RAM Speed and Timings Explained: MHz and CAS Latency

RAM speed and timings describe how fast a memory module moves data and how long each internal operation takes. Speed is the data rate in MT/s (mega-transfers per second); timings are the latency values, in clock cycles, that control the delays inside the module. A kit labelled DDR5-6000 CL30 states both at once: a 6000 MT/s data rate and a CAS latency of 30 cycles.

What Is RAM Speed?

RAM speed is the data transfer rate of a memory module, measured in mega-transfers per second (MT/s). A DDR5-6000 module completes 6,000 million transfers per second on each data pin:

• What it sets: peak bandwidth – data rate x 64-bit channel width / 8, so DDR5-6000 delivers about 48 GB/s on one channel (roughly 96 GB/s in dual channel).
• The unit: JEDEC labels memory in MT/s, not MHz, because that count reflects the actual transfers that drive bandwidth.
• Half the story: speed is one of two performance factors – timings are the other, and the two together set the real result at the controller, covered in how RAM works.

Best for: bandwidth-bound work (rendering, video, compiles) – a higher MT/s rate moves more data per second.

What Is the Difference Between MHz and MT/s?

The difference is that MHz measures the clock frequency while MT/s measures data transfers, which DDR memory completes at twice the clock rate. DDR = Double Data Rate: the module transfers on both the rising and the falling edge of every clock cycle:

• The 2x rule: a DDR5-6000 module runs a 3000 MHz clock but completes 6000 mega-transfers per second.
• Marketing shorthand: a kit sold as “6000 MHz” actually clocks at 3000 MHz – the headline number is the MT/s data rate mislabelled.
• What to compare: the MT/s data rate, since it reflects the real transfer count that determines bandwidth.

What Do RAM Timings Mean?

RAM timings mean the number of clock cycles the module waits to complete each internal memory operation. They are written as a series such as 30-36-36-76, listed in a fixed order, where lower numbers mean fewer cycles of delay:

The four primary timings, in their standard sequence:

• CL (CAS Latency) measures the cycles between a column read command and the first data output. CL is the most quoted timing and the largest single contributor to latency.
• tRCD (Row to Column Delay) measures the cycles between activating a row and issuing a column command. tRCD adds delay when the controller opens a new row.
• tRP (Row Precharge) measures the cycles to close an open row before another row opens. tRP affects access patterns that jump between rows.
• tRAS (Row Active Time) measures the minimum cycles a row stays open before it closes. tRAS sets the floor for how long an active row remains accessible.

Best for: reading a kit label – the four primary numbers are what manufacturers print and what buyers compare.

What Is CAS Latency and How Is True Latency Calculated?

CAS latency is the delay in clock cycles between a read command and the data appearing on the bus – but cycles are not time. True latency (ns) = CL / data-rate (MT/s) x 2000, which converts cycles into the real delay:

• Take the CAS latency. The CL number from the kit label (e.g. 30 for a CL30 kit).
• Divide by the data rate. Use the MT/s figure (e.g. 6000), not the MHz clock.
• Multiply by 2000. The result is the true latency in nanoseconds – DDR5-6000 CL30 = 30 / 6000 x 2000 = 10 ns.

• Why CL alone misleads: a faster clock makes each cycle shorter, so a higher-CL fast kit can beat a lower-CL slow kit – DDR5-7200 CL34 (9.4 ns) is quicker than DDR5-6000 CL30 (10 ns).
• Apples to apples: DDR4-3200 CL16 also lands at 10 ns, the same true latency as DDR5-6000 CL30 despite half the data rate.
• Compare in ns: the formula is the honest way to rank kits, and the same maths drives the DDR4 vs DDR5 latency comparison.

How Do XMP and EXPO Profiles Work?

XMP and EXPO profiles work by storing pre-tested speed, timing, and voltage settings on the module that the motherboard applies with one BIOS setting. Out of the box, memory runs a conservative JEDEC speed (typically DDR5-4800) even when the kit is rated higher:

• Why it is needed: without a profile enabled, a DDR5-6000 CL30 kit runs at ~4800 MT/s – slower than its label.
• Both on one stick: most 2026 kits carry XMP and EXPO together, so the same module hits its rated speed on either platform.
• When to set it: in firmware during the build, the step that follows physically installing the RAM.

Best for: anyone who buys rated memory – the profile is the one toggle that turns the speed on.

Speed vs Timings: Which Matters More?

