General discussion
-
Topic
-
RAM – Stuck in a rut?
LockedThe state of the art in desktop PC’s today calls for processor speeds in the multiple GHz, and bus speeds in excess of 300 MHz, with 400MHz being fairly typical.
So why the hell are chipset designers STILL using dynamic RAM (D-RAM)?!! Even with DDR memory using both edges of the clock, valuable cycles are still being wasted to refresh the memory. L1 and L2 cache memory uses Static RAM, which is very much faster than D-RAM and doesn’t require refresh cycles.
On a typical system from 5 years ago, running a 200MHz socket 7 CPU, 128kb of cache was pretty common on many motherboards. Of course the graphics card back then would have had 2Mb or if you were rich, 4 or 8Mb. System RAM would have been typically 32 or 64Mb running 98SE.
I have a graphics card in my newest machine which has four times as much video memory as the system ram in my old computer, yet price wise, the relative costs of the hardware 5 years ago weren’t much different. S-RAM is perhaps 5% more expensive to produce than D-RAM, but surely the performance advantage (better than 30%) justifies the cost?
We won’t see much improvement (compared to the last few years) in terms of bus speeds and processor speeds in the future until a completely different technology is used to produce systems – as the sheer physics of semiconductors and the law of diminishing returns will soon converge at a maximum upper limit, but by redesigning chipsets to use faster S-RAM as system memory instead of slower D-RAM, performance of typical desktop systems can be hiked by a third for a very small outlay in terms of memory cost.
I have always held the belief that computers could be much better and faster and more ‘clever’ than they are today if designers threw away the rule book and started from first principles, rather than just keep racking up a few extra MHz on top of the original (and archaic) IBM pc architecture (and segmented memory model) of over twenty years ago.
Discuss!
