Jump to content
165 posts in this topic

Recommended Posts

Are you ready to rock? .. I hope :dance_24:

 

Just another version of "Nebukadnezar" hit the scene. Now supports 100% correct clocking with any CPU, the sound issues are solved, video too and it runs tight on any AMD and INTEL CPU.

No longer FSB setup since it simply runs totally new and advanced techniques on clocking. Old {censored} code needed this value was replaced with whole new one which are fast, more compatible, offering you a accuracy of ~1Hz.

 

Grap a copy now from the IRC channel. :(

 

overview of other improvments in "Nebu" so far:

 

VMWare support

VMware is supported on AMD without necessary boot options, with Intel machines you need to use -vmware bootoption to force the AMD/Vmware code to be run.

 

FULL AMD support with GUI

The new kernel provides everything needed to run a full featured OSx86 on a AMD SSE3 compatible processor. To run GUI you need to get the rufus patched bins, they include special opcode for a function in kernel which enables the AMD support in them.

 

free FSB selection

Because of the heavy demand from the overclocking and AMD64 people I added a free selectable fsb option and skipped the conecpt of only common values as bootflags. You can still use the old ones if you have them already in use.

 

CPU autodetection

Kernel detect now if you run AMD or INTEL and will select proper code for each manufactor. No need for diffrent kernels anymore.

 

Multiple Core Problems fixed

Some machine crashed because of wrong detection of cores available, this issue has been fixed.

 

Video Problem solved

We solved a minor issue with video playback on AMD but this is still experimental.

 

hope you enjoy it! And there is still space for improvements. :D

Just finished a test with the latest - vv - kernel. I ran GeekBench on both my current system and the one in the test partition. There are some interesting speed differences.

 

current system - Jas-10.4.7-install + jas-10.4.8-combo-update:

System Information

Geekbench Version: Geekbench 2006 (build 230)

Geekbench Platform: Mac OS X x86 (32-bit)

Geekbench Compiler: GCC 4.0.1 (Apple Computer, Inc. build 5363)

OS: Mac OS X 10.4.8 (Build 8L2127)

Model: Developer Transition Kit

Motherboard: ADP2,1

Processor: Intel® Pentium® 4 CPU 3.00GHz

Processor ID: GenuineIntel Family 15 Model 4 Stepping 1

Logical Processor Count: 2

Physical Processor Count: 2

Processor Frequency: 2990 MHz

Bus Frequency: 400 MHz

Memory: 1536 MB

 

Integer Performance

Emulate 6502

single-threaded scalar 105.4 (rate: 1.0, result: 199.3 MHz)

multi-threaded scalar 93.9 (rate: 0.9, result: 177.5 MHz)

Blowfish

single-threaded scalar 239.4 (rate: 1.0, result: 98.8 MB/sec)

multi-threaded scalar 282.8 (rate: 1.2, result: 116.6 MB/sec)

bzip2 Compress

single-threaded scalar 174.0 (rate: 1.0, result: 27.1 MB/sec)

multi-threaded scalar 211.8 (rate: 1.2, result: 32.8 MB/sec)

bzip2 Decompress

single-threaded scalar 175.3 (rate: 1.0, result: 65.2 MB/sec)

multi-threaded scalar 214.6 (rate: 1.2, result: 77.3 MB/sec)

 

Floating Point Performance

Mandelbrot

single-threaded scalar 112.0 (rate: 1.0, result: 793.7 Mflops)

multi-threaded scalar 213.5 (rate: 1.9, result: 1.5 Gflops)

 

This result is from the test system build according to the posted test build instructions + vv:

System Information

Geekbench Version: Geekbench 2006 (build 230)

Geekbench Platform: Mac OS X x86 (32-bit)

Geekbench Compiler: GCC 4.0.1 (Apple Computer, Inc. build 5363)

OS: Mac OS X 10.4.8 (Build 8L2127)

Model: ACPI

Motherboard: ACPI

Processor: Intel® Pentium® 4 CPU 3.00GHz

Processor ID: GenuineIntel Family 15 Model 4 Stepping 1

Logical Processor Count: 2

Physical Processor Count: 1

Processor Frequency: 3000 MHz

Bus Frequency: 400 MHz

Memory: 1536 MB

 

Integer Performance

Emulate 6502

single-threaded scalar 51.3 (rate: 1.0, result: 96.9 MHz)

multi-threaded scalar 78.4 (rate: 1.5, result: 148.1 MHz)

Blowfish

single-threaded scalar 126.9 (rate: 1.0, result: 52.3 MB/sec)

multi-threaded scalar 237.2 (rate: 1.9, result: 97.8 MB/sec)

bzip2 Compress

single-threaded scalar 88.2 (rate: 1.0, result: 13.8 MB/sec)

multi-threaded scalar 180.3 (rate: 2.0, result: 28.0 MB/sec)

bzip2 Decompress

single-threaded scalar 102.5 (rate: 1.0, result: 38.1 MB/sec)

multi-threaded scalar 193.5 (rate: 1.8, result: 69.7 MB/sec)

 

Floating Point Performance

Mandelbrot

single-threaded scalar 87.6 (rate: 1.0, result: 621.0 Mflops)

multi-threaded scalar 179.3 (rate: 2.0, result: 1.3 Gflops)

Just finished a test with the latest - vv - kernel. I ran GeekBench on both my current system and the one in the test partition. There are some interesting speed differences.

