I bought a used Corsair Dominator Titanium 96GB DDR5-6000 CL30 kit made up of four 24GB modules for an AM5 system with an ASUS X870E-E motherboard and a Ryzen 9850X3D. The kit appears on the motherboard's memory QVL, and I'm running it with EXPO I enabled while leaving everything else at stock settings. Four DIMMs aren't ideal for the memory controller, but the kit was considerably cheaper than most 96GB alternatives.
Karhu RAM Test ran for about 13 hours and reached 5250% coverage before reporting an error. TestMem5 with the anta777 Absolut configuration completed two cycles in roughly 2.5 hours without errors. The system is mainly used for gaming with an RTX 5090, although I also need the extra memory for virtualization. My previous 64GB G.Skill DDR5-6000 CL30 kit has been reliable for years, and the system was stable before installing this used memory.
Would you consider this configuration stable enough for gaming, or should I return the kit and look for memory that can pass longer stress testing without errors?
3 Answers
For a gaming system, passing several hours of TestMem5 is encouraging, but a reproducible Karhu error means I’d avoid treating it as fully validated. Try the simplest troubleshooting sequence first: reseat the DIMMs, test with a fan over the memory, run the kit at a slightly lower memory speed, and then test each pair or module if necessary. If it only passes after reducing speed or substantially tuning voltages, decide whether the extra capacity is worth the compromise. Since this was a used purchase, returning it is reasonable if it cannot pass at its advertised EXPO settings with sensible temperatures.
A failure after more than 5,000% coverage is still a failure, so I wouldn’t call the configuration completely stable yet. Memory temperature is one of the first things I’d investigate, especially with four modules and a powerful GPU heating the case. Put a fan directly over the DIMMs, monitor their temperatures, and repeat the test. If the error disappears, cooling was probably the issue; if it remains, the EXPO settings or the memory controller may need adjustment.
Four high-density DDR5 sticks put considerably more stress on the AM5 memory controller, even when the modules are listed on the motherboard’s QVL. I’d check DIMM, memory-controller, and SoC voltages, along with temperatures, rather than assuming EXPO is guaranteed to work perfectly in this configuration. It’s also worth testing under a combined CPU and GPU load, since the RTX 5090 can raise case temperatures substantially. A compute test such as Linpack may help reveal instability that ordinary gaming does not.
I haven’t changed any voltages beyond enabling EXPO, and I didn’t run a simultaneous GPU load during the memory test. I’ll monitor temperatures and try a combined-load test before deciding whether to return the kit.
That makes sense. The system was stable with the previous kit, so my main goal here is confirming whether this used memory is reliable. I’ll repeat the testing with HWiNFO monitoring and additional airflow over the DIMMs.