F5F Stay Refreshed Hardware Desktop Is the Ryzen 9900X3D showing poor performance on a single core?

Is the Ryzen 9900X3D showing poor performance on a single core?

Is the Ryzen 9900X3D showing poor performance on a single core?

Pages (2): 1 2 Next
G
Gapyyyyy
Member
64
12-05-2021, 09:02 AM
#1
Hi - My single-thread benchmark scores are extremely low, while multi-thread scores remain normal. I've run OCCT, 3DMark CPU Profile, and CPU-Z multiple times, but all show significantly reduced single-thread results compared to the average for that CPU. The stability test in OCCT is error-free, temperatures stay under 80°C, and clock speeds match those of other users. The only observation is that in OCCT, the latency/bandwidth benchmark displays unusually high latency. L1 cache is 2.63ns, L2 cache 5.86ns, L3 cache 77.21ns, and memory latency is 98.11ns. Could this be the cause? I also noticed HWINFO occasionally reporting Core Utility usage in the millions (e.g., 100,000,000%), which isn't clear if it's relevant. I've tried enabling/disabling PBO and AXMP, updating the BIOS and drivers, but nothing changed. Is this a hardware issue or something fixable?

PC details:
MSI X870E Gaming Plus WIFI (7E70v1A41)
9900X3D
32GB DDR5 6000Mhz CL32
5080 (591.44)
Windows 11 26200.7462 (KB5072033)

Thank you for your help.
G
Gapyyyyy
12-05-2021, 09:02 AM #1

Hi - My single-thread benchmark scores are extremely low, while multi-thread scores remain normal. I've run OCCT, 3DMark CPU Profile, and CPU-Z multiple times, but all show significantly reduced single-thread results compared to the average for that CPU. The stability test in OCCT is error-free, temperatures stay under 80°C, and clock speeds match those of other users. The only observation is that in OCCT, the latency/bandwidth benchmark displays unusually high latency. L1 cache is 2.63ns, L2 cache 5.86ns, L3 cache 77.21ns, and memory latency is 98.11ns. Could this be the cause? I also noticed HWINFO occasionally reporting Core Utility usage in the millions (e.g., 100,000,000%), which isn't clear if it's relevant. I've tried enabling/disabling PBO and AXMP, updating the BIOS and drivers, but nothing changed. Is this a hardware issue or something fixable?

PC details:
MSI X870E Gaming Plus WIFI (7E70v1A41)
9900X3D
32GB DDR5 6000Mhz CL32
5080 (591.44)
Windows 11 26200.7462 (KB5072033)

Thank you for your help.

S
Skumbag_Robin
Junior Member
12
12-06-2021, 08:02 AM
#2
Depends on the specific problem at hand. I believe the main concern is a weak MoBo and insufficient VRMs. In the chip review, you can find the link: A MSI X870E Carbon Wifi MoBo was tested. It features 18+2+1 VRM conf, rated at 110A with SPS. On the other hand, your MSI X870E Gaming Plus Wifi has 14+2+1 VRM conf, rated at 60A with Dr.MOS (SPS is significantly better than Dr.MOS). This MoBo performs well with Ryzen 5 chips, but it falls short for Ryzen 9. For example, when running R7 9800X3D paired with AsRock X870E Nova Wifi (20+2+1 VRM conf, 110A with SPS), the performance is much better. It seems the issue might lie with the MoBo, though it could also be related to other factors. This is the only thing I can think of at the moment.
S
Skumbag_Robin
12-06-2021, 08:02 AM #2

Depends on the specific problem at hand. I believe the main concern is a weak MoBo and insufficient VRMs. In the chip review, you can find the link: A MSI X870E Carbon Wifi MoBo was tested. It features 18+2+1 VRM conf, rated at 110A with SPS. On the other hand, your MSI X870E Gaming Plus Wifi has 14+2+1 VRM conf, rated at 60A with Dr.MOS (SPS is significantly better than Dr.MOS). This MoBo performs well with Ryzen 5 chips, but it falls short for Ryzen 9. For example, when running R7 9800X3D paired with AsRock X870E Nova Wifi (20+2+1 VRM conf, 110A with SPS), the performance is much better. It seems the issue might lie with the MoBo, though it could also be related to other factors. This is the only thing I can think of at the moment.

R
ricby
Senior Member
681
12-08-2021, 03:56 AM
#3
If the problem were power-dependent, it would likely affect the multi-thread score too. The MSI site also confirms my motherboard is an SPS. Could the high cache or memory latency be the cause?
R
ricby
12-08-2021, 03:56 AM #3

If the problem were power-dependent, it would likely affect the multi-thread score too. The MSI site also confirms my motherboard is an SPS. Could the high cache or memory latency be the cause?

S
StyleTrick
Senior Member
744
12-08-2021, 07:48 AM
#4
I'm not confident about the consistency of the OCCT cache/memory latency tests, as I've noticed inconsistent results. It might help to compare with another test to verify the numbers. You mentioned adjusting BIOS settings—did you confirm the memory speed is 6000 MT/s? Regarding the clock speeds, what values are you observing in single-threaded benchmarks? (Based on what you've seen, they should be around 5.5-5.6GHz unless it's using a different CCD.) I suspect the issue isn't with the motherboard power delivery, as multithreaded tests would show problems.
S
StyleTrick
12-08-2021, 07:48 AM #4

I'm not confident about the consistency of the OCCT cache/memory latency tests, as I've noticed inconsistent results. It might help to compare with another test to verify the numbers. You mentioned adjusting BIOS settings—did you confirm the memory speed is 6000 MT/s? Regarding the clock speeds, what values are you observing in single-threaded benchmarks? (Based on what you've seen, they should be around 5.5-5.6GHz unless it's using a different CCD.) I suspect the issue isn't with the motherboard power delivery, as multithreaded tests would show problems.

