MSI H110M Grenade Micro ATX Motherboard Review

[bit_user]

I managed to find a Samsung 960 evo m.2 and install a fresh copy of windows 10 but in a benchmark I'm only hitting 5 gigs of transfer speed what gives?

Um, 5 gigs of what?

First, what size?

Second, here's performance data from the 256 GB model (elsewhere in the article is 1 TB):
http://www.tomshardware.com/reviews/samsung-960-evo-nvme-ssd-review,4802-3.html

Where does yours deviate?

Finally, consider that at QD=1 and QD=2, the PCIe speed is a non-issue. At QD=4, you take a slight hit on sequential read, but that queue depth & higher are pretty rare for desktop usage.

For sequential writes and all random access, the drive is a bottleneck at all queue depths.

So, based on these benchmarks, there's simply no justification for upgrading just to improve over the x4 PCIe 2.0 M.2 slot, for most NVMe drives, including yours.

 

[XENO_1]

I managed to find a Samsung 960 evo m.2 and install a fresh copy of windows 10 but in a benchmark I'm only hitting 5 gigs of transfer speed what gives?

Um, 5 gigs of what?

First, what size?

Second, here's performance data from the 256 GB model (elsewhere in the article is 1 TB):
http://www.tomshardware.com/reviews/samsung-960-evo-nvme-ssd-review,4802-3.html

Where does yours deviate?

Finally, consider that at QD=1 and QD=2, the PCIe speed is a non-issue. At QD=4, you take a slight hit on sequential read, but that queue depth & higher are pretty rare for desktop usage.

For sequential writes and all random access, the drive is a bottleneck at all queue depths.

So, based on these benchmarks, there's simply no justification for upgrading just to improve over the x4 PCIe 2.0 M.2 slot, for most NVMe drives, including yours.

The size is 250gb I just ran the bench again here are the results Max speed: PCIe 5,000 MB/s another test with more in depth results.

* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 1807.680 MB/s
Sequential Write (Q= 32,T= 1) : 1321.629 MB/s
Random Read 4KiB (Q= 32,T= 1) : 651.397 MB/s [159032.5 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 561.587 MB/s [137106.2 IOPS]
Sequential Read (T= 1) : 1355.085 MB/s
Sequential Write (T= 1) : 1183.964 MB/s
Random Read 4KiB (Q= 1,T= 1) : 46.469 MB/s [ 11345.0 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 189.376 MB/s [ 46234.4 IOPS]

Test : 1024 MiB [C: 8.8% (20.4/232.3 GiB)] (x5) [Interval=5 sec]
Date : 2017/02/21 22:56:54
OS : Windows 10 Professional [10.0 Build 14393] (x64)

 

[bit_user]

The size is 250gb I just ran the bench again here are the results Max speed: PCIe 5,000 MB/s another test with more in depth results.

I don't even understand this, because PCIe 3.0 is 985 MB/s per lane. So, that can't be right, even for buffer reads over x4 PCIe 3.0. And it's too low to measure anything in Mbits/sec.

What does the T indicate? Is it threads?

And why Q=32? Are you running a database server, or something? You should only be concerned about Q=1 to Q=4 (at the most), for desktop usage. But I'd really only worry about 1 and 2.

Sequential Read (Q= 32,T= 1) : 1807.680 MB/s

This is the biggest place you're hurting. And it's still about 68% of the benchmark in that article. Of course, it's a moot point (see above).

Sequential Write (Q= 32,T= 1) : 1321.629 MB/s

Actually, better than the article. But, it looked like there was something funny going on there, and only one drive did better than yours.

Random Read 4KiB (Q= 32,T= 1) : 651.397 MB/s [159032.5 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 561.587 MB/s [137106.2 IOPS]

Read is about the same, but for write they don't have the same test.

Sequential Read (T= 1) : 1355.085 MB/s
Sequential Write (T= 1) : 1183.964 MB/s

Okay, read is a bit worse.

Random Read 4KiB (Q= 1,T= 1) : 46.469 MB/s [ 11345.0 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 189.376 MB/s [ 46234.4 IOPS]

Okay, random read is a little worse, and again I can't tell about writes.

I wouldn't worry about it. The difference between your board and one with a PCIe 3.0 slot is generally going to be smaller than the difference between 960 Evo vs. 960 Pro. So, one way of looking at it is that if you really put such a high premium on SSD performance, why didn't you get the Pro?

Another way of looking at it is that the OS hides the raw performance of your drive, in many cases. It buffers writes and performs read-ahead. So, the amount of time that you're really experiencing the limit of what the drive can deliver is quite small. In a side-by-side test, there would only be a tiny number of cases where you could notice the difference. But it really depends a lot on what you do. If you do video editing or play games with long load times, you'd hit those cases a little more often. For most other desktop usage patterns, such cases will be rare, with the main one just being OS cold boot.

 

[XENO_1]

No not trying to run a server its just a generalized benchmark test but as you mentioned I'm only running at 68% and I do plan to play games and do some lite editing so I would like to push as much performance as this rig allows.

 

[bit_user]

No not trying to run a server its just a generalized benchmark test but as you mentioned I'm only running at 68% and I do plan to play games and do some lite editing so I would like to push as much performance as this rig allows.

Well, my whole point about database servers is that it's the only way you'd reach the 68% level. QD=32 is an extreme that normal desktops will rarely ever see. Even gaming & video editing.

If you like to push things to an extreme, that's fine. But then why did you buy the EVO, instead of Pro, or even ask if you should upgrade? I sort of feel like I just wasted my time looking over your results.

I think you sort of missed my point that, even in a side-by-side comparison, the difference would be hard to notice. But, go ahead and upgrade if you want.

 

[XENO_1]

In regrades to upgrading I was just looking at my options and why didn't I get the pro? Budget and not many were stock and sorry if I wasted your time.

 

[bit_user]

In regrades to upgrading I was just looking at my options and why didn't I get the pro? Budget and not many were stock and sorry if I wasted your time.

Nah, it's cool.

Going for the extreme is one thing that makes this hobby fun. Even when it doesn't make much practical sense.

I've certainly made some build decisions that were a little hard to justify, but they felt gratifying to me. In the end, being happy with your build is what really matters.

 

[bit_user]

I just ran the bench again here are the results
...
Sequential Read (T= 1) : 1355.085 MB/s
...
Test : 1024 MiB [C: 8.8% (20.4/232.3 GiB)] (x5) [Interval=5 sec]

I'm still puzzled by the sequential read result.

Did you happen to have anything else running, when you ran those benchmarks? Even a web browser? Did you check that your system was completely idle?

Do you still have the board? In the interest of science, would you be willing to rerun the benchmarks, after a clean boot (+ a couple minutes to settle out). Use Task Manager (with "Show Processes from All Users" enabled), or the Windows 10 equivalent, to confirm the system is idle, first.

If you do this, please post up all your results.

Thanks.