Benchmark:
https://openbenchmarking.org/result/1802112-TY-900PBYTEE44
Note: The 960 Pro should be around 16-20 not as much as you see here. However, the 960 Pro was formatted XFS not EXT4 and its firmware is "file system aware" so I am guessing that is the performance delta here vs other reported benchmarks. It could also be the 960 pro has had some use now and is not a blank/empty drive which also affects benchmarks somewhat.
Most other benchmarks do not show this much performance delta between optane and the new nvme 960 pro 1tb. (this is approx sata nand speeds in this sqlite test from the 960 pro).
This is a companion discussion topic for the original entry at https://level1techs.com/video/intel-optane-900p-our-initial-thoughts-and-testing
Damn are those Tidepods microwaved now ?
You guys have some strange fetish for real, a few months back there were keycaps with the letters “Tea” mixed with M&M. For about 2 weeks there are Tidepods and now they look like they went trough some thermal changes in structure.
Also whats up with those screws about a year back you disproved of anything but they are in every video and are even increasing in numbers and diverse in sizes?
You got me wondering about the technology Intel uses. Apart from beeing delayed a bunch of times and then not really beeing available in 2017, it is expensive.
Now to the “bunch of technology”:
28 memory chips for 480GB total capacity. That would be just over of 17GB per chip.
For comparison, Samsungs 960pro 2TB drive has 4 memory chips (250GB each) and one controller on it.
So either Intels solution lacks severly in the density department or is overprovisioned to make up for something.
I don’t really have any endurance numbers but the iodrive and iodrive2 we have been using for years that were built like these intel 750s( the iodrives were the exemplars of the 750 design) have > 90% endurance remaining after 5+ years of continuous 24/7 use. Even in grueling conditions. I’ve got one DB server that totally rewrites the SSD once or twice a week and it’s still at like 91%
So it’s all a bit anecdotal… But you’d need optane endurance numbers if you were using it instead of ram
Makes sense, but that makes me wonder something else, again due to the insane RAM prices.
Would an average linux user benefit from having the swap partition on a cheap Intel Optane Memory 16GB M.2 NVMe SSD in case the system starts using swap? Or in other words, how faster would the swap be compared to a normal situation where you have the swap partition on the same boot ssd.
I know RAM is always better, but since the small optanes cost less than half of a RAM upgrade of the same size (at least on my country) it would be interesting to see if it’s worth it, especially if one has an older machine: My desktop PC doesn’t support more than 16gb of RAM, for example.
As @wendell stated, if you need swap for large data sets, then any NVMe SSD would be fine. If you do not have NVMe capability, then SATA III SSDs are the way to go, obviously. In either case you do not need the fastest as it will not make that much of a difference. Most drives saturate the bus that they use within 90%.
When you talk about RAM versus any other bus in consumer computers, you are looking at orders of magnitude of difference;with RAM being faster. In the grand scheme of things, you should be looking at upgrading the platform if speeds are important and you are RAM capped.
If you are not dealing with large data sets, then your swap is recommended to not be more than 8Gib. I have been running a swap size of only 1024Mib for over a decade. Unless you use suspend, there is no need as RAM is hella fast and Density is superior in comparison to back in the day where SWAP made sense.
@wendell@nx2l@Mastic_Warrior thank you for your insights. I think I overestimated the speeds of the small optanes, and I needed some perspective on the usefulness and proper usage of swap. I guess my mind was stuck in habits of almost a decade.