D
Dillon451
Junior Member
15
12-08-2021, 03:01 PM
#5
The "low score" refers to what CPUz average means—it's just an average, which can vary slightly. Running multiple apps in the background may also reduce your score, even if it impacts the multi-thread score.
D
Dillon451
12-08-2021, 03:01 PM #5

The "low score" refers to what CPUz average means—it's just an average, which can vary slightly. Running multiple apps in the background may also reduce your score, even if it impacts the multi-thread score.

I
Ilikepie81
Member
184
12-08-2021, 09:00 PM
#6
I verified it thoroughly and indeed MoBo VRM includes SPS. However, it still features 14+2+1 VRM configurations rated at 60A. The specifications are available here: https://www.msi.com/Motherboard/X870E-GAMING-PLUS-WIFI. It's recommended to run Cinebench for CPU evaluation. OCCT may sometimes cause issues and produce unusual results. Cinebench also provides a score, which can be used to compare with other CPUs of the same model.
I
Ilikepie81
12-08-2021, 09:00 PM #6

I verified it thoroughly and indeed MoBo VRM includes SPS. However, it still features 14+2+1 VRM configurations rated at 60A. The specifications are available here: https://www.msi.com/Motherboard/X870E-GAMING-PLUS-WIFI. It's recommended to run Cinebench for CPU evaluation. OCCT may sometimes cause issues and produce unusual results. Cinebench also provides a score, which can be used to compare with other CPUs of the same model.

N
NutellaGesicht
Junior Member
41
12-15-2021, 08:10 AM
#7
I don't believe OCCT latency is particularly precise, since I've observed similar outcomes from others. I've checked L1 cache latency in AIDA64, which reported 0.8ns, but couldn't find data for L2/3 as it's the trial version. Memory performance is solid, and Win Mem Diag shows no problems. Clock speeds are consistent at 5.5 in HWInfo and 3DMark. I've also looked into how system interrupts in Task Manager affect CPU usage during games—often reaching up to 6% while I usually see around 3%, which isn't something I've noticed on my older CPU.
N
NutellaGesicht
12-15-2021, 08:10 AM #7

I don't believe OCCT latency is particularly precise, since I've observed similar outcomes from others. I've checked L1 cache latency in AIDA64, which reported 0.8ns, but couldn't find data for L2/3 as it's the trial version. Memory performance is solid, and Win Mem Diag shows no problems. Clock speeds are consistent at 5.5 in HWInfo and 3DMark. I've also looked into how system interrupts in Task Manager affect CPU usage during games—often reaching up to 6% while I usually see around 3%, which isn't something I've noticed on my older CPU.

K
Krzywy
Member
150
12-29-2021, 06:48 PM
#8
The main problem with the dual CCD X3D components is that items must be linked to the right CCD. When using OCCT/CPU-Z and other single-threaded tasks, they should stay on the non-X3D CCD, whereas games (and other workloads that gain from extra cache) should run on the X3D one. AMD hasn't been very clear about the clock speed for the X3D CCD with these parts, but it's likely similar to the 9800X3D at around 5.2GHz. This suggests a ~7% variation in boost clocks, meaning any small consistent discrepancy could point to the wrong CCD being used.
K
Krzywy
12-29-2021, 06:48 PM #8

The main problem with the dual CCD X3D components is that items must be linked to the right CCD. When using OCCT/CPU-Z and other single-threaded tasks, they should stay on the non-X3D CCD, whereas games (and other workloads that gain from extra cache) should run on the X3D one. AMD hasn't been very clear about the clock speed for the X3D CCD with these parts, but it's likely similar to the 9800X3D at around 5.2GHz. This suggests a ~7% variation in boost clocks, meaning any small consistent discrepancy could point to the wrong CCD being used.

K
kimck2000
Junior Member
47
12-29-2021, 10:04 PM
#9
I understand your concern. Would you like to replace the CPU and test a different one?
K
kimck2000
12-29-2021, 10:04 PM #9

I understand your concern. Would you like to replace the CPU and test a different one?

C
CAMOOO
Member
225
01-15-2022, 04:43 AM
#10
AMD addresses this with software that doesn't function perfectly, ensuring the 7950X3D and 9950X3D remain the top X3D options in their series. Their chipset driver paired with Game Bar in Windows is their preferred method for managing dual CCD X3D components. Another approach mentioned involves using CPPC in the BIOS, though it's generally simpler and focuses on selecting a particular CCD.
C
CAMOOO
01-15-2022, 04:43 AM #10

AMD addresses this with software that doesn't function perfectly, ensuring the 7950X3D and 9950X3D remain the top X3D options in their series. Their chipset driver paired with Game Bar in Windows is their preferred method for managing dual CCD X3D components. Another approach mentioned involves using CPPC in the BIOS, though it's generally simpler and focuses on selecting a particular CCD.

Pages (2): 1 2 Next