My SSD isn't performing like it should

[mrchetsteadman]

I ran a benchmark on my computer and here are the results;

 

UserBenchmarks: Game 119%, Desk 128%, Work 112%
CPU: Intel Core i7-8700K - 103.9%
GPU: Nvidia GTX 1080 - 132.3%
SSD: Samsung 960 Evo NVMe PCIe M.2 250GB - 213.8%
SSD: Samsung 960 Evo NVMe PCIe M.2 1TB - 228.4%
RAM: Corsair Vengeance LPX DDR4 3000 C15 2x8GB - 100.2%
MBD: Asus ROG MAXIMUS X HERO (WI-FI AC)

 

My question is, even though I don't notice an issue, why would my SSD be performing "poorly"?

 

 

Should I ignore this? Should I use a different benchmark program? Thanks in advance for all the help.

[farmeunit]

Benchmarks have little to do with real-life performance.  What about it is performing "poorly"?  If you're comparing to others there are LOTS of variables.  Motherboards primarily, but lots of things can be going on.  If you look at your speeds, they mostly fine, correct?  Just because it's a little slower than someone else's right, doesn't really mean much.

 

[Alera]

SSDs generally slow down the more files that are on them and your speeds look lower than those usually seen for those drives. Perhaps your build doesn't have enough PCI-E lanes or the storage controller can't cope? Need more information for sure.

[mrchetsteadman]

10 minutes ago, farmeunit said:

Benchmarks have little to do with real-life performance.  What about it is performing "poorly"?  If you're comparing to others there are LOTS of variables.  Motherboards primarily, but lots of things can be going on.  If you look at your speeds, they mostly fine, correct?  Just because it's a little slower than someone else's right, doesn't really mean much.

 

 

I don't personally see a difference at all. I'm going solely based on these tests. Will I lose sleep over it? Not at all. I just finished Arkham Knight on Ultra and not even a single stutter no matter how many tank battles I got into. The reason for the inquiry is to see if there is something I am missing. When I ran the tests before, my GPU was performing below average because I had G-Sync enabled which doesn't make a lick of sense. I turned off G-Sync, just for the tests, and my score jumped from a 47% to 132%. Are there settings or drivers that I am missing somewhere that is giving these readings or is it just a glitch? It's more for peace of mind really. 

 

3 minutes ago, Alera said:

SSDs generally slow down the more files that are on them and your speeds look lower than those usually seen for those drives. Perhaps your build doesn't have enough PCI-E lanes or the storage controller can't cope? Need more information for sure.

That makes sense. What other information do you need?

[Alera]

19 minutes ago, mrchetsteadman said:

 

I don't personally see a difference at all. I'm going solely based on these tests. Will I lose sleep over it? Not at all. I just finished Arkham Knight on Ultra and not even a single stutter no matter how many tank battles I got into. The reason for the inquiry is to see if there is something I am missing. When I ran the tests before, my GPU was performing below average because I had G-Sync enabled which doesn't make a lick of sense. I turned off G-Sync, just for the tests, and my score jumped from a 47% to 132%. Are there settings or drivers that I am missing somewhere that is giving these readings or is it just a glitch? It's more for peace of mind really. 

 

That makes sense. What other information do you need?

Looking at your components, I can't see there being any slowdown on the CPU/motherboard side of things.

 

Have you tried using Samsung Magician, make sure Firmware is updated, see what the health of the drive shows up as, run the benchmark test and see what numbers it spits out for you? See if they're any different from UserBenchmark, I'd also try CrystalDiskMark while you're there. That could shed some light, if the numbers are similar then perhaps it really is down to your storage use!

 

EDIT: So I did UserBenchmark for myself using my Samsung 970 EVO NVMe drive and it shows:

Read 1,959

Write 2,225

 

Samsung Magician shows:

Read 3,568

Write 2,394

 

There's a major difference in read speeds there which is very interesting.

Edited November 26, 2018 by Alera

[farmeunit]

Not sure on GSync, but I'm assuming your low percentage is fine.

