Intro
The type of RAM doesn't matter nearly as much as how much of it you've got, but using plain old SDRAM memory today will slow you down. There are three main types of RAM: SDRAM, DDR and Rambus DRAM.
SDRAM (Synchronous DRAM)
Almost all systems used to ship with 3.3 volt, 168-pin SDRAM DIMMs. SDRAM is not an extension of older EDO DRAM but a new type of DRAM altogether. SDRAM started out running at 66 MHz, while older fast page mode DRAM and EDO max out at 50 MHz. SDRAM is able to scale to 133 MHz (PC133) officially, and unofficially up to 180MHz or higher. As processors get faster, new generations of memory such as DDR and RDRAM are required to get proper performance.
DDR (Double Data Rate SDRAM)
DDR basically doubles the rate of data transfer of standard SDRAM by transferring data on the up and down tick of a clock cycle. DDR memory operating at 333MHz actually operates at 166MHz * 2 (aka PC333 / PC2700) or 133MHz*2 (PC266 / PC2100). DDR is a 2.5 volt technology that uses 184 pins in its DIMMs. It is incompatible with SDRAM physically, but uses a similar parallel bus, making it easier to implement than RDRAM, which is a different technology.
Despite it's higher price, Intel has given RDRAM it's blessing for the consumer market, and it will be the sole choice of memory for Intel's Pentium 4. RDRAM is a serial memory technology that arrived in three flavors, PC600, PC700, and PC800. PC800 RDRAM has double the maximum throughput of old PC100 SDRAM, but a higher latency. RDRAM designs with multiple channels, such as those in Pentium 4 motherboards, are currently at the top of the heap in memory throughput, especially when paired with PC1066 RDRAM memory.
DIMMs vs. RIMMs
DRAM comes in two major form factors: DIMMs and RIMMS.
DIMMs are 64-bit components, but if used in a motherboard with a dual-channel configuration (like with an Nvidia nForce chipset) you must pair them to get maximum performance. So far there aren't many DDR chipset that use dual-channels. Typically, if you want to add 512 MB of DIMM memory to your machine, you just pop in a 512 MB DIMM if you've got an available slot. DIMMs for SDRAM and DDR are different, and not physically compatible. SDRAM DIMMs have 168-pins and run at 3.3 volts, while DDR DIMMs have 184-pins and run at 2.5 volts.
RIMMs use only a 16-bit interface but run at higher speeds than DDR. To get maximum performance, Intel RDRAM chipsets require the use of RIMMs in pairs over a dual-channel 32-bit interface. You have to plan more when upgrading and purchasing RDRAM.
Memory Speed
SDRAM initially shipped at a speed of 66MHz. As memory buses got faster, it was pumped up to 100MHz, and then 133MHz. The speed grades are referred to as PC66 (unofficially), PC100 and PC133 SDRAM respectively. Some manufacturers are shipping a PC150 speed grade. However, this is an unofficial speed rating, and of little use unless you plan to overclock your system.
DDR comes in PC1600, PC2100, PC2700 and PC3200 DIMMs. A PC1600 DIMM is made up of PC200 DDR chips, while a PC2100 DIMM is made up of PC266 chips. PC2700 uses PC333 DDR chips and PC3200 uses PC400 chips that haven't gained widespread support. Go for PC2700 DDR. It is about the cost of PC2100 memory and will give you better performance.
RDRAM comes in PC600, PC700, PC800 and PC1066 speeds. Go for PC1066 RDRAM if you can find it. If you can't, PC800 RDRAM is widely available.
CAS Latency
SDRAM comes with latency ratings or "CAS (Column Address Strobe) latency" ratings. Standard PC100 / PC133 SDRAM comes in CAS 2 or CAS 3 speed ratings. The lower latency of CAS 2 memory will give you more performance. It also costs a bit more, but it's worth it.
DDR memory comes in CAS 2 and CAS 2.5 ratings, with CAS 2 costing more and performing better.
RDRAM has no CAS latency ratings, but may eventually come in 32 and 4 bank forms with 32-bank RDRAM costing more and performing better. For now, it's all 32-bank RDRAM.
Understanding Cache
Cache Memory is fast memory that serves as a buffer between the processor and main memory. The cache holds data that was recently used by the processor and saves a trip all the way back to slower main memory. The memory structure of PCs is often thought of as just main memory, but it's really a five or six level structure:
The first two levels of memory are contained in the processor itself, consisting of the processor's small internal memory, or registers, and L1 cache, which is the first level of cache, usually contained in the processor.
The third level of memory is the L2 cache, usually contained on the motherboard. However, the Celeron chip from Intel actually contains 128K of L2 cache within the form factor of the chip. More and more chip makers are planning to put this cache on board the processor itself. The benefit is that it will then run at the same speed as the processor, and cost less to put on the chip than to set up a bus and logic externally from the processor.
