I built a new Windows 11 PC with an Intel Core Ultra 7 270K Plus, ASRock Z890 Steel Legend WiFi, 32 GB of Corsair Vengeance DDR5-6000 CL30 (CMH32GX5M2B6000Z30K), an RTX 5070 Ti, and BIOS version 3.31. The system worked initially, but I began getting frequent blue screens, game crashes, and random Chrome "Aw snap" errors. Unreal Engine and D3D12 games are especially problematic, sometimes freezing both monitors before crashing. Errors have included DXGI_ERROR_DEVICE_REMOVED, texture/resource allocation failures, DPC_WATCHDOG_VIOLATION, and IRQL_NOT_LESS_OR_EQUAL. I reseated the GPU and tested each memory module individually, with no obvious fault found. The only consistently stable configuration so far is Intel 200S disabled with memory at the default DDR5-4800 speed. Intel 200S disabled with the memory running at its XMP settings of DDR5-6000, 30-36-36-76 at 1.40 V, is unstable. Intel 200S enabled with DDR5-6000 is also unstable. The motherboard appears to link its normal XMP Profile 1 selection with Intel 200S Profile 1, so I instead used the BIOS option to load and apply the XMP timings manually while leaving 200S disabled. The system then reports the memory running at 6000 MT/s, but crashes still occur. ASRock support tested the board and could not reproduce the problem, although their benchmark screenshot appears to show the integrated graphics rather than the RTX 5070 Ti. Could unstable DDR5, the CPU memory controller, or the motherboard cause graphics-driver and Unreal Engine errors such as DXGI_ERROR_DEVICE_REMOVED? Are there additional BIOS or memory tests I should run before considering an RMA?
3 Answers
XMP is technically memory overclocking, so even a kit rated for DDR5-6000 is not guaranteed to run at that speed on every CPU and motherboard combination. Since DDR5-4800 is stable while 6000 is not, the memory controller, board training, or the particular CPU's silicon is a strong suspect. It is also entirely possible for bad memory data to surface as NVIDIA, D3D12, Unreal Engine, browser, or seemingly unrelated driver crashes rather than a clean memory error. I would run MemTest86 or another bootable memory test for several passes, test with 200S disabled, and verify that the system is actually using the discrete GPU. If the kit fails at its advertised settings after BIOS updates and conservative manual settings, testing another memory kit or considering a CPU/board RMA would be reasonable.
The link between XMP Profile 1 and Intel 200S is worth investigating. 200S Boost changes more than memory speed; it can also raise internal CPU interconnect and related clocks. Look for a separate XMP option, memory preset, or manual DRAM configuration that leaves every 200S setting on Auto or Disabled. Then test DDR5-5600 and DDR5-6000 separately, using conservative command rate and memory-controller settings if the BIOS exposes them. Also make sure the BIOS is current. Passing a quick Windows memory test does not prove the configuration is stable, so MemTest86, TestMem5, Karhu, or extended system and gaming tests would be more useful.
The BIOS is already updated, and I found the manual DRAM option that applies the XMP timings without enabling 200S. I haven’t run MemTest86 yet, so that is next on my list.
The fact that it is stable at the JEDEC DDR5-4800 setting points more toward a marginal high-speed memory setup than a definite graphics-card failure. Try each DIMM in the recommended slot arrangement, then test the complete kit at 4800, 5600, and 6000. If 6000 fails even with 200S disabled and verified timings and voltage, try another known-good kit or leave it at the highest speed that passes a long memory test. Unstable RAM can corrupt graphics resources and trigger device-removed errors, so the GPU message does not automatically mean the RTX 5070 Ti is defective.
I’m trying to complete a few more tests before starting an RMA. The board’s XMP menu is confusing because selecting Profile 1 also changes the 200S profile, so I’ve been applying the DRAM settings manually instead.