Poor Low Cinebench R23 results with a Ryzen 7 5700X in stock condition.
Poor Low Cinebench R23 results with a Ryzen 7 5700X in stock condition.
This chip should match the 5800X around 15,000 but I’m capped at 12,600. Temperatures hover near 55°C, ruling out thermal throttling. Cinebench process priority is set to realtime. Hardware: Gigabyte B450M DS3H with Mobo Ryzen 7 5700X, paired with Noctua NH-D15 2x16GB Corsair Vengeance RAM at 3200mhz, a Corsair CX450M GPU, and a PSU that barely meets expectations. Any suggestions for why my multicore isn’t performing better? Thanks ahead.
PBO status: All core speeds listed by HWInfo during testing. Temperatures shown by HWInfo while the bench is operating.
Besides what others mentioned: The images I found don’t show any VRM heatsinks. The D15 cooler isn’t the typical downward-flowing type seen in most standard boards. It’s possible the issue is thermal throttling on the VRMs, not the CPU. This might not be a concern for a regular 5700X, but it would be worth checking the temperatures if the board has monitoring sensors (HWiNFO should detect them, and on my board they list [OEM and mobo model#] - [Name of VRM controller]).
When discussing performance standards, small issues can quickly add up. It reminds me of the Hansel and Gretel tale—“Nibble, nibble, little mouse, who’s been nibbling at my house?” Essentially, minor setbacks in one part can lead to major problems overall. Still, aside from board and PBO concerns, your RAM might be the culprit. I’ve noticed Ryzens often dislikes Corsair RAM lately, causing crashes and instability. It could stem from a mix of board, RAM, and PBO issues. Personally, I wouldn’t use a 5000-series on a B450; it’s risky to waste power on that PSU. It’s wise not to cut corners—this is just one reason why. Ryzens tend to perform best with Crucial Balistix. My setup includes a 5900X and 3060ti with two 16GB Crucial Ballistix DDR4-3200 CL16 modules on a B550, which works well on CB R23 with standard DOCP 2666-3200 settings, voltage and timings. Edited June 27, 2022 by An0maly_76 Revised, more info
Cinebench tends to ignore memory speed specifications. Using a tuned B die (dual rank 3800MT/s with optimized subtimings) on my setup (5900x, fixed at 4.5GHz for consistency) gave 22912. With JEDEC RAM (2133 auto settings, still dual rank) it reached 22851. The differences are minor. Fast memory definitely helps, but it's possible the clock frequency is set too high—like running a 4000MT/s kit with a 2000MHz FCLK—which can lead to performance drops. I wouldn't blame memory here. More likely, the issue lies elsewhere: a faulty motherboard causing throttling or PBO not being enabled. Ballistix RAM isn't ideal for Ryzen overclocking, even though they're decent for value, and prices have risen significantly. Your system is working well beyond what memory alone should provide.
I never emphasized RAM as essential for top performance. Still, I believe it could be among several elements causing significant slowdowns. The B-die seems to offer better results, though I usually pick parts that give good value. With recent market changes, the situation has shifted a bit. Maybe I’ll consider a B-die later if trends hold. In any case, I think most agree on OP having mismatched components that don’t quite fit well together. Edited June 27, 2022 by An0maly_76 Updated with more info
In Cinebench (the platform OP is discussing) the results differ. RAM doesn’t affect performance here because the cores operate independently (FCLK isn’t critical). The PSU and motherboard aren’t major bottlenecks, though the 5700X draws a moderate amount of current—around 70A in Cinebench IIRC. That means the case should handle it if it has good airflow. The power supply isn’t exceptional and might need an upgrade, but it shouldn’t be the main reason for the low score.
Current BIOS details? Previous CPU model might differ. Latest AMD firmware updates? Check for the newest drivers. Your single-core performance looks fine, but hyperthreading or one core might be off. For instance, if you used a Ryzen 5 2600 and adjusted settings to allocate 12 cores/threads without saving, that could affect results.
For an affordable 8x2 setup, consider purchasing some 2666 BIN crucial or micron bare PCBs with ICs ranging from D9VPP to C9BJZ (worst to best). These are around $23 on eBay and offer similar value compared to what you can find in India. The same ICs as Ballistix are available, providing comparable OC performance (4600+ C18/17 1.6V+). Ballistix is only advised for OC kits because otherwise it doesn’t deliver good value—especially since Corsair RGB RAM is now cheaper and older Ryzen CPUs no longer reject Corsair memory. This setup looks like a board VRM throttle or PBO off; I recommend enabling PBO first and testing adding a fan over VRMs to see the impact.