- May 11, 2008
- 19,586
- 1,197
- 126
There is something i am still wondering about. The 1600x in the latency tests done by hardware.fr shows a much lower latency than the rest. I have always assumed that this was a measurement error.
And now, in the anandtech article about Ryzen ThreadRipper from Ian Cutress, i see numbers that also say the same.
Why has the 1600x such a lower latency at the 8MB test ?
It cannot be clock frequency alone because the 1800x (which has the same base and boost clocks as the 1600x) has the same latency as all the other ryzens, only the 1600x seem to have a cache latency advantage when compared to all other ryzen models.
If it is purely cache arrangement, Why has the 1600 not the same benefits ?
I could make a guess that it is clock speed and cache configuration combined that causes for the cache latency numbers, but i have strong doubts about it (The difference in clockspeed is to little for such a lowered latency):That would mean that a 1600 overclocked would show very similar cache latency numbers as the 1600x. Then it is a combination of cache configuration for 3 cores in a CCX sharing 8MB and the clockspeed.
My question is, who ever has a ryzen 1600, are you willing to test the cache latency at a dataset of 8MB ?
I do not know if the program for testing is publicly available ?
Here are some graphs to compare :
And the text form anandtech :
And now, in the anandtech article about Ryzen ThreadRipper from Ian Cutress, i see numbers that also say the same.
Why has the 1600x such a lower latency at the 8MB test ?
It cannot be clock frequency alone because the 1800x (which has the same base and boost clocks as the 1600x) has the same latency as all the other ryzens, only the 1600x seem to have a cache latency advantage when compared to all other ryzen models.
If it is purely cache arrangement, Why has the 1600 not the same benefits ?
I could make a guess that it is clock speed and cache configuration combined that causes for the cache latency numbers, but i have strong doubts about it (The difference in clockspeed is to little for such a lowered latency):That would mean that a 1600 overclocked would show very similar cache latency numbers as the 1600x. Then it is a combination of cache configuration for 3 cores in a CCX sharing 8MB and the clockspeed.
My question is, who ever has a ryzen 1600, are you willing to test the cache latency at a dataset of 8MB ?
I do not know if the program for testing is publicly available ?
Here are some graphs to compare :
And the text form anandtech :
Facebook
Twitter