1

Here is a screenshot of a 256GB Samsung 840 Pro:

enter image description here

The system on right seems to be:

Intel Core i5-2400
Intel DH67GD
8GB 1600MHz CL9 DDR3

The system on the left is PowerEdge R730 with a Windows Server 2012 R2 standard running inside Windows Server 2012 R2's Hyper-V and it has a DELL PERC H730 and two Xeon processors. The virtual machine itself has 32GB of RAM, and 8 virtual processors.

Why is there such a huge gap in performance different between both machines? What should I look at?

  • 2
    Performance will be affected by hardware and operating system. Also it may be degraded further if using virtual machine. Another point to be noted is that Samsung SSD performance degrade over time. This is a known issue. pcworld.com/article/2887255/… – arundevma Jul 13 '15 at 11:18
  • Both are connected to SATA III ports I presume? Why can't you tell us, if you want an answer, we need that information. – Ramhound Jul 13 '15 at 11:23
  • Im sure the one on the right is connected to SATA III. That R730 has SAS ports which Im not sure at what speed it is at. – user466124 Jul 13 '15 at 11:26
  • Im sure the one on the right is connected to SATA III. Seeing that this is a 2015 server, Im pretty sure this is a SAS-3 (12 Gbit/s). That R730 has SAS ports which Im not sure at what speed it is at. – user466124 Jul 13 '15 at 11:34
  • Anyone have ideas on what to start looking at? – user466124 Jul 14 '15 at 11:07
1

Im still looking around for information but some have caught my eye and might help you too @Tonny.

It seems that the controller PERC H730 Mini is perfectly ready for SAS-3 (12) BUT I have two SAS drives ST600MM0006 (RAID1) and four SAS drives ST2000NM0023 (RAID5) and they seem to go only up to SAS-2 (6). It could be that the controller itself then falls back everything to SAS-2. The iDRAC tells me that all drives are just hitting SAS-2 while it reports that the controller itself can hit SAS-3.

BTW, after enabling that "Read Ahead" Im hitting this:

enter image description here

Intrestingly, I changed to passthrough and Im getting slightly worst performance then after enabling "Read Ahead":

enter image description here

My last test would be passthrough on SCSI controller. It is on a IDE controller because RemoteFX isnt compatible. Sadly I cannot get it to boot off a passthough SSD attached to a SCSI controller in Windows Server 2012 R2's Hyper-V. Limitation?

Well the best I can do is boot off a VHDX on a IDE controller and passthrough the SSD to a SCSI controller. These are the results:

enter image description here

Intresting results. Write speeds improve and read speeds SLIGHTLY go down.

Another test

I directly attached the SSD to the Hyper-V host and then ran CrystalDiskMark on the actual Hyper-V host (no virtualization)

enter image description here

Speeds are not that impressive so I would tend to start to believe that the issue is the controller (or its driver).

| improve this answer | |
  • That wouldn't explain the terrible write performance. – James P Jul 14 '15 at 12:21
  • I edited my answer a bit: It wouldnt explain much at all. – riahc3 Jul 14 '15 at 12:23
  • Hyper-V generally seems to have poor disk write performance especially for virtual disks, I think that is likely to be a factor. – James P Jul 14 '15 at 12:26
  • Well, I was going to that next: Do a passthough of the drive directly thru to the virutal machine and boot off that. This way Hyper-V is barely involved. – riahc3 Jul 14 '15 at 12:29
  • More comments and tests added. – riahc3 Jul 15 '15 at 7:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy