Your access to this site has been limited by the site owner
If you think you have been blocked in error, contact the owner of this site for assistance.
If you are a WordPress user with administrative privileges on this site, please enter your email address in the box below and click "Send". You will then receive an email that helps you regain access.
Block Technical Data
| Block Reason: | Access from your area has been temporarily limited for security reasons. |
|---|---|
| Time: | Fri, 20 Oct 2023 18:49:51 GMT |
About Wordfence
Wordfence is a security plugin installed on over 4 million WordPress sites. The owner of this site is using Wordfence to manage access to their site.
You can also read the documentation to learn about Wordfence's blocking tools, or visit wordfence.com to learn more about Wordfence.
Click here to learn more: Documentation
Generated by Wordfence at Fri, 20 Oct 2023 18:49:51 GMT.
Your computer's time: .
How to calculate tRC timing in cycles?
Hi. I was looking for some additional ram for my aging desktop PC and tried to find modules that match the timings perfectly.
At first I thought I found a perfect match with some Kingston modules, but on a closer look things are a bit weird.
Wikipedia, among other sources states that tRC = tRAS + tRP
Running CPU-Z on my desktop machine would confirm that for 1200MHz mode (50 = 35 + 15):
However, that doesn't fit with the Kingston datasheets, e.g for HX424C15FB3K2/8. Datasheet screenshot:
In the datasheet Kingston declares the module runs at at 15-15-15 timings (assuming tCL-tRCD-tRP notation based on convention mentioned at Wikipedia pages). However at the same time the datasheet declares the Row Cycle Time (tRC) to be 45.75ns(min.) and the Row Active Time (tRAS) to be 29.125ns(min.). The frequency of 1200Mhz dictates a 0.83ns cycle, which means the tRC translates to roughly 55 cycles and tRAS to 35 cycles. This does not make any sense. Plugging the numbers into the equation we get 55 = 35 + 15, which is obviously false. One of these three values (tRAS, tRP, tRC), or the equation must be wrong, because they don't add up.
When I sent this question to Kingston support, they replied that they checked the datasheets and everything is OK. Is the equation for tRC wrong? How to properly calculate the tRC?
- Add to quote
- ShareOnly show this user
- Add to quote
- ShareOnly show this user
An answer to both of these questions (why is tRC rated higher than it should be? why is tRFC rated higher than it should be?), comes down to the same answer: process variation.
Producing DRAM is an incredible feat of engineering. Billions of capacitors are arranged on a die with many functional portions of those capacitor arrays (features) being smaller than the smallest wavelength of visible light. Violet light has a wavelength of 380nm, which is now more than 10 times larger than the smallest features that exist in DRAM designs. This means that it is literally impossible to see DRAM designs, even with the strongest optical microscopes. Companies like Samsung, Micron, and Hynix can now use only electron microscopes if they want to visually validate that their designs were successful. Think about how incredible that is, that we as a species have such a strong need for DRAM that we require perfection from billions upon billions of capacitors that can't be seen with any optical technology.
Anyway, to vastly simplify the way that these capacitors are produced, various chemicals (Argon, Phosphorous, Dielectrics) are deposited onto silicon in plasma form at extremely high temperatures. This production process has literally hundreds of steps. It's not only one person's jobs but hundreds of people's jobs at each of these companies just to optimize the number of processing steps and the order of these steps. Each step has the opportunity to introduce variation onto the wafer.
Imagine a wafer as a big pizza pie that you want to cut into square pizza pieces. Obviously some of the pizza at the edge is going to be wasted, because you need perfect squares. Similarly in semiconductor manufacturing most of the edge of the wafer is wasted, even if it was patterned and processed. However, there are much more subtle variations across the wafer that become extremely significant for product design and product marketing. Imagine that you have a machine that deposits sauce onto the pizza using a squirt in the middle. Even if you have a subsequent step that attempts to smooth the sauce around the whole pizza, you are going to have a pattern of variation. Some of the pizza is going to have slightly (maybe millimeters more) sauce height in some areas than others. Usually in semiconductor engineering these patterns appear as rings on the wafer when subsequently analyzed. In the case of DRAM engineering we might be talking about some ring of the wafer receiving a layer of 10 — 20 additional atoms per square nanometer of Argon.
Now imagine what that ring on the wafer means when you cut it up into square pizza pieces. Some pieces which are on the edge of the ring have a corner which has slightly more Argon than the rest of the piece, and some pieces are entirely covered by this additional amount, which is elevated compared to most of the dies on the wafer.
Now let me get closer to the point. There are many components of a DRAM die: combinational logic which controls addressing, sense amplifiers used to amplify the data trapped in the capacitors during a read, combinational logic used during refresh operations, the capacitors themselves, various interconnects at different layers of the die, and most importantly test structures used to determine if the die even operates as a DRAM or not. All of these components react differently to different variations in process steps: a little bit more Phosphorous and the sense amplifies might be a bit more sensitive, but the refresh logic sees some slowdown due to a compounding effect on feature electrical resistance later down in the processing line.
Considering this all together: each DRAM die is a unique snowflake. Half of the capacitors might need to be refreshed every N cycles, while the other half absolutely need to be refreshed in N — 3 cycles. A little "chunk" of the die (usually in a semicircle pattern) might have sense amplifiers that can detect bits within 3 cycles while the rest of the die needs 8 cycles. All this is to say that any formulas that you see that relate DRAM timings together is based on theoretically ideal conditions of a single capacitive memory device operating in isolation. You have billions and billions of memory devices, in the case of 16 GiB of memory 137,438,953,472 memory devices (that's not a typo). In addition you have all of the surrounding components that I described earlier all attempting to operate under the same strict timing conditions.
The result is that you often have a die in which some timing, like precharge time, can be pushed very hard, because all of the process steps related to producing the sense amplifiers was perfect. However, cycle time must be relaxed because one third of the capacitors hold 12% less charge than the other two thirds.
As I alluded to earlier, this is where product marketing comes in. After you slice your pizza into square pieces, you sell the ones that were least affected by process variation as $10 dies, you sell some iffy ones as $5 dies, and you sell the rest as $1 dies (and it gets much more fine grained than this). At the end of the day you sell a piece of sand with some argon and phosphorous on it ($3 in raw materials) for $3000, but you spent $2700 worth of engineering to figure out how to pattern it and test it.
So again, every die is a unique snowflake. You can use theoretical relationships between timings to determine the optimum conditions for the best DRAM cell of your 137,438,953,472 memory cells (and they all have to work!), but each timing will find some limitation based on the worst performing component (sense amplifier, addressing logic) related to some individual cell or some collection of cells. The slowest subcomponent in any stage of the DRAM pipeline limits the timing of that pipeline stage.
