Performance Comparisons – Mixed Burst
These are the Mixed Burst results introduced in the Samsung 850 EVO 4TB Review. Some tweaks have been made, namely, QD reduced to a more realistic value of 2. Read bursts have been increased to 400MB each. 'Download' speed remains unchanged.
In an attempt to better represent the true performance of hybrid (SLC+TLC) SSDs and to include some general trace-style testing, I’m trying out a new test methodology. First, all tested SSDs are sequentially filled to near maximum capacity. Then the first 8GB span is preconditioned with 4KB random workload, resulting in the condition called out for in many of Intel’s client SSD testing guides. The idea is that most of the data on an SSD is sequential in nature (installed applications, MP3, video, etc), while some portions of the SSD have been written to in a random fashion (MFT, directory structure, log file updates, other randomly written files, etc). The 8GB figure is reasonably practical since 4KB random writes across the whole drive is not a workload that client SSDs are optimized for (it is reserved for enterprise). We may try larger spans in the future, but for now, we’re sticking with the 8GB random write area.
Using that condition as a base for our workload, we now needed a workload! I wanted to start with some background activity, so I captured a BitTorrent download:
This download was over a saturated 300 Mbit link. While the average download speed was reported as 30 MB/s, the application’s own internal caching meant the writes to disk were more ‘bursty’ in nature. We’re trying to adapt this workload to one that will allow SLC+TLC (caching) SSDs some time to unload their cache between write bursts, so I came to a simple pattern of 40 MB written every 2 seconds. These accesses are more random than sequential, so we will apply it to the designated 8GB span of our pre-conditioned SSD.
Now for the more important part. Since the above ‘download workload’ is a background task that would likely go unnoticed by the user, we also need is a workload that the user *would* be sensitive to. The times where someone really notices their SSD speed is when they are waiting for it to complete a task, and the most common tasks are application and game/level loads. I observed a round of different tasks and came to a 200MB figure for the typical amount of data requested when launching a modern application. Larger games can pull in as much as 2GB (or more), varying with game and level, so we will repeat the 200MB request 10 times during the recorded portion of the run. We will assume 64KB sequential access for this portion of the workload.
Assuming a max Queue Depth of 4 (reasonable for typical desktop apps), we end up with something that looks like this when applied to a couple of SSDs:
The OCZ Trion 150 (left) is able to keep up with the writes (dashed line) throughout the 60 seconds pictured, but note that the read requests occasionally catch it off guard. Apparently, if some SSDs are busy with a relatively small stream of incoming writes, read performance can suffer, which is exactly the sort of thing we are looking for here.
When we applied the same workload to the 4TB 850 EVO (right), we see an extremely consistent and speedy response to all IOs, regardless of if they are writes or reads. The 200MB read bursts are so fast that they all occur within the same second, and none of them spill over due to other delays caused by the simultaneous writes taking place.
Now for the results:
From our Latency Percentile data, we are able to derive the total service time for both reads and writes, and independently show the throughputs seen for both. Remember that these workloads are being applied simultaneously, as to simulate launching apps or games during a 20 MB/s download. The above figures are not simple averages – they represent only the speed *during* each burst. Idle time is not counted.
The important metric here is reads since writes would be in the background in this scenario. The 970 PRO offers the most impressive result here, nearly matching Intel Optane SSDs (see larger comparison charts below). The newer 970 EVOs now fall half way between the older 960 EVO and PRO models.
Now we are going to focus only on reads, and present some different data. I’ve added up the total service time seen during the 10x 400MB reads that take place during the recorded portion of the test. These figures represent how long you would be sitting there waiting for 4GB of data to be read, but remember this is happening while a download (or another similar background task) is simultaneously writing to the SSD. This metric should closely equate to the 'feel' of using each SSD in a moderate to heavy load. Total read service times should hopefully help you grasp the actual time spent waiting for such a task to complete in the face of background writes taking place.
The 970 EVO's once again fall neatly between the older 960 EVO and PRO models, but look at that 970 PRO result! To put that impressive 3.7s figure into perspective, consider that the intel Optane 900P scored 3.4s in this same metric.
Crazy long comparison charts follow (sorted by random read performance to keep things consistent chart-to-chart):
$.10 per GB or GTFO
$.10 per GB or GTFO
It’s hard to take comments
It’s hard to take comments like yours seriously.
People considered $1.00/GB a breakthrough affordability price point right up until it was reached. Then people considered it offensively expensive and said $0.50/GB was the affordability point. We passed that point as well and much the same thing occurred.
Now here we are with people complaining that disks are not $0.10/GB. If this new disk -were- that cheap, I suspect that you’d be complaining that it’s not $0.05/GB. Please recognize that you’re unlikely to ever be satisfied at any price point. Setting realistic expectations goes a long way toward fulfillment.
In the OP’s defense, $0.10/GB
In the OP's defense, $0.10/GB is a bit of a running joke on our podcast. Ryan want's it to happen yesterday, and I keep reminding him that we're just not there yet.
If not yesterday, how about
If not yesterday, how about tomorrow? 🙂
haha, glad someone got the
haha, glad someone got the reference.
I think this is not limited to just Ryan wanting it to be $.10/GB, I am pretty sure everyone wants that(by everyone, I mean customers).
So, 970 series is slightly
So, 970 series is slightly better than the 960 series. That’s what I got from this review.
