UPDATED! - My Tests are In! The new 81.26 drivers DO use both cores! More than 70% 2x core usage.

[formulav8]

This was originally talked about the the cpu forum. I thought it would fit in here too.

I made a basic app that monitors the cpu's usage. I installed each of the 77.72, 78.01 and 81.26 drivers and played BF2 for a few minutes each. The 77.72 and 78.01 drivers shows very little total usage above 50%.

Edit: I also wanted to say that the monitor was set to every 1.5 seconds and to not report 0% usages either.

My system is a 3800+ X2 @ 2600mhz, PNY 6800 NU @ 16x/6x pipes using stock speeds. Resolution was 1024x768 and 1GB dual channel BH5 memory @ 172mhz or so.

But when I installed 81.26 drivers and played BF2 the Dual Core usage was more than 70%! So, without a doubt the new drivers DO play with both of the cpu's cores.

This should somewhat prove that the performance increases with the drivers does have something to do with the using of both cpu cores and not just driver optimizations in general.

Hopefully this info helps put some insight into the matter You can download the monitor program and the cpu usage logs below. Post your results if you use the app if you want too

Here is the download link for the 3x BF2 cpu logs.

And here is the link to the CPU Load Monitor application I made. The app was made with Visual Studio.NET 2003 so, you will need the 1.1 version of the DotNet framework installed. It will most likely not run properly if you do NOT have 2x cpu's. So single cpu's do not use

-----------------------------------------------------------------------------------------

UPDATE: Below are my benchmark results with each driver versions.

OK, I ran a couple benches. I used versions: 77.72, 78.01, and 81.26 drivers. Below are my results.


*************************

*81.26 Driver Results*

3DMark2001: 22,191

Aquamark: 63,645

------------------------

*78.01 Driver Results*

3DMark2001: 21,346

Aquamark: 60,734

------------------------

*77.72 Driver Results*

3DMark2001: 21,215

Aquamark: 60,845

*************************

So, although the results are not huge, there ARE almost a 5% performance increase in the use of the 81.26 drivers which uses some of the 2nd cpu core.

So, a free performance boost for Dual Cpu/Core users is not to shabby at all



Jason

 

[ArchAngel777]

Originally posted by: formulav8
This was originally talked about the the cpu forum. I thought it would fit in here too.

I made a basic app that monitors the cpu's usage. I installed each of the 77.72, 78.01 and 81.26 drivers and played BF2 for a few minutes each. The 77.72 and 78.01 drivers shows very little total usage above 50%.

Edit: I also wanted to say that the monitor was set to every 1.5 seconds and to not report 0% usages either.

My system is a 3800+ X2 @ 2600mhz, PNY 6800 NU @ 16x/6x pipes using stock speeds. Resolution was 1024x768 and 1GB dual channel BH5 memory @ 172mhz or so.

But when I installed 81.26 drivers and played BF2 the Dual Core usage was more than 70%! So, without a doubt the new drivers DO play with both of the cpu's cores.

This should somewhat prove that the performance increases with the drivers does have something to do with the using of both cpu cores and not just driver optimizations in general.

Hopefully this info helps put some insight into the matter You can download the monitor program and the cpu usage logs below. Post your results if you use the app if you want too

Here is the download link for the 3x BF2 cpu logs.

And here is the link to the CPU Load Monitor application I made. The app was made with Visual Studio.NET 2003 so, you will need the 1.1 version of the DotNet framework installed. It will most likely not run properly if you do NOT have 2x cpu's. So single cpu's do not use


Jason

Thanks Jason, this rocked.

 

[otispunkmeyer]

cheers man!

 

[formulav8]

No probs, glad it was interesting somewhat

In the log for the 77.72 drivers, the highest total core usage was 54%. In the 81.26 drivers, the highest total core usage was 73%. That is more than a 25% increase in core usage. Which would very well make some kind of increase in performance in cpu bound situations.

I just though it was kinda interesting



Jason

 

[mwmorph]

any personal benchies?

 

[formulav8]

Hi, no benches. I was only interested in whether or not the new drivers were generally optimized or if they really did make use of the 2nd core.

I honestly do not know how to bench BF2. I would be more than happy to if someone can tell me the easiest way how


Jason

 

[yhelothar]

Get fraps!!