Trefi тайминг сколько ставить
User Rating: 5 / 5
- DDR4 Overclocking Wiki
- DDR4 Overclocking Guide
- DDR RAM Overclocking Terminology FAQ
Article Index
- Overclocking Guide for DDR4 RAM
- PCBs and ICs
- Voltages
- Timings
- Preparations
- Overclocking
- Stability Testing
- DDR4 Profiles
- All Pages
This easy, comprehensive guide illustrated with images is for the overclocking DDR4 using several brands such as Corsair and G.Skill with different IC’s as an example. This guide is to supplement existing DDR4 overclocking guides whose knowledge was combined with the experience of overclocking the various DDR4 kits listed below(more to come).
Listed below are the specs of the motherboards, DDR4 kits and CPUs.
HisEvilnes:
- Motherboard:
ASUS Maximus Hero XI Z390.
ASUS ROG Strix X370-F Gaming.
MSI MPG Z490 Gaming Plus.
ASUS ROG Strix B450-F Gaming.
ASUS ROG Strix X570 Crosshair VIII Hero. - CPU:
Intel i7 8086K.
AMD Ryzen 5 1600X.
AMD Ryzen 5 2600.
Intel i5 10600K.
AMD Ryzen 7 3700X.
Intel 9700K
AMD Ryzen 7 5800X. - DDR4:
CMT32GX4M4C3466C16, single-rank Samsung B-die, 3466C16 XMP 32GB.
F4-3200C14D-16GTRS, single-rank Samsung B-die, 3200C14 XMP 16GB.
F4-3466C16-8GTZR, single rank Hynix C-die(CJR), 3466C16 XMP 16GB.
F4-3600C15D-16GTZ. single-rank Samsung B-die, 3600C15 XMP 16GB.
PVS416G440C9K, single rank Samsung B-die, 4400C19 XMP 16GB.
HX426C16FB3K2/32, dual-rank Micron E-die, 2666C16 XMP 32GB. - Topology Z390: T-Topology based on motherboard QVL.
Topology X370: Daisy-Chain based on motherboard QVL.
Topology Z490: Daisy Chain based on motherboard QVL.
Topology B450: Daisy Chain based on motherboard QVL.
Topology X570: Daisy Chain based on motherboard QVL.
Nanogenesis:
- Motherboard: Asrock Fatal1ty Z370 Gaming K6.
- CPU: Intel 8700K.
- DDR4: BLS16G4D32AESB.M16FE, dual-rank Micron E-die.
- Topology: T-Topology based on motherboard QVL.
There are several programs you need for stability testing, benchmarking, verification and a readout of your RAM IC’s and PCB.
- AIDA64: RAM Benchmark, AIDA64 download section
- Prime95: RAM Stress test, Prime95 download section
- MemTest2 & MemTestHelper2, MemTest2 download section / MemTestHelper2 download section
- CPU-Z: Validation, CPU-Z download section
- Thaiphoon: Component readout, Thaiphoon Burner download section
Dual rank 2x vs Dual Rank 4x.
When purchasing a DDR4 RAM kit consider your capacity in GB as well as how many DIMMs it will populate. For overclocking a DDR4 kit of 2 is best however if you need more or just want more capacity with 32GB or more you can either end up with a 4 x 8GB kit or a 2 x 16GB kit. Most consumer CPU’s will support dual rank like the Intel 9900k and the AMD 3600X. This means if you use more than 2 DDR4 sticks, you will share each channel. This will put more stress on the internal CPU IMC(Internal Memory Controller). Thus for extreme overclocking, Dual Rank DDR4 kits are better. When using Dual Rank kits, make sure you populate the 2nd and 4th DIMM slots or reference the motherboard manual.
Another consideration to make when using 4 DIMM slots on dual-channel motherboards is the topology of the motherboard traces. The traces leading to your DIMM slots can either be Daisy chained or use T-Topology. This is however not commonly advertised to guesstimate check the motherboard QVL. When the highest validation on the motherboard for RAM is with a kit of 4, then most likely that motherboard is using T-Topology. If the motherboard highest validation on the motherboard for RAM is with a kit of 2 sticks, then it is most likely Daisy chained. This is also a good indication of the motherboard quality. Even when using a RAM kit of 2 sticks, T-Topology is better for extreme overclocking(4000mhz or higher). But again for extreme overclocking it is always better to use two sticks of RAM on dual-channel for the best result.
PCB and IC»s
This is the bread and butter of overclocking your RAM kit. The better your ICs and PCBs are, the better your overclocking potential will be. To verify what your PCB and IC’s use Thaiphoon Burner and read one of your RAM sticks. Open Thaiphoon Burner and select Read then select any of the installed RAM sticks by selecting an SPD.
Now you will get a readout of your specific RAM stick, and this should be the same on all RAM sticks. To gauge the potential overclocking reference the example Thaiphoon Burner illustration.
- This will tell you the amount of IC ranks on your PCB. You want one rank, so the ICs are located on one side of the PCB.
- This will tell you the PCB quality. AO is the preferred PCB for extreme overclocking.
- This will tell you the manufacture of the ICs.
- This will tell you the version of the IC and the total count per PCB.
- This will tell you the size in MB for each IC. 1024 MB means one rank of 1024MB ICs per PCB.
So, in short, I have Samsung B-die in a single rank configuration of 1024 Mb IC’s and an A0 PCB. With a total of 8GB per stick with four sticks, a total capacity of 32GB.
There are several PCBs for DDR4 some will be better for overclocking than others.
- A0: The JEDEC original stock PCB, single rank(so only ICs on one side of the PCB) with a stock speed of DDR4-2133. The IC’s are evenly spread under the heat spreader. And is the favourite PCB for extreme overclocking.
- A1: JEDEC stock PCB for DDR4-2400 and has an ECC sibling called D1. Not very popular with overclocker there for rarely used on high-end RAM kits.
- A2: JEDEC stock PCB for DDR4-2666. Commonly used with RGB RAM kits. A good PCB that has some issues with 2 DIMM motherboards.
- B*: Respective JEDEC spec as with the A-series. These are not good overclockers and are dual-rank(IC’s on both sides of the PCB) PCBs.
- C0: JEDEC spec DDR4-2400 with larger IC’s. Little is known about overclocking potential.
There is a wide variety of ICs by the three main manufacturers of IC’s. Listed below are known ICs that are known to overclock. But some will do better than others. If your ICs are not listed below, they will most likely overclock poorly. Reference the linked DDR4 Overclocking Wiki for more specific details for specific IC’s. Below is a list of ICs with a good overclocking potential mileage that may vary on the specific IC. For a more accurate list of DDR4 IC’s please visit this Reddit article for a more in-depth picture of all available IC’s: Reddit DDR4 IC listing.
