Jump to content
Collin(OD)

M.2 Drives... The more you know...

Recommended Posts

I have so many inputs...


 

On 7/22/2018 at 2:48 AM, snicker66(OD) said:

A M.2 is a form factor of a Hardrive, its a type of solid state hardrive. The M.2 is A directly attached to the motherbaord or it attached via a PCIe Slot on the motherboard providing alot more read\write speed to the drive

 

a traditional Solid state drive has on average 500 Mbps read write speed where a M.2 has 2000+ Mbps read write speed, which means is and read and write to the drive very quickly giving you a lot better performance of the computer because a PC after all is only as fast as its slowest part.

 

I want proof of the highlighted bits please.

Share this post


Link to post
Share on other sites
Posted (edited)
22 minutes ago, Collin(OD) said:

I have so many inputs...


 

 

I want proof of the highlighted bits please.

https://www.pugetsystems.com/labs/articles/Overview-of-M-2-SSDs-586/#PerformanceRange

 

The above link has a LOT of useful information regarding m.2 drives, especially the section on performance ranges:

 

M.2 Drive Throughput
by Connection
Theoretical
Maximum Throughput
Est. Real-World
Maximum Throughput
SATA III 6.0 Gb/s (750 MB/s) 4.8 Gb/s (600 MB/s)
PCI-E 2.0 x2 8 Gb/s (1 GB/s) 6.4 Gb/s (800 MB/s)
PCI-E 2.0 x4 16 Gb/s (2 GB/s) 12.8 Gb/s (1.6 GB/s)
PCI-E 3.0 x4 32 Gb/s (4 GB/s) 31.5 Gb/s (3.9 GB/s)

 

 

Edit:  This article was published in 2014 so take that into consideration, but considering the tech really hasn't changed much since then, the numbers are still pretty accurate.

Edited by CompFreak(OD)

Share this post


Link to post
Share on other sites

Those are estimates. They're not accurate. At all. Real world application has SATA and m.2 SSDs being basically identical. Especially in gaming. NVMe m.2 drives are another story. Maybe that's what that graph is talking about?

 

By graph I meant table.

Share this post


Link to post
Share on other sites
52 minutes ago, Collin(OD) said:

Those are estimates. They're not accurate. At all. Real world application has SATA and m.2 SSDs being basically identical. Especially in gaming. NVMe m.2 drives are another story. Maybe that's what that graph is talking about?

 

By graph I meant table.

To answer your question, Yes. M.2 is actually a "form factor", that can be rated on a multitude of speeds. As per the table, an M.2 in SATA mode is nearly identical to that of a standard SSD connected to a SATA cable (max. real-world throughput of about 550 MB/s for regular SSD, and an M.2 in SATA mode is about 600 MB/s). Where as an "nvme m.2" (using pci-e lanes) can reach speeds much higher. This leads to many confusions, where people think all m.2 drives are the same; They are far from that. "Nvme m.2" is the  name for an SSD the rough size of a stick of gum, that uses pci-e lanes fed directly to the cpu, instead of being routed through the SATA controller of a motherboard (resulting in much better throughput and response times). 

 

So in essence, both of you are correct. Even more confusingly, "nvme m.2" drives are even backward-compatible, in that even though they are designed to be able to run in "PCI-e 4x mode", they will still work in "SATA" mode. This is one of the "uncommon" things that you must check to make sure is adjusted in your BIOS settings, before installing your operating system (changing modes will cause windows not to boot), to give you the best performance your hardware has to offer. It is similar to the "IDE" vs. "AHCI" modes for the drives attached to SATA cables, in that "AHCI" mode will allow for higher throughput on SATA devices, and "IDE mode" is more for legacy hardware (slower but more compatibility among older hardware). In any case, all SSD's should be run in "AHCI mode" if attached via SATA cable to achieve best performance.

 

Please see below article from 2017, discussing what I'm talking about here:

http://www.velocitymicro.com/blog/nvme-vs-m-2-vs-sata-whats-the-difference/

Share this post


Link to post
Share on other sites

*Troll face* Have you tried turning it off and back on again?

