My home server is getting a bit old, and I'm beginning to ponder what I might replace it with. Skylake is getting good reviews as a desktop platform, but I've seen some comments that suggest it may not be a good server choice, one example is here:
I have some questions about this for those who understand these things more than I: The first one is: shouldn't it still be much better than previous consumer-grade platforms? Secondly, I'd like to better understand the issue and options. It seems the above comment is specifically about the bandwidth of the link between the chipset and the CPU. But if I understand correctly, the x16/x8 PCI express slots connect directly to the CPU, bypassing the DMI link. Since Skylake CPUs have integrated graphics, and you don't need high-end graphics for a server, couldn't one simply connect their HBAs into the PCI express slots, which should have ample bandwidth? That's how my current server is configured: I have an IBM M1015 in the x16 slot (which has an HP SAS expander connected to it), and a SASLP-MV8 in a x4 slot. It seems to me with the Z710 being capable of running x8/x8/x4, it should be able to handle a lot of bandwidth before even considering the bandwidth to the ports accessed through the chipset. Am I missing something here?
I'd really like to get into something that can better handle transcoding. My current server can handle a single SD stream fine, but it chokes on a blu-ray. Skylake should be able to handle this with ease, maybe even multiple streams. But I'm missing why it shouldn't also speed up my multi-disk parity calculations.
"The new DMI 3 link offers bandwidth equivalent to four lanes of PCIe Gen3. That's not nearly enough to sustain simultaneous I/O operations across all 26 of the Z170's high-speed ports. Heck, it's not even close, which is kind of a big deal since the system's memory sits beyond that DMI link, hanging off of the CPU's integrated memory controller. If this were a server architecture, I'd be worried about that fact."
I have some questions about this for those who understand these things more than I: The first one is: shouldn't it still be much better than previous consumer-grade platforms? Secondly, I'd like to better understand the issue and options. It seems the above comment is specifically about the bandwidth of the link between the chipset and the CPU. But if I understand correctly, the x16/x8 PCI express slots connect directly to the CPU, bypassing the DMI link. Since Skylake CPUs have integrated graphics, and you don't need high-end graphics for a server, couldn't one simply connect their HBAs into the PCI express slots, which should have ample bandwidth? That's how my current server is configured: I have an IBM M1015 in the x16 slot (which has an HP SAS expander connected to it), and a SASLP-MV8 in a x4 slot. It seems to me with the Z710 being capable of running x8/x8/x4, it should be able to handle a lot of bandwidth before even considering the bandwidth to the ports accessed through the chipset. Am I missing something here?
I'd really like to get into something that can better handle transcoding. My current server can handle a single SD stream fine, but it chokes on a blu-ray. Skylake should be able to handle this with ease, maybe even multiple streams. But I'm missing why it shouldn't also speed up my multi-disk parity calculations.