- Samsung B-die: The best ICs you can get for Intel or AMD builds. Scales well with voltage and does very well on tight timings and frequency 4000Mhz +. Save voltage for DRAM 1.5v and is the only IC that can safely run this voltage 24/7.
- Samsung E-die: Good ICs for Intel builds. Does not scale well with higher voltage but can reach frequencies higher than 4000Mhz. Save daily voltage for DRAM is 1.45v.
- Micron E-die: Good ICs for Intel and AMD. Scales well with voltage and can reach frequencies higher than 4000Mhz. Save daily voltage for DRAM is 1.45v.
- Hynix AFR(A-Die): Good ICs for Intel builds. Scales well with voltages and can reach frequencies up to 4000Mhz. Save daily voltage for DRAM is 1.45v.
- Hynix CJRC-Die): Good ICs for Intel and AMD builds. Unknown voltage scaling but can reach speeds up to 4000Mhz. Save daily voltage for DRAM is 1.45v
* Manufacturers do bin their ICs so adjust for expectations accordingly. A DRAM kit XMP rated for 3600 Mhz C18 will struggle to reach frequencies above 4000 Mhz unless you are really lucky. On the other hand, a DRAM kit XMP rated for 3600 Mhz C16 is of the finest binning and should only be limited by the maximum frequency your motherboard can handle.
Heat Spreaders & Temperatures.
For overclocking this will be vital as well as airflow to remove heat from the heat spreaders. RAM sticks without a heat spreader you will either have to install aftermarket heat spreaders or pass on overclocking. Having heat spreaders is also a good indication if they are meant for overclocking and the quality of the heat spreaders how well they will overclock. So expect to overclock results based on the quality of your RAM kit heat spreaders if your RAM kit comes with internal temperature probes you want to aim to keep them below 50c. ICs can take more heat, but some instability occurs after 50c that you will need to compensate for in more voltages or looser timings.
Voltages.
There are up to 3 voltages important for overclocking DDR4. For Intel System Agent, VCCIO and, DRAM voltages. And AMD SoC and DRAM voltages. These voltages help to stabilize the uncore domain of the CPU that ties into RAM as well as the RAM itself. The difference between Intel VCCIO and VCCSA and AMD SoC is that VCCIO requires an increase in voltage when increasing memory frequency in Mhz. And VCCSA requires a voltage increase when tightening timings.
Voltages exceeding 1.45v is only recommended for Samsung B-die for daily use up to 1.5v. To consider is that DDR4 kits with a rated 1.5v also have PCB’s that can run that voltage as a daily driver. Extreme voltages for daily use is only wise if you can water cool your DDR4 kit.
VCCSA also is known as System Agent voltage is for the IMC and PCIe domain. Going higher than 1.25v can damage the IMC and motherboard traces and is only recommended with high-end motherboards and short benchmarking and record attempts. System Agent voltages are commonly tied to tighter timings but lower speeds in Mhz. Using more System Agent voltage than needed for a overclock can cause instability.
VCCIO is voltage for the IMC and using more than 1.25v VCCIO can damage your IMC and motherboard traces and is only recommended for high-end motherboards, and short benchmarking and record attempts. More VCCIO is needed for higher speeds in Mhz. Using more VCCIO than needed for a overclock can cause instability.
SoC voltages help to reach higher FCLK(Infinity Fabric), MCLK(Memory Frequency) frequencies, and tighter timings. Extreme voltages are not recommended and are commonly used for HWbot extreme overclocking scores. And will require a beefy cooling solution if not exotic cooling.
To put these numbers into perspective here are some number from overclocking a range of DDR4 kits. Covering both AMD and Intel with their respective timings, frequency, voltages, and capacity.
Timings.
This guide will cover the main timings, and two secondary timings will get you little gains as well as you will either like working with 2nd, 3rd and 4th timings or you will hate it. For this guide, only primary timings and 2 of the secondary timings that will lower your Latency are covered. In due time this guide will be expanded to cover all timings.
- tCL: Aso known as CAS, is the first timing listed on every memory kit.
- tRCD: RAS to CAS delay is the 2nd timing listed on every memory kit.
- tRP: Row Precharge Time is the 3rd timing listed on every memory kit.
- tRAS: RAS Active Time is the 4th timing listed on every memory kit.
- CR: Command Rate is listed as CR, T or N in the BIOS.
- tRFC: Refresh Cycle Time is a secondary timing listed as tRFC under secondary timings. IF listed different look under secondary timings as the only three-digit number.
- tREFI: Refresh Interval is a tertiary timing listed as tREFI commonly listed as tertiary timing but considered by most a secondary timing as this is adjusted alongside RFC. IF listed differently look under tertiary timings for a five-digit number.
All the primary Timings will allow for a performance gain in Read, Write and Copy speeds while tRFC and tREFI will lower the Latency of your memory kit. Read, Write and Copy represent raw performance in applications, thus more FPS or lower render times. While reduced Latency will make Windows more snappy as well as games being snappier, the is especially useful when gaming in a competitive sense.
Preparations.
The easy way is to save any overclock on other components like the CPU and GPU and then reset to default this will ensure that any instability while testing can only come from the RAM overclock. As RAM overclocking is time-consuming, it is wise to exclude any other factors of possible instability. Also, ensure that any overclock is saved in the BIOS to profile and well as any GPU to a profile. Also, make sure you have a baseline of the XMP profile from Aida64 or MaxxMEM2 with any CPU overclock applied since the memory OC will show a significant increase with or without a CPU overclock. And you also want a baseline to measure against when you revert all other overclocks. You can write down or save a screen capture of the Read, Write, Copy and Latency of the memory performance and use it as a reference when overclocking your memory. Reset the motherboard to default settings from within the BIOS before proceeding.
Tips for AMD.
Runnings a 1:1 ratio with MCLK and FCLK is best for Ryzen. Only use higher RAM Mhz speeds when your infinity fabric can keep up. When maxing out the infinity fabric go for tighter timings as de-syncing will decrease performance.
Gear Down Mode is useful for reaching higher frequencies and can be seen as CR 1.5(T1.5/N1.5)but will lock out the option to select CR2. As well as forcing tCL to even numbers, for example, C15 is set to C16 with GDM.
CLDO_VDDG can in some cases help with stabilizing FCLK overclock on the Ryzen 3000 series. This only applies when manually adjusting both SoC and CLDO_VDDG voltages and should be done as a pair. With an SoC voltage of 1.10v, the CLDO_VDDG voltages will drop depending on the SoC LLC by roughly
40mv to 1.01mv if left on Auto. So adjusting CLDO_VDDG manually you will have to ensure the SoC voltage drop under load is 40mv or less.
Tips for Intel.
Increasing Ringbus clock speed will increase performance; however, make sure the CPU overclock is stable before overclocking the DRAM and after.
