As with DDR to DDR4, memory bandwidth, latency and latency are also decisive for the classification of the speed of the current generation DDR5 memory in addition to the effective memory clock, which should correctly be specified in MT/s . ComputerBase provides an overview with a comparison to DDR4.
Table of contents
- 1 Bandwidths, latencies and latency times of DDR5
- The latency time results from latency and clock
- DDR5-4800 to DDR5-7000
- Low latency memory kits
- Conclusion
- Sharp latencies in the test
Bandwidths, latencies and latency times of DDR5
Already the “Double Data Rate Ssynchronous Dynamic Random Access Memory” or DDR-SDRAM (test), which came onto the market in 1999 but only gained wide acceptance in 2002, was mainly advertised for its effective memory speed of up to 400 MT/s according to JEDEC specifications.
But the number of transfers that can be carried out per second is only one aspect when assessing the performance of a RAM module (it together with the width of the connection gives the bandwidth). The other is latency, i.e. the time lag between the request for data and its provision. This is usually expressed via the Column Address Strobe Latency (CAS Latency, “CL”). It indicates how many cycles elapse between request and response. There are also other primary, secondary, and tertiary timings that can be critical to performance, but we won't go into detail about them in this article – other articles on these aspects are linked at the end.
 </figure>
</figure>
<p class=)
The two essential questions for evaluating DDR5 are: How fast does data get from the main memory to the processor or vice versa (latency)? And how many of them make it at once (bandwidth)?
The number of memory channels plays an additional role for the effective bandwidth and the length of the conductor tracks and the processor architecture for the effective latency. For example: With what number of processor clock cycles can the processor write data to RAM or read data from RAM once? Or how is the data transported within the processor? This means that the processor clock and the cache clock (or Infinity Fabric clock) also have an impact on the effective memory latency. The RAM module as such has no influence on either of these.
DDR5 main memory was approved according to the JESD79-5A standard with the following specifications:
- 3,200 to 8,400 MT/s
- DDR5-3200 to DDR5-8400
- 1.1 volt operating voltage
- Modules with 8 to 64 GB
- On-Die-ECC
A CAS latency is not part of the norm. It has increased significantly compared to DDR4 for products on the market. If fast DDR4 modules offer CAS latencies of 17, 16, 15 or even 14, with DDR5 it's more like 40 clock cycles at the beginning, the minimum so far is CL28.
Is So DDR5 has become significantly slower?
Image 1 of 2
Specifications of DDR5 
DDR5 advantages The latency time results from latency and clock
Not necessarily, because latency is specified in clock cycles and DDR5 clocks much higher – so a clock cycle takes less time.
The effective latencies and latency times that are primarily relevant for gamers and are of interest with regard to the minimum FPS are calculated as follows:
CAS latency ÷ effective memory clock × 2,000 ns
The editors have compiled a corresponding overview of all combinations of memory clock and memory bandwidth as well as latency and latency time, the so-called clock cycle time, that are already available, and provides an overview of the most common DDR5 memory modules. A fast DDR4-3200 module with CL14 was used for comparison.
DDR5-4800 to DDR5-7000
With DDR5, the rule is currently modules with CL40 bis down to CL36 – that is more than 2 times higher than the fastest DDR4 modules and means that their latency of under 9 ns is not reached – not even by modules with 7,200 MT/s.
