SSD for Lenove server vs 10k drives

Yes with SSD 6Gb and eight drive in PCIe V3 we get about over 2500MB/s @ Yes Byte not bit, but it must be PCIe Gen 3 controller and slot of Mobo

You should be able to see what controller that you have, in Windows in should list under Storage Controller category. Or you can call DATOptic folks they can help!

We use Intel SSDSC2BP480G4R5 and DATOptic RM8-S12
https://www.newegg.com/Product/Product.aspx?Item=N82E16820167192
http://www.datoptic.com/ec/2u-eight-12gb-sas-sata-hot-swap-bay-with-hd-mini-sas-sff-8644-interface.html with option of ARC-1883X

Samsung 850 is good too... Our customer only needs 480GB SSD for their application

I realize now you meant between the Intel and Samsung. Intel is good in terms of reliability, but I find them behind in terms of technology and outright speed and massively overpriced most of the time.

For comparison:
Intel 730 480GB: 128TBW (267TBW/TB)

I measure the TBW per TB because ultimately, you're going to write twice as much data to a 1TB drive than you will to a 500GB drive, so the extra TBW will only compensate for the additional space if you're going to use all of the drive. If you go overboard, using 1TB drives when you could get away with 500GB drives, you'll obviously extend the longevity considerably.

I'm going to take a wild stab and guess you're running a Lenovo x3650 M4 (based on your 16 drive statement), that'd make your storage controller (assuming it's the one it shipped with) a Lenovo ServeRAID M5110, which is PCIe 3.0 x8 (for a total bandwidth cap of 7.7GBps). Assuming you go for SM863 or PM863s, that'd put you at 4.1GBps theoretical maximum bandwidth (read) for a 16 drive RAID 10 or 8.1GBps for RAID 0.

Even with an SSD though, with almost any load you're going to hit IOPS limits first. You'll be lucky to get half the theoretical max on a typical file server workload, and even less with a load like a database server. Not to mention the bottleneck of transport bandwidth (2x aggregated 10Gbps links is still only 2.5GBps, and you lose about 10% overhead to even the most efficient network link, so 2.25GBps).

I keep telling my customers though, its worth it to use an SSD even if you bottleneck the theoretical bandwidth all the way down to the same speed as an HDD. Let's say you run 2x 1TB SATA 3Gbps HDD and 2x 1TB SSD 6Gbps in two identical servers, both with SATA 3Gbps controllers and you access both sets of data over gigabit network connections (so 125MB/s bottleneck, real world closer to 110MB/s on Intel NICs and 90MB/s on Realtek NICs). Your HDD based server array in RAID 0 can hit 200MB/s theoretical max (100MB/s disk transfer limit, limited by the platters), and your SSD based server array can hit 1GB/s (600MB/s limited by the 3Gbps SATA interface), but both are limited to the NIC transfer speed, so when transferring a large, single file (lets say a 1TB VHD), the performance is nearly identical. However, you can enumerate the files in the folder with the VHD instantly on the SSD, on the HDD it might take 10 seconds. When you add additional workloads (like will be present in just about any server environment), the HDD's arm movement is going to slow the disk down to a crawl (4K QD32 on the HDD array drops to 3MBps) where the SSD array is unphased (4K QD32 on the SSD array is still way over the bottleneck point at 800MBps+). Not to mention the lack of moving parts and significant heat lowers vibration in the chassis, helps the system run cooler, and drops power consumption putting less stress on the power supply and electric bill.