 

For example, in Rainbow Six Siege, I get 140+ FPS, but my Freesync is limited to 70 FPS due to monitor limitations.  So it will LOOK slower because it's less frames, even though it COULD do 140 if I took Freesync off.  I'm assuming that's how it works, but I'm not an expert.  Again, benchmarks mean little because of variables.  They're fine for showing general stats, but hard to accept for complete "truth".

 

You would need to get a high refresh rate monitor for your numbers to be "more accurate".  Again, that's my assumption and may not be fact.

 

 

[Mindovermaster Moderator]

Benchmarks are only for bragging rights. They don't show day-to-day operations. Between FF and any game, theres a big difference in performance. You aren't gaming 24/7/365...

[xendrome]

15 minutes ago, Mindovermaster said:

Benchmarks are only for bragging rights. They don't show day-to-day operations. Between FF and any game, theres a big difference in performance. You aren't gaming 24/7/365...

That's not necessarily true, they can also show under-preforming hardware which may be the case here, hence his post.

[Mindovermaster Moderator]

5 minutes ago, xendrome said:

That's not necessarily true, they can also show under-preforming hardware which may be the case here, hence his post.

Be that as it may. But it read/writes under different circumstances as posted above.

[+Biscuits Brown MVC]

If you really want blazing performance (that you will never see) with an NVME drive, you really should run Linux because the Linux I/O Scheduler greatly out performs the Windows variant.  Of course, that's at the cost of running Linux as a daily driver and that's not something everyone can do depending on their own PC usage. 

[CrashG]

As Alera's post shows, different benchmarks yield different results. Try CrystaldiskInfo for a third opinion

 

Edited November 27, 2018 by CrashGordon

[mrchetsteadman]

16 hours ago, Alera said:

Looking at your components, I can't see there being any slowdown on the CPU/motherboard side of things.

 

Have you tried using Samsung Magician, make sure Firmware is updated, see what the health of the drive shows up as, run the benchmark test and see what numbers it spits out for you? See if they're any different from UserBenchmark, I'd also try CrystalDiskMark while you're there. That could shed some light, if the numbers are similar then perhaps it really is down to your storage use!

 

EDIT: So I did UserBenchmark for myself using my Samsung 970 EVO NVMe drive and it shows:

Read 1,959

Write 2,225

 

Samsung Magician shows:

Read 3,568

Write 2,394

 

There's a major difference in read speeds there which is very interesting.

 

-----------------------------------------------------------------------
CrystalDiskMark 6.0.2 x64 (C) 2007-2018 hiyohiyo
                          Crystal Dew World : https://crystalmark.info/
-----------------------------------------------------------------------
* 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) :  2882.722 MB/s
Sequential Write (Q= 32,T= 1) :  1526.146 MB/s
Random Read 4KiB (Q=  8,T= 😎 :  1186.476 MB/s [ 289667.0 IOPS]
Random Write 4KiB (Q=  8,T= 😎 :  1132.452 MB/s [ 276477.5 IOPS]
Random Read 4KiB (Q= 32,T= 1) :   423.172 MB/s [ 103313.5 IOPS]
Random Write 4KiB (Q= 32,T= 1) :   366.112 MB/s [  89382.8 IOPS]
Random Read 4KiB (Q=  1,T= 1) :    45.981 MB/s [  11225.8 IOPS]
Random Write 4KiB (Q=  1,T= 1) :   173.870 MB/s [  42448.7 IOPS]

Test : 1024 MiB [C: 42.7% (99.3/232.3 GiB)] (x5)  [Interval=5 sec]
Date : 2018/11/27 8:19:34
OS : Windows 10  [10.0 Build 17134] (x64)

 

-----------------------------------------------------------------------
CrystalDiskMark 6.0.2 x64 (C) 2007-2018 hiyohiyo
                          Crystal Dew World : https://crystalmark.info/
-----------------------------------------------------------------------
* 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) :  1778.604 MB/s
Sequential Write (Q= 32,T= 1) :  1556.647 MB/s
Random Read 4KiB (Q=  8,T= 😎 :  1410.898 MB/s [ 344457.5 IOPS]
Random Write 4KiB (Q=  8,T= 😎 :  1397.156 MB/s [ 341102.5 IOPS]
Random Read 4KiB (Q= 32,T= 1) :   463.742 MB/s [ 113218.3 IOPS]
Random Write 4KiB (Q= 32,T= 1) :   419.083 MB/s [ 102315.2 IOPS]
Random Read 4KiB (Q=  1,T= 1) :    45.330 MB/s [  11066.9 IOPS]
Random Write 4KiB (Q=  1,T= 1) :   176.270 MB/s [  43034.7 IOPS]

