Meaning: For 32-Bit OS it doesn't make a difference whether you use discrete graphics or shared memory. Having 4GB installed on X3100 enquipped XPS M1330 results in 3,58GB usable memory with latest BIOS versions if I'm correct. This is wrong for 32-Bit non PAE System (= most systems including Vista 32 Bit). It has to take RAM from what the OS can see, not the unused portion. Interesting for you is that it's certainly taken from main memory, but you don't end up having less memory avaible compared to somebody with a discrete graphics card such as GeForce 8400M GS in 32 Bit wrote: Unless you are experienced enough to know where to look for to tell the difference.ĭepending on that, it will take the memory for the video card from the OS or from the physical before the OS)? This doesn't work for Windows Vista: Since SP1 your system will always report 4GB if you install 4GB regardless whether you can access that. The easiest way to tell whether your BIOS support it: Install 4GB of memory and boot Ubuntu 64-Bit Live DVD. The 64-Bit build will run away regarding performance if you don't have SSE2 for 32-Bit and have no source code to recompile using gcc ' -msse2 -mfpmath=sse' switch.īy now, it isnt clear just about the chipset, if it supports or not 4 Gb and memory remapping. If your scientific floating point number application has a SSE2 build, then 32-Bit and 64-Bit OS and application will perform equally fast as well. If your application is only available as 32-Bit build, the performance on 32-Bit and 64-Bit will perform equally fast. In practise this means for applications using the floating point numbers: The performance gain is tremendous and can be done on both 32 and 64 Bit architectures. In fact two additions (= 2x2 of those 64-Bit numbers) can be added by a single instruction: However since SSE2 those 64-Bit double precision floating point numbers can be computated inside the 128-Bit long SSE2 registers. 32-Bit compilers default to generate code for the x87 floating point unit, which is stack based and rather slow.

The interesting and most used float point format in scientific calucations is 64-Bit double precision. I've used Matlab + Simulink and Maple as well, however I don't know whether there is a SSE2 version of those applications. Unless you use very special software, you'll loose performance. Well, only IF there is a performance gain (and not a loss) at all. Ubuntu 64-Bit is not only the simpler case, the performance gain 32 -> 64-Bit Ubuntu is bigger than going Windows 32 -> 64Bit. I will try Ubuntu 64-bt also, good to know that it is a simpler case! Some of the softwares I need ( like Matlab, Maple) have 64-bit version, but I believe that it will be a headache for other compabilities. I will use 32-bit OS for now, I am quite short in time to play around with this right now, but I plan to test 64-bit OS.

Intel gma 3100 vista 32 driver