IEGD driver performance decreased after updating the driver

[TSchu20]

Hi all,

after updating the IEGD driver from Version 8 something to 10.3.1 I experience a 2D drawing performance decrease of about 20%.

Has anyone similiar experiences? Is there anything that can be tuned?

IEGD version 8:

PassMark(TM) PerformanceTest 7.0 Evaluation Version (http://www.passmark.com http://www.passmark.com)

Results generated on: Thursday, July 28, 2011Benchmark Results

Test Name: This Computer

Graphics 2D - Solid Vectors: 6.0Graphics 2D - Transparent Vectors: 1.7Graphics 2D - Complex Vectors: 75.5Graphics 2D - Fonts and Text: 100.3Graphics 2D - Windows Interface: 119.1Graphics 2D - Image Filters: 146.2Graphics 2D - Image Rendering: 8.12D Graphics Mark: 366.6PassMark Rating: 784.5

System information: This Computer

CPU Manufacturer: GenuineIntelNumber of CPU: 1Cores per CPU: 1CPU Type: Intel Pentium M 1.40GHzCPU Speed: 1396.5 MHzCache size: UnknownO/S: Windows XP (32-bit)Total RAM: 2038.4 MB.Available RAM: 1769.8 MB.Video settings: 1680x1050x32Video driver: DESCRIPTION: Intel Corporation 915GM/915GMS/915GME/910GML/910GMLE Embedded Graphics Controller Function 0 MANUFACTURER: Intel Corporation BIOS: Intel Video BIOS DATE: 3-10-2008Drive Letter: CTotal Disk Space: 6.0 GBytesCluster Size: 4.0 KBytesFile system: FAT32

IEGD 10.3.1:

PassMark(TM) PerformanceTest 7.0 Evaluation Version (http://www.passmark.com http://www.passmark.com)

Results generated on: Thursday, July 28, 2011Benchmark Results

Test Name: This Computer

Graphics 2D - Solid Vectors: 4.3Graphics 2D - Transparent Vectors: 1.6Graphics 2D - Complex Vectors: 57.0Graphics 2D - Fonts and Text: 41.8Graphics 2D - Windows Interface: 86.0Graphics 2D - Image Filters: 145.8Graphics 2D - Image Rendering: 130.62D Graphics Mark: 291.2PassMark Rating: 623.2

System information: This Computer

CPU Manufacturer: GenuineIntelNumber of CPU: 1Cores per CPU: 1CPU Type: Intel Pentium M 1.40GHzCPU Speed: 1396.5 MHzCache size: UnknownO/S: Windows XP (32-bit)Total RAM: 2038.4 MB.Available RAM: 1700.8 MB.Video settings: 1680x1050x32Video driver: DESCRIPTION: Intel Corporation 915GM/915GMS/915GME/910GML/910GMLE Embedded Graphics Chipset Function 0 MANUFACTURER: Intel Corporation BIOS: Intel Video BIOS DATE: 3-1-2010Drive Letter: CTotal Disk Space: 6.0 GBytesCluster Size: 4.0 KBytesFile system: FAT32

[Kirk_B_Intel]

Hello,

At a driver level, we do a pretty good job of monitoring perfomance of key functions to make sure that performance maintains or improve from release to release. That is why we have a Validation team and release criteria that we set for every release that include performance.

What we can't do is look at every performance tool, or understand the ramifications of all changes made to the driver. We do not use PassMark as part of our validation suite so we have not monitored whatever combination of functions it uses to evaluate performance.

Because we have not measured any significant decrease in performance in the driver version to version, my guess is that we have affected a system change that has impacted whatever narrow window of functionality that PassMark monitors. Likely what has happened is that we have added additional 3D capability to the driver between V8 and V10 and Windows is now attempting to use that new 3D capability in the driver for its 2D functions. That is not under our direct control, and is not always a smart thing to do. As graphics operations become more complex, the use of an accelerator is usually a good thing to speed it up. THere is a crossover however where the overhead of transfering a function from the CPU to the GPU and back is greater than the time savings of using the GPU. This is normally the case with 2D (and things like Windows Presentation Foundation 2D functions) where using the 3D acceleration slows down the 2D operations yet Microsoft seems to insist on using them and using them very inefficiently. You might be able to verify this by going into your system performance settings and setting back the overall hardware acceleration usage a notch or two and see how that impacts PassMark performance on IEGD 10.4. My guess is that it will improve thus showing what MIcrosoft is trying to do under the covers.

Hope this helps, Kirk

[TSchu20]

Thanks Kirk, it turned out to be caused by a different cleartype setting and not by updating the driver.

[Kirk_B_Intel]

Interesting. I had not heard of 2D performance impacts of Clear Type. That is an interesting one I will need to watch out for in the future. Might also be a good one to add to our FAQ...

Thanks for the heads up! Kirk