Share this post


Link to post
Share on other sites
5 hours ago, CompFreak(OD) said:

To answer your question, Yes. M.2 is actually a "form factor", that can be rated on a multitude of speeds. As per the table, an M.2 in SATA mode is nearly identical to that of a standard SSD connected to a SATA cable (max. real-world throughput of about 550 MB/s for regular SSD, and an M.2 in SATA mode is about 600 MB/s). Where as an "nvme m.2" (using pci-e lanes) can reach speeds much higher. This leads to many confusions, where people think all m.2 drives are the same; They are far from that. "Nvme m.2" is the  name for an SSD the rough size of a stick of gum, that uses pci-e lanes fed directly to the cpu, instead of being routed through the SATA controller of a motherboard (resulting in much better throughput and response times). 

 

So in essence, both of you are correct. Even more confusingly, "nvme m.2" drives are even backward-compatible, in that even though they are designed to be able to run in "PCI-e 4x mode", they will still work in "SATA" mode. This is one of the "uncommon" things that you must check to make sure is adjusted in your BIOS settings, before installing your operating system (changing modes will cause windows not to boot), to give you the best performance your hardware has to offer. It is similar to the "IDE" vs. "AHCI" modes for the drives attached to SATA cables, in that "AHCI" mode will allow for higher throughput on SATA devices, and "IDE mode" is more for legacy hardware (slower but more compatibility among older hardware). In any case, all SSD's should be run in "AHCI mode" if attached via SATA cable to achieve best performance.

 

Please see below article from 2017, discussing what I'm talking about here:

http://www.velocitymicro.com/blog/nvme-vs-m-2-vs-sata-whats-the-difference/

Thanks

Share this post


Link to post
Share on other sites

Yes I think all 3 of us knew that... I was only addressing the statement that m.2 are much faster than normal SSDs. They're not. They're exactly the same in SATA mode, which is how most of them are run by most people. If you want a faster drive, you want NVMe. Unfortunately, NVMe is only actually faster in very large read/write sessions. Think 4K videos. Also, m.2/NVMe drives can get very hot and have even seen thermal throttling depending on case airflow.

 

M.2 is dope, though.

Share this post


Link to post
Share on other sites
Posted (edited)
2 hours ago, Collin(OD) said:

Unfortunately, NVMe is only actually faster in very large read/write sessions

This statement is false. Having experimented with my own testing / timing with a stopwatch, I can, in fact, testify that with a single standard SSD, booting the same system from BIOS to a fresh install of windows, timed in at 29 seconds. With two SSD's in RAID 0, that dropped to 20 seconds, and with a properly set up NVME m.2 (in pci-e 3.0 4x mode), that time has dropped to 6 seconds. After some research to support why I was witnessing this decrease in loading times, I have found that Nvme SSD's are not only excelling at large file transfers, but their random access times are far superior (compared to a standard ssd), resulting in brute performance when accessing many random files (i.e. loading your OS, games, and the like). This is where the "IOPS" figures come into the scene. IOPS stands for "Input / Output Per Second". A regular ssd typically has a rating of 70,000 IOPS, where as Nvme ssd's typically can sustain upwards of 750,000 IOPS on file sizes of only 4 kilobytes (not exactly a large 4k video now, is it?). You can find much of the information here:

https://www.cs.utah.edu/~manua/pubs/systor15.pdf

 

What you are witnessing when you think that Nvme drives aren't realistically "faster", could actually be either a bottleneck that stems from your cpu, or possibly just drivers loading (that are programmed to have a set time to allow for the hardware to initialize before allowing the system to continue to the next thing).

 

2 hours ago, Collin(OD) said:

They're exactly the same in SATA mode, which is how most of them are run by most people.

Which is why I mentioned the BIOS settings that must be adjusted according to the hardware you have installed.

 

2 hours ago, Collin(OD) said:

Also, m.2/NVMe drives can get very hot and have even seen thermal throttling depending on case airflow.

While there have been noticeable thermal issues associated during the early days of m.2 drives, I personally believe it was mostly due to poor motherboard designs putting the m.2 ports directly covered by the graphics card, where there is little to no airflow, no matter what case airflow you have. Many newer designs incorporate the m.2 slot much lower on the board (near where SATA ports are typically located), or have them located under a large heat-sink that looks like a cover. Even some ASUS boards come with a PCI-e adapter card to connect multiple m.2 drives with a single Pci-e card that you can position away from other "hot" components, and in addition, some even aligning the m.2 port so that the drive vertically sticks out from the motherboard, to take better advantage of your case's airflow. Moral of the story, don't install something that gets hot, between your graphics card and motherboard... The throttling is a safety feature so that the drive doesn't damage itself or other components.

 

* I have split the topic stemming from @MelodicRose(OD)'s original plea for assistance in her PC upgrade, so that we don't derail her topic any further.