Test : 1024 MiB [D: 59.3% (552.7/931.5 GiB)] (x5)  [Interval=5 sec]
Date : 2018/11/27 8:25:17
OS : Windows 10  [10.0 Build 17134] (x64)
 

Here are the numbers from Samsung Magician. I really have no idea what it says lol.

 

250 GB Drive

 

1 TB Drive

16 hours ago, farmeunit said:

Not sure on GSync, but I'm assuming your low percentage is fine.

 

For example, in Rainbow Six Siege, I get 140+ FPS, but my Freesync is limited to 70 FPS due to monitor limitations.  So it will LOOK slower because it's less frames, even though it COULD do 140 if I took Freesync off.  I'm assuming that's how it works, but I'm not an expert.  Again, benchmarks mean little because of variables.  They're fine for showing general stats, but hard to accept for complete "truth".

 

You would need to get a high refresh rate monitor for your numbers to be "more accurate".  Again, that's my assumption and may not be fact.

 

 

My refresh rate is 120 Hz on my AW3418DW. And to be clear, I do not see an issue at all personally but it's always good to know everything is working the way it should.

[adrynalyne]

Those benchmarks are about right for that drive. 

[DevTech]

14 hours ago, mrchetsteadman said:

 

 

Here are the numbers from Samsung Magician. I really have no idea what it says lol.

 

250 GB Drive

 

1 TB Drive

My refresh rate is 120 Hz on my AW3418DW. And to be clear, I do not see an issue at all personally but it's always good to know everything is working the way it should.

You haven't given your system specs, so this whole thread is like how many angels dance on a pinhead...

 

- Verify both drives are getting a full allocation of 4 PCIe lanes each from the motherboard - many mobos steal PCIe lanes between NVME and SATA so sometimes you need to leave certain SATA ports unused.

 

- There is a certain amount of disagreement concerning the Samsung NVMe device drivers as to which are faster, 2.0 or 3.0 so you should check that in Device Manager - they will be listed near the bottom in Storage Devices (something like that) NOT where you expect in Disk Drives.

 

 

 

 

[Mindovermaster Moderator]

17 minutes ago, DevTech said:

You haven't given your system specs, so this whole thread is like how many angels dance on a pinhead...

 

Umm, yeah he did... well, partly.

[DevTech]

1 hour ago, Mindovermaster said:

Umm, yeah he did... well, partly.

Thanks, I didn't notice his original benchmark list as a system spec.

 

So, let's proceed then....

 

1. In BIOS, set M.2_1 to "PCIE mode"

 

2. set M.2_2 to X4 which will DISABLE SATA ports 5 and 6

 

3. if anything connected to SATA 5, 6 move elsewhere!

 

4. verify you have latest BIOS of Oct 2

 

5. Check benchmarks before and after, but you probably want to remove all aspects of Intel RST drivers

 

6. Install the latest Intel chipset drivers from the ASUS site, then install more recent ones form Intel site if available

 

7. Install Samsung NVMe drivers 3.0

 

8. Win 10, turn off hybernation

 

9. Win 10, assign a fixed Page file around 12 gigs

 

10. Win 10, turn off fast boot

 

11. If you ever reinstall Windows, find super secret Samsung utility to RESET the drives so you can over-provision them. That will speed up the drives when they get close to full. I over-provision ALL SSD drives anyways to increase longevity.

 

12. Given the cost of these components, NOT using a Samsung 970 Pro for the O/S drive is a glaring deficiency.

 

 

[DevTech]

On 11/26/2018 at 7:40 PM, Zag L. said:

If you really want blazing performance (that you will never see) with an NVME drive, you really should run Linux because the Linux I/O Scheduler greatly out performs the Windows variant.  Of course, that's at the cost of running Linux as a daily driver and that's not something everyone can do depending on their own PC usage. 

