Your CPU score is very high. Looks like you have GPU PhysX enabled in the nvidia driver just like I did. Turn it off and it will drop to 20K or so I bet.
OKay been playing around... The funny thing is the CPU is running at 3.8GHz and not 4.0GHz and my score sky rocketed. I am using the new Nvidia drivers... wonder if that helped at all here.
You guys realize leaving GPU PhysX enabled in the nvidia driver seriously distorts your CPU score, right?
The nvidia driver helps the CPU do its tests since they're both PhysX-based. I get a difference of about 30,000 points in CPU score between GPU PhysX enabled or disabled.
If it was turned off before when you updated your driver it should turn it back on. Your GPU went up under 3k but your CPU went up just over 6k.
Then that's why your CPU score jumped. Note in my comment I pointed out the GPU helps the CPU do its test since it's PhysX based, meaning if you overclock your graphics card, your CPU score is going to go up.
I see... things I never knew because I never looked. So fail! I really need to dig into the current graphics again & get myself versed in current tech
Heh... just for reference, that is why many people (including myself) hold the opinion that nvidia graphics cards "cheat" on the 3DMark Vantage test. My personal opinion is that the Vantage score from an nvidia card where PhysX acceleration is enabled is only valid for PhysX-accelerated games, such as Arkham Asylum. For a true comparison score to AMD's hardware in non-PhysX games, the benchmark has to be run with PhysX acceleration disabled.
Well the extra points should go to NVIDIA in general for taking the time to make PhysX work on their chip. I see it is an advantage that comes with the ship. Maybe not all games use it, but some do and you will benefit from it.
NVIDIA may have taken the time to make PhysX work on their cards, but that's because they bought out Ageia and all Ageia's intellectual property, effectively cutting AMD off from it.
If AMD had bought out Ageia, NVIDIA cards wouldn't be able to accelerate PhysX, and we'd be in the same situation only reversed. How is that different?
Either way, the arguments are tired, and at the end of the day, very few products explicitly rely on PhysX. Essentially, your Vantage score with PhysX enabled does not give you an accurate idea of how your PC will run most DX10 games.
Actually many games use PhysX but most do not push it as a top item. The fact still stands that NVIDIA paid for a technology and benefit from it in tests. If Futuremark didn't want the increase they could adjust the test to not utilize PhysX.
Also if you want to push your system for max score (which everyone who runs Vantage does for top scores) they leave PhysX on not off.
Of course they do, because it makes their processor score artificially higher.
Observe the difference between these two shots:
Both benchmarks run on the same system one after the other. Notice how the GPU score is practically unaffected, but oops, my CPU somehow got five times as fast...
Here's some clarification of what's going on. The first test (Jane Nash) makes use of PhysX to animate the water and the canopy on the boat. The scene is graphically complicated, so PhysX acceleration and graphics rendering end up in a fight for GPU time. The end result is that the scene really doesn't render any faster than when GPU PhysX acceleration is disabled.
The CPU tests, which are also PhysX based, are supposed to calculate your CPU power as a function of how well your processor handles PhysX calculations. However, since the scene is supposed to be rendered by the CPU, the GPU remains idle, apart from throwing the picture onto the monitor. When PhysX acceleration is disabled, the test runs properly, with the GPU remaining idle while the CPU crunches its numbers.
When PhysX acceleration is enabled, the NVIDIA driver hooks into the scene and does all the PhysX calculations for the CPU using near-100% of the GPU core, since the GPU core is otherwise idle. In my mind, this makes it valid to point out that in a PhysX-based game, the advantage you would see is not as large as 3DMark Vantage would seem to indicate.
Edited in an attempt to reduce wall of text effect.
Derek is right. The amount of GPU horsepower available for PhysX on Vantage's CPU tests are about 10x the amount available in a regular game, since the GPU is not otherwise working during the Vantage CPU test, but it's at full load during a game.
I understand why the score is inflated.. but what I am saying is all final test are ran with it on. But overall the score effect is usually less than 1000 points in the total. When your dealing with 44k scores that isn't a huge difference even as it inflates by 3000+ points.
The overall difference is about 3500 points on a C2Q / GTX285. It's probably even larger for your system, with its large amount of additional CPU crunch.
So the last 24 hours have been fun.... Been trying to push my system to its limits & in doing so I can't get the bastard to finish a full Vantage run. I know this thing has more than it is showing currently
Not being able to finish a run is a good indicator it's unstable... I tend to worry about making sure my PC isn't going to randomly crash before I start benchmarking it.
Then again, I only overclock aiming at long-term stable settings. I've never really been one to push it to the maximum possible just for a benchmark.
