REAL WORLD FILE TRANSFER COMPARISON
For our file transfer comparison, included storage media consisted of the Kingston DC1000 1.6TB NVMe Data Center SSD, Samsung 960 Pro 2TB NVMe M.2 SSD and OWC Mercury Electra 2TB SATA 3 SSD. Our data samples consisted of 15GB folders of video, music, photos and operating system and all movement was simply copying from one folder to another on the drive. The Optane was a little different. We wanted to see if we could capture any evidence of caching in the transfer to and from the disk.
We then transferred the four data samples onto the disk and then off of the disk. We reset the disk caching system and once again transferred the same files off of the disk. Lastly, we simply moved the data on the hard drive from on place to another, similar to all other drives. The external SSD used for these transfers was the Glyph ATOM RAID 2TB Portable SSD which connected by USB Gen 2, and our previous report validated its transfer speeds up to 950MB/s; the fastest we have ever seen.
The results, for the most part, are like we expected with the NVMe drives being top, followed by SATA 3, caching and the HDD being slowest. If you look at the Optane results, moving data from the disk in both HOT and COLD states, however, we start to see evidence of caching, even in data transfers as large as 15GB. Having just moved the 4 file sets to the Optane drive and then off again, we found the initial disk transfer speeds moving off the disk to be very fast considering.
To better understand these times, we disabled Intel Optane Memory, formatted and placed data on the drive, restarted, switched on Intel memory again and enabled the cache where it reformatted and required another restart…and then we tested again. Although not a large difference, we can see that there is between the two. More importantly though, both hot and cold results are much faster than the HDD itself, this of course because the HDD had to endure reading and writing to and from its own medium though.
SYSTEM STARTUP, SHUT DOWN & FEEL
This chart details system startup with the hard drive, the hard drive cached with Intel Optane Memory, my own main test bench with an OWC 2TB Mercury Electra 6G SATA 3 SSD, and that same system cached.
When we first received the Intel Optane Test Bench, all was installed and ready to go. On first start, it started around 11 seconds which was amazing considering that it is caching an SSD. The comparison is great considering the start time of the HDD alone was 3x that at 32 seconds. This is not a realistic scenario, however, and we need to consider this is a fresh OS, optimized and installed with Intel Optane Memory to provide such. This is what you will get when you buy a new oem PC system with Intel Optane onboard.
And then there are those that will upgrade their gear themselves, but retain the OS on their HDD or SSD; or maybe they will build their own new system with all the right stuff. Simply because of the lightning low 4K read speeds even in the cached drive, any system will realize a significant upgrade from a HDD alone. From our own observations over the past few days and using this system consistently (yup I am working on it for this report), there seems to be a jump from start times, to application loading, web browser loading, Acrobat Pro and vetting files, as well as an immediate shut down which is great. In using our SSD run Test Bench with Optane, we never realized much change whatsoever, but the start time did improve. Our initial boot is longer because the Z270 Apex Test Bench is in use for just about everything and has so much more loaded into it at startup. I don’t game so cannot provide opinion on any gaming scenarios.
REPORT ANALYSIS
Disk caching is nothing new to consumer SSDs and NVELO looked to have a very promising niche back in 2011 when we tested their Dataplex caching software with several SSDs. NVELO was then bought out by Samsung who repositioned it within their own SSDs, leaving a void in the market (HDD caching) that may then have been filled quite well. Remember, 79% of desktop PCs today remain to be hard drive based, and there is no shortage of new laptops and PCs being fitted with hard drives. Who knows, Intel may have just created a dent in the overall SSD industry transition by realizing that PC performance and large volume storage could be very affordable. After all, a 32GB Intel Optane Memory module with software will only be $77.
Looking at Intel Optane Memory directly, we will see a ton of new consumer systems available soon enough, already powered by a large capacity HDD and Intel Optane performance. It is a given as Intel is still the leader in the CPU market and it will be a very cheap upgrade to team both in new PCs. This will work equally well for those that build systems, understanding of course that Optane will only work with a new and compatible 200 series motherboard, Intel 7th Gen CPU, a HDD, SSHD or SATA 3 SSD, and Windows 10. The unfortunate group left out of this equation are those who still have antiquated systems, the SSD still being their only hope and promise.
Would you consider testing these with the built in Windows 10 Ready Boost alongside the plain HDD?
