My 9600GT "Benchmarks"

[Medion]

Background:

These benchmarks aren't going to seem very impressive. Due to me playing World or Warcraft for too darn long, I've fallen way behind on gaming. I'm now forcing myself to finish my backlog of games before buying anything new. I'm pretty close to buying something new, and as I do, I'll be adding to this list.

Testing Methods:

This is going to seem very different from how just about anyone else benches games. I benchmark the way that I play. Meaning, I don't use synthetic benchmarks, and I actually do the benchmarks with VSYNC on and refresh set to 60hz. My goal is to bring my games to a solid 60fps. With VSYNC off, this distorts the framerate, as a 90fps spike can bring your results up, countering a 30fps dip. I'm more concerned with the effect those dips have on my gameplay experience. Locking the max at 60 gives me a more accurate playing framerate. So a FPS close to 60 (55+) denotes a smooth experience.

NVIDIA Control Panel:

I think I've got a firm grasp on the basics, but if anyone sees anything odd here in my settings, please, educate me. I go for max quality, but if I have a setting wrong, I'd like to be corrected on it. With that said, the following are my NV CP settings. I've ommitted the levels off AA/AF, and the Transparency AA setting, since those will be the settings I reduce on a per game basis, as needed.

Antialiasing - gamma correction: Off
Antialiasing - mode: Override application setting
Conformant texture clamp: Use hardware
Error reporting: Off
Extension limit: Off
Force mipmaps: Trilinear
Maximum pre-rendered frames: 3
Multi-display/mixed-GPU acceleration: Single display performance mode
Texture filtering - Anisotropic sample optimization: Off
Texture filtering - Negative LOD bias: Clamp
Texture filtering - Quality: High Quality
Texture filtering - Trilinear optimization: Off
Threaded Optimization: Auto
Triple Buffering: On
Verticle Sync: Force on

In addition to this, I'm using Rivatuner's D3DOverrider to force vsync and triple buffering in games that have issues with this.

The Games:

Every game below is running 1680x1050, 16xAA, 16xAF, Transparency AA set to Multi-Sample, as well as maxed in-game settings (LoD set to OFF for World of Warcraft).

Devil May cry 4 (DX9) - Avg: 49.13 - DX9 performance test, again vsync capped at 60fps (hence the lower numbers), everything on super high, 16xCSAA, 16xAF, transparency set to multi-sample.

Devil May cry 4 (DX10) - Avg: 47.39 - DX10 performance test, again vsync capped at 60fps (hence the lower numbers), everything on super high, 16xCSAA, 16xAF, transparency set to multi-sample.

Fable: The Lost Chapters - Min: 54 Avg: 59.23 - Ran a lap around Oakvale (post burning)

Need for Speed: Most Wanted - Min: 49 Avg: 59.31 - Tested using a custom race, medium traffic, sprint, seaside and power station, maxed out Mustang.

Tomb Raider Anniversary - Min: 59 Avg: 59.89 - Played in Croft manor

Unreal Tournament 2004 - Min: 47 Avg: 59.19 - Played a game of Onslaught - Torlan (with new vehicles)

World of Warcraft - Min: 56 Avg: 59.89 - Ran a L1 newbie from the human start to Stormwind. All of those trees are a good way to get a performance hit from Transparency AA.

Conclusion:

Given that I don't currently own any bleeding edge games, nor will I ever own Crysis, today's low/mid range GPUs are ideal for me. However, with Bioshock, Witcher Enhanced, Devil May Cry 4, GRID, and Burnout Paradise on my "to buy" list, I'm definitely going to have to dial-down the AA some in the near future. Given that I don't mind playing without AA, I've set my upgrade metric as 8xAF. When I have to go below that setting, it's time for an upgrade. My goal is 2 years with this GPU.

[Medion]

Added DMC4 benchmarks.

[MisterMister]

Dude, you do know that the 9600GT KILLS in SLI, right? It scales like no other. After adding a second 9600GT to my set up, my 3Dmark06 score went from ~12,000 to 17,420 (click on the score in my signature for a compare chart in the FM ORB)!

That's an insane jump in performance for just adding a second mid-range graphics card! In fact that score bests even PCs running faster CPUs, more RAM (then me) and 9800GTXs!

As for the games, I play every game you mentioned with the exception of Crysis. So far I've run:

Bioshock, COD4, BF2, BF2142, Grid, AoC, Gears of War, Supreme Commander, The Orange Box, FEAR, DOOM3, UT3, CSS, and the latest iteration of MS Flight Simulator.

I play ALL of those games at their MAXIMUM settings (including maxed out AA/AF) at my screen's native resolution of 1680x1050. Even so, Im still getting a 60FPS average frame rate in pretty much all of them (via FRAPS and in-game monitoring). Not once do I experience choppiness, slowdown, skipped frames, or anything else which would detract from game play.

Given that your motherboard supports SLI, and that the 9600GT is going for a super cheap average of $135 USD, its well worth the investment for you to get a second one, IMO. That way you wont ever have to compromise on your graphics settings.

BTW, I DO run the Crysis Benchmark Tool. With that, I can run it smoothly at full detail, and at 1680x1050 but with AA turned off.

Plus, given that we have very similar set ups (I'm running a Q6600 too, but with half as much system RAM), I don't see why you cant/wont get similar performance figures. So it's not like I'm running a super cast CPU compared to yours or something.

BTW, since you currently own a 9600GT, this is a good read for you:

http://www.techpowerup.com/reviews/N...9600_GT/1.html

Keep in mind all of my testing and benchmarking was done BEFORE I found out about that "trick".

[Medion]

I appreciate the input, but I have no desire to go SLI. For one, my mobo doesn't support it (you were confused, I have the 650i Ultra, not the 650i SLI). Also, I currently run all of my games at or near 60fps, with maxed settings, 16x/16x. So I'd be gaining almost nothing in the games I currently play.

Again, thanks for the input though.

[MisterMister]

IC. Yes, I didnt know there was a non SLI version of that board. You learn something new every day it seems.

Anyway, did you give that link a once over? Its pretty sweet that the 9600GT is one of the ONLY PCI-E cards which actually gain a performance increase by OCing the PCI-E bus clocks in the BIOS. That's pretty bad ass IMO.


http://www.techpowerup.com/reviews/N...9600_GT/1.html

[Medion]

Originally Posted by MisterMister

IC. Yes, I didnt know there was a non SLI version of that board. You learn something new every day it seems.

Anyway, did you give that link a once over? Its pretty sweet that the 9600GT is one of the ONLY PCI-E cards which actually gain a performance increase by OCing the PCI-E bus clocks in the BIOS. That's pretty bad ass IMO.


http://www.techpowerup.com/reviews/N...9600_GT/1.html

Yea, I've seen that link before. I tend to avoid overclocking until I'm starting to get a desire to upgrade. Like I said, I'm not only impressed at the overall performance of my system, but I'm flat out shocked at how this 9600GT performs. I was expecting a marginal boost from the 8600GT at best. There will come a time I'll feel the urge to upgrade. When that time comes, I'll probably OC as much as I can.