Samsung 950 Pro Review: The next step in SSD performance

[Steve]

Since the arrival of the XP941 last year, and later on the SM951 NVMe, we have been waiting for a more affordable mainstream version. Recently Samsung delivered just that with its new SSD 950 Pro Series. Made exclusively in the M.2 2280 form factor, the new 950 series comes in either 256GB or 512GB capacities -- priced at $190 ($0.74/GB) and $350 ($0.68/GB), respectively.

Featuring the same UBX 3-core controller found in the SM951, the 950 Pro receives native PCIe and NVMe support. Surprisingly, the SM951-NVMe was paired with Samsung's 16nm 64Gbit MLC NAND flash where as the mainstream 850 Pro series received the more advanced Samsung 3D V-NAND (32-layer 86Gbit MLC) flash.

This is where the 950 Pro has been upgraded, as the UBX controller has been paired with Samsung's 3D V-NAND (32-layer 128Gbit MLC).

Therefore technically there is nothing "new" about the 950 Pro series, it just combines all the best bits that Samsung has brought to market so far.

Read the complete review.

 

[HardReset]

I'd like to hear more about excellent Samsung warranty. Samsung 840 fiasco (not Evo, 840) is still unresolved and according to Samsung, slow drives are not enough "broken" for warranty.

And reliability is expected to be great given Samsung's history. Just remembering Samsung 840 Evo fiasco, this is easy to disagree.

 

[crocography]

I would have really liked to have seen many more real world benchmarks... I really don't understand why in 2015 we are still using synthetic benchmarks.

 

[cliffordcooley]

There is no harm in using synthetic benchmarks. They still serve to give a general idea of what to expect. Which is all any benchmark is good for.

 

[Theinsanegamer]

I bought one of these for my alienware 15r2. The two second boot time is just so fantastic. And pictures really do not do m.2 drives justice, they are tiny compared to 2.5 inch drives.

Now, if only they would hurry up with the 1TB model.

 

[Squid Surprise]

Alas I bought the Intel 750 back in June when it was the only option

Also, Page 9, the last 2 benchmarks read "higher is better" when they should say "lower is better" (Adobe and WOW)

 

[Steve]

I would have really liked to have seen many more real world benchmarks... I really don't understand why in 2015 we are still using synthetic benchmarks.

I think you will find we run far more real-world tests than most tech sites and synthetic benchmarks are still an accurate means of gauging true SSD performance.

Moreover it is very difficult to find real-world situations that will show the performance differences between various SSDs that you can accurately measure. That said I am always open to suggestions.

I bought one of these for my alienware 15r2. The two second boot time is just so fantastic. And pictures really do not do m.2 drives justice, they are tiny compared to 2.5 inch drives.

Now, if only they would hurry up with the 1TB model.

Yep a 1TB model really would be something. I know what you mean about the size, there is an M.2 and mSATA SSD in comparison to a few screws

https://static.techspot.com/articles-info/969/images/Image_11S.jpg

Alas I bought the Intel 750 back in June when it was the only option

Also, Page 9, the last 2 benchmarks read "higher is better" when they should say "lower is better" (Adobe and WOW)

The 750 is still a cracking SSD Thanks for pointing out the mistakes, they are fixed now.

 

[crocography]

I think you will find we run far more real-world tests than most tech sites and synthetic benchmarks are still an accurate means of gauging true SSD performance.

Moreover it is very difficult to find real-world situations that will show the performance differences between various SSDs that you can accurately measure. That said I am always open to suggestions.

Yes I know you do, and I do appreciate your site tremendously. (visit it everyday for tech news)

But I work heavily in the graphics industry where I never need to run synthetic benchmarks for anything, and what I find with most synthetic benchmarks is that they do not translate into real world experience for me. Whether it is video or image processing I find the synthetic benchmarks are always more than unrealistic in their actual difference in (what would be) my normal workflow. In real world experiences I find the technology to be much more similar in time.

