Welcome to the MacNN Forums.

booboo · Apr 4, 2003, 04:45 PM

Sorry if I've been asking a lot of questions lately...

But here's one more :-)

PC133 CL3 is apparently OK for my Dual 533 but is there any performance advantage in using CL2 RAM?

proux · Apr 4, 2003, 08:09 PM

People have said 5% max in some applications, but overall probably no noticable difference

bradoesch · Apr 4, 2003, 09:19 PM

These numbers designate the latency of the RAM. CL2 uses 2 clock cycles until the RAM is ready to be read/written to, and of course CL3 uses 3 clock cycles. My iMac for example has a 400 MHz processor. So it does 400 million clock cycles every second. So, what's the difference between taking 2 or 3 clock cycles? I don't think I could tell the difference, but I don't know for sure.

Brad

schalliol · Apr 5, 2003, 12:57 AM

Well the percentage between 2 & 3 clock cycles is pretty big.

booboo · Apr 5, 2003, 04:45 AM

OK, so is the consensus that the cpu access the RAM as fast as the RAM will allow - i.e. dependent on the RAM speed, rather than at a pre-determined rate?

Uh?

I say that because all these little '5%' boosts here and there probably add up to something worthwhile...

bradoesch · Apr 5, 2003, 12:05 PM

Originally posted by schalliol:
Well the percentage between 2 & 3 clock cycles is pretty big.

I don't understand what you mean. Can you explain this further?

schalliol · Apr 5, 2003, 02:20 PM

Originally posted by bradoesch:
...it does 400 million clock cycles every second. So, what's the difference between taking 2 or 3 clock cycles?
Brad

If what Brad's saying is true, it would go to the RAM 200 times per second in the 2 cycle case where as it would only go to the RAM 133.3 times per second in the 3 cycle case. Or only 60% of the 2 cycle case.

bradoesch · Apr 5, 2003, 04:08 PM

Originally posted by schalliol:
If what Brad's saying is true, it would go to the RAM 200 times per second in the 2 cycle case where as it would only go to the RAM 133.3 times per second in the 3 cycle case. Or only 60% of the 2 cycle case.

I think I see where I've confused people now. The latency of the RAM only matters right before you're going to access the RAM. Once it's being written/read, the latency makes no difference. Sorta like the time it takes to turn on the tap. The difference between it taking 2 seconds of 3 seconds to turn the tap on makes little difference if you're gonna fill a pitcher of water.

schalliol · Apr 5, 2003, 06:20 PM

I gotcha, so you just have to wait a little big longer to get into the RAM. I could see how that could be about 5% notice.

bradoesch · Apr 6, 2003, 04:25 PM

Originally posted by schalliol:
I gotcha, so you just have to wait a little big longer to get into the RAM. I could see how that could be about 5% notice.

Exactly.

booboo · Apr 7, 2003, 08:42 AM

Originally posted by bradoesch:
I think I see where I've confused people now. The latency of the RAM only matters right before you're going to access the RAM. Once it's being written/read, the latency makes no difference. Sorta like the time it takes to turn on the tap. The difference between it taking 2 seconds of 3 seconds to turn the tap on makes little difference if you're gonna fill a pitcher of water.

So if something is read from memory - even if it is 200MB this counts as 1 access...?

But what about if the computer is running a lot of background processes, each of these will be accessing RAM, during the above 200MB access, so there will be cumulative latencies - or am I just making a mountain out of molehills?

bradoesch · Apr 7, 2003, 09:03 PM

Originally posted by booboo:
So if something is read from memory - even if it is 200MB this counts as 1 access...?

That's correct.

Originally posted by booboo:

But what about if the computer is running a lot of background processes, each of these will be accessing RAM, during the above 200MB access, so there will be cumulative latencies - or am I just making a mountain out of molehills?

As far as I know, there isn't cumulative latencies. Once the RAM is in use then it all operates the same.