Performance issues reported on 3DMark assessments.
Performance issues reported on 3DMark assessments.
Full Specs: CPU: 7950X3D GPU: RTX 4090/RTX 5090 RAM: Corsair 64GB DDR5 6000 (CMP64GX5M2B6000Z30) Motherboard: Asus ROG Strix X870E Wi-Fi PSU: Seasonic GX-1200 (12VHPWR plug in use for GPU) I'm having an issue with my TimeSpy scores that appear to have followed a motherboard swap. Back in 2023 when I first benchmarked this system on my old motherboard (An Asus TUF X670E+Wifi), I got the following score: https://www.3dmark.com/spy/42589707 (CPU: 16000~, GPU: 34000~). Basically in-line with expected results. I swapped out to a new motherboard (Strix X870E) and RAM a little over a month back and foolishly did not benchmark at the time. When I tested today, however, I discovered that while the CPU scores met expectations, the GPU scores had dropped significantly: https://www.3dmark.com/spy/53049328 (28000~) When I run benchmarks that stress the GPU more heavily, the results appear normal. When I stress the CPU specifically, the results appear normal (Cinebench R23 gets 37000 pts, Passmark returns 63K pts on the CPU test, Unigine Superposition 4K Optimized score 32K~). When I run Furmark, I get full power draw (450W). However in TimeSpy, the biggest indication that something is wrong is that my GPU usage is very low and the power draw is similarly reduced (300-350W). GSync and VSync are both disabled in Nvidia's settings for these settings. I additionally note that when reducing settings in video games such as Cyberpunk 2077, I will see similarly low GPU usage and poor framerates. The biggest indication that there is an issue is that when I plug in my new 5090, I get the same exact scores: https://www.3dmark.com/spy/53047294 Timespy Benchmark (4090) TimeSpy GPU Usage Timespy Benchmark (5090) Firestrike Benchmark CPU Benchmark What I have tried: Ensuring temperatures were within normal ranges Using Process Lasso to force 3DMark onto the cache cores alone or onto the frequency cores alone on the High Performance power profile. Additionally tried disabling Process Lasso and enabling Game Bar/Performance power. Enabling EXPO 1, EXPO 2, Disabling Expo Updating motherboard BIOS Using Default BIOS settings, swapping CMOS battery Enabling PBO and applying a stable undervolts (-20 max) Enabling CPB Ensuring Resizable Bar is enabled Ensuring max bandwidth from PCIE slot (Correct slot, no interfering PCIE SSDs hogging lanes) Clearing GPU drivers in safe mode with DDU, updating drivers, downgrading drivers. Updating Windows 11 Clearing AMD Chipset drivers with Revo uninstaller and reinstalling Unplugging monitors from motherboard DP Unplugging memory sticks one at a time, swapping around to different slots Disabling Memory Integrity in Defender Honestly, I feel like this is a hardware fault, but I've no idea what it could be at this point. Any ideas?
It seems the outcome is fairly typical. Around 15 participants achieved higher scores, which might explain the confusion. I assume the results align with expectations. What were your predictions?
The typical score in TimeSpy (Non-Extreme) for the 4090 is between 34-36K, not 28K. Results on the 5090 vary, but most reviewers mention around 46-48K as the expected outcome. I won’t get the same GPU score for both the 4090 and 5090. The GPU utilization is noticeably lower than anticipated. Asking a friend with the same setup didn’t encounter this problem.
Hey, same situation here—just not using Project Lasso. I’ve been looking into the details and your analysis. What else could be useful? You’re running a 5900x with 5090 and afterburner, plus CTR on the same board. Thanks!