Nothing escapes you
/s
Nothing escapes you
/s
That’s pretty much it.
That's pretty much it.
LOL @ prices.
LOL @ prices.
Some misinformation going on
Some misinformation going on here. IEEE1667 finally made it to the Samsung 960 EVO and Pro with the latest firmware. Or actually, the two latest in terms of the Pro but the previous got pulled.
The sad thing is that Samsung keeps complaining about UEFI firmware issues with most motherboards making it impossible to get IEEE1667/Microft Edrive to work with the 960 EVO and Pro as boot drives which is likely what they are being used as 99% of the time.
It’s not working with neither my Asus Maximus IX Apex or my Asus Maximus X Apex both running the latest BIOS/UEFI Firmware version and when contacting Asus about the problem they claim they don’t know anything about such a issue.
It works perfectly when the drive is being used as a secondary drive. So there seems to be something going on with the NVMe module in the UEFI firmware and how it loads in terms of Windows 10.
The big question is.. How does this all work with these new drives? Do they magically work without a UEFI Firmware fix like Samsung keeps claiming is needed for the 960 EVO and Pro and if so how is it that these new ones don’t require the same fix from motherboard manufactures?
Hopefully PC-Per and others can do some digging here.
Seems to be an issue with the
Seems to be an issue with the BIOS chipset manufacturers like Megatrends, Phoenix, etc, and Samsung has stated they are working with them to resolve the issue.
I consider IEEE1667 broken
I consider IEEE1667 broken for 960 until the community reports that it is working (especially after the firmware back and forths). Same goes for the 970. I'm taking Samsung at their word for this launch, but that will change if the community feedback is the same as it was for the 960. We have a limited sample size of systems that it may or may not work on, so this particular niche use case is better left to those more experienced in using it.
Would like to see a test
Would like to see a test measuring write speed using full disk encryption. The results would probably be similar to the saturated write on a full drive, but given that this is not an unusual setup these days, it might be interesting.
Modern SSDs encrypt to the
Modern SSDs encrypt to the disk regardless. Enabling encryption at the host level just changes the key.
Page “Performance Focus –
Page “Performance Focus – Samsung 970 EVO 250GB, 500GB, 1TB”, under sequential 250GB graph reads “1TB shows a cached (burst) write speed of 1.5GB/s, with sustained (saturated) writes falling off to ~300MB/s.”.
Probably needs to be changed to “250GB shows a cached […]”.
Thanks for the catch. Fixed!
Thanks for the catch. Fixed!
Im a little disappointed you
Im a little disappointed you guys weren’t able to do the same with the new WD Black . For friends,family, and myself i usually dont go higher than purchasing the 250GB. I then always use for storage a regular HDD. I see that the 1TB is on here , i already purchased the WD 250 and installed it last week. Would been cool to see both the WD 250GB VS the new Evo 250GB since there price the same. Like a budget to budget which should you get kinda deal since the better performance is usually is seen on the high capacity drives which is what was tested.
Were you able to use WD Black
Were you able to use WD Black 250 GB as a Windows 10 boot drive?
That was all we were sampled.
That was all we were sampled. I did ask for lower capacities…
What was the NVME driver
What was the NVME driver version used? Tom’s claims to get lower perf on the newest 1.3 drivers.
We used Samsung 3.0 drivers.
We used Samsung 3.0 drivers.
$0.01 per GB or GT*O
/s
$0.01 per GB or GT*O
/s
Any chances of a HP EX920
Any chances of a HP EX920 review? A few reviewers actually put it slightly faster than the 970 EVO at some tasks, especially real world testing and low queue depths all the while being much cheaper.
Allyn, I love how hyped you
Allyn, I love how hyped you are on storage, it’s ridiculous and awesome! I will not purchase a drive without it getting your “Editor’s Choice” stamp. ( my own money, sometime server budgets say otherwise ) Keep it up man!
Yes, we will soon be
Yes, we will soon be crowd-funding a prototype that will clone multiple copies of Allyn Malventano, for exclusive competitions against AI robots falsely claiming comparable knowledge, experience and analytical capabilities — kinda like famous chess matches with Russian masters of times past. My money is on Allyn (and clones), every time!
Hi guys, really appreciate
Hi guys, really appreciate your work. It would be really interesting to compare Samsung 970 with Seagate Nytro 3730 SSD Dual 12 Gb/s SAS 3D eMLC 400gb. Maybe in raid combination too. Just a thought that tickle senses. 🙂 Keep up the good work.
Alan, money no object someone
Alan, money no object someone is giving you a SSD for free. What do you choose Optane P900 or 970 Pro?
Money no object, I’d probably
Money no object, I'd probably do the 900P, but only in a >480GB capacity and if I had a spare slot to support it. M.2 is way more convenient for client storage.
Great review. Thank you for
Great review. Thank you for the context on scenarios wrt write speed after fast write. It shaped my purchasing decision. I always check w pcper Allyn before making a storage decision! I’m going to try one of these puppies on my trusty Z77 board which actually has an NVMe UEFI for M.2 boot. I’ll post results re: IEEE1667.
So, Allyn, upgrading from a
So, Allyn, upgrading from a 950pro 256 to a 970evo 1 gig much of an improvement besides the capacity? I have 3 950 pros in a soc force Mobo.