[SOLVED] Very informal and unscientific comparison of HDD vs NVME

punkncat

Champion
Ambassador
J st wanted to throw a "comparison" of sorts out there.

Main system is a Ryzen 7 2700X with 32GB of CL14 RAM running 3200mhz. Opening a base windows load, nothing changed from stock install, clone running 2 threads and 4GB of RAM each.

One of them on a Samsung 960 EVO NVME and the other is a Toshiba 2.5" 5400 RPM (laptop) HDD.

Using Oracle VM Virtualbox, from desktop, click start to desktop.

The Toshiba HDD took roughly 78 seconds. The Samsung NVME took roughly 20 seconds. This was a super loose count, but shows very plainly that if you are still running ANY system off a HDD for OS you should strongly consider SSD.

I am going to continue some informal tests on both 3.5" drives of the "green" and WD "yellow", as well as a couple of SSD like the good old HD Blue and some other Samsung. Just for the sake of doing it.
 
Solution
I....

huh?

I was pointing out that NVMe is a protocol; HDDs can use NVMe. And my second point was that if you're testing raw performance you need to pass the drive through for a VM which isn't the same as having the virtual disk image on the drive. That being said, I realize you meant NVMe SSD (which can be in M.2 but also U.2, etc) and intend to host multiple VMs on the drive. I think you'll find a good SATA SSD is plenty for a half dozen VMs.

punkncat

Champion
Ambassador
Lol, wow...so was rather surprised to find that between the NVME drive that my host OS is on, and the SSD that I use for games that loading the virtual machine and performance of both are nearly the same. It makes sense when you take into consideration the overhead the NVME has on it already.
I am going to have to break out my timer and do some actual timed testing between these. By and large the solid state and hard drive loads are performing almost identically within this environment (in relation to drive type) despite the drive it's loaded on.

I am also quite surprised at how well the host system runs even while all the virtual machines are on. Right now I have four Win 10 machines, 2 cores, 4GB RAM, on four different drives and running alongside host, all with active windows open and I note no specific issues. Pretty neat.
 

punkncat

Champion
Ambassador
Played around a bit more last night and got six VM and the host running. The NVME is running host and 1 VM, 2VM on a Samsung SSD, 2VM on a WD Green 3.5" HDD, and 1VM on the Toshiba 2.5" HDD.
I loaded each machine in and then just went to random video on the tube in scaled mode. The NVME and SSD VM ran great even in spite of having 2 machines running on each. No hitching, no issues, and quick enough on the desktop to be totally usable. The green drive ran well enough with two machines as well. Video ran smooth but you could see some limitations on the read/write side with both machines working. The machine loaded on the 2.5" drive was a different story. It was giving the impression of frame drop and really running poorly. I can make an assumption that it's bandwidth as well as drive speed showing a weakest link type thing. I might play with that issue some later and see if throwing RAM at it helps.

Fun stuff to fool around with. My previous VM host was an older i5 so I didn't have the cores to play with.
 
I....

huh?

I was pointing out that NVMe is a protocol; HDDs can use NVMe. And my second point was that if you're testing raw performance you need to pass the drive through for a VM which isn't the same as having the virtual disk image on the drive. That being said, I realize you meant NVMe SSD (which can be in M.2 but also U.2, etc) and intend to host multiple VMs on the drive. I think you'll find a good SATA SSD is plenty for a half dozen VMs.
 
Last edited:
  • Like
Reactions: punkncat
Solution

TRENDING THREADS