ASUS Maximus Tweaks will tighten RTLs and I-OLs values. Mode 1 works up to 4000Mhz and Mode 2 works above 4000mhz.
Overclocking.
You are starting to overclock by simply booting into your BIOS and look for the DRAM, VCCIO, VCSSA and SoC(AMD) voltages. Make sure no XMP profile is selected and set the RAM profile manual. Also, check if MRC Fast Boot is turned off in the Timings section of your BIOS. Can be found in some high-end motherboard—Example BIOS from an ASUS Maximus Hero XI.
To start overclocking your DRAM you want to adjust your DRAM voltage to 1.45v for Samsung B-die and 1.40v for other dies like CJR, Intel VCCIO and VCCSA voltages to 1.15v or AMD SoC voltage to 1.10v. The following steps are in numerical order for overclocking gain.
Intel.
- Increase DRAM voltage.
- Increase VCCIO & VCCSA voltages. Both can be equal unless going for a low frequency but tighter timings then only VCCSA should be increased; otherwise, instabilities might occur.
- Increasing DRAM VTT voltage(not mistaken for CPU VTT) & VPPDDR voltage. Works best with stabilizing extreme overclocks that are over 4000Mhz.
- Increase DRAM voltage.
- Increase SoC voltage. Ryzen 3000 series when adjusting frequency stay in the 0.95
Leaving all the timings to Auto you just want to increase the memory frequency with the voltages entered and see how fast your RAM kit can run. Depending on your XMP profile you want to start higher so with a 3200mhz XMP kit you want to start with 3600Mhz. If your RAM kit XMP profile is rated higher, you want to start 400Mhz higher than the base XMP profile. Make sure all the timings are set to Auto, which includes the Primary Timings, Secondary Timings etc. Just want to get the highest possible memory frequency and then work on the timings. With AMD Ryzen you want to set the FCLK half of your memory speed, so when using, for example, 3600mhz then set the FCLK to 1800mhz.
After selecting a frequency then save these settings in the BIOS and reboot. Your PC might reboot a couple of times adjusting the timings since you changed the frequency. If your RAM kit can run the speed, it will boot right into Windows. If it fails, it will show a failure to post a message. In the case of a failure, you can increase DRAM voltage for Samsung B-die to 1.50v and 1.45v for other dies. And adjust some of the settings listed below.
- VCCIO and VCCSA voltages for Intel from 1.15v to 1.2v or 1.25v.
- SoC voltage for AMD from 1.10v to 1.20v or 1.25v. The exception is the Ryzen 3000 series SoC voltage has very little effect above 1.15v when it comes to frequency. With the AMD 5000 series, SoC voltage is better left on Auto and apply LLC 3 to the SoC and adjust the transient response of the DRAM and SoC by changing the Switching Frequency to 400Mhz if the motherboard allows it.
- Ensure ProcODT for AMD is within the frequency range if so go 1 step up or down to the closest frequency range.
- For AMD as well, check if Gear Down Mode is enabled to help stability but remember you won’t be able to set uneven tCL.
- Or dial back the memory frequency 1 step from the previously selected memory frequency while keeping the same voltages.
When it does boot into windows use Aida64 Cache & Memory Benchmark or MaxxMEM2 to gauge if your overclock is stable by checking the Read, Write, Copy and Latency. For Intel, you want a Latency between 55ns and 60ns without changing the timings and for AMD timings between 75ns to 80ns. And always an increase in Read/Write/Copy speeds. The first indication of instability is the memory having to correct for errors that will lower Read/Write/Copy speeds. You might also be able to boot into Windows, but when running a benchmark, your PC could shut down or BSOD which also is an unstable memory overclock. When using a Ryzen CPU, make sure you keep adjusting your FCLK to half the memory frequency. With booting or stability problems on Ryzen adjusting ProcODT and CLDO VDDP voltage will help stabilize memory overclocks. Important to note that CLDO VDDP is very sensitive, a 10mv range can be the difference between booting/unstable/stable.
Repeat this process to get the highest possible frequency on your RAM kit. Stop increasing the frequency as soon as it fails to boot into Windows or when you see a decrease in Read/Write/Copy speeds. Then you want to increase the DRAM voltage to 1.50v only for B-die or 1.45v for other dies if not already done so.
- VCCIO and VCCSA voltages for Intel from 1.15v to 1.2v or 1.25v.
- DRAM VTT voltage for Intel can be increased around 0.700v
If you have not done so already. Also, something to look for the timings that are automatically set by your PC and make sure that those do not get overly loose. When you get a high frequency but your timings are so loose in the range of 25-25-25-50 or more you might want to consider running a lower frequency. With higher VCCIO, VCCSA or AMD SoC voltages you can try to keep the frequency that failed to post or did not return a better benchmark. But in general DRAM voltage gives a bigger boost and VCCIO, VCCSA and SoC help to stabilize overclocks that boot but won’t pass a stress test. Below is a table of expected VCCIO and VCCSA voltages for AMD they are mainly tied to SoC voltage.
| Memory Frequency | — | VCCIO/VCCSA* |
| 3000 — 3466 Mhz | 1.10v — 1.15v | |
| 3466 — 3800 Mhz | 1.15 — 1.20v | |
| 3800 — 4200 Mhz | 1.20v — 1.25v | |
| 4200 Mhz and higher | 1.25v — 1.30v |
* Combining VCCIO and VCCSA works best for higher frequencies and their timings, for running 3200mhz with 14-14-14-32 as an example it is wise to drop VCCIO to 1.15v as it might introduce instability. For AMD SoC voltage a bit of similar advice when it comes to the Ryzen 3000 series. SoC voltage should not exceed 1.15v when only adjusting the frequency but can exceed 1.15v SoC to 1.20v
1.25v SoC when tightening the timings of your DDR4 overclock.
* FCLK should be synced with the memory frequency, and in all of the BIOS versions, I’ve seen this was done automatically. De-Syncing FCLK from MCLK could increase Latency and lower Read/Write/Copy speeds.
When you have established the highest possible memory frequency, it is time to tighten the timing. Here you start with the Primary Timings. To consider is that the best RAM kits run timings of 14-14-14-32 @ 3200mhz or 16-16-16-38 @ 3600mhz with 1.35v. This gives you a good indication of whatever max memory frequency you have for what memory timings you want to aim for and what is possible. Make sure the Command Rate is set to CR1/T1/N1 unless you are using a memory kit of 4 sticks then use CR2/T2/N2. If your timings on Auto went to CR2 regardless of the number of sticks keep these settings since using CR2 is slightly less demanding to overclock. You can consider using CR1/T1/N1 but expect less Mhz and loosen timings only really well-binned ICs do well on CR1.