 

[...]

 

we encountred false results by geekbench earlier, try xbench. Also these result might relay on diffrences in timing which now works a whole other way and relays on exact linear hardware values.

 

We tested with diffrent benchmarks on serveral system and the speed is just fine and much better than 10.4.4.

 

Another thing to remmber, don't compare earlier builds with mixed up clock and that one, they will give you incredible results.

 

For further questions contact me on IRC ;)

we encountred false results by geekbench earlier, try xbench. Also these result might relay on diffrences in timing which now works a whole other way and relays on exact linear hardware values.

 

We tested with diffrent benchmarks on serveral system and the speed is just fine and much better than 10.4.4.

 

Another thing to remmber, don't compare earlier builds with mixed up clock and that one, they will give you incredible results.

 

For further questions contact me on IRC :D

It is to laugh - I mentioned the differences I saw with xbench on the kernel irc and someone told me to ignore them and use GeekBench :2cents:

It is to laugh - I mentioned the differences I saw with xbench on the kernel irc and someone told me to ignore them and use GeekBench :2cents:

 

Lol, yeah but is absicly a clockin issue you can't compare diffrent clocked builds, I pushed bench results yesterday but just speeding the clock a few mili seconds, but who wants faked results? With a little manipulation we could speed up results from 100-200 up to 800-900 with any noticeable time diffrence but this is not for real and simply useless :D

Ok I am raring the new AMD SSE3 BETA iso now and will be releasing it to testers,this build includes this kernel with all the fixes :( Enjoy all you lucky testers.

new 4.5GB release ?

or a ppf small patch to this one came out at 4of november? (yours)

cause many ppl donwloaded the intel 10.4.8 hope for a ppf patch

mine is at 70% it should finish tommorow

keep up the good work

Lol, yeah but is absicly a clockin issue you can't compare diffrent clocked builds, I pushed bench results yesterday but just speeding the clock a few mili seconds, but who wants faked results? With a little manipulation we could speed up results from 100-200 up to 800-900 with any noticeable time diffrence but this is not for real and simply useless :(

 

Ya, I don't really care. It was just something interesting to look at and I posted it in case someone who knew might find it useful. I can't tell any difference in the performance by just using it.

 

On the progress side. My display resolution problem is now fixed. I have a Radeon 9250 grafics card which should use the ATIRadeon8500.kext. I noticed that the GMA9500 was loading in it's place. I trashed the GMA9500 extensions and copied the radeon8500 and Callisto extensions from my main system. Now I have display resolutions adjustable from System Preferences -> Displays. great joy abounds.

to semthex

i joined the chat on irc but too many complicated discussion between advanced users.

i'm a beginner and would like to know how to do with intel platforms, i.e. the setup of a test system is explained for amd, but does it apply to intel?

if anyother with intel cpu can send me a pm in order to keep clean this 3d, i would be happy and not disturb the more advanced discussion (i will learn myself in i have only some tips).

Thank you and bye.

Hi, no problem it's currently high tide in there. If you want detailed discribtion how to setup, just read the FAQ and the linked guide to set everything up.

Stay a while inthe channel and I am sure you will get all needed infos very fast :-)

Thanks to those involved with this project. I've been pondering on how much millage SSE2 users will actually have when an SSE2 kernel is available. It would be a task to have to hack upcoming applications so that it can run on an SSE2 chip, not to mention the stability of applications when the translator is available.

 

I'm thinking about sticking with 10.4.7 on my SSE2 machine, buying a 965 board and C2D and install 10.4.8 on it. If anyone can give us a reason for being optimistic on the SSE2 kernel, please let us hear about it.

Definitely a sick job, Semthex!

 

I'm hard pressed to find anything that's fubar yet. Clock is flawless, sound is flawless, speed is amazing, Titan works great... just wow.

 

Now all I have left to do is resize my partitions so that this one has 30GB and the old one has 10GB.

if i dont have IRC, where can i download this new kernel. I am eager to test.

 

i hope someone could email it to me or something.

Go to IRC channel #kernel to get the link.

 

Click on the Chat link in the title bar above and then click on the Forum link to have a Java IRC client take you to IRC.

Guest
This topic is now closed to further replies.
×
×
  • Create New...