With Intel Alder Lake, a new memory standard is finding its way into PCs: DDR5 RAM. Since the twelfth generation from Intel also benefits from high RAM clock rates and optimized timings, the question arises as to what the best approach to DDR5 is in order to achieve success. Community member Alexander provides the answer. The author of this article is registered on the ComputerBase forums with the username Esenel and is known for his expertise in the field of RAM-OC. In the community benchmark tests, it usually takes first place, also thanks to the faster CPU and GPU, but also thanks to the RAM settings that have been pushed to the optimum. The memory was made available on loan by the respective companies at the request of the editors.
DDR5: RAM OC odyssey through the beta BIOS jungle
With Intel Alder Lake as the 12th Gen Core (test) and DDR5, a new playground has entered the PC for enthusiasts. However, first attempts at walking with the new technology are still adorned with obstacles and odysseys. Depending on the memory chips installed on the DDR5 modules, the BIOS used plays a very important role and ultimately decides whether the PC starts correctly even in JEDEC mode or not. The “laws” known from DDR4 do not seem to apply, at least at the start. Where the well-loved Samsung memory chips (B-Die) have been setting the bar for Intel systems for years with DDR4, the picture for Samsung memory chips with DDR5 darkens at the start. Only with these chips did the problem arise that booting was not even possible in the JEDEC profile – XMP 3.0 was also affected. But the Hynix memory chips were not exempt from this problem either. In this context, however, the use of ASUS by Peter “Shamino” Tan is commendable: Within a very short time, three revised BIOS variants were available during the test period and were able to eliminate the problems. With the Micron memory chips, such behavior could not be determined from the start.
Three DDR5 RAM kits, three chip manufacturers in the test
Three RAM kits with three different memory chips currently available were available for this test, although none of the manufacturers guarantees the permanent use of the same memory supplier.
Corsair Dominator Platinum RGB DDR5 (Image: Corsair) G.Skill Trident Z5 RGB (Image: G.Skill) TeamGroup T-Force DELTA RGB DDR5 (Image: TeamGroup) The key technical data of the modules read as follows:
The settings for the following test course
The XMP 3.0 profiles of the memory kits and an optimized overclocking profile were used for the test course. The basis for comparison is the JEDEC specification of the platform for two DIMM modules on the Asus ROG Maximus Z690 Apex, DDR5-4800 MHz two-DIMM slot motherboard, and the much more common DDR5-4400 MHz specification for two DIMM modules on one Four slot board. Both were imaged with Corsair’s modules. Testing the JEDEC, XMP 3.0 and OC profiles
OC requires higher voltages
In order to achieve the clock rates and timings in the overclocked state, it was necessary to increase the voltage further. With DDR5 there are currently five different voltages that play a role in overclocking the main memory: VDD, VDDQ, IVR TX VDDQ, IMC VDD and IVR SA. An explanation of all of these voltages is beyond the scope of this review, so the focus will be on two voltages—VDD and VDDQ. The other voltages can run on “Auto”. The Micron memory chips installed on the Corsair Dominator Platinum RGB (Special Edition) scale little or not at all via the voltage. With a moderate voltage of 1.30 volts VDD and 1.35 volts VDDQ, a maximum is quickly reached at DDR5-5400 or DDR5-5600 MHz. The Corsair modules did not have any problems with the temperatures.
The G.Skill modules run (too) hot
In complete contrast to the G.Skill modules. The assumption is that their new heatspreader does not dissipate the power loss well – of course, that shouldn’t be the case. However, even in XMP 3.0 operation using a separate fan for RAM cooling at 2,000 rpm, the temperature of the RAM kit does not drop below 50 °C. Obviously there is a problem at the transition from the chips to the cooler. In the end, the Teamgroup modules with Hynix memory chips offered the highest OC potential, also due to the temperature. Visually, the cooling system is not the big eye-catcher compared to the other RAM kits, but it shines with good heat dissipation and the chips scale well over the voltages. Even with a voltage of 1.435V VDD, the temperatures were a low 42 °C.
Z690 BIOS versions: a tale of woe
The Asus ROG Maximus Z690 Apex, which was explicitly designed for RAM OC with only two DIMM slots, was used for the test. The BIOS version 0231 of the delivery status was skipped, like version 0702, which the press received for the Alder Lake tests on November 4, 2021, since Asus directly provided another optimized BIOS 007(0). This was intended for more stability with Micron memory ICs and remained on the first BIOS chip even during stability and performance tests. Luckily the Apex has two. Asus ROG Maximus Z690 Apex (Picture: Asus) Because when Teamgroup’s kit with the Hynix ICs was used, BIOS 007(0) was still used for initial testing, but showed that it didn’t work well with it. Luckily, Asus was already 100 BIOS versions ahead via “Shamino” and BIOS 0093, presumably better for Hynix ICs, found its way to BIOS chip number 2. This enabled me to test the stability in OC operation with TestMem5 and validate Karhu RAM test. The performance measurements were also carried out with it. The change to the memory from G.Skill with ICs from Samsung then showed again: BIOS 0093 does not work. “BIOS Generator” Shamino then made 0010 available as an “XMP Hotfix” for the Trident Z series and this version enabled clean operation in the XMP and the selected OC profile. This version was also used for the performance measurements of the Trident-Z modules. While this article was on target, two more BIOS versions were released for the Apex, 0811 for more stability when using XMP and a special BIOS with version number 0021. You can see: The testing doesn’t end for the time being.
Stability Testing: Why the Effort?
A separate Windows installation was used for all stability tests. Luckily. Because once a setting that was much too sharp was tried and Windows 11 was already gone, which resulted in three days of debugging. Because every time while Karhu was running and an application was opened at the same time, an error occurred, even if the computer had been running error-free for three hours beforehand. It is therefore always advisable to use a separate hard disk with Windows for such tests. A finished setting was only declared stable if it lasted 20 cycles through TestMem5 with the profile 1usmus_v3 (approx. 2 h running time) and 10,000% Karhu (approx. 3 h). Individual settings were also checked with HCI MemTest (3 h). Alder Lake has shown that TM5 reacts very quickly to instabilities. Why do you take on this enormous effort at all? On the one hand, of course, for the fun of it, but mainly because most of the performance lies in the reduction of the secondary and tertiary timings of the main memory. It’s usually relatively useless to just load the XMP and hope for a huge leap in performance. If you want to see results, you have to get into the details – and that takes time. However, the test results on the following pages will show that this is not always crowned with success. On the next page: test results and benchmarks
#Overclocking #DDR5RAM #brings #XMP #high #clock #optimized #timings