The fourth level, is being referred to as L3 cache. This cache used to be the L2 cache on the motherboard, but now that some processors include L1 and L2 cache on the chip, it becomes L3 cache. Usually, it runs slower than the processor, but faster than main memory.
The fifth level (or fourth if you have no "L3 cache") of memory is the main memory itself.
The sixth level is a piece of the hard disk used by the Operating System, usually called virtual memory. Most operating systems use this when they run out of main memory, but some use it in other ways as well.
This six-tiered structure is designed to efficiently speed data to the processor when it needs it, and also to allow the operating system to function when levels of main memory are low. You might ask, "Why is all this necessary?" The answer is cost. If there were one type of super-fast, super-cheap memory, it could theoretically satisfy the needs of this entire memory architecture. This will probably never happen since you don't need very much cache memory to drastically improve performance, and there will always be a faster, more expensive alternative to the current form of main memory.
Memory Redundancy
One important aspect to consider in memory is what level of redundancy you want. There are a few different levels of redundancy available in memory. Depending on your motherboard, it may support all or some of these types of memory:
The cheapest and most prevalent level of redundancy is non-parity memory. When you have non-parity memory in your machine and it encounters a memory error, the operating system will have no way of knowing and will most likely crash, but could corrupt data as well with no way of telling the OS. This is the most common type of memory, and unless specified, that's what you're getting. It works fine for most applications, but I wouldn't run life support systems on it.
The second level of redundancy is parity memory (also called true parity). Parity memory has extra chips that act as parity chips. Thus, the chip will be able to detect when a memory error has occurred and signal the operating system. You'll probably still crash, but at least you'll know why.
The third level of redundancy is ECC (Error Checking and Correcting). This requires even more logic and is usually more expensive. Not only does it detect memory errors, but it also corrects 1-bit ECC errors. If you have a 2-bit error, you will still have some problems. Some motherboards enable you to have ECC memory.
Older memory types
Fast Page Mode DRAM
Fast Page Mode DRAM is plain old DRAM as we once knew it. The problem with standard DRAM was that it maxes out at about 50 MHz.
EDO DRAM
EDO DRAM gave people up to 5% system performance increase over DRAM. EDO DRAM is like FPM DRAM with some cache built into the chip. Like FPM DRAM, EDO DRAM maxes out at about 50 MHz. Early on, some system makers claimed that if you used EDO DRAM you didn't need L2 cache in your computer to get decent performance. They were wrong. It turns out that EDO DRAM works along with L2 cache to make things even faster, but if you lose the L2 cache, you lose a lot of speed.
The type of RAM doesn't matter nearly as much as how much of it you've got, but using plain old SDRAM memory today will slow you down. There are three main types of RAM: SDRAM, DDR and Rambus DRAM.
SDRAM (Synchronous DRAM)
Almost all systems used to ship with 3.3 volt, 168-pin SDRAM DIMMs. SDRAM is not an extension of older EDO DRAM but a new type of DRAM altogether. SDRAM started out running at 66 MHz, while older fast page mode DRAM and EDO max out at 50 MHz. SDRAM is able to scale to 133 MHz (PC133) officially, and unofficially up to 180MHz or higher. As processors get faster, new generations of memory such as DDR and RDRAM are required to get proper performance.
DDR (Double Data Rate SDRAM)
DDR basically doubles the rate of data transfer of standard SDRAM by transferring data on the up and down tick of a clock cycle. DDR memory operating at 333MHz actually operates at 166MHz * 2 (aka PC333 / PC2700) or 133MHz*2 (PC266 / PC2100). DDR is a 2.5 volt technology that uses 184 pins in its DIMMs. It is incompatible with SDRAM physically, but uses a similar parallel bus, making it easier to implement than RDRAM, which is a different technology.
Despite it's higher price, Intel has given RDRAM it's blessing for the consumer market, and it will be the sole choice of memory for Intel's Pentium 4. RDRAM is a serial memory technology that arrived in three flavors, PC600, PC700, and PC800. PC800 RDRAM has double the maximum throughput of old PC100 SDRAM, but a higher latency. RDRAM designs with multiple channels, such as those in Pentium 4 motherboards, are currently at the top of the heap in memory throughput, especially when paired with PC1066 RDRAM memory.
DIMMs vs. RIMMs
DRAM comes in two major form factors: DIMMs and RIMMS.
DIMMs are 64-bit components, but if used in a motherboard with a dual-channel configuration (like with an Nvidia nForce chipset) you must pair them to get maximum performance. So far there aren't many DDR chipset that use dual-channels. Typically, if you want to add 512 MB of DIMM memory to your machine, you just pop in a 512 MB DIMM if you've got an available slot. DIMMs for SDRAM and DDR are different, and not physically compatible. SDRAM DIMMs have 168-pins and run at 3.3 volts, while DDR DIMMs have 184-pins and run at 2.5 volts.
RIMMs use only a 16-bit interface but run at higher speeds than DDR. To get maximum performance, Intel RDRAM chipsets require the use of RIMMs in pairs over a dual-channel 32-bit interface. You have to plan more when upgrading and purchasing RDRAM.
