The Samsung 950 Pro PCIe SSD Review (256GB and 512GB)
by Billy Tallis on October 22, 2015 10:55 AM ESTPerformance Consistency
Kicking things off, our performance consistency test saturates the drive with 4kB random writes for a full hour, with a queue depth of 32, the maximum supported by the AHCI protocol used by SATA and most PCIe drives. This puts the drive's controller under maximum stress and writes enough data to exhaust all free space and spare area on the drive. This is an unrealistic workload for any client use, but it provides a worst-case scenario for long-term heavy use, and it sheds light on how different SSD controllers behave and if their performance will hold up as they fill up.
The average of the last 400 seconds of the test gives us a steady-state IOPS rating that is usually very different from what the manufacturer specifies for a new, empty drive. We also quantify the consistency of the drive's random write performance, and provide plots of the performance over the course of the test.
Once steady state is reached, performance is determined more by the controller's algorithms than the interface speed, so it's not too surprising to see the 950 Pro performing similarly to other Samsung drives.
The consistency metric shows a surprising disparity between the two 950 Pros, with the 256GB performing much better.
 |
|||||||||
| Default | |||||||||
| 25% Over-Provisioning | |||||||||
Comparing the graphs of the two 950s shows that the inconsistency of the 512GB drive comes from frequent jumps in performance above a solid baseline. This pattern holds even for the test with overprovisioning. Graphing the power consumption over time (not shown) reveals that the periods of lower performance have lower power. If the lower performance were due to periodic background garbage collection, then we would expect power consumption to be at least as high as when the drive is performing well. Instead, it appears that the 512GB drive is experiencing thermal throttling.
 |
|||||||||
| Default | |||||||||
| 25% Over-Provisioning | |||||||||
With most of its time spent thermally limited, our 512GB sample's low average is explained. It appears that the thermal throttling mechanism is bumping the drive down to one of several discrete performance levels, rather than a continuous performance mechanism.
Facebook
Twitter
RSS
142 Comments
View All Comments
MHz Tweaker - Monday, November 16, 2015 - link
I have an IT service company and have been doing IT for a few decades. I am with you on how incredibly annoying it is to work on some of the hodge podge of hardware that comes in for service. Specifically there are those that refuse to upgrade early P4's or Sempron single core systems with 512 or 256 megs of RAM. Certainly these should have failed by now. They barely run XP let alone modern security needed to protect them. Internally all our workstations are Z97, X99, x79, Z170's with 4690K's, 4790K's, 5930K, 3930K and 6700K CPUs. Every machine has a Samsung or Intel SSD for a boot drive and a Hitachi 4, 5 or 6TB drive for work and storage. All systems also have 16 or 32GB RAM. I am just amazed at how much time the average person wastes waiting on an entry level PC to do things. I mean many of us spend in excess of 40 hours a week at a keyboard. I get much more done when I keep my equipment in tip top shape and fresh. Nothing has a chance to break down. Many companies throw tons of money at "image" but internally the infrastructure is held together with bubble gum and band aids.Oh BTW, I got our first 950 Pro 512MB Yesterday :-)
Deders - Thursday, October 22, 2015 - link
Until Samsung announce again that SSD's are not meant to use sleep, their excuse for computers with 830/840's freezing for 30 seconds after waking up.BurntMyBacon - Friday, October 23, 2015 - link
@Deders: "Until Samsung announce again that SSD's are not meant to use sleep, their excuse for computers with 830/840's freezing for 30 seconds after waking up."Is this a known issue or are you just speculating about what might (probably?) occur in the future?
I am using quite a few 830s and have deployed quite a few more all (up to this point) without this issue. In my experience, they've been some of the most reliable drives I've used. I know of a few 840s without issues as well, but I've tried to avoid them as much as practical as I'm still not convinced the industry fully understands the long term consequences of simultaneously increasing leakage and cutting down the margin of error to get that third (fourth, fifth, etc.) bit in. I suspect it'll be fine on some (larger) process nodes, but the 840EVO issues suggests that there is a lower limit to how small you can get while trying to get that extra bit (and storing long term). In any case, it would be nice to know if there is something I should be looking out for here and whether practical mitigations exist.
bill.rookard - Sunday, October 25, 2015 - link
Agreed. I have several systems with the 830s in them and have zero problems with them. I avoided the 840s because those went with the TLC and I was iffy about them, and the 830s are rock solid and very fast.Chaser - Tuesday, October 27, 2015 - link
Boot times matter to me too. Makaveli must have a remote control while he's making his bagels.carl0ski - Monday, May 16, 2016 - link
I'd say the most annoying part is the Windows boot logo is a fixed time to complete.So even if Windows 7 8 10 boot up faster it still waits for the pretty video.
Turn it off from msconfig and windows 7 on 850 will boot in 3 seconds instead of 10 seconds
beginner99 - Thursday, October 22, 2015 - link
Most people that buy such a product usually are enthusiast and play around like over clocking. 10 sec more per boot is then pretty annoying.Makaveli - Thursday, October 22, 2015 - link
I'm from the era of overclocking with jumpers on a motherboard and far longer post times. If waiting 10 seconds for a machine to boot is too long while trying to get a stable overclock then do it in windows. Or does everyone in this generation have ADD?AnnonymousCoward - Thursday, October 22, 2015 - link
Oh, so time is only valuable to those with ADD? You're a fool.Makaveli - Thursday, October 22, 2015 - link
Go troll else where tool!