I'd be surprised if it really can utilize the dual cores.
I thought the games itself had to be coded to utilize dual cores. It would be totally awesome if the drivers can make the game utilize dualcore.

 

[Pabster]

Fraps?

 

[formulav8]

Yeah whats Fraps?

 

[yhelothar]

Originally posted by: formulav8
Yeah whats Fraps?

google is your friend

 

[formulav8]

Why not just tell me


Jason

 

[yhelothar]

Originally posted by: formulav8
Why not just tell me


Jason

it'll take more effort for me to tell you than for you to look it up, and you might find where to download it too on the way.

 

[nitromullet]

Nice job.

http://www.fraps.com/

 

[klah]

nm

 

[Pete]

FRAPS.com. It's a program that shows a framerate in a corner of the screen, can show you the average framerate over an interval of your choosing, and can be configured to automatically or manually log framerates every second or so to a log file.

Thanks for the detective work, Jason. I'd love to see if this translates into proportionally higher framerates, if you or someone else is willing to log some FRAPS time.

 

[Ackmed]

To me it dosnt matter if it uses the other core, if the frames arent any higher. If frames jump up, Ill be trading in this 3700+ for a X2.

 

[Velk]

Originally posted by: Ackmed
To me it dosnt matter if it uses the other core, if the frames arent any higher. If frames jump up, Ill be trading in this 3700+ for a X2.

I posted some bench results in the german review site thread, but as obviously a fair few people missed them :

For the example of doom3, I tested by installing the drivers and then running benchmarks in various configurations with the doom3 process either allowed single or dual cpu affinity ( to separate any performance improvements that were just from the updated drivers and not directly from dual core functionality itself ). Interestingly, regardless of system settings, the doom3 process with dual core enabled used 100% of both CPUs.

The improvements were as follows :
1600x1200, High Detail, 4xAA - 42 (single) to 48.7 (dual)
640x480, High Detail, No AA - 119.6 (single) to 142.8 (dual)
640x480, High Detail, 4xAA, windowed - 103 (single) to 119.3 (dual)

What's of interest here is that the performance improvements vary to a quite small extent from situations where it is definetly GPU limited to situations where it is almost exclusively CPU limited.

What was also kind of interesting was that AOE3 demo showed absolutely no improvement and continued to run at a mediocre 29fps (@1920x1200 max detail).

Aquamark increased from 80,944 (single) to 88,324* (dual)


* Testing method was a little bit different here - aquamark refused to allow affinity setting on it's process ( cheat prevention ? ), so for the single cpu test I ran prime95 torture test at normal priority on the second cpu to simulate single cpu.

 

[BFG10K]

I made a basic app that monitors the cpu's usage

Why not use Task Manager?

 

[Pete]

Or, better yet, System Monitor, to log CPU usage (Start > CP > Administrative Tools > Performance).

 

[ArchAngel777]

Originally posted by: BFG10K

I made a basic app that monitors the cpu's usage

Why not use Task Manager?

Task manager would require you to ALT+TAB out and since the CPU jumps all over the place when doing that, it isn't going to be accurate. Writing a program such as he did is the best way.

 

[ArchAngel777]

Originally posted by: Pete
Or, better yet, System Monitor, to log CPU usage (Start > CP > Administrative Tools > Performance).

Never tried that myself... Interesting, I'll have to check it out.

 

[BFG10K]

Task manager would require you to ALT+TAB out

Why, given it keeps a history?

 

[ArchAngel777]

Originally posted by: BFG10K

Task manager would require you to ALT+TAB out

Why, given it keeps a history?

Defaults settings log about 1 minute. Not sure if that would be enough to get an all around the board performance number. Unless there is a way to change the default setting for the speed of the graph. Whata Pete suggested would work, but I do not think the short timer of task manager would be good enough. Possibly a way to change it? I do not know if there is or not.

 

[BFG10K]

If you maximize the window it takes about 5 minutes to fill up the page @ 1600x1200 and that's with the fastest update speed.

Unless there is a way to change the default setting for the speed of the graph

View -> update speed.

 

[hans030390]

So like...does this do anything for p4 hyper thread cpus? Supposedly the HT is supposed to help with multi-threaded programs...so would these drivers apply to the p4ht cpus?

that'd be sweet if they did

2.8ghz roxorz!!! w00t!