Memory Speed
SDRAM initially shipped at a speed of 66MHz. As memory buses got faster, it was pumped up to 100MHz, and then 133MHz. The speed grades are referred to as PC66 (unofficially), PC100 and PC133 SDRAM respectively. Some manufacturers are shipping a PC150 speed grade. However, this is an unofficial speed rating, and of little use unless you plan to overclock your system.
DDR comes in PC1600, PC2100, PC2700 and PC3200 DIMMs. A PC1600 DIMM is made up of PC200 DDR chips, while a PC2100 DIMM is made up of PC266 chips. PC2700 uses PC333 DDR chips and PC3200 uses PC400 chips that haven't gained widespread support. Go for PC2700 DDR. It is about the cost of PC2100 memory and will give you better performance.
RDRAM comes in PC600, PC700, PC800 and PC1066 speeds. Go for PC1066 RDRAM if you can find it. If you can't, PC800 RDRAM is widely available.
CAS Latency
SDRAM comes with latency ratings or "CAS (Column Address Strobe) latency" ratings. Standard PC100 / PC133 SDRAM comes in CAS 2 or CAS 3 speed ratings. The lower latency of CAS 2 memory will give you more performance. It also costs a bit more, but it's worth it.
DDR memory comes in CAS 2 and CAS 2.5 ratings, with CAS 2 costing more and performing better.
RDRAM has no CAS latency ratings, but may eventually come in 32 and 4 bank forms with 32-bank RDRAM costing more and performing better. For now, it's all 32-bank RDRAM.
Understanding Cache
Cache Memory is fast memory that serves as a buffer between the processor and main memory. The cache holds data that was recently used by the processor and saves a trip all the way back to slower main memory. The memory structure of PCs is often thought of as just main memory, but it's really a five or six level structure:
The first two levels of memory are contained in the processor itself, consisting of the processor's small internal memory, or registers, and L1 cache, which is the first level of cache, usually contained in the processor.
The third level of memory is the L2 cache, usually contained on the motherboard. However, the Celeron chip from Intel actually contains 128K of L2 cache within the form factor of the chip. More and more chip makers are planning to put this cache on board the processor itself. The benefit is that it will then run at the same speed as the processor, and cost less to put on the chip than to set up a bus and logic externally from the processor.
The fourth level, is being referred to as L3 cache. This cache used to be the L2 cache on the motherboard, but now that some processors include L1 and L2 cache on the chip, it becomes L3 cache. Usually, it runs slower than the processor, but faster than main memory.
The fifth level (or fourth if you have no "L3 cache") of memory is the main memory itself.
The sixth level is a piece of the hard disk used by the Operating System, usually called virtual memory. Most operating systems use this when they run out of main memory, but some use it in other ways as well.
This six-tiered structure is designed to efficiently speed data to the processor when it needs it, and also to allow the operating system to function when levels of main memory are low. You might ask, "Why is all this necessary?" The answer is cost. If there were one type of super-fast, super-cheap memory, it could theoretically satisfy the needs of this entire memory architecture. This will probably never happen since you don't need very much cache memory to drastically improve performance, and there will always be a faster, more expensive alternative to the current form of main memory.
Memory Redundancy
One important aspect to consider in memory is what level of redundancy you want. There are a few different levels of redundancy available in memory. Depending on your motherboard, it may support all or some of these types of memory:
The cheapest and most prevalent level of redundancy is non-parity memory. When you have non-parity memory in your machine and it encounters a memory error, the operating system will have no way of knowing and will most likely crash, but could corrupt data as well with no way of telling the OS. This is the most common type of memory, and unless specified, that's what you're getting. It works fine for most applications, but I wouldn't run life support systems on it.
The second level of redundancy is parity memory (also called true parity). Parity memory has extra chips that act as parity chips. Thus, the chip will be able to detect when a memory error has occurred and signal the operating system. You'll probably still crash, but at least you'll know why.
The third level of redundancy is ECC (Error Checking and Correcting). This requires even more logic and is usually more expensive. Not only does it detect memory errors, but it also corrects 1-bit ECC errors. If you have a 2-bit error, you will still have some problems. Some motherboards enable you to have ECC memory.
Older memory types
Fast Page Mode DRAM
Fast Page Mode DRAM is plain old DRAM as we once knew it. The problem with standard DRAM was that it maxes out at about 50 MHz.
EDO DRAM
EDO DRAM gave people up to 5% system performance increase over DRAM. EDO DRAM is like FPM DRAM with some cache built into the chip. Like FPM DRAM, EDO DRAM maxes out at about 50 MHz. Early on, some system makers claimed that if you used EDO DRAM you didn't need L2 cache in your computer to get decent performance. They were wrong. It turns out that EDO DRAM works along with L2 cache to make things even faster, but if you lose the L2 cache, you lose a lot of speed.
No comments:
Post a Comment