RAM is a cornerstone of any PC build, yet it often flies under the radar unless it's boasting flashy heatsinks or RGB lighting. While the CPU dictates overall performance, optimizing your RAM can deliver noticeable speed gains. Clock speed and timings (latency) are the primary factors determining RAM performance.
You can check your RAM speed on the module label, packaging, via tools like CPU-Z, or in BIOS/UEFI. A typical spec looks like this:
DDR4 3200 (PC4 25600)
DDR4 indicates the generation, matching the 'PC4' prefix.
The number 3200 represents the data rate in megatransfers per second (MT/s)—not the raw clock speed, despite common marketing shorthand. In DDR RAM, the effective clock is half the data rate (1600 MHz here), boosted by prefetch techniques from an internal 400 MHz clock. Since DDR transfers data on both clock edges, the data rate aligns with the marketed 'clock speed' in MHz.
The PC number (25600) denotes peak bandwidth in MB/s: 3200 MT/s × 64 bits / 8 bits per byte = 25,600 MB/s. Both figures convey the same speed info in different units.
Timings measure latency—delays in clock cycles between common RAM operations. They're listed as four numbers like 16-18-18-38; lower is faster. Specs define minimums for each DDR generation.
CAS Latency (CL) is the delay from column address command to data output, after row activation. Combined with tRCD, it determines time to first data bit from an idle row.
Row Address to Column Address Delay (tRCD) measures the minimum cycles between activating a row and accessing its columns—essentially the time for RAM to 'reach' the target address.
Row Precharge Time (tRP) is the latency to close one row and open another. It often matches tRCD due to similar internal processes.
Row Active Time (tRAS) is the minimum cycles a row stays open for reliable data writes—from activation to precharge. For SDRAM, it's typically tRCD + CL.
These timings cap your RAM's speed, enforced by the memory controller—not physics. With a capable motherboard, overclocking and tightening timings (e.g., a few cycles lower) can yield gains.
From years of testing high-end builds, RAM overclocking is finicky—expect trial and error—but it accelerates memory-intensive tasks like rendering and VM performance.