The system is stable under the Prime tests and other runs but it is the graphics driver that is giving out. Most people are not able to push the GTX295's beyond 680MHz core which is causing the graphic drivers to fail when I push it over 680. I am now playing with the memory & the shader clock to see what the culprit is.
The CPU is purring at a nice 4.10GHz and is zooming along with the memory running at a nice 1640MHz
Well, you know your CPU is stable, then. You can test to find out whether your graphics is long-term stable in a similar fashion with Furmark or one of Unigine's benchmarking products.
yeah that's the problem... It passes both Furmark & the Unigine tests, but craps out on the last part of the 2nd GPU test in Vantage. Honestly I think I am going to need some voltage increases to keep her stable, but under full load the GPU is hitting temps I am concerned with.
I am very tempted to jump into water cooling for both CPU & GPU and picking up EVGA's water block for this GPU.
I have now gotten my CPU running 4.2GHz with the memory just over 1700MHz. In all it is running very stable on the CPU side and temps are shockingly cooler than I expected.
CPU 2.66GHz = 24c No Load - 41c Full Load @ 0.95V
CPU 3.80GHz = 33c No Load - 58c Full Load @ 1.125V
CPU 4.2GHz = 41c No Load - 80c Full Load @ 1.34V - Was running pretty hot for my liking & decided to not push her any further with air cooling.
Got her to finally complete the run... Had to dumb down the GPU memory clock, which overall killed the score but almost got me my 26,000 so close it hurts
Comments
Edit - sweet rig, by the way.
3DMark Score
P21729 3DMarks
CPU Score
44616
Graphics Score
18556
Slight 7% GPU OC got me upto:
3DMark Score
P23849 3DMarks
CPU Score
46098
Graphics Score
20544
q9450 3.55
Dual 88GT 650/1625/955
If it was turned off before when you updated your driver it should turn it back on. Your GPU went up under 3k but your CPU went up just over 6k.
Heh... just for reference, that is why many people (including myself) hold the opinion that nvidia graphics cards "cheat" on the 3DMark Vantage test. My personal opinion is that the Vantage score from an nvidia card where PhysX acceleration is enabled is only valid for PhysX-accelerated games, such as Arkham Asylum. For a true comparison score to AMD's hardware in non-PhysX games, the benchmark has to be run with PhysX acceleration disabled.
Either way, the arguments are tired, and at the end of the day, very few products explicitly rely on PhysX. Essentially, your Vantage score with PhysX enabled does not give you an accurate idea of how your PC will run most DX10 games.
Edit for typo.
Also if you want to push your system for max score (which everyone who runs Vantage does for top scores) they leave PhysX on not off.
Observe the difference between these two shots:
Both benchmarks run on the same system one after the other. Notice how the GPU score is practically unaffected, but oops, my CPU somehow got five times as fast...
Here's some clarification of what's going on. The first test (Jane Nash) makes use of PhysX to animate the water and the canopy on the boat. The scene is graphically complicated, so PhysX acceleration and graphics rendering end up in a fight for GPU time. The end result is that the scene really doesn't render any faster than when GPU PhysX acceleration is disabled.
The CPU tests, which are also PhysX based, are supposed to calculate your CPU power as a function of how well your processor handles PhysX calculations. However, since the scene is supposed to be rendered by the CPU, the GPU remains idle, apart from throwing the picture onto the monitor. When PhysX acceleration is disabled, the test runs properly, with the GPU remaining idle while the CPU crunches its numbers.
When PhysX acceleration is enabled, the NVIDIA driver hooks into the scene and does all the PhysX calculations for the CPU using near-100% of the GPU core, since the GPU core is otherwise idle. In my mind, this makes it valid to point out that in a PhysX-based game, the advantage you would see is not as large as 3DMark Vantage would seem to indicate.
Edited in an attempt to reduce wall of text effect.
GTX285 Benchies
The overall difference is about 3500 points on a C2Q / GTX285. It's probably even larger for your system, with its large amount of additional CPU crunch.
Then again, I only overclock aiming at long-term stable settings. I've never really been one to push it to the maximum possible just for a benchmark.
The CPU is purring at a nice 4.10GHz and is zooming along with the memory running at a nice 1640MHz
I am very tempted to jump into water cooling for both CPU & GPU and picking up EVGA's water block for this GPU.
I have now gotten my CPU running 4.2GHz with the memory just over 1700MHz. In all it is running very stable on the CPU side and temps are shockingly cooler than I expected.
CPU 2.66GHz = 24c No Load - 41c Full Load @ 0.95V
CPU 3.80GHz = 33c No Load - 58c Full Load @ 1.125V
CPU 4.2GHz = 41c No Load - 80c Full Load @ 1.34V - Was running pretty hot for my liking & decided to not push her any further with air cooling.