sloggy fps

hey, i have just installed another gig of ram and got a second graphics card for an SLI setup, but for some reason i am only getting around 50fps in Midian City (viewer side laggiest sim i know) and can't figure out why, i should be getting a good 120 - 180fps, 200 in an empty sim

Hm, 50 fps seems like a very good rate. Why are you so adamant about getting 100fps or more? The human eye can't distinguish over 30 fps anyways. I think you should be glad to get 50, most of us are battling daily with rates in the single digits...

u can tell the difference between 30 and 50. 100 looks allot damn smoother,

how can you play second life (infact how can u play anything) running slower than 30fps? it would drive me up the wall

a sim at its bestest will push out 45 fps and much more than that on the client side is what my imported friends call wankery.

I often find myself spending the money that would otherwise be poured into system upgrades on other things. For instance, single malt scotch. It's quite relaxing and nicely compensates for a fairly low frame rate.

From the Knowledge Base:



Even the best of those (nVidia GeoForce 8800) doesn't quite hit 50fps.

Sim FPS is how fast the server is updating the area and and FPS is how fast your PC is drawing the area. These are two very different things.


But.. but.. but.. Freecell runs so great! Why can't SL?!?!1?

*raises glass to Snark* I'm with you.

Wow, Chaos, nice rig. Be sure you have some single-digit frame rate folks test your new sim before you go live with it. The number of people using SL on laptops is really very large.
.

lol @ the S3 and Rage128 on that chart, who managed 2 get SL to even run on that,lol

it cost me a few months worth of wages,lol

heres the specs

CPU: AMD Athlon 64 Quad FX-74 3207 MHz (thats 2x3200, 4 processors)
Memory: 6144 MB (DDR3)
OS Version: Microsoft Windows Vista Service Pack 1 (Build 6001)
Graphics Card Vendor: NVIDIA Corporation
Graphics Card: 2xGeForce 8800 Ultra 768Mb (running in SLI mode)
OpenGL Version: 2.1.2
HD: SATA 2x320GB (Running a raid) and 1x80GB (for backup and cache drive)
32" 4ms Fujitsu Siemens TFT purpose build PC screen (not a TV!)

it is the only expensive thing i own, well my PC and my soundsystem i use for raves

Do guys get "computer envy"?

A client side FPS much higher than the server side will be little more than tweening, there is no additional information to display. That is why it is not really useful.

While I'm not arguing that Chaos is making any sense, far from it, I'm not sure that I agree with you.

If your viewer can do, say, 30 fps and the sim drops down to 10 sim fps, you're going to want your viewer to keep redrawing at the faster rate. Even if objects start rubber-banding, and they will at 10 sim fps, it's still going to look at lot better at 30 fps.

Yeah, they do. And the opposite would be a form of penis enlargement.

Yeah 10 fps is an interesting number because that is right in the range where persistence of vision stops working, so you almost have to have some kind of tweening to keep the illusion of motion going.

We can kind of put flicker off to the side for this since it's mostly tied to other hardware in this kind of environment, and just look at frames.

30 is right in the sweet spot for making things look like fluid motion, even a little higher than what movies used forever. Some newer formats run right in the range where SL servers max out, they use 48 and that is around where "motion picture" characteristics like strobing and backwards wheels mostly disappear, and people start to complain and get uncomfortable because the motion looks "too real". By 70 it is about as fluid as people can perceive, though flicker which we're ignoring for now can be detected pretty far past that.

For a CRT, nothing will actually display faster than the monitor's refresh rate. For an LCD, normally nothing will go faster than 60FPS and it will look like it has motion blur anywayz. Everything else you see above that is just the placebo effect. (^_^)

I bet you love those Dual FX74s though... I know I do. (^_^)y Now, If only I didn't skimp on the video cards.... Meh.... I'm not a gamer anywayz. (=_=)

Lots of stuff to talk about here. No particular order.




That's simply not true, Calveen. Not in any way, shape, or form. The only reason the number 30 has any significance at all with regard to FPS is because that happens to be the speed TV and video operates at in the US. It's a completely arbitrary number, more an artifact of our electrical system than anything else, and has nothing whatsoever to do with how many frames per second the eye can detect.

