If you are using an optane memory module to cache an SSD go to a mirror and repeat this over and over
I
M
We
Tall
Did
Optane memory modules are accelerators meant for spinners to get them closer to SSD performance. They will slow down every SSD there is and cease to show any improvement on spinners once the cache is saturated.
I
M
We
Tall
Did
Optane memory modules are accelerators meant for spinners to get them closer to SSD performance. They will slow down every SSD there is and cease to show any improvement on spinners once the cache is saturated.
Yes makes a difference. Now what drive is giving sub par performance? Is the speed you are referring to as subpar the spinner with the Optane cache?
Help can be better provided with specifics
All system specs
Bench runs of both drives with and without the Optane stick.
Help can be better provided with specifics
All system specs
Bench runs of both drives with and without the Optane stick.
Couldn't have been too clear if it wasn't understood. I wasnt the context of the post as much as the bouncing information.
If you have DIMM.2_1 set to PCH then its sharing bandwidth with the optane stick on the DMI bus. meaning 2 4X devices sharing 4X bus to the CPU. Set it to CPU instead and make sure all PCIE devices are set to GEN 3 instead of auto and nothing in the bottom PCIE slot and you will be golden. The statement given about 24 lanes on the PCH is misleading. Sure there are 24 lanes BUT and its a HUGE BUT those are all downstream of the PCH and the PCH connects via the DMI bus to the CPU on 4 Lanes hence the term bottle neck. Same reason why its useless to run raid0 on NVME drives downstream. They are already at or very close to the total bandwidth of the DMI bus. Everything else is downstream too. USB ports. sound etc.
This isnt the exact diagram but add in DIMM.2_2 always on the PCH and a switch on DIMM2.1 putting it down the same lanes with PCIEx16_4 or on the PCH.
If you have DIMM.2_1 set to PCH then its sharing bandwidth with the optane stick on the DMI bus. meaning 2 4X devices sharing 4X bus to the CPU. Set it to CPU instead and make sure all PCIE devices are set to GEN 3 instead of auto and nothing in the bottom PCIE slot and you will be golden. The statement given about 24 lanes on the PCH is misleading. Sure there are 24 lanes BUT and its a HUGE BUT those are all downstream of the PCH and the PCH connects via the DMI bus to the CPU on 4 Lanes hence the term bottle neck. Same reason why its useless to run raid0 on NVME drives downstream. They are already at or very close to the total bandwidth of the DMI bus. Everything else is downstream too. USB ports. sound etc.
This isnt the exact diagram but add in DIMM.2_2 always on the PCH and a switch on DIMM2.1 putting it down the same lanes with PCIEx16_4 or on the PCH.
