M.2 drive. Games it actually makes a difference on.

[Stormyskyes]

Over countless forums have told me m.2 speed is useless and the speed is increase is unnoticeable. 

 

Well this isn't always the case. A few games it helps imensily. I'm trying to create a list of games m.2 actually helps compared to ssd. 

 

So far. 

Dishonored 2. Startup time from 5mins to 3 seconds. Regular loading screen times 4 seconds down from 13 seconds. 

 

Black desert online. Startup time 12 seconds down from 58 seconds. 

Regular loading 2 seconds down from 28 seconds. 

 

Overwatch 4k. Startup time unaffected loading times from 20 seconds to 4 seconds. 

Note in 1080p in max settings the loading time is unaffected. At 3-5 seconds. 

 

Feel free to add to the list. 

 

Ones I plan to test. 

PUBG

Prey

Fallout 4

 

[Oshino Shinobu]

M.2 is just a connector. You can have SATA M.2 drives, which have no speed benefits at all over 2.5" SATA drives. What I assume you're talking about are NVMe drives, which are available with the M.2 connector. 

 

What are you comparing these speeds to? What drive takes 5 minutes to launch Dishonored 2? What condition is that drive in? How full is it? What load was it under other than launching the game? 

 

From pretty much all benchmarks I've seen, NVMe makes a negligible difference to game load times compared to SATA SSDs. I find some of the results you've listed hard to believe, especially without any details on what drives you're using, how you're actually comparing them and under what conditions. 

[Stormyskyes]

14 minutes ago, Oshino Shinobu said:

M.2 is just a connector. You can have SATA M.2 drives, which have no speed benefits at all over 2.5" SATA drives. What I assume you're talking about are NVMe drives, which are available with the M.2 connector. 

 

What are you comparing these speeds to? What drive takes 5 minutes to launch Dishonored 2? What condition is that drive in? How full is it? What load was it under other than launching the game? 

 

From pretty much all benchmarks I've seen, NVMe makes a negligible difference to game load times compared to SATA SSDs. I find some of the results you've listed hard to believe, especially without any details on what drives you're using, how you're actually comparing them and under what conditions. 

Comparing it to am 120 gb san disk ssd. With it being the only program on the disk. 

Computer specs as follows.

Gtx 1080 founders edition 

I7 7700k 4.2 ghz water cooled. 

24gb ddr4 2133 mhz ram corsair. 

Samsung evo drive and the ssd. Mentioned above. 

Ssd also running os. 

[Crunchy Dragon]

-Moved to Member Reviews-

 

Interesting thing you've got going on here.

 

Worth noting that M.2 is just a form factor, it can use either the SATA or NVMe bus.

[Oshino Shinobu]

Just now, Crunchy Dragon said:

Worth noting that M.2 is just a form factor, it can use either the SATA or NVMe bus.

Mr. Nitpick here, M.2 is a connector rather than an entire form factor. M.2, combined with the length and width specifications (noted in a 4 digit number) and the keying make up the form factor. NVMe is a communication protocol which uses the PCIe bus, rather than being a bus itself. 

 

Sincerely, Mr. Nitpick. 

[Stormyskyes]

3 minutes ago, Crunchy Dragon said:

-Moved to Member Reviews-

 

Interesting thing you've got going on here.

 

Worth noting that M.2 is just a form factor, it can use either the SATA or NVMe bus.

Yea. It is odd. Every resource says it shouldn't make a difference but with a few games it is. The ssd is good  o know as when I benchmarked it it was performing in the 99th percentile. 

Also the nvmedrive was performing at 250% expected speed. 

[Oshino Shinobu]

17 minutes ago, Stormyskyes said:

Also the nvmedrive was performing at 250% expected speed. 

Do you have Samsung RAPID mode enabled? If so, it's using RAM to increase the speeds. 

 

What are the speeds you get from benchmarking the NVMe drive?

[Revamp]

To be honest, it sounds like you're comparing a nvme drive to a regular 3.5" HDD, your 2.5" ssd is bordenline failing due to being worn out, or it's nearly maxed out capacity wise and severely hindering performance.

I previously had a 2.5" evo ssd and when I upgraded to my current 960 evo NVMe m.2, my load speeds at most increased 1-2 seconds in a select few games. and when I just recently moved all my games on to my crucial sata m.2 and the speeds are still nearly identical.

