F5F Stay Refreshed Hardware Desktop So are there DDR5 configurations with an IMC in a 1:1 proportion?

So are there DDR5 configurations with an IMC in a 1:1 proportion?

So are there DDR5 configurations with an IMC in a 1:1 proportion?

P
Puppzi
Member
209
06-21-2021, 10:57 AM
#1
The IMC in an Intel DDR5 CPU doesn't match memory bandwidth exactly, and the same applies to AMD. It seems a 1:1 relationship could suggest speeds around 6000 MT/s, but current CPU cores may not support it yet, so IMCs likely can't either. It's worth considering if running lower speeds gives a better balance with IMC performance. How much can you push an AMD or Intel IMC to reach that target? I'd aim for at least 5Ghz.
P
Puppzi
06-21-2021, 10:57 AM #1

The IMC in an Intel DDR5 CPU doesn't match memory bandwidth exactly, and the same applies to AMD. It seems a 1:1 relationship could suggest speeds around 6000 MT/s, but current CPU cores may not support it yet, so IMCs likely can't either. It's worth considering if running lower speeds gives a better balance with IMC performance. How much can you push an AMD or Intel IMC to reach that target? I'd aim for at least 5Ghz.

S
SweatLikeFazon
Junior Member
12
06-25-2021, 10:42 AM
#2
The CPU operating frequency isn't the main constraint for the IMC. In fact, the IMC clock speed could exceed the CPU's base or boost frequency without issues, provided the system remains stable. The challenge lies in DDR5 speeds being excessively fast; matching them directly to the CPU's frequency in a 1:1 ratio would cause instability. Regardless of memory speed, DDR5 has proven difficult to run in Gear 1 mode on most systems except specific Ryzen models with DDR5 5200Mt/s. Even with very low settings, I couldn't achieve stable performance on Intel platforms. Perhaps others have found more success, but I haven't. It seems most users haven't managed this effectively.
S
SweatLikeFazon
06-25-2021, 10:42 AM #2

The CPU operating frequency isn't the main constraint for the IMC. In fact, the IMC clock speed could exceed the CPU's base or boost frequency without issues, provided the system remains stable. The challenge lies in DDR5 speeds being excessively fast; matching them directly to the CPU's frequency in a 1:1 ratio would cause instability. Regardless of memory speed, DDR5 has proven difficult to run in Gear 1 mode on most systems except specific Ryzen models with DDR5 5200Mt/s. Even with very low settings, I couldn't achieve stable performance on Intel platforms. Perhaps others have found more success, but I haven't. It seems most users haven't managed this effectively.

O
okA_
Member
141
06-30-2021, 09:17 PM
#3
Darkbreeze shared their experience on the topic. They questioned whether running a Ryzen system with equal parts IMC and DDR5 memory at 5200 MT/s was beneficial, and they inquired about the actual latency experienced at that speed.
O
okA_
06-30-2021, 09:17 PM #3

Darkbreeze shared their experience on the topic. They questioned whether running a Ryzen system with equal parts IMC and DDR5 memory at 5200 MT/s was beneficial, and they inquired about the actual latency experienced at that speed.

B
buckeree3
Member
81
06-30-2021, 09:44 PM
#4
It was a simple customer system designed for a minimal purpose and it clearly justified its cost because using it in Gear 2 would have resulted in a 1:2 penalty. Regarding latency, I didn’t investigate further, but based on the standard true latency calculation, it should have been around 12.6 microseconds given the specifications. The formula used is: 1 divided by frequency (not DDR) multiplied by latency equals true latency in nanoseconds. For example, a stock of 3200 at 14 MHz would perform better than 3600 at 16 MHz, and applying the CAS latency of 32 gives approximately 12.6ns. If it had incurred a 1:2 penalty, the actual true latency would have been closer to 25.2ns. CAS latency differs from true latency, so the numbers don’t match exactly.
B
buckeree3
06-30-2021, 09:44 PM #4

It was a simple customer system designed for a minimal purpose and it clearly justified its cost because using it in Gear 2 would have resulted in a 1:2 penalty. Regarding latency, I didn’t investigate further, but based on the standard true latency calculation, it should have been around 12.6 microseconds given the specifications. The formula used is: 1 divided by frequency (not DDR) multiplied by latency equals true latency in nanoseconds. For example, a stock of 3200 at 14 MHz would perform better than 3600 at 16 MHz, and applying the CAS latency of 32 gives approximately 12.6ns. If it had incurred a 1:2 penalty, the actual true latency would have been closer to 25.2ns. CAS latency differs from true latency, so the numbers don’t match exactly.

L
langbertcar
Junior Member
2
07-02-2021, 06:47 PM
#5
And I actually revised the True latency formula that Computronix provided, simplifying it for better understanding of everyday users. The calculation involves dividing one by frequency (not double data rate, such as 3200MT/s, where frequency is half the DDR speed) and then multiplying by latency to get true latency in nanoseconds. For instance, a 3200MT/s memory with a frequency of 1600GHz divided by 14 gives approximately 8.75nS. Changing the frequency to 1.6 results in a slightly higher value, confirming that a 3200MT/s CL14 kit outperforms a 3600MT/s CL16 kit under these conditions.
L
langbertcar
07-02-2021, 06:47 PM #5

And I actually revised the True latency formula that Computronix provided, simplifying it for better understanding of everyday users. The calculation involves dividing one by frequency (not double data rate, such as 3200MT/s, where frequency is half the DDR speed) and then multiplying by latency to get true latency in nanoseconds. For instance, a 3200MT/s memory with a frequency of 1600GHz divided by 14 gives approximately 8.75nS. Changing the frequency to 1.6 results in a slightly higher value, confirming that a 3200MT/s CL14 kit outperforms a 3600MT/s CL16 kit under these conditions.