Memory Types Overview

The earliest servers used DRAM. The fastest DRAM speed was 60ns. The lower the ns rating, the faster the memory. (This can be confusing if you're accustomed to the newer method of expressing memory speeds in MHz or throughput [MBps].) A nanosecond is defined as one billionth of a seconda very short time indeed. To put some perspective on that, the speed of light is 186,282 miles (299,792 kilometers) per second in a vacuum. In one billionth of a second, a beam of light travels a mere 11.80 inches, or 29.98 centimetersless than the length of a typical ruler!

To convert access time in nanoseconds to MHz, you use the following formula:

1 / nanoseconds x 1,000 = MHz

To convert from MHz to nanoseconds, you use the following inverse formula:

1 / MHz x 1,000 = nanoseconds

Although 60ns might seem very fast, it converts to about 16.7MHz, or less than 1/100 of the speed of recent server processors. Standard DRAM was used in servers up through about 1998. A slightly faster form of DRAM, known as extended data out (EDO) DRAM, was developed in the mid-1990s and was used in some servers as well. Although EDO DRAM's clock speed was also 60ns, it had improved memory access internally, to provide slightly faster throughput.

When we compare the speed of typical server processors in the mid-1990s to DRAM or EDO DRAM memory, it's obvious that the memory was actually slowing down system performance. For example, consider a Pentium Pro processor running at 200MHz. 200MHz is equivalent to 5ns. In other words, the processor was 12 times faster than memory (60/5=12)!

With faster processors in use by 19961997, such as the Pentium II and the Pentium II Xeon, the mismatch between processor speed and memory speed became even more lopsided. For example, a 300MHz Pentium II runs at the equivalent of 3.3nsmore than 18 times faster than the 60ns memory used in early models!

By 2000, servers were using memory types based on a new standard, called SDRAM (synchronous DRAM). 100MHz memory (PC100) and 133MHz memory (PC133) were the most common. Starting in early 2001, double data rate (DDR) memory of 200MHz and 266MHz become popular, along with 800MHz Rambus DRAM (RDRAM). In 2002, standard 333MHz DDR memory arrived, and in 2003, the speeds increased to 400MHz. The fastest current memory is DDR2, now available at 400MHz, 533MHz, 667MHz, and 1000MHz.

System memory timing is a little more involved than simply converting nanoseconds to megahertz. The transistors for each bit in a memory chip are most efficiently arranged in a grid, using a row-and-column scheme to access each transistor. All memory accesses involve selecting a row address and then a column address and then transferring the data. The initial setup for a memory transfer where the row and column addresses are selected is a necessary overhead referred to as latency. The access time for memory is the cycle time plus latency for selecting the row and column addresses. For example, SDRAM memory rated at 133MHz (7.5ns) typically takes five cycles to set up and complete the first transfer (5x7.5ns = 37.5ns) and then perform three additional transfers with no additional setup. Thus, four transfers take a total of eight cycles, or an average of about two cycles per transfer.

Over the development life of the PC, memory has had a difficult time keeping up with the processor, requiring several levels of high-speed cache memory to intercept processor requests for the slower main memory. Table 5.3 shows the progress and relationship between system board (motherboard) speeds in servers and the most common types and speeds of main memory or RAM used and how these changes have affected total bandwidth.

Table 5.3. DRAM Memory Module and Bus Standards/Bandwidth (Past, Current, and Future)^[1]

Module Standard	Module Format	Chip Type	Clock Speed (MHz)	Cycles per Clock	Bus Speed (MT/s)	Bus Width (Bytes)	Transfer Rate (MBps)
Fast Page Mode and EDO DRAM
FPM	SIMM	60ns	22	1	22	8	177
EDO	SIMM	60ns	33	1	33	8	266
SDR DRAM
PC66	SDR DIMM	10ns	66	1	66	8	533
PC100	SDR DIMM	8ns	100	1	100	8	800
PC133	SDR DIMM	7/7.5ns	133	1	133	8	1,066
DDR DRAM
PC1600	DDR DIMM	DDR200	100	2	200	8	1,600
PC2100	DDR DIMM	DDR266	133	2	266	8	2,133
PC2700	DDR DIMM	DDR333	166	2	333	8	2,667
PC3200	DDR DIMM	DDR400	200	2	400	8	3,200
PC4000	DDR DIMM	DDR500	250	2	500	8	4,000
DDR2 DRAM
PC2-3200	DDR2 DIMM	DDR2-400	200	2	400	8	3,200
PC2-4300	DDR2 DIMM	DDR2-533	266	2	533	8	4,266
PC2-5400	DDR2 DIMM	DDR2-667	333	2	667	8	5,333
Rambus RIMM
RIMM1600	RIMM-16	PC800	400	2	800	2	1,600
RIMM2100	RIMM-16	PC1066	533	2	1,066	2	2,133
RIMM2400	RIMM-16	PC1200	600	2	1,200	2	2,400
RIMM3200	RIMM-32	PC800	400	2	800	4	3,200
RIMM4200	RIMM-32	PC1066	533	2	1,066	4	4,266
RIMM4800	RIMM-32	PC1200	600	2	1,200	4	4,800

^[1] Key: MT/s = megatransfers per second; MBps = megabytes per second; ns = nanoseconds (billionths of a second); FPM = Fast Page Mode; EDO = extended data out; SIMM = single inline memory module; DIMM = dual inline memory module; RIMM = Rambus inline memory module; SDR = single data rate; and DDR = double data rate.

Generally, a system works best when the throughput of the memory bus matches the throughput of the processor bus. Compare the memory bus transfer speeds (bandwidth) to the speeds of the processor bus shown in Table 5.4, and you see that some of the memory bus rates match some of the processor bus rates. In most cases, the type of memory that matches the CPU bus transfer rate is the best type of memory for systems with that type of processor.

Table 5.4. Processor Bus Bandwidth^[1]

CPU Bus Type	Clock Speed (MHz)	Cycles per Clock	Bus Speed (MT/s)	Bus Width (bytes)	Bandwidth (MBps)
133MHz Pentium III Xeon FSB	133	1	133	8	1,066
333MHz Athlon MP FSB	166	2	333	8	2,667
400MHz Opteron FSB	200	2	400	8	3,200
667MHz Itanium 2 FSB	166	4	667	8	5,336
800MHz Xeon FSB	200	4	800	8	6,400

^[1] Key: FSB = front side bus; MBps = megabytes per second; and MT/s = megatransfers per second.

Because the processor is fairly well insulated from directly dealing with main memory by the L1 and L2 caches, memory performance has often lagged behind the performance of the processor bus. Systems using RDRAM, SDRAM, DDR, and DDR2 SDRAM have memory bus performance equaling that of the processor bus. When the speed of the memory bus equals the speed of the processor bus, memory performance is optimum for that system.