Google Earth 3D GPU performance thread

I'm interested in knowing what kind of performance people get while rendering Google Earth 3D buildings with different GPUs (especially the mobile ones). This information may be useful to people who would like to have good performance for GE with 3D buildings enabled, when they choose their next computer.

The most challenging model I found was Main St. in Disney World's Magic Kingdom (the street leading up to the castle). To get the thread started, here are my system specs and the numbers I got (measured with fraps), while moving along that street toward the castle:

GPU: nVidia Go 7700 (not from the latest generation, but still pretty good).
Dedicated VRAM: 512 MB
CPU: Intel Core 2 Duo 2.0 GHz
RAM: 2 GB
O.S.: Windows XP SP2
nVidia driver version: 8.6.0.2
GE Version: 4.3 Beta
3D Engine: OpenGL
Resolution: 1440x900 (Full screen)
At low settings: 20 fps
With anisotropic filtering on High, maximum terrain quality, and 2x anti-aliasing: 12 fps

Hovering around the castle at a higher altitude gave me slightly better numbers: 25 fps at low settings and 15 fps at high settings. Surprisingly, I also noticed that with DirectX the fps counts drop to less than half what you get with OpenGL.

Anybody else care to report their fps counts?

 

i havent used fraps to get results on google earth but I do know that running fraps can cause a 10-15 fps drop.

 

Is there any better way to measure fps when the application (in this case, GE) doesn't do it by itself? I only know about fraps. In any case, if everybody uses the same tool to measure the fps for this comparison, the counts will still provide a measure of the relative performance of the GPUs, even if the absolute numbers are off by 10-15 fps, as you suggest.

Also, fraps performance hit is probably proportional to the frame rate (meaning that, if the rate was, for instance, 10 fps, fraps wouldn't cause it to drop to 0). In my case, the drop feels more like at most 5 fps, corresponding to about 20-25 % of the frame rate.

 

There's no way FRAPS causes 10-15 fps drop. Maybe when recording video, but not just displaying fps. It's a simple calculation that uses very little resources to determine.

 

Nobody? Have people even looked at GE's Disney World 3D model? It is not a game but it is still pretty impressive.

The reason I am interested in this is that I'm going to buy a new notebook and I'd like to make sure I get good performance (at least 15-30 fps) rendering GE's more complex 3D models at high settings. You're generally not allowed to install software when test-driving notebooks at stores (at least not the ones around here), so I was hoping people would contribute the fps counts they get with their rigs. I'm especially interested in knowing the performance of the nVidia 8400M GS/GT, as these are quite common in the class of notebooks I am considering. (I did have a chance to do a quick test on a borrowed notebook with this GPU, but I didn't have it long enough to try a lot of different settings, and the fps count I got was so low that I suspect the configuration was sub-optimal).

Other things that it would be interesting to compare would be the impact on the fps counts of Windows XP x Vista (w/ and w/o Aero and Aero Glass) and OpenGL x DirectX9 x DirectX10 (in addition to low x high settings).

Would anyone be interested in helping with this benchmarking?

 

I will go look at the model later, but i do have the same findings as you. If I get time, I will install the FireGL V5600 drivers for the HD2600 and see how that runs. I am running XP by the way.

One problem would be that Geforce/HD cards and their drivers are not OpenGL optimised.

 

Yeah, fraps only kills the framerate when you're recording video.

 

O.k., I see people here aren't too interested in Google Earth's 3D models...

So, instead of trying to compare GE's performance on various GPUs, how about this: could somebody with an nVidia 8400M GS just tell me what fps count you get moving down Disney World's Main St. toward the castle, at low and high settings (and perhaps also using OpenGL and DirectX, for comparison)?

As I mentioned above, I did this test on a borrowed computer running Vista, but I only got 4 fps at low settings and 2 fps at high settings, which seems too low, because the 8400M GS is not supposed to be 5 times slower than a Go 7700, for which I got 20 and 12 fps, respectively. There was probably something wrong with the notebook's configuration, but I didn't have it long enough to try different things to maximize the fps count.

Can anybody with an 8400M GS confirm these fps counts, or tell me if you were able to do better?

 

Why don't you get something with a 8600 or 9500-9600?

 

Actually that does sound about right. the x700 series is far superior to the x400 series.

For instance, if you compare the 6600 to the 8400, the 6600 will actually outperform the 8400.

