Overclocking
Besides the XMP profile, we of course still test RAM kits manually and stably overclocked in our reviews. As already read out in the SPD, the PMIC is limited to 1.435 V, which even the Maximus Z690 Apex motherboard cannot bypass. Interestingly, the PMIC is the same chip that could deliver more voltage on the ADATA XPG Lancer modules, but that seems to be due to on-board settings of the PMIC.
In fact, the modules clock very high with their 16 Gbit B-Die ICs and limited voltage, with up to DDR5-7000 in the Gear 2. Above 6666 Mbps, however, memory training becomes a gamble, insofar as the settings are sometimes stable, sometimes throw a few errors after a few hours, and sometimes only survive a few seconds in TM5. The higher the temperature of the memory modules, the more constant the training, but preheating the RAM with a hairdryer at every system start is not really practical.
Ultimately, the maximum overclocking amounts to DDR5-6666 with tCL 36, tRCD 37, tRP 38, tRAS 76 and tCR 2, with entirely passive cooling. Although the modules reach 95 °C after a few hours in the stress test, they are still completely stable and the installed components are also designed for these temperatures according to JEDEC. If you actively cool the modules, the temperatures can be lowered well below 50 °C, but then the setting actually becomes unstable. So the Samsung 16 Gbit B-Die ICs apparently need a fairly high operating temperature to run error-free at the performance limit.
Even if other RAM modules with Hynix 16 Gbit A-die ICs are even faster or can be clocked higher, the Samsung 16 Gbit B-die memory chips can at least be tightened in the important tRCD, tRP, tRRD_s/t and tFAW timings and thus make up for a part of the performance deficit. The remaining subtimings and especially the refresh settings tREFI and tRFC cause a clear and measurable disadvantage, as we will see in the benchmarks.
Test systems
A pre-selected Asus Maximus Z690 Apex is used as the motherboard and an i9-12900K with the best out of over 60 binned i9 CPUs serves as the CPU. As always, the rest of the test hardware is available as an overview:
Test systems |
|
---|---|
Hardware: |
|
Cooling: |
|
Housing: |
|
Periphery: |
|
Measuring devices: |
|
For better clarity, the following abbreviations are used in the diagrams:
- 12900K: Intel Core i9-12900K CPU
- 51/0/49: Multipliers for P-cores x51, E-cores 0 (deactivated), cache x49 at BCLK 100 MHz
- TZ5: Trident Z5 module from G.Skill
- XPGL: XPG LANCER RGB DDR5 modules from ADATA
- DPR: Dominator Platinum RGB DDR5 modules from Corsair
- FB: Fury Beast DDR5 modules from Kingston
- DR: Dual-Rank, if not specified Single-Rank
- 3600c14: configuration with 3600 Mbps and tCL timing 14, see screenshots for all timings
- 1T, 2T: Command rate 1T or 2T
- *: completely manually set subtimings, see screenshots
- G1, G2: Gear 1, Gear2 – ratio between IMC clock and RAM clock, Gear 1 = 1:1, Gear 2 = 1:2
- H16M: SK Hynix 16 Gbit M-Die memory chips
- S16B: Samsung 16 Gbit B-Die Memory Chips
- ❄: active cooling with a 120 mm 2000 rpm fan
Timings of the DDR4 comparison config:
- 1 - Unboxing and Design
- 2 - Dimensions and RGB lighting
- 3 - Heatsink test and SPD information
- 4 - Teardown and PCB analysis
- 5 - Overclocking and test systems
- 6 - Synthetics (1/2) – Linpack Xtreme, AIDA64, Geekbench 3
- 7 - Synthetics (2/2) – SuperPI 32M, Pyprime 2.0 2B, Timespy CPU
- 8 - Gaming QHD and FHD – ACC, SoTR, CS:GO
- 9 - Zusammenfassung und Fazit
15 Antworten
Kommentar
Lade neue Kommentare
Neuling
Veteran
Veteran
Veteran
Urgestein
Mitglied
Veteran
Veteran
Veteran
Mitglied
Veteran
Mitglied
Veteran
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →