F5F Stay Refreshed Hardware Desktop Unusual outcomes (3700x plus 3000mhz memory)

Unusual outcomes (3700x plus 3000mhz memory)

Unusual outcomes (3700x plus 3000mhz memory)

B
113
09-04-2019, 03:16 PM
#1
I tried various settings today to boost performance. My memory isn't great—Hynix CJR die, Vengeance LPX CMK8GX4M1D3000C16 8gbx2 XMP at 3000mhz. Check the outcomes: CPU-Z 5415, CB20 4621, 2133mhz CAS 15 xmp off CPU-Z 5420, CB20 4583, 3000mhz CAS 22 xmp off CPU-Z 5320, CB20 4638, 3000mhz CAS 16 xmp on. I have detailed memory specs for the last one. How can I get similar CB20 results at 2133mhz CAS 15 versus 3000mhz CAS 16? Also, a higher CPU-Z score helps. My system is Ryzen 3700X with an Asus X570 Tuf Gaming Wi-Fi.
B
BasiComplexity
09-04-2019, 03:16 PM #1

I tried various settings today to boost performance. My memory isn't great—Hynix CJR die, Vengeance LPX CMK8GX4M1D3000C16 8gbx2 XMP at 3000mhz. Check the outcomes: CPU-Z 5415, CB20 4621, 2133mhz CAS 15 xmp off CPU-Z 5420, CB20 4583, 3000mhz CAS 22 xmp off CPU-Z 5320, CB20 4638, 3000mhz CAS 16 xmp on. I have detailed memory specs for the last one. How can I get similar CB20 results at 2133mhz CAS 15 versus 3000mhz CAS 16? Also, a higher CPU-Z score helps. My system is Ryzen 3700X with an Asus X570 Tuf Gaming Wi-Fi.

E
Ebkon
Member
166
09-04-2019, 10:59 PM
#2
they function as CPU stands rather than memory stands
E
Ebkon
09-04-2019, 10:59 PM #2

they function as CPU stands rather than memory stands

R
Redacting
Member
207
09-05-2019, 03:39 AM
#3
Because you're testing for CPU performance, not latency
R
Redacting
09-05-2019, 03:39 AM #3

Because you're testing for CPU performance, not latency

A
alejandrobo1
Posting Freak
877
09-05-2019, 04:22 AM
#4
I believe it might be due to the memory usage they require. I thought the CPU-Z benchmark tested L1 or L2 cache, yet I assumed CB20 would be more demanding. Thanks.
A
alejandrobo1
09-05-2019, 04:22 AM #4

I believe it might be due to the memory usage they require. I thought the CPU-Z benchmark tested L1 or L2 cache, yet I assumed CB20 would be more demanding. Thanks.

P
PandaBlue11
Member
51
09-06-2019, 08:06 PM
#5
Cinebench scores around R15 and R20 show minimal dependence on RAM speed. This makes them ideal for demonstrating near-perfect CPU performance improvement. For more accurate results with Prime95, larger calculations are necessary—using one core per physical core (disabling HT/SMT). This is the first benchmark I recall that is mostly limited by memory bandwidth, yet the FFTs still require substantial data to fill the L3 cache. With Aida64, the photoworxx sub-test appears sensitive to RAM speed, though I haven’t analyzed its scaling behavior in detail.
P
PandaBlue11
09-06-2019, 08:06 PM #5

Cinebench scores around R15 and R20 show minimal dependence on RAM speed. This makes them ideal for demonstrating near-perfect CPU performance improvement. For more accurate results with Prime95, larger calculations are necessary—using one core per physical core (disabling HT/SMT). This is the first benchmark I recall that is mostly limited by memory bandwidth, yet the FFTs still require substantial data to fill the L3 cache. With Aida64, the photoworxx sub-test appears sensitive to RAM speed, though I haven’t analyzed its scaling behavior in detail.

P
Petzku1618
Member
181
09-25-2019, 01:58 AM
#6
Got it! I'll use aida64 next time.
P
Petzku1618
09-25-2019, 01:58 AM #6

Got it! I'll use aida64 next time.

_
_Lukeee
Junior Member
16
10-02-2019, 05:26 PM
#7
Aida64 includes RAM bandwidth and latency tests that can better reflect changes in your configuration. The paid edition offers complete features, while the trial version may have limited capabilities.
_
_Lukeee
10-02-2019, 05:26 PM #7

Aida64 includes RAM bandwidth and latency tests that can better reflect changes in your configuration. The paid edition offers complete features, while the trial version may have limited capabilities.

S
shanleighrose
Member
181
10-08-2019, 02:49 PM
#8
Thank you! I understand how to use the trial version effectively.
S
shanleighrose
10-08-2019, 02:49 PM #8

Thank you! I understand how to use the trial version effectively.