For AMD Gear Down Mode can be enabled as well as adjusting ProcODT and CLDO VDDP voltage for AMD CPUs to help with stability, CLDO VDDG voltage can also be manually adjusted but make sure that SoC voltage is 40mv or more above CLDO VDDG voltage. As mentioned in the above paragraph regarding adjusting the frequency on the AMD 3000 series SoC should not exceed 1.15v when ONLY adjusting the frequency. However, when adjusting the timings tighter you can exceed the 1.15v for SoC as it will help stabilize your overclock in the 1.20v
1.25v range for SoC this mostly applies to the AMD 3000 series. As SoC voltage has a wider impact on frequency and timings on the AMD 1000 and 2000 series as a whole. With the AMD 3000 series the impact has become far less but it could benefit from a slight boost in SoC voltage. As an example, an XMP kit of 3600C15 did run fine on 1.15v SoC using the XMP profile but when tightening the timings to 3466C14-14-14-28 SoC voltage had to be increased to 1.20v on the SoC or it would be unstable. With the AMD 5000 series, SoC has become very sensitive and changing it leads to instability very quickly. And with the 5000 series as an example with the 5800X, the SoC voltage is left on Auto and sits around 1.088 volts and changing the Switching Frequency and LLC in the VRM options yields great results for higher overlocks.
You can consider a wider range of overclocked profiles and pick your daily driver from there, but it will take considerable more time to test and stress testing each profile for 100% stability. And it must be said that stress testing is vital even if done for hours with Linpack, Prime95 at times you might find instabilities down the road, rare but it can still happen. In that case, some minor voltage tweaks could work but consider you might have to loosen or lower the frequency when it does happen.
Some important notes on adjusting the timings you want to make sure the 1st timing is only 1 or 2 steps away from the 2nd and 3rd timings. Only B-Die can run the same timings on the first three timings, for example, 16-16-16 for the first set. Other IC’s you want to keep the first timings as low as possible but only loosen the 2nd and 3rd timing, as an example 16-18-18 for the first set. The 4th timing you want to double the 3rd timing and if needed to loosen + 4 steps ideally. Examples included in this guide are the most common used timings used for most IC’s. It is, however, possible to run different sets like 14-15-14-31. However, this does not always work with a wide range of motherboards and DDR4 RAM kits. For this purpose, examples like 14-14-14-32, 19-19-19-38 and 17-19-19-38 are used since these always work on their respective DDR4 kits and all motherboards. Below are some more primary timings to consider as an example.
Example B-die Timings.
- 14-14-14-32
- 15-15-15-34
- 17-17-17-34
- 19-19-19-38
Example RAM Timings.
- 14-15-15-32
- 14-16-16-32
- 16-18-18-36
- 17-19-19-38
You want to keep making the timings tighter until the PC fails to boot or you start seeing a drop in Read/Write/Copy speeds in Aida64 or MaxxMen2. Adjusting Primary Timings will also lower the Latency, so there should also see a drop in Latency. The Latency of about 45ns is good, but you can get the Latency to drop below 40ns. In short, you want the Primary Timings as tight as you possibly can. If you want to make sure that these timings are stable you can run a demanding benchmark like Fire Strike Ultra in 3DMark. And do a 2 hours custom 512FFT to 4096FFT stress test before working on Secondary and Tertiary Timings.
The final timings you want to adjust are tREFI and tRFC these works best if adjust as a pair be can be done separately. The tRFC timing you want to lower as much as you can a tRFC of 300 or lower is very good, but this will also depend on the motherboard and RAM kit quality. Starting with a tRFC of 450 is a safe starting point and work the tRFC down as low as you can. tREFI is the opposite of all timings, and you want this value as high as you possibly can, some overclocks have a tREFI of 65000 or higher, but those are on the high-end motherboards like the Maximus Hero XI Extreme. And an easy tREFI to have stable is 25000. Adjusting tRFC and tREFI just like the Primary Timings by adjusting them until Windows fails to boot and/or the Aida64 or MaxxMem2 benchmarks show a drop in Latency but not an increase in Read/Write/Copy speeds.
Stability Testing.
Testing the stability of your memory is key, and this will be much harder than CPU and GPU overclocking as it is more time-consuming. This is different from your initial stability testing in the overclocking process and takes hours to complete.
- Prime95: At least a 2-hour or longer stress test with big custom FFTs.
- MemTest2: At least 350% coverage on each instance.
Besides the synthetic testing, you want to run some demanding programs like games that make use of your RAM or applications such as video rendering if it is work-related. And look for stuttering in gaming, buffering speed while editing and rendering time. All in all, you want to see how your PC is doing after the overclock and that everything is fully stable.
When using Prime95, you want to run Custom Torture Test with large FFTs as shown below. 512 FFTs to 4096 FFTS and make sure you use at least 75% of your total memory capacity. So with a memory capacity of 16Gb, you want to at least run 14Gb or more. With a memory capacity, of 32Gb, you want to run at least 24Gb or more. Make sure that you leave some memory for Windows; Windows 10 needs at least 800Mb of memory. And you might want to consider running a hardware monitoring app like HWiNFO64 keeping an eye on voltages and temperatures. You want to complete at least 2 hours with no errors or longer stress test.
With overclocking any DDR4 kit comes trial and error this could present itself in BSOD, Watchdog and WHEA errors. To ensure your Windows stay healthy run CMD in Admin mode and running sfc /scannow to scan and fix corrupted files. This only takes a couple of minutes but it will ensure OS stability.
For the sake of certainty, a 2nd stress test was applied to the memory overclock using MemTest2 with MemTestHelper2 and is a highly recommend stress testing tool besides Prime95. The program MemTest2 is limited to only 2048MB per thread, and this is not the total capacity; it is the capacity used for each thread. When using an 8086k/8700k, you have six cores with 12 threads, and when using a 9900k, you have eight cores with 16 threads. So make sure you use all available threads. MemTestHelper should run at least a 350% coverage per instance, but more coverage in % per instance is better. When using MemTestHelper2, you can dial in the total capacity for RAM used for testing here you want to make sure you use 75% or more of your total capacity. Here you also want to make sure you run a hardware monitoring app like HWiNFO64 to keep an eye on voltages and temperatures.
On previous-generation CPU’s from AMD on X370/B350 or X470/B360, some memory issues can occur when even running an XMP profile. You can find a separate article linked below that does not involve overclocking the RAM but adjust some minor settings to fix the problems that might occur with AMD Ryzen CPUs with QVL and none QVL DDR4 and can be found here: How to stabilize DDR4 with infinity fabric.
DDR4 Profile.
Listed here are some DDR4 RAM kit profiles with some of the timings. This is useful in case you have the same DDR4 kit while the silicone lottery still applies you can copy and paste some of the timings.
