From: weick@master.lds-az.loral.com (Dana Weick) Subject: Re: 70ns or 60ns RAM with a 486/66: does it matter? Date: Fri, 9 Jul 1993 22:54:19 GMT
In article <1993Jul9.183937.27460@mnemosyne.cs.du.edu> zevans@nyx.cs.du.edu (Zack Evans) writes:
>>In article <mr2C9rpJC.1rv@netcom.com> mr2@netcom.com (Jongyoon Lee) writes:
>>>Chris Metcalf (metcalf@CATFISH.LCS.MIT.EDU) wrote:
>>>: I'm about to buy a 486/66 with 16M of RAM, 256K 20ns cache. Is it worth $50
>>>: or so to get 60ns RAM instead of 70ns RAM? I'll summarize in a day or two.
>>>
>>>It depends on the motherboard. If the motherboard can be configured
>>>for different speed of SIMMs, you might benefit from faster memory.
>>>However, a lot of motherboards require only 70ns or 80ns SIMMs. I use
>>>a 486DX-66 and it requires 80ns SIMMs. So I can use my 4x9@70ns SIMMs
>>>safely.
>>
>>Hold on...if the motherboard is running at 33Mhz, then each tick is
>>reciprocal 33Mhz seconds long, right? That's 30.030303... ns as far as
>>I can see, which means there will always be an advantage to faster RAM
>>until you start comparing 30ns with 29ns SIMMs. (I know that there
>>aren't many instances where a CISC chip will want to read from memory
>>on two sucessive ticks, but the principal is the same, I think).
A 60ns DRAM can't be accessed every 60ns.
A 33Mhz 80486 CPU frequently requests data on two sucessive clock ticks.
Even if 30ns DRAM *could* be accessed once every 30ns it would not
be possible to obtain zero wait states on a 33Mhz 80486 motherboard.
Little things like address decoding and transmission delays get in