WAN traffic on the basis of raw bandwidth.

Latency is a different matter, and Ethernet interfaces will always add latency in reverse proportion to their clocked speed, so faster is better. However, does it matter? Usually, no, because the latency that even a sluggish 10base-T card introduces is a small fraction of a millisecond, whereas overall network connection latencies over WANs are always of the order of several milliseconds at least, and more often than not many dozens of milliseconds, and in some cases hundreds. So, the tiny amount of latency added by even a slow Ethernet NIC is totally insignificant in reality.

Network latency cannot be divorced from the issue of round-trip times (RTTs) though, and this is where it gets complicated, because the RTTs that one commonly sees quoted as the result of pings and traceroutes and other generic tools do not really have much in common with real application RTTs. The problem is that application RTTs are determined very strongly by the delays incurred in processing the game-specific content of packets within the software running at both the client and the server ends, and this is very often determined primarily by the design of the relevant items of software.

The only time when generic networking tools show up bad latency is when there is something anomalous happening out on the Internet (short-term spikes), or when your ISP or the server end is being attacked and massively congested (that can last hours or days), or when you have a local fault and are dropping packets. These things are easy to identify and never persist for long.

In contrast, the most common cause of persistent game lag bar none is high response latency from congested client or server software or both, as a result of non-scalable design. In a nutshell, designing software systems for high scalability tends to be rather rare, right across the online industry. It is almost unheard of to design software primarily on the basis of returning a response within a specified time window, rather than "when the server gets around to it". Not surprisingly then, almost all online system go into bad lag states as soon as their user populations increase significantly. SL is no exception.

In the present case though, LL know that they have a huge designed-in latency problem at the client end, owing to the rendering and communications being done on the same thread. They're working on it, we're told. For the new engine, that is, not the current one.

We haven't heard anything specific about latency problems *by design* in the software at the server end, but we know that there is a practical latency problem inherent in their hardware design, since sims have a static allocation of computers available to them in SL's statically tiled grid. In other words, if more than a certain amount of local activity occurs on a sim, or if too many visitors arrive, then severe congestion is unavoidable, by design.

Of course, this inherently implies dreadful non-scalability within each sim as the world population rises, unless "everybody stays at home" which would be patently ridiculous. Philip knows about it, but has no solution other than saying that they're "researching the problem". It's not particularly clever to do your researching this late in the day though. The issues are pretty well known anyway, to anyone who's been working in the high parallelism industries.

So you see, the issue goes significantly beyond what speed your NIC runs at.

Yes, on another thread someone was complaining that people 'whine' about their performance, but provide no details about their setup. So I posted virtually every bit of information I could find about my computer. The response to that was that someone else came along and requested even more details (some of which I had no idea how to obtain). I reported what I could, and asked how I could obtain some of the more arcane ones.

No response at all from anybody.

im running SL on a 2MB DSL connection with a 64bit 4200+ dual core AMD with 1GB RAM and 300GB HD and a 6800GT card.

I was having lag issues so i turned everying to lowish and things seemed fine but since last update for whatever reason Ive turn on high quality water and shiny objects and LAG has vertually gone regardless of where I am.

I think part of it is to do with the SIM your on, what server runs that sim and the number of people drawing off that server I think your actual DLS speed has just as much impact as your pc.

I'm all up for the SL 2.0 update.

Also worth considering; not all ethernet cards of the same type work as well as others. 2 different 100BASE-T cards of different quality will give different transfer rates on the same computer. Another thing, is a 10Mbps cable connection really 10Mbps? Most likely not. It depends a lot on how many people are on your cable network. Same thing with ADSL, except the bottleneck is at the switching station.

my new comp is a amd athlon 64 bit 2gbs of ram and the 7900 gtx and my shitty 1.8 celron with 512mb and 9250 radeon played better i am way dissapointed

Someting is amiss with your setup - My amd64 with 1gb ram and a 6600 flys.