CMT32GX4M4C3466C16.
This kit from Corsair comes with a single-rank Samsung B-die and an A0 8 layer PCB and even comes with RGB. And two final stable overclock profiles were tested with a 3rd still in the works. The highest possible profile of 4266Mhz 17-19-19-38-CR2 with 400tRFC, 50000 tREFI with DRAM voltage of 1.465v, VCCIO and VCCSA voltages of 1.25v but failed the longer stress test of 2 hours and returned an error. Possible due to the fact B-die has instability issue’s when reaching a temperature over 50c. In stress testing temperatures reached 58c on the inner 2 DIMMs.
The 2nd overclock profile was an attempt to get tight timings first and foremost and see what the highest possible frequency would be. This profile runs 3600Mhz 15-15-15-30-CR2 with tRFC 350, 50000 tREFI with a DRAM voltage of 1.45v and VCCIO and VCCSA voltages of 1.175v. This overclock profile passed the 2-hour Prime95 Custom FFTs test.
The 3rd profile is used as a daily driver and was the best possible trade-off in voltages, memory frequency and timings. This profile runs 4000Mhz 17-17-17-34-CR2 with tRFC 300, 60000 tREFI with a DRAM voltage of 1.45v and VCCIO and VCCSA voltages of 1.2v. This overclock passed a 2-hours Prime95 Custom FFTs and MemTestHelper2 350% coverage stress test.
The 4th profile is on an X570 ASUS Crosshair VIII Hero combined with a 5800X and runs a frequency of 3800Mhz with 16-16-16-32 timings, Command Rate 1 with a 1:1 Infinity Fabric of 1900Mhz with a DRAM voltage of 1.45.
F4-3466C16-8GTZR.
This kit from G.Skill comes single rank Hynix C-die(CJR) and an A1 10 layer PCB with RGB. And has these stable overclock profiles tested. The highest possible frequency runs 3800Mhz 18-20-20-40-CR2 with tRFC 500, 50000 tREFI with a DRAM voltage of 1.45.
The 2nd overclock profile is an attempt to lower the timings to gain performance with memory frequency being secondary. This profile runs 3200Mhz 15-17-17-34-CR2 with tRFC 450 and 50000 tREFI with a DRAM voltage of 1.40v.
F4-3200C14D-16GTRS
A G.Skill Royal kit that comes with single rank Samsung B-die on an A0 10 layer PCB with RGB. This kit has several stable overclock profiles tests. The 1st profile is on an X370 motherboard with a Ryzen 2600. Opting for real tight timings on the XMP frequency of 3200Mhz 12-12-12-28 CR1 with tRFC of 300 and a DRAM voltage of 1.5v.
The 2nd profile is Intel-based and centres around getting the highest possible frequency with the tightest timings. This profile runs 4000Mhz 16-16-16-34 with a 300 tRFC, 65535 tREFI with a DRAM voltage of 1.475v.
F4-3600C15D-16GTZ.
The go-to G.Skill kit when it comes to DDR4 overclocking with Samsung B-die has no RGB just a beefy heatsink. The highest possible profile and this was on a mid-range Z490 motherboard was 4200Mhz 15-15-15-30-CR2 with tRFC of 310, and 65000 tREFI with a DRAM voltage of 1.50v.
The 2nd profile was set to get very tight timings with the highest possible frequency. Even on a mid-range Z490, this is far easier and the overclock was locked in at 3800Mhz 13-13-13-28-CR2 with tRFC of 310, 65000tREFI and a DRAM voltage of 1.5v.
BLS16G4D32AESB.M16FE.
This kit comes with a dual-rank Micron E-die and a B1(Dual Rank IC PCB) PCB without RGB. And has two stable overclock profiles tested. The highest possible frequency profile runs in 3700Mhz 16-19-19-38-CR1 with tRFC 560, 28800 tREFI with a DRAM voltage of 1.45v.
The 2nd overclock profile is a to try to get the timings as tight as possible with memory frequency a secondary. This profile runs 3333Mhz 14-17-17-34-CR1 with tRFC 515, 11430 tREFI with a DRAM voltage of 1.45v.
Тайминги оперативной памяти: как узнать и настроить
Тайминг ОЗУ — это набор параметров, которые определяют, насколько быстро модуль оперативной памяти может получать доступ к данным и доставлять их. Тайминг оперативной памяти устанавливается производителем и определяется в спецификациях устройства. Они сказываются на эффективности оперативной памяти и могут повлиять на скорость работы системы. Более быстрые показатели способны привести к более быстрому доступу к информации и их передаче, но они также имеют увеличенную стоимость.
Важны ли тайминги оперативной памяти?
Тайминги ОЗУ важны, поскольку они могут повлиять на общую производительность персонального компьютера. Они имеют непосредственное отношение к задержке между моментом, когда делается запрос на доступ к данным из RAM чипов, и моментом, когда данные фактически доступны для использования.
Однако важно отметить, что такой фактор, как скорость оперативной памяти, также может влиять на эффективность.
Как узнать тайминги оперативной памяти?
Самый простой способ узнать тайминги — проверить спецификации. Эта информация доступна пользователю на упаковке продукта или на сайте производителя. Если же доступа к упаковке нет, используется уже собранный персональный компьютер или ноутбук, то очень часто можно найти характеристики скорости в системном БИОС или прошивке UEFI.
Проверка таймингов оперативной памяти
Чтобы получить доступ к этим данным, необходимо перезагрузить компьютер и войти в утилиту настройки BIOS/UEFI, нажав определенную клавишу во время процесса загрузки. Стандартный вариант требует нажатия либо F2, либо F10 или Delete (на выбор). Когда вы окажетесь в утилите настройки, найдите раздел с надписью «Память» или «ОЗУ» — это должно дать доступ к необходимой информации.
Если персональный компьютер собирался самостоятельно или была проведена замена планки оперативки, можно использовать сторонние программные утилиты, чтобы узнать тайминги памяти. CPU-Z — один из самых популярных вариантов, который может предоставить подробную информацию об аппаратном обеспечении устройства.
Расшифровка таймингов
Такие характеристики обычно представляются в виде ряда чисел, например «CL16-18-18-38» или «16-16-16-39». Эти цифры относятся к различным аспектам производительности оперативной памяти таким, как частота и тайминги оперативной памяти. Первое число, часто обозначаемое как «CL» или «CAS Latency», указывает на тактовые циклы, требующиеся RAM для ответа на запрос от ЦП.
Меньшее число CL указывает на более высокую производительность — например, модуль ОЗУ с CL 14 будет быстрее отвечать на запросы, чем модуль с CL 16. Остальные числа во временной последовательности относятся к другим аспектам производительности ОЗУ, например, ко времени для чтения или записи данных. Эти числа обычно менее важны, чем число CL, но они все же могут влиять на общую производительность.
