What are you looking for in a Motherboard Review?

[r9]

Formfactor and if it's for Intel K chip to be Z chipset that's about it.
Motherboard is at the very bottom of my list for resource allocation.
GPU > CPU > RAM > SSD > PSU > Mobo.
Sorry but I just fall for the 29 phase power and 34 sata and usb ports nonsense. If I need more performance I just buy the next CPU up.

 

[Wirko]

Regarding AM5 motherboard reviews: I think it would be useful to test two fast transfers to/from M.2 and USB at the same time, fast enough to overload one or both CPU-chipset-chipset buses. Do the buses still operate close to their maximum data rate at all times, or are there any interruptions or speed drops?

 

[ir_cow]

@Wirko, are you asking to transfer data from M2 to USB, or just bandwidth test of USB and M2 at the same time?

 

[Wirko]

@Wirko, are you asking to transfer data from M2 to USB, or just bandwidth test of USB and M2 at the same time?

Not sure... I consider a M.2 to USB transfer as two separate transfers, one going from M.2 through DMA controller to main memory, the other from memory through DMA to USB. So both examples you mentioned should be equal. Or is it not that simple?

 

[ir_cow]

Not sure if its simply to setup or not honestly. I use DiskSPD and perform a read test of a 32GB file to see the bandwidth of the USB port or M2 socket. I'm not after the performance of the drive. just the bandwidth a drive "could" provide.

I have no idea how to do a read test through the USB port to a M.2 socket. Maybe someone has a script I can use.

 

[Count von Schwalbe]

Copy a big file (throughput test) from USB to M.2 should do it. Or run a bench on each separate, and then at the same time to see if a shared bus decreases either drive bench when both are run concurrently.

 

[ir_cow]

Copy a big file (throughput test) from USB to M.2 should do it. Or run a bench on each separate, and then at the same time to see if a shared bus decreases either drive bench when both are run concurrently.

Copy is limited to the drive write. It will give much lower numbers. But I could batch script multiple drive read test at the same time. That may(?) show the bandwidth limitation of the chipset?

 

[Count von Schwalbe]

Copy is limited to the drive write. It will give much lower numbers. But I could batch script multiple drive read test at the same time. That may(?) show the bandwidth limitation of the chipset?

Crumbs, didn't think about that. I think one (read throughput) bench each should saturate it though, assuming you have some fast drives.

 

[Wirko]

Crumbs, didn't think about that. I think one (read throughput) bench each should saturate it though, assuming you have some fast drives.

Something like that, yes.

Even writing to SSD should be plenty fast. Sustained writing exceeds one half of maximum interface bandwidth in most PCIe 4.0 x4 drives W1zzard tested recently; it's about 4,500 to 5,500 MB/s. It's impossible if you try to fill the whole drive but what he probably does is rewrite the same file many times over, in which case the writing occurs in the pSLC cache only.