I don't have figures for your exact comparison, but these charts will paint a good picture for you:

charts

Based on these charts the 6600GT is a hair better then the 8400GS.

The 7600GT (7700 would be a touch higher i would think) is roughly double the power of the 8400.

I would venture a guess that the 7700 GT would be 2.5x the power of an 8400GS.

Having double the power can translate to more then double fps, it all depends on teh situation and what your bottlenecks are.

I've seen people improve their fps by 10-20% simply my overclocking a cpu.

 

Yeah, I would really recommend something stronger then an 8400 if you're looking to game on a laptop. The 8400 is garbage.

 

i would recommend AT LEAST a 8600/3650 if you want to keep a laptop for more than 3 years

 

Thanks, Ayle, Kermit, and Kaz

The reason I'm asking about the 8400M is because I'd like to buy a smaller and lighter notebook (more specifically, a 13.3 incher) than my current one, and, as you probably know, these don't come with the more powerful GPUs you're suggesting. Besides, I'm really not a gamer, but Google Earth's 3D models are important to me. So if I get a decent frame rate for it on the 8400M, that would do.

Kermit's guess that the Go 7700 should be about 2.5 x faster than the 8400M GS is consistent with the benchmarks I have been seeing (like the ones here). That's why the 5 x difference I got in that quick test I mentioned seems too large. However, since these benchmarks are averages, and different applications may have different bottlenecks, I was hoping somebody could replicate the test, and tweak the configuration (except for overclocking) to see what's the highest fps they can get with the 8400M GS.

 

All right, I see people here are REALLY not interested in Google Earth's 3D models. (Before anyone suggests this, I have already posted on GE's forums, but over there the problem is people don't seem to know much about GPUs.)

So, here's Plan C: Since most of you are gamers, can anybody just tell me what game is approximately equivalent in rendering load to GE with 3D models enabled? What game gives approximately the same fps count on your rig, in low and high settings, as moving down Magic Kingdom's Main St., toward the castle, in Google Earth?

Since there are lots of discussion threads and web pages comparing the performance of mobile GPUs for various games, if I know what game is equivalent in rendering load, I can have an idea of what GPU will give me an acceptable performance for Google Earth.

Anybody?

 

You can get the LG P300 with 8600GS. But they are hard to find...

 

Your internet connection is more limiting to Google Earth than your graphics card.

 

Here are info and some screenshots about few shader effects (Bloom, Ambient Occlusion SSAO, Depth of Field, Palettes, and Post Proccess shaders) available for Google Earth. Makes it look a lot nicer, but you need pretty good graphics card.

 

Thanks, Johnny. I hadn't heard of this notebook before. I'll check it out.

Sure. I'm assuming you have waited for the 3D model of the test area to finish downloading, before measuring the fps.

Looks nice, ronkotus, but this would certainly require an even larger/heavier notebook. I'll be happy if I get good 3D rendering performance for the standard Google Earth, without these fancier effects.

 

For those of us not very familar with Google Earth, and the 3d related features walk us through what you do.

Do you start by typing in an address to take you to Main Street? Your looking at Main Street in Disney World (in Florida) not Disneyland (in California) correct? (Disneyland on Google Earth doesn't have a 3d Magic Castle ). Are you starting at a particular destination (like starting at the sky location, accessed by clicking on the planet/saturn symbol)? Are you using the flight simulator on Google Earth, or are you zooming in and out of the landscape using the tools on the upper right side of the screen?

 

Avid Gamer:

Thanks for your interest! Make sure you have the "3D Buildings" layer enabled. Starting from any location, just type "Disney World" in the search box, hit the search button, and you should fly immediately over there. You will see a yellow Mickey-Mouse shaped icon labeled "Walt Disney World in 3D". Click on it, then select the "Magic Kingdom" thumbnail to go there. You will see the castle straight ahead and Main St. leading to it. Using the navigation controls on the right (NOT the flight simulator), click on the button labeled with a "+" to drop down to ground level, then grab and pan the image to go to the beginning of Main St. You should see a representation of two men pulling a cart labeled "Family Fun Day". You can't miss it.

Give Google Earth enough time to finish downloading the 3D model, then move ahead along the street toward the castle by clicking on the "^" arrow above the hand icon.

I used FRAPS to measure the fps. The anisotropic filtering, terrain quality and OpenGL/DirectX settings can be found under Tools -> Options -> 3D View. Make sure you stop other programs or services that may have a negative effect on performance (such as Windows Search).

Let me know if you need any other info.