Speed and timings interact, and for most workloads a higher data rate outweighs tighter timings when true latency stays similar:

• Bandwidth tasks win on speed: video editing, content creation, and integrated-graphics gaming gain more from the higher MT/s rate.
• Latency tasks win on timings: competitive gaming and simulation benefit from a low true latency – a high data rate paired with a low CAS number.
• The balanced target: DDR5-6000 CL30 for AMD Ryzen 7000/9000, DDR5-6400 CL32 for many Intel builds – capacity is a separate question, covered in how much RAM you need.

Best for: a single buying rule – pick the lowest true latency you can at the platform sweet-spot data rate, not the biggest headline number.

What Is the AMD and Intel RAM Sweet Spot?

The sweet spot differs by platform because AMD rewards tight timings at a 1:1 fabric ratio while Intel scales to much higher data rates:

What Are Secondary and Tertiary Timings?

Secondary and tertiary timings are additional latency parameters beyond the four primary numbers that fine-tune memory behaviour. The primary CL-tRCD-tRP-tRAS appear on the label, but a module runs dozens more set automatically by firmware:

• tRFC (Refresh Cycle Time) sets the cycles the module needs to complete a refresh before normal access resumes. Lower tRFC improves performance in memory-heavy tasks because refresh blocks access less often.
• tRC (Row Cycle Time) sets the minimum cycles between successive activations of the same row, equal to tRAS plus tRP. tRC governs how quickly the controller reuses a row.

• Set per profile: a labelled DDR5-6000 CL30 kit ships with matched secondary timings, so the rated profile is tuned as a whole.
• Manual tuning: tightening secondaries can lower true latency further, but it is an enthusiast step, not a purchase comparison.
• What still decides a buy: the primary timings plus the data rate remain the figures to compare across kits.

How Do Speed and Timings Affect Real Performance?

Speed and timings affect real performance because memory bandwidth and latency change frame rates in gaming and completion times in productivity:

• Gaming uplift: independent testing (TechPowerUp and others) shows raising a DDR5 kit from 4800 to 6000 MT/s adds about 5-12% average frame rate at 1080p, where the CPU feeds the GPU most.
• Resolution scaling: the gain narrows at 4K, where the graphics card, not memory, becomes the limit.
• The AMD reason: Ryzen 7000/9000 gains at 6000 MT/s because that rate keeps the fabric clock synchronised 1:1, avoiding a latency penalty.

Best for: 1080p gamers and productivity users – faster memory pays off most when the CPU is the bottleneck.

How Do You Read a RAM Kit Label?

A RAM kit label is read by identifying the standard, data rate, capacity, and timings printed on the module sticker or product listing. A label reading “DDR5-6000 CL30 32GB (2x16GB) 1.35V” states each attribute in order:

• Standard + data rate: “DDR5-6000” = fifth-generation memory at a 6000 MT/s data rate (a 3000 MHz clock).
• Timings: “CL30” is the CAS latency; the full string 30-36-36-76 is CL-tRCD-tRP-tRAS in cycles.
• Capacity + voltage: “32GB (2x16GB)” is two modules for dual-channel; “1.35V” is the voltage the XMP or EXPO profile sets.

The table below decodes each field of a typical DDR5-6000 CL30 specification, including the calculated true latency and bandwidth:

Last Thoughts on RAM Speed and Timings

RAM speed and timings together define memory performance. Speed, measured in MT/s, sets bandwidth; timings, measured in clock cycles, set the delays inside the module. The MHz figure on marketing labels is half the MT/s data rate, because DDR transfers on both clock edges. True latency in nanoseconds, calculated from the CAS latency and the data rate, reveals that a higher CAS number does not always mean a slower module. XMP and EXPO profiles unlock the rated speed that otherwise defaults lower, and for current platforms DDR5-6000 CL30 balances both halves of the equation – running AMD Infinity Fabric 1:1 while Intel scales higher on bandwidth.

Key Takeaways:

• RAM speed is the data rate in MT/s; timings are the latency delays measured in clock cycles.
• MT/s equals twice the MHz clock, because DDR memory transfers on both clock edges.
• Timings list CL-tRCD-tRP-tRAS, with CAS latency contributing the largest single delay.
• True latency in nanoseconds equals CAS latency divided by the data rate, times 2000 – so compare kits in ns, not by CL.
• XMP (Intel) and EXPO (AMD) profiles apply the rated speed and timings that otherwise default to a slower JEDEC setting.
• The AM5 sweet spot is DDR5-6000 CL30 (Infinity Fabric 1:1); Intel scales to higher data rates via Gear 2.

Frequently Asked Questions (FAQs)