Ryzen 7 Pro 4750G with Forza Horizon 4 runs smoothly.
Ryzen 7 Pro 4750G with Forza Horizon 4 runs smoothly.
Hello everyone. I’m reaching out for some advice. I own an AMD Ryzen 7 Pro 4750G with 32GB of 3200MHz RAM (upgraded from 3000MHz). It’s running smoothly overall. The system is set to dynamic graphics, which defaults me to 1280x720 at 30fps on a low preset with minimal motion blur. My ADATA NVMe drive should handle the load without bandwidth problems. Temperatures stay below 60°C, and I’m not sure if there’s a BIOS limit I should check.
I’ve tried this setup without issues while driving, but once I start the first major event in the game (the post-opening race), it crashes. It freezes on the “GO!” screen, then sometimes goes black with a static-like flash and no sound. A restart is the only solution. The game is available via Steam, and my Windows 10 Pro 20H2 is up to date with the latest patches (KB5001330, KB4023057, released 15 April 2021).
I’ve checked for background monitoring tools or antivirus interference—my Bitdefender is cleared, and no other apps are running. I read that my specs technically fall short, but others have managed to run FH4 on similar APUs with lower specs. There’s no UMA buffer adjustment mentioned as a fix, and I’m not sure if the GPU is the main bottleneck.
Any suggestions? Is this a RAM or driver issue? Should I tweak BIOS settings or look into alternative options? Thanks in advance for your help!
It seems like there might be some confusion about memory settings and APU performance. The issue could stem from incorrect UMA buffer configurations or mismatched memory speeds. For 720p, 4Gigs should suffice, but 1080p is approaching limits recently. Adjust the UMA size in BIOS to 4GB and align the Infinity Fabric speed with your RAM speed (like 3200MHz memory clock matching 1600MHz fabric). If problems continue, consider removing two memory sticks and using a 2x8GB setup with manual UMA buffer settings. Enable or disable XMP as needed to optimize performance. For full APU benefits, ensure you have at least 4200MHz RAM.
Thank you for your message. The UMA buffer size is currently set automatically on my Gigabyte Aorus B550M Pro. It seems changing it doesn’t save the preference, even after several attempts. I’m unsure if another setting needs to be turned off. With 32GB RAM, a 4GB UMA should fit comfortably, as 1GB UMA equals about 8GB VRAM and 2GB UMA about 16GB VRAM. In the Task Manager, my dedicated GPU memory shows 512MB, but it lists 16.2GB shared and 15.7GB total. The XMP and OC settings are set to 3200MHz, with infinity fabric at 1600MHz and core clock at 2100MHz. I’ll back up the current configuration and reinstall to see if it works.
What is this motherboard? UMA buffer size is the system memory that is reserved for the APU to use as VRAM. 4Gigs is 4 gigs. You should have setting spanning al the way down to 16GB (half of your total system memory). If it goes back to whatever value it wants, then either you forgot some setting somewhere on Auto or you have another problem. Check for new BIOS version for that motherboard on the manufacturer's site. Back in the beginning of the year a lot of them had a BIOS update, concerning memory stability.
It's a Gigabyte Aorus B550M Pro-P. The UMA buffer didn't cause any issues. Reading the motherboard manual really helped. I changed the graphics to forced mode, set UMA to UMA_specified, and increased the buffer size to 8GB. (Note: My earlier explanation about UMA was just a guess.) When auto-detect tried 1080P ultra, it exceeded the APU's limits. I switched to 1080P low with no motion blur, ran the benchmark, and got around 60fps with 80-90% GPU usage. After 30 seconds, it crashed—freezing, black screen, then static grey, then back to black. The game audio kept playing. I restarted the PC and backed up the game. I saw reports about game stutters after a Windows update (KB5001330). I've uninstalled that update now and plan to reinstall FH4 to test. Will update you soon.
If the issue continues, use two memory sticks placed in the 2nd and 4th positions, counting from left to right. You might also want to test with a Furmark and see if freezing occurs there.
Apologies for the slow response. I reinstalled Windows 10 Pro from scratch using a brand new ISO. Only essential work applications were added; the system is fully updated, there are no Radeon drivers installed, and everything is functioning properly. The RAM was reset to the original 2133MHz without any XMP settings. I executed sfc /scannow once, and it resolved all issues. After that, I tested Euro Truck Simulator 2. Given its age, the APU should handle the game without major problems. I set it to 1080p High resolution, which worked until it crashed. In the Control Panel, I reviewed the reliability history and noticed these events occurred around 23:25 (almost immediately after each other) and then at 23:26 ETS2. Source: Windows Summary Hardware error Date 2021/04/25 23:25 Status Not reported Description A hardware issue halted Windows properly. Problem Event Name: LiveKernelEvent Code: 141 Parameter 1: ffff800c889f2460 Parameter 2: fffff8078f4d5230 Parameter 3: 0 Parameter 4: c7c OS version: 10_0_19042 Service Pack: 0_0 Product: 256_1 OS Version: 10.0.19042.2.0.0.256.48 Locale ID: 7177 Source Windows Summary Hardware error Date 2021/04/25 23:25 Status Not reported Description A hardware problem stopped Windows correctly. Problem signature Problem Event Name: LiveKernelEvent Code: 117 Parameter 1: ffff800c8f1d9050 Parameter 2: fffff8078f4d5230 Parameter 3: 0 Parameter 4: 0 OS version: 10_0_19042 Service Pack: 0_0 Product: 256_1 OS Version: 10.0.19042.2.0.0.256.48 Locale ID: 7177 Source Windows Summary Hardware error Date 2021/04/25 23:25 Status Not reported Description A hardware issue stopped Windows correctly. Basically LiveKernelEvent 141, 117 and then 141 again (all in seconds from each other) followed by ETS2 crashes. In this instance, APU settings were default, so the UMA buffer size was at 512MB (as shown in Task Manager for Dedicated GPU Memory). I ran sfc /scannow once more and found no issues. The RAM is rated for 3000MHz, so I’m unsure if it should be set to that speed or not; currently I stick with the base 2133MHz. I’ll test Furmark today.**
Updated: Ran Furmark at 1080p using 2xMSAA, RAM configured at 3000MHz with XMP activated. UMA buffer configured to 4GB. Issues persist as described previously. LiveKernalEvent 141, 117, then 141 all occurred within seconds. I plan to lower the RAM to 2133 and test again.