Did My own SL Benchmarking

Got bored the other day & decided to benchmark all the computers I could get a hold of (friends, family, fiance, mine). I set the graphic settings to the lower or middle setting since we already know turning on all the bells N whistles drops the FPS.

Anywho... The Settings & Computers:

Desktop A AMD Athlon64 3000+ (single core @ 1.8 GHz)
2 GB PC2700 RAM (333 Mhz)
XFX GeForce 6800 w/ 128 MB RAM

Desktop B AMD Athlon XP 2200 (2.2 Ghz)
512 MB PC2100 RAM (266 Mhz)
ATI Radeon 9200 w/ 128 MB RAM

Desktop C AMD Duron 1.3 Ghz
512 MB PC2100 RAM (266 Mhz)
MSI GeForce 4 TI 4800 w/ 128 MB RAM

Laptop Dell Inspiron 1150
Intel Celeron @ 2.6 GHz
512 MB PC2700 RAM (64 MB Shared for video)
Intel Extreme Graphics 2
++++++++++++++++++++++++++++++++++++++++++++++++++++

The Settings:
Network- Max Bandwidth & Disk Cache set to 500
Graphics- "Performance" settings all OFF
Avatar Rendering- Normal
Terrain Detail- some
Object/Tree/Avatar Detail- All 3 sliders set to the CENTER position
draw distance- 128
Fog Distance- 2
Drop Draw Distance- 10
Bumpiness- 10
Max Particles-256
Outfit Composite Limit- 5

Each system was tested with the same Avatar and at the same spot (Gibson 147,40,39)

Results:

Desktop A= 59-61 FPS

Desktop B= 18.5-20 FPS

Desktop C= 27-30 FPS

Laptop = 6.4-7.7 FPS (9.5 FPS with clouds & grass turned off)

So what's this prove? I dunno! Newer is better? Older computers need settings turned down? Just was really bored & thought I'd see what would happen. Not that any of the results are surprising. Next I'm gonna try the 64-bit system with settings turned on. I wouldnt even bother trying it on the laptop.

Why did I chose that particular spot (it's the platform on a post)? Because Gibson was on the Mainland, empty, and has tons of geometry.

At least now if someone is looking to build/buy a better computer for SL, maybe this'll be some sort of help. If anyone with a Mac or the Linux version would like to try the same settings & post their results, that'd be sweet too!

I'll post up my own bechmarking scores too later. You think you can do that again, on the preview grid?

Ooo! I wish! Only 2 of those were mine. My fiance hates SL (she'd rather argue with emo kids on blogs and chat with her friends on AIM ..... yet I still love her! )

O.O I wan't computer A!

Even more results with Desktop A and Graphics Settings:

With ALL Graphic Settings ON- 0.4 FPS Couldnt even TP anywhere!!!

With Local Lighting OFF- 13-16 FPS (18-19 with Clouds OFF)

With Local Lighting and Disable Far-Clip OFF- 15-16 FPS

With Local Lighting, Disable Far-Clip, Shadows OFF- 18 FPS

With Local Lighting, Disable Far-Clip, Shadows, Anistropic Filtering OFF- 18.8- 19.3 FPS

All results were with Avatar Render set to Bump Mapped & Cloth and Terrain Detail set to Full and Ripple Water ON.

(Gibson 147,40,39)


The Settings:
Network- Max Bandwidth & Disk Cache set to 500
Graphics- "Performance" settings all OFF
Avatar Rendering- Normal
Terrain Detail- some
Object/Tree/Avatar Detail- All 3 sliders set to the CENTER position
draw distance- 128
Fog Distance- 2
Drop Draw Distance- 10
Bumpiness- 10
Max Particles-256
Outfit Composite Limit- 5


My specs:
CPU: 0.13 micron Intel Pentium 4 (2194 Mhz)
Memory: 1024 MB
OS Version: Microsoft Windows XP Service Pack 2 (Build 2600)
OpenGL Vendor: NVIDIA Corporation
OpenGL Renderer: GeForce FX 5500/PCI/SSE2
OpenGL Version: 2.0.1



FPS - 9.2

HEY! Xander! I think I saw ya at Gibson when I did my last benchmark.

you list an Athlon XP 2200 running at 2.2ghz. Do you have that overclocked? if its a 2200+ the most common versoin with that model number ran at 1.8ghz as far as i remember

Oops! yer right! I forgot about the old naming conventions- Athlon XP 2200 supposedly working as fast as an Intel at 2.2 Ghz.

