the ram...why so slow.

[Red*Fox]

why are these companies like corsair making memory so slow, i hearn the memory runs at 800 mhz...is this true? and if it is true why not make the memory like 2-3 ghz for servers?

[fos]

It would turn your computer into a radio generating microwave oven, for one thing.

fos....

[bhobjj]

And it is not just microwaves.
One of the challenges of microchip production is to filter out various frequences that can Interfere with other devices. There are standards for certain radio frequencies that these chips are not allowed any detecable emissions. There are lots of trade secrets and patents on filtering for chips.

[Red*Fox]

hmm my phone runs at 5.8 ghz and my wireless at 2.4 ghz

im sure it would help seervers alot to increase the speed of ram, itd probably help alot...lol my coumuters are already microwaves :p

[AndreL]

hmm my phone runs at 2.1 ghz and my wireless at 2.151 ghz :pActually, it's 2.4GHz for both. That is why they now have 5.8GHz phones.

[Red*Fox]

ok, so why do they let processors run at 2 -4 and they wont let ram run at 2-3 ghz? im sure they'd put off alot of heat but im sure itd be worth

lol i e-mailed corsair and they said almost exactly what fos said " Because itll turn your computer into a microwave" :p

[AndreL]

ok, so why do they let processors run at 2 -4 and they wont let ram run at 2-3 ghz? im sure they'd put off alot of heat but im sure itd be worth it.I guess that might be why... Processors have fans, RAM chips do not. Some have heatsinks now. Maybe someday we'll have RAM blocks with fan... Then we can all go back to those thinner desktop cases, where the components are closer to the top, and keep our coffee warm in the morning and maybe fry a few eggs along the way. :smiley4:

[Red*Fox]

they already make watercooling for ram : http://www.koolance.com/shop/product_info.php?cPath=29_56&products_id=251

ima buy a watercooling kit this cristmas when i get some money.

[danieldk]

RAM clock speed and CPU clock speed are different things, they are not an indication for speed.

E.g., if a CPU is 1 MHz it means that it has one million clock cycles per second. Traditionally one instruction could be executed per clock cycle, meaning that a CPU could execute one million instructions per second. To give you one example why the clock frequency is not a good indication of speed: some operations can be done by a CISC CPU in one cycle, taking more cycles on a RISC CPU. You can do the math, this will blast the whole megahertz measurement, even if some operation only takes two instructions on a CISC CPU and three on a RISC CPU.

Now, why can't the CPU clock be compared with a memory clock? First of all, most modern memory modules transfer 64 bits in one cycle. Second, modern DDR chips can transmit data both on the rising and falling edges of a cycle (effectively making a 100MHz DDR chip perform like a 200MHz SDRAM module). Third, boosting the clock frequency does not help much if memory latency is the bottleneck (often it is).

PS. Of course, CPUs and memory have short frequency radiation, but that's filtered out very well. See http://ask.slashdot.org/article.pl?sid=05/07/02/203218 for an interesting discussion.

[fos]

146.76 Mhz is a common harmonic produced by personal computers. The background noise caused by this harmonic is considerable when using communications on this frequency.

fos....

[Red*Fox]

good info thanks all:smiley4:

[tom_servo]

Computer designers have come up with neat tricks to work around the relative slowness of RAM, like wider data bus to RAM (dual-channel), and big cache in the CPU (which is much closer, and is SRAM, so it can therefore be clocked closer to the main CPU clock speed).

There are 2 big reasons that RAM isn't lots faster then it is. The first is that static RAM (SRAM) is fast (in some CPUs the static RAM cache runs at the CPU clock speed, or 1/2 the CPU clock speed), but requires at least 4 transistors (or more for some designs, like dual-ported) per bit, while SDRAM (synchronous dynamic RAM, what you usually think of as RAM in your computer) is slower, but usually has 1 transistor, and one capacitor per bit. The disadvantage is that since the capacitors loose charge over time, the data in a bit of DRAM must be read, then written back to that bit periodically (usually more then 10 times a second). This makes it somewhat slower (and higher power usage), but cheaper per byte. As caching can usually hide this slowness pretty well, you usually (depending on the type of job being done) end up ahead in price/performance with SDRAM, and an SRAM cache.

The other problem is that even at the relatively slow several hundred Mhz that RAM runs at, the physical distance between it and the CPU is huge. On my motherboard the traces (wires embedded in the circuit board) from the CPU to the RAM (and some other components) have an interesting zig-zag pattern in them on the inside of turns. These inside traces would be shorter then the traces on the outside of turns if they didn't have this zig-zaging, and so signals would arrive on these traces first. The timing can be that touchy. It amazes me that computers work at all. :)



Alain