Celeron G3900 with Asrock Z170 motherboard with 7 NVMe SSDs?

[PandaBear]

I'm trying to see if this sounds crazy or if it is actually reasonable, please forgive my stupid questions because I'm still using Phenom II X4 and OCZ Vertex 2 at home so I'm a bit obsolete to begin with.

Recently I got a project that I need a build to test some NVMe SSD prior to deployment. This will be a machine that just run a write once, read for 10 minutes test throughout the entire SSD capacity to screen out the bad apples. I need as many PCIe 3.0 4x interface as I can so that means the best choice I can get would likely be an Asrock Z170 board with 4 PCIe3.0 slots with PCIe 4x to M.2 adapters, and 3 M.2 slots to support 7 M.2 SSDs.

To take advantage of all that I'd likely need a Skylake CPU (for the DMI 3.0 and Z170's PCIe 3.0), and the cheapest one would be a celeron. However I'm not sure if it would be good enough to read 7 M.2 drive at full speed (say around 2GB/S each) if we are just discarding the data anyways, or should I use something with quad core and hyper threading so it would work ok for 7 SSD read.

Any suggestion?

 

[monkeydelmagico]

Seems a bit more like a server type of workload. I would be inclined to get as many cores as you can afford. Based on the workload described even a slow-ish 4 core like an i5-6400 would be cheap insurance against processor bottleneck.

HSIO/Flex-IO on the Z170 should allow for the configuration but I will admit I have never seen such in real life.

 

[PandaBear]

It would be even less than server type of workload as I do not need any processing of the data and send out to ethernet, so in theory I can cut the traffic by half but double the amount of drives I load. I'm just wondering if the CPU side's cache size and frequency matters on these kind of work....

 

[PliotronX]

Amount of PCI-E lanes may need to be considered, even though the z170 boards can feature a bunch of x16 slots, they might operate on fewer lanes and likewise with Celerons.