Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[TheBoy]

I get confussed easy.  When I started reading the thread I was going to mention that a lot more boards are now coming out with DDR3 (my latest board has 4x DDR2 and 2x DDR3)  so therefore would it be worth considering the DDR3 againt cost of purchase and appliction/performance.

I am still running DDR2 till DDR3 prices settle.

But after seeing the chart that Cem listed I am not sure anymore. :-/

Same as dual AGP/PCIe boards, you get a compromise between the 2, rather than max performance from one.

[PaulW]

My rule of thumb is always have 2gig per CPU core, so dual core will be 4gig, quad core 8gig.

XP32 won't support more than 3gb tho.

thats because linux needs it

Linux memory management is awful.  Actually, Linux resource management is awful (non existent)

does hell   *never* had any memory issues on both of my laptops, desktop or server. (admittedly I did have some issues when running Debian to begin with, but that was quite some time ago)

makes more sense to match memory sizes per core than mix & match, purely on a performance level!

had 3gig in my laptop when bought, since upping to 4gig it is noticeably faster.  Even then, memory usage with just 3gig there was only around 1gig with no swap during normal running, at most, 2.5gig used during compiling big apps (even then swap was hardly used)

Noticed similiar performance increases on my brothers desktop (dual core) running Vista64, he had 3gig prior, upped to 4gig.

[TheBoy]

My rule of thumb is always have 2gig per CPU core, so dual core will be 4gig, quad core 8gig.

XP32 won't support more than 3gb tho.

thats because linux needs it

Linux memory management is awful.  Actually, Linux resource management is awful (non existent)

does hell   *never* had any memory issues on both of my laptops, desktop or server. (admittedly I did have some issues when running Debian to begin with, but that was quite some time ago)

makes more sense to match memory sizes per core than mix & match, purely on a performance level!

had 3gig in my laptop when bought, since upping to 4gig it is noticeably faster.  Even then, memory usage with just 3gig there was only around 1gig with no swap during normal running, at most, 2.5gig used during compiling big apps (even then swap was hardly used)

Noticed similiar performance increases on my brothers desktop (dual core) running Vista64, he had 3gig prior, upped to 4gig.

The current poor resource management in Linux is one of the inhibiting factors for it to become more than just a poor man's webserver.  I think the most we trust Linux with is DNS, servicing about 80,000 queries per second, but we've had to put more servers in since moving from proper Unix to Linux to handle the load (even though the hardware is newer and faster). We trusted it with News, but hacked too often.


On every x86/x64 architecture I have seen, there is no performance reason to match memory to cores. In fact, on any chipset I have used, there is not that level of ability.


As to using 2.5Gb running little more than a compiler, that is disgraceful , makes even Visual Studio look efficient on memory usage