Comparison of 7950x3D CCD0 and CCD1 using Cinebench R23 with HWinfo64 data
Comparison of 7950x3D CCD0 and CCD1 using Cinebench R23 with HWinfo64 data
General update on direct-die and disabled CCDs. The 7950x3D and X670e Aorus Master appear to struggle with disabled CCDs, particularly CCD0 during voltage control. CCD0 showed improved performance but CCD1 didn’t provide much voltage control. Direct-die enabled me to adjust the SoC voltage, possibly freeing up more power for the CCDs. I achieved 154.6W draw and nearly 5GHz consistently (~4.985GHz). While I didn’t capture the full test data, using the UEFI’s special Vcore load line feature helped—set it to 'turbo' for optimal results.
High voltage load profile, 1.1V SoC, 90°C/+200/-25,-15/ PBO, liquid metal (replacing kryosheet) direct-die NH-D15. 37531 R23 where it first reached 5.25GHz across all cores. Probably near TDP limits because it hasn't even hit 85°C. Seems no option to bypass this setting.
I found my numbers a bit different from yours. I got 35800 on a TR PA120 with two P12 Max modules, operating at 84.5°C max, no direct die, stock TR paste. All CCD0 cores hit around 5.25 and CCD1 around 5.75. They run in CBR at about 4.9/4.975 CO:CCD ratios – CCD0 -30/CCD -20 with PBO +200, auto LLC. It's hard to compare CO undervolt and SoC voltage offset since yours uses variable increments, but your voltage seems lower at max 1.15V VID compared to my 1.275. Still, why are the temperatures so similar even with worse cooling?
Adjusted SoC voltage to 1.2V for testing; observed a reduction in power consumption by 4-8W, freeing up energy for the CCDs. Lower than 1.1V hasn't been explored yet. Gigabyte UEFI includes a Vcore load line option. I'm also applying a 10mV higher negative offset on CCD0's cores compared to CCD1. Most advanced configurations were avoided until after direct-die, leaving uncertainty about stability with the added copper, solder, and TIM.
The 3D v-cache mod provides strong insulation for CCD0, which is what I aimed to verify. A roughly 25°C temperature rise was observed when comparing CCD0 to CCD1 core temperatures. The same effect appeared even with the lid on your unit. The expected idea was that the 7950x3D would behave similarly thermally to a standard 7950x, which aligns with what I saw—your 7950x3D achieved 37531 at 154W, performing impressively.
When you rerun the tests with reduced top and C-state data, the power consumption fits well. It’s unlikely your typical wattage will exceed 135W. Alternatively, the 7950x3D was undervolted due to direct-die effects, which actually improves efficiency by making better use of available TDP and maximizing headroom to stay close to thermal limits. This explains the nearly 2000-point higher score I see. Also worth mentioning is a run average boost clock of 4.97GHz.
Hello~ I understand this discussion was shared earlier, but I’m here to assist if you notice my query later. I recently assembled a PC with a 7950X3D and a crosshair X670E, I have limited experience with overclocking. My goal is to keep the processor at its stock speed, but I’m seeing a low score on Cinebench R23 (around 30K). I only know my CPU temperature stayed near 85°C during CR23 runs without changing any BIOS or OC settings on the motherboard (BIOS version 2007). I’m curious if this is typical for a 7950X3D without enabling PBO/OC settings, since many users report scores above 36K or even 37K while not adjusting those parameters. My PC store returned the CPU to them, who confirmed it’s functioning within limits and suggested possible silicon quality issues. With all that in mind, do you think my score is due to luck rather than hardware? Appreciate your support ahead of time.
Hello~ I understand this discussion was shared earlier, but I’m here to assist if you notice my query later. I recently assembled a PC with a 7950X3D and a crosshair X670E, I have limited experience with overclocking. My goal is to keep the processor at its stock speed, but I’m seeing a low score on Cinebench R23 (around 30K). I only know my CPU temperature stayed near 85°C during CR23 runs without changing any BIOS or OC settings on the motherboard (BIOS version 2007). I’m curious if this is typical for a 7950X3D without enabling PBO/OC settings, since many users report scores above 36k or even 37k while not adjusting those parameters. My PC store returned the CPU to them, who confirmed it’s functioning within limits and suggested possible silicon quality issues. With all that in mind, do you think my score is due to luck rather than hardware? Appreciate your support ahead of time.
Expectations are around 34k-36k depending on cooling and binning. Sharing a sample run will help compare wattages, voltages, boost clocks, temperatures, etc. After running, I cleared HWinfo64 data right away and took a screenshot once it finished. If there were issues, it might be the cooler mount or defective indium solder between die and IHS.
Thank you for your message. I returned my PC to the shop that assembled it. They explained that the low CR23 score was caused by an outdated chipset driver from the ArmoryCrate update on my ASUS motherboard. After reinstalling Windows and manually downloading the correct chipset driver from AMD, the CR23 score improved as seen in the photo. However, I’m still puzzled because the clock speed under "Cores x GHz" shows 4.33Ghz instead of the expected 4.2Ghz. Do you know why it’s displaying 4.33Ghz? I reviewed your detailed CR23 benchmark post—your photos consistently show a base clock of 4.2Ghz for 7950x3d. P.S.: They mentioned all motherboard settings are default, with no PBO, curve optimization, or c-state disabled, yet I suspect there might be something missing.