Схема таймингов памяти
Часто такой показатель оперативной памяти представляются с помощью схемы, которая выглядит следующим образом — CL-tRCD-tRP-tRAS. Вот какое значение таймингов оперативной памяти в каждом из этих терминов:
- CL. Задержка CAS.
- tRCD. Задержка от RAS к CAS. Это относится к задержке между активацией строки в памяти и доступом к столбцу.
- tRP. Время предварительной зарядки строки необходимое для деактивации строки после того, как данные были прочитаны или записаны в нее.
- tRAS. Активен для времени предварительной зарядки. Это относится ко времени, которое требуется для деактивации строки и активации новой строки.
Все эти показатели имеют очень важное значение для полноценной работы устройства.
Штатные тайминги
Тайминги оперативной памяти могут сильно различаться в зависимости от конкретного модуля и производителя. Однако есть несколько «штатных» моментов, с которыми сталкивается большинство пользователей:
- CL16. Это наиболее распространенное время для модулей оперативной памяти DDR4, и оно указывает на относительно высокую производительность. CL, или задержка CAS, является наиболее важным моментом, который следует учитывать при покупке ОЗУ. Он представляет собой количество тактов, которое требуется оперативной памяти для ответа на запрос от ЦП. Чем меньше число CL, тем быстрее будет работать оперативная память. Модуль ОЗУ CL16 обычно имеет время отклика 10 нс.
- CL18. Модули оперативной памяти DDR4 с синхронизацией CL18 немного медленнее, чем модули CL16, но они также более доступны по цене. Обычно они имеют время отклика 11,25 нс.
- CL14. Модули оперативной памяти DDR4 с синхронизацией CL14 считаются высококачественными и, как правило, более дорогими. Они обеспечивают высокую производительность со временем отклика 8,75 нс.
Нужно разобраться в том, какие средние показатели в наиболее актуальных поколений RAM.
Средние тайминги DDR4
DDR4, или Double Data Rate 4 — это память DDR четвертого поколения, которая предлагает более высокую пропускную способность и скорость, чем ее предшественница DDR3. DDR4 предназначена для работы с современными процессорами и материнскими платами, обеспечивая более высокую производительность и эффективность компьютеров и других электронных устройств.
Модули памяти DDR4 имеют разные характеристики, влияющие на их производительность, и выбор правильных таймингов может значительно повлиять на скорость и стабильность вашей системы. Следует разобрать средние тайминги DDR4 и как выбрать правильные тайминги для конкретной системы.
Память DDR4 работает с базовой тактовой частотой 2133 МГц и может повышаться до 4800 МГц, обеспечивая более высокую пропускную способность, чем память DDR3. Память DDR4 поставляется в различных конфигурациях таких, как одноранговая, двухранговая, четырехранговая и восьмиранговая, каждая из которых имеет свои преимущества и недостатки.
Память DDR4 также имеет различные тайминги, такие как задержка CAS (CL), tRCD (задержка RAS-CAS), tRP (время предварительной зарядки строки), tRAS (задержка между активным и предварительным зарядом) и tRC (время цикла строки). Эти тайминги определяют, сколько времени требуется контроллеру памяти для доступа к конкретной ячейке памяти и завершения операции чтения или записи.
Средние тайминги DDR4 — это среднее значение 4 основных таймингов:
- задержка CAS (CL);
- tRCD;
- tRP;
- tRAS.
Средние тайминги выражаются 4 числами, например 16-18-18-36, которые представляют четыре тайминга по порядку.
Задержка CAS (CL)
Время, необходимое контроллеру памяти для доступа к ячейке памяти после получения команды. CL является наиболее важным параметром времени для производительности памяти DDR4, а более низкие значения CL указывают на более быстрое время доступа. Типичные значения CL для памяти DDR4 находятся в диапазоне от 14 до 19, где 14 — самый быстрый, а 19 — самый медленный.
tRCD (задержка RAS to CAS)
Время, необходимое контроллеру памяти для доступа к ячейке памяти после активации строки памяти. tRCD — второй по важности показатель времени для производительности памяти DDR4, а более низкие значения tRCD указывают на более быстрое время доступа. Типичные значения tRCD для памяти DDR4 находятся в диапазоне от 14 до 19, где 14 — самый быстрый, а 19 — самый медленный.
tRP (время предварительной зарядки строки)
Время, необходимое контроллеру памяти для предварительной зарядки строки памяти перед доступом к другой строке. Наименее важный показатель для производительности памяти DDR4, и более высокие значения tRP не влияют на производительность так сильно, как более высокие значения CL или tRCD. Типичные значения tRP для памяти DDR4 находятся в диапазоне от 14 до 19, где 14 — самый быстрый, а 19 — самый медленный.
tRAS (задержка от активации до предварительной зарядки)
Время, необходимое контроллеру памяти для активации строки памяти и последующей ее предварительной зарядки. tRAS также является важным параметром времени для производительности памяти DDR4, а более низкие значения tRAS указывают на более быстрое время доступа. Типичные значения tRAS для памяти DDR4 находятся в диапазоне от 30 до 42, где 30 — самый быстрый, а 42 — самый медленный.
Чем выше рабочая частота ОЗУ, тем выше время CL. Также отметим, что в таких случаях производительность поднимается за счёт увеличенной частоты при более менее приемлемых таймингах. Для наглядности стоит посмотреть на сравнительную таблицу самых популярных вариантов.
| Поколение RAM | Частота | CL | tRCD | tRP | tRAS |
| DDR4 | 2400 | 14 | 15 | 15 | 35 |
| DDR4 | 2666 | 15 | 17 | 17 | 35 |
| DDR4 | 3000 | 15 – 16 | 16 – 18 | 16 – 18 | 35 – 36 |
| DDR4 | 3200 | 16 | 18 | 18 | 36 |
| DDR4 | 3600 | 16 – 18 | 18 – 22 | 18 – 22 | 38 – 42 |
Значения для модулей памяти DDR4 могут различаться в зависимости от производителя и конкретной модели модуля памяти.
Подбор таймингов памяти DDR4
При выборе DDR4 необходимо учитывать несколько факторов, таких как характеристики процессора и материнской платы, предполагаемое использование системы и бюджет. Вот несколько рекомендаций, которые помогут выбрать правильные DDR4:
- Проверьте характеристики процессора и материнской платы. Характеристики процессора и материнской платы играют решающую роль в определении максимальной поддерживаемой скорости DDR4. Перед покупкой DDR4 проверьте характеристики процессора и материнской платы, чтобы убедиться, что память совместима.
