Difference in 'Dimm Type'...

JoeWells

Member
Aug 30, 2000
27
0
0
I was over at Crucial.com looking at some of the lastest memory prices and seeing if I wanted to upgrade and I noticed that they have 2 different DIMM Types for each memory simm. For instance, they have two different listings for PC 133 memory to fit my board. The only difference between them is one says it is Unbufferred and one says it's Registered under DIMM type.

What does that mean and what's the difference? Which one is better? I also noticed that there are listings for memory that are CL=2 and CL=3, which one of these is better?


As a side note question, I'm big into gaming so I always assume that the more memory the better. But is there a limit you can reach where adding more won't really make that big a difference in game performance? For instance, I have 256 megs of RAM. Would going up to 512 megs really make my games run that much better to justify the upgrade?

Thanks for the knowledge,

Joe Wells
 

TheCorm

Diamond Member
Nov 5, 2000
4,326
0
0
Unbuffered supposedly can have more errors that registered, i'm not sure how or why :confused:

I think mine is unbuffered, never have memory related problems.

CL Is to do with Latency, the length of time to access information in the memory, I'm not sure about those ratings though.

Regarding the amount of memory, there aren't really applications or games around at the moment that require that much memory, 128MB is adequate and 256 is plenty for a hardcore gamer, save your money instead, what Graphics card do you have? an upgrade there would increase game speed the most.

Corm
 

medic

Diamond Member
Oct 9, 1999
5,160
0
0
As far as CAS2 or CAS3, CAS2 will be faster slightly and you should also check into the ns. rating of the actual chips used on the module.

pm had a good answer to the other question a few weeks ago:

Registered SDRAM (aka "buffered" SDRAM - the terms are interchangeable when used to describe SDRAM) are intended for workstations and servers. Basically all pins except the data and the clock pins) are buffered through a register. This reduces the loading by "buffering" the address and control inputs. The main advantage is that you can put more than 4 DIMMs on a memory controller - up to 8, IIRC. There's a one clock latency penalty (from the register), but sometimes this can be masked.

ECC SDRAM is a whole different thing. ECC is a type of fault detection/correction circuit that is used to detect and fix memory errors on the fly. Alpha and Beta particle impact ionization in a memory bit can cause a bit flip (0->1 or 1->0) which causes a memory error. With the old parity memory, your system could detect that a bit had bit flipped on memory read, but would then print "Parity Error" and freeze up - which prevents this bad data from being written, but erases all current data. ECC can determine one flipped bit in 64-bits and can detect an error in two bits (which causes a parity error shutdown as before).

Registered memory is definitely intended for servers/workstations or systems with a lot of RAM (>2GB). ECC is argueably intended for anyone who cares a lot about data integrity on their system - which is not necessarily someone on a workstation.
 

Noriaki

Lifer
Jun 3, 2000
13,640
1
71
CL stands for CAS Latency and since it's a latency lower is better.

TheCorm is right, 256MB is plenty, if you run Win2000 you might notice a differnt up to 384MB...but in Win98/ME 512 won't make any difference over 256MB.

Registered and Buffered mean the same thing. And it has nothing to do with Error handling (that's ECC) registered or buffered DIMMs have a register on them, it's an extra step in data arbitration, it lets you have many many many DIMMs on a board. (Seen those server boards with 16 DIMM slots? They can have so many thanks to registered/buffered DIMMs)
Unless you specifically know that you want registered, chances are about 98% that you don't. You want unbuffered/unregistered "regular" SDRAM.

Since you'll probably ask what ECC is next it's Error Checking and Correction it allows to correct single bit errors in transfer and detect dobule bit errors. It's good for making sure you don't get errors in your data.

For your main PC you don't really need ECC (it's for like mission critical things, your chances of getting a bad data transfer are small..but small isn't good enough in some cases) and you almost definately don't need registered/buffered.
 

Neoplasia

Member
Dec 8, 2000
169
0
0
Thanks. I new CAS2 was faster than CAS3, I just wasn't positive it it meant the same as CL2 and CL3. Fwiw I'm getting a couple sticks of Corsair CAS2, not skimping on quality this time 'round.
 

medic

Diamond Member
Oct 9, 1999
5,160
0
0
Registered and Buffered mean the same thing but they go about it a little differently. I had a site somewhere that had a good explanation of it (will look for it) But basically buffered is an older tech used mainly on EDO and FPM modules for servers mainly and it was needed as the older chipsets couldn't access enough memory without the buffers.
Registered is a much faster and newer version and is connected to the system clock, transfering data (I believe) on the rising edge of the clock cycle to at least 2 registers for a 128MB module to 3 for 256MB and above.