SSD Benchmark Variations

[turdwagon1800]

     I have a question concerning the comparison of benchmark results between the Samsung 850, and the Samsung MZHPV PCIe M.2. I have the M.2 drive set up for windows, and programs. I have an HDD for storage. I am not currently using the 850 for anything. At the moment it sits vacant in my PC. Today I ran crystal disk on both of the SSDs (850 and PCIe M.2), and I was a little taken back by the results. The M.2, with it's PCIe interface scored much higher in sequential read and write times, but surprisingly the 850 scored slightly higher than the M.2 drive in the 4K Q32T1, and virtually the same in the 4K. 

     My question is, is this normal? I was expecting the M.2 PCIe SSD to blow the SATA 850 out of the water on all fronts, but this was not the case. I have searched the forum, and the internet, but I haven't been able to find an answer to this question. Also, as a side note, would it be better for me to be using the intel rapid storage driver rather than the microsoft driver for my SSD? I have the controller set to AHCI in bios now. I also have a radio switch in bios which has an on/off setting for IRT (intel rapid storage?), which I just turned on. I bench-marked again after turning that on, and there is no difference. I also checked in device  manager and confirmed that it is still using the microsoft SSD driver. What is this intel switch in my bios?

   Here are the specs, and a side by side of the two benchmarks ...

Asus Z-170k (motherboard)

i7 6700k (CPU)

R9-390x (Graphics)

16GB Corsair Vengence DDR4

Samsung MZHPV PCIe SSD (C:\ Drive)

Toshiba HDD 7200 rpm (Storage)

Samsung 850 SSD (Vacant)

Windows 10

 

 

[Mark77]

Perhaps secure-erase both of them and try doing an equal run of CrystalDiskMark on both?  You've filled one with 80gb of stuff, and the other is blank/empty.

 

Having said that, there are certainly parts of the benchmark, like "4K", for which the interface itself doesn't make a lot of difference.   Just because you can get the data to the controller faster, doesn't mean that it has enough  channels to the physical flash to actually conjure up additional IOPS. 

 

 

[TrifectaIII]

Well I may be totally wrong here, but I  believe that makes sense. The sequential speeds are going to be bottlenecked by the sata interface on the 850, and no such bottleneck exists on the pcie m.2 drive. For random 4k performance, you are going to be bottlenecked by the controller of the drive, which is probably similar in performance on both drives.

 

In other words, there is nothing inherent about a pcie ssd that would make it faster at random read/write performance, all pcie really does is remove the sata bottleneck on max speeds (which you only see with sequential tests).

[Enderman]

just because it has M.2 connection does not mean that all different types of speeds will be faster

its not an NVME drive

 

as you can see with the 850, the 4k tests are far below 550MBps which is the sata bottleneck

if its not using the full bandwidth of sata, why would giving it more bandwidth (pcie M.2) increase anything?

 

its being bottlenecked by the controller and nand, not by sata

and thats why your M.2 SSD still has 4k speeds as bad as the sata drive

[turdwagon1800]

5 minutes ago, Mark77 said:

Perhaps secure-erase both of them and try doing an equal run of CrystalDiskMark on both?  You've filled one with 80gb of stuff, and the other is blank/empty.

 

Having said that, there are certainly parts of the benchmark, like "4K", for which the interface itself doesn't make a lot of difference.   Just because you can get the data to the controller faster, doesn't mean that it has enough  channels to the physical flash to actually conjure up additional IOPS. 

 

 

Sweet.. thanks for the answer. I actually hadn't considered what your saying about the controller. Forgive me for my ignorance pertaining to the hardware. I was under the assumption that each SSD utilized it's own controller. Is this not the case? Is there like a master controller for all of the SSDs in the system? Also, what is a secure erase, and how would I go about doing that? Are you meaning I should wipe them clean?

[turdwagon1800]

5 minutes ago, TrifectaIII said:

Well I may be totally wrong here, but I  believe that makes sense. The sequential speeds are going to be bottlenecked by the sata interface on the 850, and no such bottleneck exists on the pcie m.2 drive. For random 4k performance, you are going to be bottlenecked by the controller of the drive, which is probably similar in performance on both drives.

 

In other words, there is nothing inherent about a pcie ssd that would make it faster at random read/write performance, all pcie really does is remove the sata bottleneck on max speeds (which you only see with sequential tests).

That actually makes total sense... Thank You! I feel better now 

[Mark77]

10 minutes ago, turdwagon1800 said:

Sweet.. thanks for the answer. I actually hadn't considered what your saying about the controller. Forgive me for my ignorance pertaining to the hardware. I was under the assumption that each SSD utilized it's own controller. Is this not the case? Is there like a master controller for all of the SSDs in the system? Also, what is a secure erase, and how would I go about doing that? Are you meaning I should wipe them clean?

 

Yes, each SSD uses its own controller/SoC.  Basically a SSD is an entire embedded computer in and of itself.  Just because you can get data to the SSD quicker (through a PCI-E interface) doesn't mean the system is any faster.

 

Think of it intuitively like taking an old computer built in 2006 , with some of the first PCI-E slots and an AMD 64-bit X2 dual core CPU, and sticking one of Linus's 10gig-E network NICs into it.  The bottleneck will be in the computer itself, not in the external interface.   Now, as flash technologies and controllers become more sophisticated over time, a greater amount of capability of the M.2 PCI-E interface will be unleashed over time. 

[turdwagon1800]

6 minutes ago, Mark77 said:

 

Yes, each SSD uses its own controller/SoC.  Basically a SSD is an entire embedded computer in and of itself.  Just because you can get data to the SSD quicker (through a PCI-E interface) doesn't mean the system is any faster.

 

Think of it intuitively like taking an old computer built in 2006 , with some of the first PCI-E slots and an AMD 64-bit X2 dual core CPU, and sticking one of Linus's 10gig-E network NICs into it.  The bottleneck will be in the computer itself, not in the external interface.   Now, as flash technologies and controllers become more sophisticated over time, a greater amount of capability of the M.2 PCI-E interface will be unleashed over time. 

Thank you for the explanation. I wasn't thinking about it that way before, but this makes since.