The eye does not operate by frames anyway. There can be no such thing as a maximum number of frames per second the eye can detect. You might as well ask how many drops per second flow through a water pipe. The question itself can't be applied because pipes don't deal in drops; they deal in continuous water flow. Likewise, the eye deals in the continuous flow of light. The concept of frames has nothing to do with how we see.

The question of how many FPS it takes to make a given scene look fluid is more answerable, but again, it has nothing to do with how our eyes work. There are many factors that go into fooling our brains into believing a series of still images is actually a moving picture. One of the major ones is motion blur. TV, video, and movies can operate at relatively low frame rates, and still create the illusion of fluid motion, because each moving object in each frame is blurred. Take away the blur, and you'd notice right away that the motion looks choppy.

This is why video games require much higher FPS to look fluid. Realtime rendering engines are generally not sophisticated enough to incorporate motion blur. Most games at 25 or 30 FPS (the speed of movie film and video, respectively) look very stop-and-go. It takes a lot more FPS for a game to seem more real.

For a good article on this subject, go to http://www.100fps.com/how_many_frames_can_humans_see.htm




As I said above, it's not just the frames, but what's in them. A fog simulation, or a water simulation, can look perfectly good at 20, 10, or even 5 FPS. If there's not much discernible difference from frame to frame, it doesn't really matter how fast they change.




Once again, it's what you're drawing, not just how fast you're drawing it. There can be no concrete assignment of "lower than 10 and it stops working" or "higher than 70 can't be perceived". That's just not how our eyes work.



My system is similar to yours (not identical, but very similar), and I'll get anywhere from 25-175 FPS, depending on what's going on. Double digits are far more common than triples. That's at 512 draw, all settings maxed, and with 16XQ anti-aliasing. I've gotten it into the 200's in empty sims before, while in SLI mode, but since I run dual monitors, I don't usually use SLI.

If you honestly expected to see a constant FPS of 120-180, you were being very unrealistic in your thinking. You'll get peaks of that, sure, but you won't get anywhere near that steadily.

If you really want to grab every last frame you can, there are a few other factors to consider:

What is the size of your monitor? Larger ones will cause you to run more slowly than smaller ones.

Does your particular system work better with SL in full screen mode, or in a window? This tends to vary from machine to machine. I've never figured out any rhyme or reason to it.

How fast are your hard drives? You said they were SATA II, but you didn't quote the RPM's. It makes a difference to how fast you can access your cache.

Is your OS 32-bit or 64-bit? Vista comes in both flavors. The reason I ask is you mentioned you have 6GB of RAM. Unless you're running a 64-bit OS, most of that is unusable. All 32-bit operating systems have a hard limit of 4GB memory (including video memory). With those two 8800's in there, you've got 1.5GB of that locked in video. That leaves just 2.5GB available for system RAM.

Since you're running Vista, have you given it time to "settle" yet? Vista is slow as molasses for the first few weeks. Somehow it monitors your behavior in the beginning to learn how you use your computer, and then it gets faster based on its observations. I don't understand what it's actually doing, but I can tell you from experience that those first few weeks absolutely suck.

Have you disabled all the Vista crap that slows everything down? There's a whole giant laundry list here. Google "make vista run faster" and you'll find tons of tips.


I could go on and on, but the only really important question is can you max your settings, apply a good amount of anti-aliasing, and still get a "good" frame rate most of the time. By the sound of it, you are. 50 FPS in a built up area is plenty. I have no idea what this Midian City is you mentioned, but if it's really the laggiest sim you've ever been to, and you're still getting 50 FPS there, I'd say your machine is doing great.




Tweening it may be, but that doesn't mean it's not useful. The higher the FPS you get (in a non-motion-blurred environment), the smoother the whole experience will feel, period. It doesn't matter if the client is running faster than the simulator or not. It doesn't matter if the tweening is "real" or not. What matters is how things look and feel on your end.




I have issues with those statistics. For one thing, they're very incomplete, as they don't tell you anything other than the video card series. Within the 8800 series alone, there are no less than six different major models (not counting modifications from nVidia's various partners). I would imagine that most people who buy and "8800" are buying the less expensive versions. How many actually have the Ultra, like Chaos has?

