Suggestions for a FSXmarkCPU benchmark to compare CPU architectures and cache sizes

[SAAB340 28]

I’ve long wanted to know how much different CPU architectures and cache sizes affect FSX. What’s needed is a benchmark that takes clock frequency, GPU, PCIe bus, RAM, and storage out of the equation.

I think I’ve come up with such a benchmark.

 

By making sure everyone is using the same clock frequency and RAM speeds regardless of CPU we can get rid of them as a factor.

 

By closing down FSX fully between each run (Ctrl+C) and excluding the first run we can get rid of most of the effects different storage might have.

 

Finally we need settings and a benchmark flight path that is fully CPU bond at these settings and not affected by the GPU or the PCIe bus it’s connected with.

 

A basic requirement is a quad-core (or higher) without hyper threading or turbo modes enabled.

The CPU speed for the benchmark is 2.4GHz.

The RAM speeds and timings are 1066MHz 6-6-6-20 1T

You might think that these speeds are very slow but they are chosen as I believe pretty much every quad-core (or higher) system should be able to do these speeds. Even without having to resort to overclocking. From the Intel Q6600 and the AMD Phenoms all the way up to i7 IvyBridge today. We’re after the difference between architectures and cache sizes, not the highest FPS.

 

The first 4 minutes of the FSXmark07 flight path using the same settings I used for my texture loading tests apart from the screen resolution seems to be un-affected by the difference between a 8800GTS512 on a x8PCIe2.0 bus and a GTX470 on a full x16PCIe2.0 bus. The screen resolution I’ve used is 1680x1050x32. A resolution most of us should be able to use.

 

I’ve rebuilt the cfg from scratch and just added these 4 lines for the benchmark

[bufferPools]

UsePools=0

[JOBSCHEDULER]

AffinityMask=14

 

You should still use AM=14 if you have a hexa or octa core.

 

I’ve also made sure all add on sceneries are deactivated and that I’m not using any additional AA profiles through nvidia inspector (by doing a clean re-install of the gfx driver)

 

The results omitting the first run are as follows:

GTX470 on x16PCIe2.0 Avg FPS 22.033

5290, 240000, 13, 31, 22.042

5299, 240000, 13, 31, 22.079

5241, 240000, 13, 32, 21.838

5247, 240000, 13, 32, 21.863

5362, 240000, 13, 33, 22.342

 

8800GTS512 on x8PCIe2.0 Avg FPS 22.259

5393, 240000, 13, 31, 22.471

5391, 240000, 13, 32, 22.463

5302, 240000, 13, 31, 22.092

5369, 240000, 13, 32, 22.371

5256, 240000, 13, 32, 21.900

 

There is a 1% difference between the average results form 5 benchmarks (to the slower setups advantage) but the difference between the individual benchmarks is way greater so we can easily say that the GPU does not affect.

 

It would be highly appreciated if a few other persons could try this as well so we could see if the benchmark indeed produces the same result for the same CPU architectures before we write up something more official.

 

My CPU is a Intel i7 860 Lynnfield.

 

Any feedback/suggestions?

[dazz 69]

Do we need to run this on a fresh FSX install Lars?

[SAAB340 28]

Do we need to run this on a fresh FSX install Lars?

No Dario. No fresh install needed. But deactivate all add-on scenery you have within FSX.

[SAAB340 28]

Is no one interested to find out what you get for the extra money you pay for a i7 over a i5 or a SandyBridge over an IvyBridge? Just to make it clear. You won't have to reinstall FSX to do this.

 

Here is a better step by step guide if it's helpful:

 

• Make sure you have the FSXmark07 flight installed. It can be downloaded from the Avsim Library. Just search for: fsxmark07.zip
  It is called "FSX Dynamic Benchmark Package" and is located in Category: Flight Simulator X - Utilities
  
• Set your computer to run at 2.4GHz CPU, 1066MHz 6-6-6-20 1T RAM. This will most likely mean that you underclock your system so everyone should be able to safely do this. Make also sure Turbo and HyperThreading is disables
  
• Backup your cfg and than delete it.
  
• Start FSX and let it rebuild the cfg
  
• Untick all add-on scenery you have under Settings, Scenery library...
  
• Set the settings as per the settings used here but change the screen resolution to 1680x1050x32 and set Realism to Hard under Settings, Realism...
  
• Close FSX and open the cfg. Add these lines to the top:
  [bufferPools]
  UsePools=0
  [JOBSCHEDULER]
  AffinityMask=14
  
• Make sure you disable any additional AA in nvidia inspector or similar. (One way of doing this is to do a clean install of the driver)
  
• Start FRAPS and set it up with Benchmarking Hotkey "P", tick to include FPS, Frametimes and MinMaxAvg, tick and set benchmark to stop after 240 seconds.
  
• Start FSX and load the FSXmark07 flight. Run the benchmark by pressing "P" and wait until FRAPS has finished benchmarking after 4 minutes.
  
• Exit FSX fully by pressing CTRL+C
  
• Reload FSX and carry out another 5 benchmarks exiting FSX fully with CTRL+C between each benchmark.
  
• Post the results from the 5 last benchmarks in this thread. Remember to say what CPU you are using
  
• Replace the cfg with your previously backed up cfg.
  
• Re-enable your add-on sceneries in Settings, Scenery library...
  
• Reset your previous additional AA (in nvidia inspector)
  
• Reset your computer to your previous CPU and RAM speeds.
  

Done

Thanks for contributing

[dazz 69]

I tried yesterday but couldn't boot at 2.4GHz for some reason. Hopefully I'll have the time later today to try again

[SAAB340 28]

I tried yesterday but couldn't boot at 2.4GHz for some reason. Hopefully I'll have the time later today to try again

Thanks for trying. Hopefully you can manage to boot at 2.4GHz. I hope this isn't an i7 SandyBridge "feature". Any one else who tried booting an i7 downclocked to 2.4GHz?