Yeah, thats just silly. Nothing more advanced than the Windows 10 Kernel. Nothing more flexible than the Windows ecosystem. Your proposal would be well advised for masochists and fans of self defeating behavior.

 

[mrchetsteadman]

On 11/28/2018 at 1:14 AM, DevTech said:

Thanks, I didn't notice his original benchmark list as a system spec.

 

So, let's proceed then....

 

1. In BIOS, set M.2_1 to "PCIE mode"

 

2. set M.2_2 to X4 which will DISABLE SATA ports 5 and 6

 

3. if anything connected to SATA 5, 6 move elsewhere!

 

4. verify you have latest BIOS of Oct 2

 

5. Check benchmarks before and after, but you probably want to remove all aspects of Intel RST drivers

 

6. Install the latest Intel chipset drivers from the ASUS site, then install more recent ones form Intel site if available

 

7. Install Samsung NVMe drivers 3.0

 

8. Win 10, turn off hybernation

 

9. Win 10, assign a fixed Page file around 12 gigs

 

10. Win 10, turn off fast boot

 

11. If you ever reinstall Windows, find super secret Samsung utility to RESET the drives so you can over-provision them. That will speed up the drives when they get close to full. I over-provision ALL SSD drives anyways to increase longevity.

 

12. Given the cost of these components, NOT using a Samsung 970 Pro for the O/S drive is a glaring deficiency.

 

 

I'm printing these steps out and doing everything listed. Thanks! So you would recommend a 970 Pro? I'll probably pick one up today.

[farmeunit]

If you have money to blow, sure, get the 970 Pro.  What you have is fine.  You'll hardly notice the difference day to day.

[DevTech]

1 hour ago, farmeunit said:

If you have money to blow, sure, get the 970 Pro.  What you have is fine.  You'll hardly notice the difference day to day.

For my minor note on his config to warrant actual action, it would need at lot more information about objectives and various use-cases for that computer, but of course there is a HOBBY aspect to all of this as well

 

 

If by some chance, a new boot drive is purchase, it should be "over provisioned" by leaving about 5% to 10% of the space unallocated. This only works if you manually create the O/S partition before install. And you can't make an "empty" partition, it must be NEVER TOUCHED FOR ALL TIME for the drive to use it as a source of SPARE WRITE CELLS. And you can't shrink or grow or reformat to get this bonus... It's very easy to do this and also very easy to get it wrong by not fully understanding what "unallocated (i.e. never ever allocated)" means...

 

[farmeunit]

Samsung Magician works by over-provisioning after installation.  Doesn't that not work for NVMe drivers?  I have only used it on SATA drivers.  Crucial does it the same way, also.  Install Windows, install Magician, click the button to Over-Provision.

 

I agree with your hobby comment, but I wouldn't say it's a "glaring deficiency" if he's on a budget.  It makes it sound like he'll see a big improvement in performance which in 99% of use cases, he likely won't, except in benchmarks.  If he's in the 1%, great.  If not, it's a waste of money.

[DevTech]

46 minutes ago, farmeunit said:

Samsung Magician works by over-provisioning after installation.  Doesn't that not work for NVMe drivers?  I have only used it on SATA drivers.  Crucial does it the same way, also.  Install Windows, install Magician, click the button to Over-Provision.

 

I agree with your hobby comment, but I wouldn't say it's a "glaring deficiency" if he's on a budget.  It makes it sound like he'll see a big improvement in performance which in 99% of use cases, he likely won't, except in benchmarks.  If he's in the 1%, great.  If not, it's a waste of money.

I'll check the magic thing out... it should not be possible but I guess Samsung could have a secret API into their own drive...

 

I have a Pro and a non-Pro both fast, but with the Pro is very easy to see the difference  on anything with a lot of writes like installing apps, Windows Updates, etc so I can't agree it isn't worth the money and would word it the opposite to yours by saying do without the Pro ONLY when you have an extremely tight budget.