Topic on Glossary talk:Graphics card
This page shows the changes between two versions of a post by Mirh in the topic "Custom drivers" on Glossary talk:Graphics card.
You can see other versions of this post at its history page.
Line 15: | Line 15: | ||
For anything from Ivy Bridge and above instead (dx11 igps), really, don't bother. Some slight [https://old.reddit.com/r/pcmasterrace/comments/3hjonw/psa_you_can_overclock_intel_hd_graphics_for/ overclocking] or setting Windows to high performance energy profile is the most optimization you can reasonably get. | For anything from Ivy Bridge and above instead (dx11 igps), really, don't bother. Some slight [https://old.reddit.com/r/pcmasterrace/comments/3hjonw/psa_you_can_overclock_intel_hd_graphics_for/ overclocking] or setting Windows to high performance energy profile is the most optimization you can reasonably get. | ||
− | <small>EDIT: in fact, I think I just now figured out the functioning basis of all the hacks that have ever been. Intel older igps were so shitty, that despite them having in-hardware support for some features (such as [https://www.intel.com/content/dam/doc/white-paper/vertex-processing-selection-capability-paper.pdf SM3.0] and [https://web.archive.org/web/20110607033741/http://software.intel.com/en-us/articles/common-misconceptions-of-intel-processor-graphics/ T&L]), you could actually get them offloaded to the cpu and ''increase'' your performance. Because whatever the obvious overhead this operation has, it was still faster than tasking the potato gpu with yet another operation (or maybe vertex processing is just straightaway quicker on a 200€ cpu of the time, compared to a 2€ graphics adapter). | + | <small>EDIT: in fact, I think I just now figured out the functioning basis of all the hacks that have ever been. Intel older igps were so shitty, that despite them having in-hardware support for some features (such as [https://www.intel.com/content/dam/doc/white-paper/vertex-processing-selection-capability-paper.pdf SM3.0] and [https://web.archive.org/web/20110607033741/http://software.intel.com/en-us/articles/common-misconceptions-of-intel-processor-graphics/ T&L]), you could actually get them [https://web.archive.org/web/20120817231356/http://techlegend.wordpress.com/2010/06/21/55 offloaded] to the cpu and ''increase'' your performance. Because whatever the obvious overhead this operation has, it was still faster than tasking the potato gpu with yet another operation (or maybe vertex processing is just straightaway quicker on a 200€ cpu of the time, compared to a 2€ graphics adapter). |
Anyway, for these reasons the driver has a registry whitelist codenamed "Enable3DContexts" (I don't even want to think instead to that can of software worms Gen3 was). In my testing with a X4500 and the latest [https://web.archive.org/web/20160608190242/http://downloadmirror.intel.com/22166/a08/Win7Vista_64_151719.zip 2869] stock drivers, I found out that after [http://trixarian.net/reg_modders.zip setting] Mass Effect exe to ''1'', frametimes improved by more than 10% (benchmarks and a bunch of newer applications seemed to have the [https://web.archive.org/web/20130411234401/http://9xxssf.info/topic/33-igrm/ mysterious] value of ''2'' OOTB, mileage may vary on what's the best one). | Anyway, for these reasons the driver has a registry whitelist codenamed "Enable3DContexts" (I don't even want to think instead to that can of software worms Gen3 was). In my testing with a X4500 and the latest [https://web.archive.org/web/20160608190242/http://downloadmirror.intel.com/22166/a08/Win7Vista_64_151719.zip 2869] stock drivers, I found out that after [http://trixarian.net/reg_modders.zip setting] Mass Effect exe to ''1'', frametimes improved by more than 10% (benchmarks and a bunch of newer applications seemed to have the [https://web.archive.org/web/20130411234401/http://9xxssf.info/topic/33-igrm/ mysterious] value of ''2'' OOTB, mileage may vary on what's the best one). |