- Учитывайте сценарий использования системы. Предполагаемое использование системы также может влиять на выбор DDR4. Например, если вы собираете игровой ПК, более быстрая синхронизация может обеспечить лучшую производительность в ресурсоемких играх. С другой стороны, если вы создаете рабочую станцию для таких задач, как редактирование видео или 3D-моделирование, больший объем памяти может оказаться более важным, чем более быстрое время.
- Бюджет. Память DDR4 с более быстрыми характеристиками может быть дороже, чем с более медленными показателями. Подумайте о своем бюджете и выберите подходящие варианты DDR4.
- Оптимизация вручную. Некоторые материнские платы позволяют вручную настраивать DDR4 для повышения производительности. Всегда проверяйте руководство к материнской плате и следуйте рекомендациям производителя при настройке.
При выборе DDR4 учитывайте характеристики процессора и материнской платы, предполагаемое использование системы и бюджет. Кроме того, нормальная латентность памяти DDR4 позволяет вручную настраивать timing, но это требует технических знаний и может привести к аннулированию гарантии. Выберите DDR4, которая обеспечит наилучший баланс производительности, совместимости и доступности для вашей системы.
Средние тайминги DDR3
Временные значения DDR3 зависят от конкретного модуля и частоты, на которую он рассчитан. Вот типичные характеристики для DDR3 на разных частотах:
| Поколение RAM | Частота | CL | tRCD | tRP | tRAS |
| DDR3 | 1333 | 9 | 9 | 9 | 24 |
| DDR3 | 1600 | 11 | 11 | 11 | 28 |
| DDR3 | 1866 | 13 | 13 | 13 | 32 |
| DDR3 | 2133 | 15 | 15 | 15 | 35 |
Это наиболее распространенные значения таймингов, но разные производители могут предлагать несколько разные значения таймингов в зависимости от конкретного модуля. Кроме того, некоторые модули могут поддерживать XMP (Extreme Memory Profile), который может обеспечить еще более точные значения времени для повышения производительности, но требует ручной настройки в BIOS.
Задержка памяти DDR3
Задержка DDR3 относится к времени, которое требуется для доступа и извлечения данных. Задержка DDR3 может варьироваться в зависимости от нескольких факторов, включая скорость, тайминги, установленные в BIOS системы, а также конкретный ЦП и материнскую плату, используемые в системе.
Как правило, DDR3 имеет задержку CAS (строб доступа к столбцу) 9, 10, 11 или 12 тактов. Общая задержка рассчитывается путем добавления задержки CAS к дополнительной задержке, вносимой другими синхронизациями, такими как RAS (строб доступа к строке) и tRCD (задержка от RAS к CAS).
Общая задержка DDR3 может варьироваться от 9 нс для DDR3-1600 с малой задержкой до 15 нс и более для DDR3-1066 с большей задержкой. Однако важно отметить, что хотя меньшая задержка может привести к более быстрому доступу к данным, более высокая частота и большая емкость также могут улучшить общую производительность системы.
Какие тайминги лучше для DDR3
Для большинства пользователей DDR3 с таймингами 9-9-9-24 или 8-8-8-24 на частоте 1600 МГц является хорошим балансом между производительностью и стоимостью. Эти тайминги для DDR3 относятся к задержкам CAS, tRCD, tRP и tRAS соответственно.
Однако если вы создаете высокопроизводительную систему для игр или других ресурсоемких приложений, вам может потребоваться более быстрая DDR3 с более низкими таймингами, например, 7-7-7-20 или 6-6-6-18 при частоте 1866 МГц или более. Эти более быстрые комплекты, как правило, будут дороже, но они могут обеспечить заметный прирост производительности в определенных сценариях.
Латентность памяти DDR3
Задержка DDR3 относится ко временной задержке между запросом данных из модуля и доставкой этих данных в ЦП. Измеряется в наносекундах (ns) или тактовых циклах (CL). DDR3 имеет несколько временных параметров, влияющих на ее задержку, включая задержку CAS (CL), tRCD, tRP и tRAS.
Более низкие значения DDR3 приводят к более высокой производительности и меньшей задержке. Однако реальный выигрыш в производительности зависит от конкретной конфигурации системы и рабочей нагрузки. При выборе DDR3 для системы важно учитывать характеристики ЦП, материнской платы и разные тайминги оперативной памяти DDR3, чтобы обеспечить совместимость и оптимальную производительность.
Можно ли ставить оперативную память с разными таймингами?
В компьютерную систему можно установить RAM с разными таймингами. Однако когда устанавливается ОЗУ с разными таймингами, система будет использовать тайминги самого медленного установленного модуля ОЗУ. Это означает, что если у вас есть модуль RAM с таймингом 16-18-18-36 и другой модуль с таймингом 14-16-16-32, система будет работать с обоими модулями RAM на 16-18-18-36.
Должны ли совпадать тайминги оперативной памяти
Рекомендуется использовать модули RAM с одинаковыми таймингами, чтобы обеспечить наилучшую производительность и стабильность системы. Если вам нужно обновить систему и если у оперативной памяти разные тайминги — это не должно вызывать серьезных проблем. Чтобы добиться наилучших показателей, стоит со временем заменить модули на те, что будут иметь одинаковые тайминги.
Какие тайминги лучше для оперативной памяти?
Наилучшие тайминги для RAM зависят от нескольких факторов, включая конкретный тип и скорость используемого ОЗУ, а также требования к производительности компьютерной системы. Более низкие тайминги ОЗУ лучше влияют на производительность, потому что они позволяют памяти быстрее получать доступ к данным и передавать их. Однако это также означает, что модули RAM должны иметь возможность надежно работать на более высоких скоростях, чего может быть сложнее достичь с более производительной RAM.
Высокие тайминги оперативной памяти
Обычно относятся к более медленным таймингам, которые могут ограничивать производительность вашей ОЗУ. Это означает, что RAM потребуется больше времени для доступа и передачи данных, что может привести к снижению общей результативности системы.
Некоторые возможные причины использования более высоких таймингов ОЗУ могут включать:
- Стабильность. В некоторых случаях увеличение таймингов вашей RAM может сделать ее более стабильной, особенно если вы разгоняете свою плату или другие системные компоненты.
- Совместимость. Некоторые старые или более дешевые системы могут не поддерживать высокую производительность с низкими таймингами, поэтому увеличение таймингов может помочь обеспечить совместимость с вашим оборудованием.
- Стоимость. Модули ОЗУ с меньшим значением времени могут быть дороже, чем модули с более высоким значением времени, поэтому выбор более высоких значений времени может быть более выгодным для некоторых пользователей.
Однако важно иметь в виду, что использование высоких таймингов ОЗУ почти всегда приводит к более низкой производительности, чем использование более низких таймингов, поэтому вам следует тщательно взвесить преимущества и недостатки использования более высоких таймингов для вашего конкретного случая использования.