4800 MT/s 0.42 ns 16.67 ns DDR5-4800 CL38 15 .83ns DDR5-4800 CL36 15.00ns DDR5-5200 CL40 325MHz 2600MHz 5200MHz 41.6GB/s
5200MT/s 0.39ns 15.38ns DDR5-5200 CL38 14.62ns DDR5-5200 CL36 13.84ns DDR5-5600 CL40 350MHz 2800MHz 5600MHz 44.8GB/s
5600MT/s 0.36ns 14.29ns DDR5-5600 CL38 13.57ns DDR5-5600 CL36 12.86ns DDR5-6000 CL40 375MHz 3000MHz 6000MHz 48.0GB/s
6000 MT/s 0.34 ns 13.33 ns DDR5-6000 CL38 12.76 ns DDR5-6000 CL36 12.00 ns DDR5-6400 CL40 400 MHz 3200 MHz 6400 MHz 51.2 GB/s
6400 MT/s 0.32ns 12.50ns DDR5-6400 CL38 11.88ns DDR5-6400 CL36 11.25ns DDR5-6800 CL40 425MHz 3400MHz 6800MHz 54.4GB/s
6800MT/s 0, 30ns 11.76ns DDR5-6800 CL38 11.17ns DDR5-6800 CL36 10.56ns DDR5-7200 CL40 450MHz 3600MHz 7200MHz 57.6GB/s
7200MH/s 0.28ns 11 .11ns DDR5-7200 CL38 10.55ns DDR5-7200 CL36 10.00ns DDR4-3200 CL14 400MHz 1600MHz 3200MHz 25.6GB/s
3,200 MT/s 0.62 ns 8.75 ns DDR4-3200 CL18 11.25 ns *) Connection to the memory controller; **) Clock compared to SDRAM
Low latency memory kits
Above all, G.Skill has already presented some so-called low-latency memory kits with DDR5-5600 CL28 and DDR5-6000 CL30. The same applies to Corsair: The company will soon be adding DDR5-6600 CL32 to its Dominator Platinum RGB series and effective latencies below 10 ns for the first time.
5600MT/s 0.36ns 10.00ns DDR5-6000 CL30 375MHz 3000MHz 6000MHz 48.0 GB/s
6000 MT/s 0.34 ns 10.00 ns DDR5-6600 CL32 412.5 MHz 3300 MHz 6600 MHz 52.8 GB/s
6600 MT/s 0.31 ns 9.69 ns DDR4-3200 CL14 400MHz 1600MHz 3200MHz 25.6GB/s
3,200 MT/s 0.62 ns 8.75 ns DDR4-3200 CL18 11.25 ns *) Connection to the memory controller; **) Clock compared to SDRAM
These modules do not yet come close to the latency of a DDR4-3200-CL14 bar, but the difference is no longer large – and the bandwidth is more than twice as high.
Conclusion
Not only did the latencies increase sharply with the change from DDR4 to DDR5 for the time being, so did the latencies. However, because DDR5 effectively clocks much higher, i.e. a clock cycle takes less time than DDR4, the surcharges on the latency time are much lower than the first glance at the key data suggests. The first low-latency modules from DDR5 are already only just behind DDR4-3200 CL14, but are exorbitantly expensive at well over 10 euros per GB and difficult to obtain. DDR4-3200CL14 is already available for less than 5 euros per GB.
Sharp latencies in the test
The topic of RAM OC and thus the optimization of latencies as well Primary, secondary and tertiary timings are very popular within the ComputerBase community. It has therefore repeatedly been the focus of numerous tests in recent years.
- Overclocking DDR5 RAM: That's what XMP 3.0, high clock and optimized timing bring
- RAM-OC with Intel CPUs in the test: Core (X) meets overclocked main memory
- Gigabyte Aorus RGB memory Tested: DDR4-4400C19 overclocked with Hynix DJR
- T-Force Xtreem DDR4-3600C14 tested: AMD Ryzen 3 3300X on RAM steroids
- CPU gaming performance tested: Intel Core i9-9900K vs. AMD Ryzen 9 3900X with RAM-OC
- Samsung B-Die in test: Corsair Vengeance LPX DDR4-4600 on Ryzen 3000
- Micron E-Die in the test: Crucial Ballistix Tactical Tracer RGB on Ryzen 3000
Users and gamers who deal intensively with the topic of RAM and RAM-OC and the optimization of timings and latencies can be found in the ComputerBase forum.
- AMD Ryzen RAM-OC Community by cm87
- Intel Core i-Series – RAM Overclocking by Rage
The overview of the DDR5 memory kits listed above will continue to be expanded with new frequencies and latencies in the future. The editors are always happy to receive information about new memory kits or important specifications that have not yet been listed.
Was this article interesting, helpful or both? The editors is happy about any support from ComputerBase Pro and disabled ad blockers. More about ads on ComputerBase.