 

(I have NEVER seen a 5minute load time on dishonored lol)

 

[JoostinOnline]

53 minutes ago, Stormyskyes said:

Over countless forums have told me m.2 speed is useless and the speed is increase is unnoticeable. 

For a boot drive.  Nobody thinks there's no difference when loading games with large files.  It won't generally won't give you a higher frame rate, but it will absolutely help with copying those textures to your VRAM.

[Stormyskyes]

6 minutes ago, Revamp said:

To be honest, it sounds like you're comparing a nvme drive to a regular 3.5" HDD, your 2.5" ssd is bordenline failing due to being worn out, or it's nearly maxed out capacity wise and severely hindering performance.

I previously had a 2.5" evo ssd and when I upgraded to my current 960 evo NVMe m.2, my load speeds at most increased 1-2 seconds in a select few games. and when I just recently moved all my games on to my crucial sata m.2 and the speeds are still nearly identical.

 

(I have NEVER seen a 5minute load time on dishonored lol)

 

Well dam. Me and all my friends have that same problem with dishonored 2. 

[Mira Yurizaki]

Also before you go around with application loading tests, you have to remember a few things:

• Repeated loading is poisoned by OSes keeping data in RAM as "cache" in the off chance you launch it again. So if you really want to measure loading performance, you have to load it once and reboot the computer before testing it again. On top of that, you have to turn off Superfetch (or the equivalent) to ensure the OS isn't going to precache it once it notices you've been reloading the same app over and over.
• CPU performance plays a role too, as the application isn't just moving stuff from storage to RAM, but it has to initialize the data and itself before the user can interact with it. This initialization is CPU dependent, and it may not be doing it in parallel with data loading.
• Loading screens do not necessarily mean all the application is doing is loading stuff into RAM. If the loading process is buggy, it may appear to be loading longer, but in fact it may be stalled somewhere and is not touching anything for periods of time.

The proper way to analyze application loading performance is to use the OS loggers to record storage bandwidth utilization, read operations per second, and CPU utilization. You can get a grasp of the whole picture that way.

[Stormyskyes]

9 hours ago, M.Yurizaki said:

Also before you go around with application loading tests, you have to remember a few things:

• Repeated loading is poisoned by OSes keeping data in RAM as "cache" in the off chance you launch it again. So if you really want to measure loading performance, you have to load it once and reboot the computer before testing it again. On top of that, you have to turn off Superfetch (or the equivalent) to ensure the OS isn't going to precache it once it notices you've been reloading the same app over and over.
• CPU performance plays a role too, as the application isn't just moving stuff from storage to RAM, but it has to initialize the data and itself before the user can interact with it. This initialization is CPU dependent, and it may not be doing it in parallel with data loading.
• Loading screens do not necessarily mean all the application is doing is loading stuff into RAM. If the loading process is buggy, it may appear to be loading longer, but in fact it may be stalled somewhere and is not touching anything for periods of time.

The proper way to analyze application loading performance is to use the OS loggers to record storage bandwidth utilization, read operations per second, and CPU utilization. You can get a grasp of the whole picture that way.

I highly doubt with 24 gb of system ram and 12 gb of video ram that the rams being eaten by the os.

[Mira Yurizaki]

1 hour ago, Stormyskyes said:

I highly doubt with 24 gb of system ram and 12 gb of video ram that the rams being eaten by the os.

From what I understand, when data from RAM is no longer needed, the OS simply marks it as "available." The OS doesn't actually remove the data from RAM. But if for some reason the data is needed again, it's already there in RAM so the OS reactivates it. That's what I mean by my first point. If you load the app, close it, then reload it again (or at the very least, reload the same level), you're not going to observe its actual loading behavior. The OS will just flip memory pages from inactive to active, which is much faster.

 

VRAM likely doesn't do this behavior because there's no real reason to do it.

[genexis_x]

19 hours ago, Stormyskyes said:

Comparing it to am 120 gb san disk ssd

Model? Low capacity SATA SSD is usually slower. Why don't use a high capacity + fast SATA drive like Samsung 850/860 EVO and Crucial MX500 for testing?

 

19 hours ago, JoostinOnline said:

For a boot drive

And also everyday tasks/normal usage