If someone has an 8800GS, they've only got 1/2 the memory, roughly 1/3 of the bandwidth, about 1/4 of the texture fill rate, and slower clock speeds across the board. Surely those factors would likely slow the SL frame rate down considerably.

It would be nice if a better chart were available, showing a more detailed breakdown by specific card model, not just by series.

It also doesn't say anything about settings. Someone running an old 7800 at 64 draw and minimal options will likely a higher frame rate than someone with an 8800 at 512 draw, and maxed settings. That doesn't mean the 7800 is outperforming the 8800 though, obviously.

The chart needs a lot more information in it before it can be in any way definitive.

I never said 'definitive.' I was simply reporting what LL claims we should see on average.

Oh, I wasn't trying to shoot the messenger, Ann. Sorry if it seemed like I was picking on you personally. That wasn't my intention. I was just pointing out that I think that chart needs some work. I realize it's LL who needs to do it, not you.

It would probably happen faster and more efficiently if I did it.

It's realy quite disjointed for the purposes of practical discussion, alternating between a recognition that a video display can't display real motion and then seemingly forgetting that, off to other interesting but irrelevant effects that can't be enhanced by a higher frame rate because they are simply not there to display. It touches on where these are aggregated but then fails to connect the dots. All in all quite disappointing.

It seems to me that the basic problem here is the disconnect between expectations and the reality of SL. And I'm not saying it's foolish to have high expectations; after all, what SL resembles technology-wise more than anything else is a PC game, and historically anything other than the most cutting-edge games have offered framerates in the high double-digits to low triple digits on suitably powerful hardware. There are reasons why SL should not be expected to behave like a PC action game, but none of those reasons are apparent on the surface. It must be a disheartening experience to purchase some of the best, fastest hardware available just to run SL in the 40s and 50s if you're accustomed to the way PC game performance scales with shiny new hardware.

I have to disagree, Incoherendt. The article fulfills its purpose well enough. I'm sorry if you were expecting some sort of in-depth scientific study on the subject of frame rates or something, but that was not what I indicated the article would be. It merely serves to offer some rudimentary explanations for why the very question of how many frames per second is "enough" or "not enough" is not a valid one.

I'm not sure what you're defining as "interesting but irrelevant" either. Just to summarize quickly, the article only hits on a handful of points, all of which are relevant. The first point was about motion blur, which is certainly relevant. The next point, the concept that "brightness eats darkness" is also very important in any discussion of how we perceive framed motion. One example of this that was cited was the fact that people can identify in full detail an image on a photograph flickered in just 1/300 of a second. I would think you'd find that one to be VERY relevant, since you implied earlier that you believed people couldn't detect anything faster than 1/70 of a second. The next point was about afterimages, again relevant. The final list of points was a bit more abstracted, but it was entirely on topic. It hit on some factors we don't normally think about that can affect how we perceive frames, like body position, physical movement, the colors of the imagery in question, synchronicity, etc., as well as some factors we don't fully understand, like the various roles the cells within the eye play to create what we call vision. So what out of all that do you really think was "irrelevant"?

Admittedly, all of these points are presented in a very oversimplified manner, but none of them are off topic. They're all very relevant.

As for what you think is "not there to display", are you sure you read the same article I did? With the exception of motion blur, the potential absence of which was mentioned precisely to illustrate why games need higher FPS than movies, every other thing the article mentions is there. Brightness is there, darkenss is there, the colors green and blue are there, and occasionally even images of airplanes are there.

The only thing a little off topic was the author's personal theory on a possible reason for why some people don't like CD's. That took up all of three sentences. OK, so there were three irrelevant sentences out of the whole page. Terribly disappointing, got it.

pardon if i'm skipping the whole thread (i am - recent suggestion was to set all local vendors to 'phantom' to help save local reztime space. this seems to subtly change the local space to something more friendly...

Chaos,

It's a bit of a shame that you don't have more "average" hardware, because if you did, when you rebuild Felonhall, you'd be more conscious of what an "average" visitor is likely to to experience.

Median city is laughable with my hardware, presumably due to there being way too many textures and those textures possibly being too large?