But I do understand there needing to be a simple standard. And I really do apologize for knocking your site if that is what it sounded like.

 

[Steve]

Yes I know you do, and I do appreciate your site tremendously. (visit it everyday for tech news)

But I work heavily in the graphics industry where I never need to run synthetic benchmarks for anything, and what I find with most synthetic benchmarks is that they do not translate into real world experience for me. Whether it is video or image processing I find the synthetic benchmarks are always more than unrealistic in their actual difference in (what would be) my normal workflow. In real world experiences I find the technology to be much more similar in time.

But I do understand there needing to be a simple standard. And I really do apologize for knocking your site if that is what it sounded like.

No there was nothing wrong with either of your comments and we appreciate the feedback for sure. I was just explaining why you don't see as many real worlds tests as you would from one of our CPU or GPU reviews for example. I am always on the look out for applications we can use to accurately show off SSD performance, they just don't seem easy to come by.

 

[EClyde]

Whatever, I like it. I own it and things are faster.

 

[BlueDrake]

Could someone add PCIe / M.2 or something to the data? It's literally saying the same name, and to a general user it's confusing. I'm by no means clueless, but at a glance it just says "Samsung SSD 950 Pro 512GB" on both. Making it like "Wait.. how are you comparing it to itself?" xD

Small nitpick because I had to stop and check to make sure, I wasn't going crazy when the copy test had it's issues. Since I couldn't at first glance tell, if something was just mislabeled or such. Aside from the obvious bars.

 

[Steve]

Could someone add PCIe / M.2 or something to the data? It's literally saying the same name, and to a general user it's confusing. I'm by no means clueless, but at a glance it just says "Samsung SSD 950 Pro 512GB" on both. Making it like "Wait.. how are you comparing it to itself?" xD

Small nitpick because I had to stop and check to make sure, I wasn't going crazy when the copy test had it's issues. Since I couldn't at first glance tell, if something was just mislabeled or such. Aside from the obvious bars.

Hey mate, I am not sure what you are talking about here. Are you getting confused by the SATA Samsung SSD 840 Pro and the M.2 Samsung SSD 850 Pro?

 

[BlueDrake]

Hey mate, I am not sure what you are talking about here. Are you getting confused by the SATA Samsung SSD 840 Pro and the M.2 Samsung SSD 850 Pro?

Derp.. :\ I'm really bad at reading apparently. Just ignore my sleep deprived comment.

850 Pro vs 950 Pro, I somehow read them both as the same number.

 

[Steve]

Derp.. :\ I'm really bad at reading apparently. Just ignore my sleep deprived comment.

850 Pro vs 950 Pro, I somehow read them both as the same number.

Easy done mate, they are very similar

 

[Phr3d]

I tend to agree upon the benchmarks comment, as they've been gamed a bit in the past, and your 'unknown' mass copy test seems to bear this out. Dunno if drivers or what, but on paper the Pro should easily whomp the 951, and on synthetics it does so, but IMO the urrp-urrp on the not-so-well studied test shows issues - particularly in the price realm, paying a Lot for maximum performance that is getting harder to discern.
Be very interested to see this on a Win7 rig, but guessing those days are over now? (my win10 experience has been notably less than roses and unicorns).

 

[mystik]

I tend to agree upon the benchmarks comment, as they've been gamed a bit in the past, and your 'unknown' mass copy test seems to bear this out. Dunno if drivers or what, but on paper the Pro should easily whomp the 951, and on synthetics it does so, but IMO the urrp-urrp on the not-so-well studied test shows issues - particularly in the price realm, paying a Lot for maximum performance that is getting harder to discern.
Be very interested to see this on a Win7 rig, but guessing those days are over now? (my win10 experience has been notably less than roses and unicorns).

I wonder about that myself... Is the performance going to be better on Windows 10 than Windows 7 because of the drivers?