The holidays are getting closer, the holidays are getting closer … Maybe not just such traditional holidays, but slowly a great event for equipment enthusiasts is approaching, because on the horizon there are already Intel Rocket Lake processors (Core 11 Gen ). In turn, the recent debut of AMD Ryzen processors 5000 meant that I decided to refresh the test procedure a bit before this duel – new programs will come, some old ones will fall computer games and their place will be taken by newer titles. Oh, a classic update, but as always there are a few dilemmas and one concerns the choice of DDR4 memory. Before I make the final decision, I would like to hear your opinion on what should be included in the test platform.
RAM memory is an important component of any test platform, especially in the case of processors, so they must be chosen carefully so as not to complicate the interpretation of results.
“src =” http://www.purepc.pl/image/felieton/2021 / 01 / 19 _ what_memory_ddr4_chcielibyscie_w_processor_tests_total_question_of_new_measuring_procedure_nc0.png “> )
There is one goal and there are four balls, so read carefully and think about the possible scenarios I have analyzed briefly, because some dependencies come out only after a few (a dozen) tests and could escape with purely theoretical considerations. The survey is posted at the bottom of the page, open until the end of February, then I start with the processor tests and complaints will not be considered.However, it will be a valuable tip, as you see the issue of selecting components for hardware platforms. Honestly, a much bigger problem than choosing DDR4 memory turns out to There is a choice of computer games that use more than a modest six cores . Although in recent years we have had a real festival “winyj rdzyniuf”, the developers have not yet adapted their engines to the capabilities of modern equipment, which results in a waste of its potential. Moving to the newer API (Vulkan / DirectX 12 has also flipped some calculations from the CPU to the graphics chip, which further complicates the situation. Apart from a few titles, there is simply nothing to test for multi-threading …
“src =” http://www.purepc.pl/image/felieton/2021 / 01 / 19 _ what_memory_ddr4_chcielibyscie_w_processor_tests_total_question_of_new_measuring_procedure_nc1.png “> )
Option 1 – DDR4 Settings 3200 MHz
DDR4 clock speed for all processors would be 3200 MHz, representing the group of the most frequently chosen modules with the widest availability (you can check it on the skinflint), and at the same time corresponding to the fastest controllers officially used in processors. DDR4 kits – 3200 with various latencies are standard nowadays for all platforms that do not have rigid restrictions in the above scope, e.g. through the motherboard, therefore such settings seem to be the most universal. there would be additional tests of ed The performance of processors with slower modules, which will ultimately hit platforms with top-down blockades (e.g. Intel Core i3 – 10100 and DDR4 – 2666 on H 410). Similarly, additional performance tests would be performed with faster modules for processors, which allow the use of high-clocked DDR4 without major problems and additional costs (surcharges to the motherboard).
Option 2 – DDR4 settings depending on CPU controller
Timing DDR4 memory would be selected based on the official maximum supported by the controller installed in the processor. For example, AMD Ryzen 7 5800 X would work with DDR4 modules – 3200, while Intel Core i7 – 10700 K would get DDR4 – 2933, because such parameters were selected by manufacturers using the JEDEC guidelines. Rigid adherence to the specifications, at least theoretically, guarantees the highest stability, but does not take into account the full potential of the platform and causes the so-called the advantage of distorting the results. Why? Since the DDR4 memory clock significantly affects the performance of processors, therefore, by introducing a variable, even directly resulting from the specifications of the controllers, we distort the measurement methodology in the assumptions aimed at unifying the conditions for all tested components.
Option 3 – DDR4 Settings 3600 MHz
DDR4 memory clock for all processors would be 3600 MHz, responding to the so-called sweet spot – a frequency that provides a noticeable increase in performance over DDR4 – 3200, although the sets with the selected timing do not yet belong to the most expensive ones, and most of the DDR4 – 3200 can achieve such MHz (the issue of timings remains open). Oh, DDR4 memory overclocking is possible on most modules without any special knowledge and regulating delays further. However, there is a risk that some older platforms and / or processors may be unable to work stably with these settings (AMD AM4 and Ryzen series models 5000 ). In addition, the base results for Intel processors, paired with boards that prevent the installation of faster DDR4, will be really badly shot. Therefore, additional tests with lower divisors will be necessary (e.g.: 2666 MHz), nevertheless high output settings will no longer have higher dividers.
Option 4 – How much did the factory give (well almost)
The DDR4 memory clock would be adapted to the capabilities of the controller in the processor, the capabilities of the platform and the modules themselves. For example – AMD Ryzen 3000 and 5000 could work with DDR4 clock speed – 3800 in synchronous mode (asynchronous allows higher frequency, but at the expense of performance), while Intel Core 6 – 10 generation 4200 – 4400 MHz depending on timings. Here it was also advisable to calibrate the first, second and third order lags, that is, higher driving school. The problem is the inability to find a golden mean for the AMD and Intel platforms, as the memory controllers work very differently in both cases. It also introduces additional variables that affect performance scores. Besides, such results strongly distort the average, because most users limit the interference with the memory parameters to enable the XMP profile.
Survey – What DDR4 memory would you see in the CPU test routine?
The form is available only for logged in users.
Source: PurePC