Are you asking for a comparison of Ready Boost using a USB to that of Optane. You do realize that the USB has a much lower low 4k read speed than the Optane dont you… to the tune of 150-200 times. We are an SSD site but I have tested Ready Boost on my own in Windows. IMO it is more gimmic than worth the trouble. With Optane… in this case, the startup system files are readily available in the cache to equal the startup of an SSD. I might think that alone would dissuade any thought of comparison. Thanks for taking the time to comment.
My apologies, I thought it was possible to dedicate a SSD 32 GB or smaller as a Ready Boost drive. If that is not the case then disregard. Thanks for the review. I hope Micron releases a more hardware agnostic version in the future (for AMD boards with NVMe).
No need to apologize.. stuff like this makes us think outside the box. This actually let me in another direction to confirm something i wasn’t sure of with respect to new Optane.
Les, I would have like to have seen the Intel DC P3700 in the Real World File Transfer Tests, having said that the Kingston DCP1000 is beyond insane at File Transfers, the Intel Optane excels at 4K low queue depth Reads and the Lowest Latencies i’ve seen to date
“on a green PCB”
Interesting. If the PCB is environment friendly then it is bigger news to me then optane memory.
I wonder how reliable Optane is. Anandtech.com reported that when it failed it took some data with it. I have no issue recommending SSHD over plain HDD but what is worrying me is required software component.
So it performs admirably as a read cache, but what about as a write cache? Would the low latency be good enough to overcome the limited sequential write speeds if faced with continuous cache eviction?
My system runs quite a bit faster than the Intel upgrade:
9,068 MB / s Read, 14,048 MB / sec Write.
4K Random Read 1,556 MB / sec, 4K Random Write 1,068 MB / sec.
https://www.romexsoftware.com/en-us/primo-cache/
Romex Primo Caching Software accelerates all read / write operations with RAM and SSD caching, turbo charging any CPU to run just as fast as the RAM can go…
People keep trying to fix the Hard Drive speed bottle neck in hardware,
when it is much easier to fix in software, and RAM caching…
Primo Cache works similar to the Intel device. Say you have 32GB of RAM,
set aside 16 GB for a super sized RAM cache – all read / writes work at RAM speeds.
Primo Cache pairs ANY SSD device to the RAM cache, for a second level of persistent caching. All the stuff you use all the time, is copied to the SSD.
On boot up, the SSD reloads all your commonly used info into the RAM cache,
and the RAM cache dynamically updates itself to constantly keep you at top speed.
Go ahead and hook up your favorite 8 TB Seagate Drives to your system…
PrimoCache ensures your Big Iron hard drives run at RAMming speed.
Drop down box with sub-pages please. Clicking 1-9 is so annoying
There’s an option being overlooked by both Intel and MS:
Readyboost filters small (~4K), random, oft used files onto media with lower latency than the main drive.
The 2 drives then read/write files they are best and fastest at handling, at the same time.
ie: A sort of ‘optimised for drive characteristics’, RAID 0…
Now look at the random 4K read performance of Optane versus SATA SSDs and even NVME SSDs:
https://www.google.co.za/search?q=optane+4k+random+write&safe=off&source=lnms&tbm=isch&sa=X&ved=0ahUKEwib_bSaqNHTAhWqB8AKHdxABmcQ_AUIDCgD&biw=1536&bih=798#safe=off&tbm=isch&q=optane+4k+random+versus+sata+ssd&imgrc=_
Writes are not as impressive as random writes go into the DRAM cache on the SSDs, but:
> This info can be lost in a power outage, so safer.
> The low write speeds are only valid until the DRAM cache is full.
> There should be an increase in SSD life as info is written to flash in 2-4 MB blocks
nowadays.
I think it’s worth testing to see if Readyboost does a better job than Intel’s RST due to this filtering/Optimised RAID 0..?
Info on overriding MS’s Readyboost settings, to test this:
https://hatsoffsecurity.com/2015/05/31/force-enabling-readyboost-windows-78/
But will anyone? The ‘not invented here’ force is strong in humans! 🙂
Romex:
Everyone knows that with your software installed you basically end up using a HDD benchmark to benchmark RAM.
Untick ‘Direct IO’ in Atto and MS’s Super/prefetch makes your block cache look stupid.
Can your software do predictive caching, without wasting RAM by caching what’s already cached by prefetch?
ie: Write software that switches on Superfetch with SSDs and add your SSD-saving ‘deferred writes’ and I’ll buy it!
I will say that your caching of HDDs to SSDs is very good and universal, so if one wants to cache any HDD onto any SSD, or even RAIDed SSDs; PrimoCache is the best option.