Edited by CompFreak(OD)

Share this post


Link to post
Share on other sites
15 minutes ago, CompFreak(OD) said:

This statement is false. Having experimented with my own testing / timing with a stopwatch, I can, in fact, testify that with a single standard SSD, booting the same system from BIOS to a fresh install of windows, timed in at 29 seconds. With two SSD's in RAID 0, that dropped to 20 seconds, and with a properly set up NVME m.2 (in pci-e 3.0 4x mode), that time has dropped to 6 seconds. After some research to support why I was witnessing this decrease in loading times, I have found that Nvme SSD's are not only excelling at large file transfers, but their random access times are far superior (compared to a standard ssd), resulting in brute performance when accessing many random files (i.e. loading your OS, games, and the like). This is where the "IOPS" figures come into the scene. IOPS stands for "Input / Output Per Second". A regular ssd typically has a rating of 70,000 IOPS, where as Nvme ssd's typically can sustain upwards of 750,000 IOPS on file sizes of only 4 kilobytes (not exactly a large 4k video now, is it?). You can find much of the information here:

https://www.cs.utah.edu/~manua/pubs/systor15.pdf

 

Anecdotal evidence is irrelevant. My quite old W10 installation takes ~13 seconds to cold boot to system ready with several apps loaded on startup on a Samsung 850 Evo. I read the article. It's interesting, but it presents a test bench that is unrealistic for the vast majority of users and uses benchmarks that are synthetic, thus also irrelevant to the average user. Synthetic benches cannot be compared and projected to real world applications. Feel free to break the topic.

Review 1 showing little difference in real world applications:
https://arstechnica.com/gadgets/2016/10/samsung-960-pro-review-the-fastest-consumer-ssd-you-can-buy/3/

 

Review 2 showing little difference in real world applications:

https://techreport.com/review/30813/samsung-960-pro-2tb-ssd-reviewed/5

Share this post


Link to post
Share on other sites
23 hours ago, CompFreak(OD) said:

Real world application has SATA and m.2 SSDs being basically identical. Especially in gaming. NVMe m.2 drives are another story.

I'm a bit confused here... Are you literally arguing against your own statements? Your own evidence supports that Nvme drives are at the top of almost every test they ran against competing Nvme's and SATA ssd's.. Which is your position here?

 

4 hours ago, Collin(OD) said:

it presents a test bench that is unrealistic for the vast majority of users and uses benchmarks that are synthetic, thus also irrelevant to the average user. Synthetic benches cannot be compared and projected to real world applications.

As you can attest, it is difficult to find a suitable article to support each of our claims. The articles that you have provided are using an "unrealistic" testbench as well... z97 platform with an intel 4th generation cpu, running windows 8.1 (as per review 2). The Asus z97-pro motherboard used in your "evidence", is representative of the transition period when pci 3.0 first released in spring of 2014, and of a time when the tech was still maturing. You say "real world" differences, and then site articles that use PCMARK 7 & 8 (both synthetic benchmarks) as your "real world" evidence (as per review 1).

 

I can say that I'm not swayed by your "evidence"... How about a side-by-side comparison video between an Nvme, SATA, and HDD (published in 2016, using "current" hardware) loading into the OS, a few games, and compressing some files in 7zip? 

 

https://www.techspot.com/news/67222-storage-real-world-performance-nvme-vs-sata-vs-hdd.html

 

If you want to argue that 5 seconds in one load time, or 20 seconds in another, is "very little difference", well maybe to you it is.. It's all in your own perspective. The fact remains that even you have admitted:

6 hours ago, Collin(OD) said:

M.2 is dope, though.

And nothing you have provided here, has shown otherwise. Now, if you want to argue price per performance, I won't argue that SATA ssd's are at a much lower cost, and the performance gains that an Nvme drive will give you doesn't support making the jump (for the "average user"), but I've never been anything other than impressed with my choice to make that leap, and I'm pretty sure you are too.

 

 

Share this post


Link to post
Share on other sites
54 minutes ago, CompFreak(OD) said:

I'm a bit confused here... Are you literally arguing against your own statements? Your own evidence supports that Nvme drives are at the top of almost every test they ran against competing Nvme's and SATA ssd's.. Which is your position here?