And speaking of Benchmarks, I tried the preview grid. Gibson was still there so I gave it a shot.

On the "Desktop A" system the best I achieved was 30-38 FPS. Thats standing still. If you walk around it drops to about 10-15 fps.

Thats with Shaders ON
Object/Avatar/Environment Details set on Medium
Lighting: All Local & Shadows
Terrain: High
Object/Flexible/Tree/Avatar mesh: Middle setting
Anistropic: Off

Didnt get to try it on the others. And forget about it on the laptop

After reading this thread I had to try the test too! Using the same location and graphics settings, I had the same FPS as Desktop B.

2.4 gig Pentium 4
1GB RAM
Radeon X1600 Pro 256MB

Considering my PC is better than Desktop B in all of those areas, I'd have expected a better framerate, even if it was only a few FPS. Maybe I need to do some tweaking.

Could it be related to the ATI video card? If you notice, Desktop B & C = B has an ATI 9200 while C has an old GeForce 4 and a slower processor yet C got better framerates. I know the 9200 is a bit newer than the GeForce 4 so even I expected better results.

(Gibson 147,40,39)

The Settings:
Network-
Max Bandwidth & Disk Cache set to 1000
Graphics-
Anistropic Filtering: ON
Local Lighting: OFF
Shadows: ON
Ripple Water: ON
Disable Far Clip: ON
Shiny Objects: ON
Avatar Rendering: Bump Mapped and Cloth
Terrain Detail: Full
Object/Tree/Avatar Detail- All 3 sliders set to furthest right
draw distance- 128
Fog Distance- 4.0
Drop Draw Distance- 0
Bumpiness- 10
Max Particles- 8192
Outfit Composite Limit- 5

Other extra settings:
Local Tree Lighting: ON
Clouds and everything else: ON



My specs:
AMD 4400+X2 OC to 4800+
MSI 7800GTX 256MB OC
2GB OCZ Platnium Rev 2 Ram 2-3-2-5 1T Timings
74GB 10K RPM WD Raptor

FPS: 67-69

Could be interesting if this could be scripted and the result uploaded to a database, like 3DMark...

After a while you could see the bottlenecks on your system (hey, he got 10 more FPS on the same system, only he has 2 Gb and I have only 512 Mb) or as a guide for when you are going to buy a new PC.

But it also would need a very stable sim, where only 1 person at a time can enter. Else there would bee too many things influencing your frame rate. Maybe LL has an old server that could be used only for this purpose. Just make sure there's always the same content (some buildings, some scripts, some trees, a bit of rippling water, always the same time of day and when 1.9.1 is being released, some local lights). It doesnt matter if it's an old server, just make sure that the settings are always the same.

Teleport there, run a bench of different tests, upload them to the database, and teleport back.

And this testing done with same avatar, aka blank or heavy arse primmed one ? , got one that pretty much kills client rendering vs one that has zero attachments, huge diffrence.

Mouselook, 3rd person?

THe script could force a shape and outfit (and maybe some attachements and bling) on the avatar. As long as it's the same with every benchmark.
I think the benchmark should be done in 3rd person. That way the client has to render exactly 1 avatar. Because for a fair benchmark you should at least include 1 avatar, but it would be difficult to have more then 1 avatar present and keep the test conditions the same.

And it was just an idea to develop a benchmark, coz I think it could be useful. I havent really thought about the details...

Mine:

+3700 San Diego Core (Overclocked)
Geforce 7800GT (Overclocked)
2gb Corsair XMS RAM (MONKEY!)

As long as I'm not rezzing and its not crazy in the area I'm in I expect about 70-80ish FPS

Aye, I ran my tests with the same avatar, no attachements, no bling, Mouselook & 3D View. Mouselook added 1-5 FPS. Naturally folks should realize adding all that stuff will slow down framerates.

RAM seems to be a big help. Went from 1 Gig to 2 and saw a nice increase in FPS. Once they start taking advantage of GPU's it should be even better!