 

As you can attest, it is difficult to find a suitable article to support each of our claims. The articles that you have provided are using an "unrealistic" testbench as well... z97 platform with an intel 4th generation cpu, running windows 8.1 (as per review 2). The Asus z97-pro motherboard used in your "evidence", is representative of the transition period when pci 3.0 first released in spring of 2014, and of a time when the tech was still maturing. You say "real world" differences, and then site articles that use PCMARK 7 & 8 (both synthetic benchmarks) as your "real world" evidence (as per review 1).

 

I can say that I'm not swayed by your "evidence"... How about a side-by-side comparison video between an Nvme, SATA, and HDD (published in 2016, using "current" hardware) loading into the OS, a few games, and compressing some files in 7zip?

 

https://www.techspot.com/news/67222-storage-real-world-performance-nvme-vs-sata-vs-hdd.html

 

If you want to argue that 5 seconds in one load time, or 20 seconds in another, is "very little difference", well maybe to you it is.. It's all in your own perspective. The fact remains that even you have admitted:

And nothing you have provided here, has shown otherwise. Now, if you want to argue price per performance, I won't argue that SATA ssd's are at a much lower cost, and the performance gains that an Nvme drive will give you doesn't support making the jump (for the "average user"), but I've never been anything other than impressed with my choice to make that leap, and I'm pretty sure you are too.

 

 


Tons of people are still using Z97 and Haswell / Devil's Canyon processors. Like.. probably a lot more people than are using Kaby Lake or Coffee Lake processors. I think your timeline is off. PCIe 3.0 debuted in 2010? PCIe 3.1 was 2013/2014. Haswell / Devil's Canyon processors were end of 2013/2014.

 

These are all relevant:

 

b9d013c1eaec8e7682ace67bfb7f883a.png

 

 

Also, PCMark simulates the actual applications. Such as specific games or applications. Not just arbitrary reading/writing

 

I said M.2 is dope. I didn't say NVMe is dope. M.2 != NVMe. NVMe is silly. 

Share this post


Link to post
Share on other sites
Posted (edited)
3 hours ago, Collin(OD) said:

Tons of people are still using Z97 and Haswell / Devil's Canyon processors. Like.. probably a lot more people than are using Kaby Lake or Coffee Lake processors.

Proof? 

 

When discussing the relevance of one of the newest technologies when it comes to the PC scene, do you not think it prudent to provide evidence using the latest technologies and testing methods? 

 

3 hours ago, Collin(OD) said:

Also, PCMark simulates the actual applications.

"simulates the actual applications" verses Real time, side-by-side loading times... Which do you think is more accurate?

 

3 hours ago, Collin(OD) said:

I think your timeline is off. PCIe 3.0 debuted in 2010?

Yeah, that may very well be, but when did it become readily available in the market? the first pci-e 3.0 board, not until 2011, and before it became "mainstream"? My thought is around the end of 2013 or beginning of 2014, Like you said:

3 hours ago, Collin(OD) said:

Haswell / Devil's Canyon processors were end of 2013/2014.

Seems kind of silly to sell motherboards without compatible processors utilizing the 1150 socket and it's ability to take advantage of the newest technologies, doesn't it? 

 

For example, did you know that Pci-e 4.0 was released in October of 2017? No? That's because even though they've released the specs and standards, the PC market isn't ready for it yet. Why do you think Nvidia isn't scheduled to release a new graphics card until summer of 2019? You think it could possibly have to do with the fact that because there is very little competition with their flagship Video Cards and they are waiting for motherboards / chipsets to support the new standards? Just maybe.... But that is all theoretical at the moment. Moore's law still applies (for the most part, only giving way to the marketability of advancements), and providing "evidence" from 4 years ago just isn't exactly relevant in today's cutting edge of technology... Hell, OCZ doesn't even exist anymore, they were purchased by Toshiba for $35 million back at the end of 2013. Anyways, As you have yet to provide any "relevant" evidence, there isn't much point in beating a dead horse here, your argument (which is now the "sillyness" of Nvme?) is invalid, and the only (we'll call it "dated") evidence you can muster, still shows Nvme drives topping 95% of the tests. Keep reading your "evidence" to the final conclusions page:

performance.png

 

Strange... that looks like every single one of the Nvme Drives in that entire lineup, have at least a 63% better overall score than the closest SATA drives.... Interesting, isn't it?

Just gonna leave these here:

11 hours ago, Collin(OD) said:

If you want a faster drive, you want NVMe.

3 hours ago, Collin(OD) said:

NVMe is silly. 

Edited by CompFreak(OD)
  • Like 1

Share this post


Link to post
Share on other sites

 

 

Thanks LinusMediaGroup!!

  • Upvote 1

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×