$77 gets you 32 GB, the largest capacity available. For essentially that price, you could have a 240 GB SSD that could hold the OS, all your small programs (web browsers, etc.), and several games. In 32 GB, you'll get bits and pieces of some scattered things, but nothing like the benefits of a real SSD.
Nor does it make sense to use Optane as a cache drive. Even a cache SSD to pull accesses away from a hard drive never worked very well, in spite of SSDs being multiple orders of magnitude faster than hard drives in some important things. If Optane is barely faster than a NAND SSD, the cache doesn't have much of a point.
So what about those performance numbers? Intel is claiming 1200 MB/s read and 280 MB/s write. That cheap 240 GB SSD can probably beat the latter at sequential writes. It won't beat the read numbers, but even if that's your priority, a 250 GB Samsung 960 Evo for $130 offers 3200 MB/s read and 1500 MB/s write. Even if that's out of your budget, a 128 GB OCZ RD400 will get you 2200 MB/s read and 600 MB/s write for $95.
Intel does highlight that Optane can deliver high performance at lower queue depths than NAND. And it is true that consumer applications usually have pretty low queue depths on NAND-based SSDs. But the reason for that is that the SSD can keep up with the workload, which doesn't allow a deeper queue to form. If you had a theoretical infinitely fast SSD that returned whatever you requested instantly, it would be impossible to have a queue depth higher than 1.
Optane's reduced latency would shine in situations where one read has to be read and processed to know what the next one will be. But that's not storage-heavy consumer workloads, where needing to load a long list of art assets allows as high of a queue depth as you want.
Optane's vaunted write endurance seems to have vanished. Three drive writes per day isn't bad, but in terms of total write bandwidth before the drive fails, that 240 GB NAND SSD is going to beat a 32 GB Optane drive, and probably by a lot. That makes Optane unsuitable as a sort of extended DRAM.
Not that extended DRAM had much of a point, anyway, now that you can get 16 GB of DDR4 for $100. Optane offers performance a little better than NAND, but when you need something better than NAND, the solution is usually going to be DRAM. DRAM is multiple orders of magnitude faster than Optane in latency timings and sequential write throughput. And unlike Optane, DRAM's write endurance is effectively infinite.
And then comes the power consumption of around 1 W at idle. In a desktop, that merits a shrug. In a laptop, adding 1 W of idle power consumption is a problem of the sort that people dim their screens to try to counter. Many NAND-based SSDs are far lower than that, and can replace a hard drive anyway, thus offering a net reduction in power consumption. A 32 GB Optane is not going to be your only drive in a laptop, as Windows by itself likely wants more space than that.
Comments
This sounds like it only is useful for people with special need, like read a lot of data really fast for some reason but don't care about writing speed...
I guess you could raid it to get an insane reading speed but I rather have a slower read speed with a lot faster write speed since I myself use SSD for my OS and games and you can raid those as well, in fact you could throw in even more drives for a lower price and still have a really fast reading speed. Not that I have bothered to raid SSD since my 4 old Intel X-80s around 7 years ago or so when I replaced my SCSI setup with SSDs.
I think they are on the track for a breakthrough but some moronic boss asked them to release what they have before the specs is good enough. But until the get the writing speed up to close the reading speed it is a no go for me.
I agree, it doesn't sound very appealing.
Your link says that they plan to offer more conventional SATA or M.2 SSDs for consumers, which is not at all the same as having a product you can buy today. But compared to, say, the Samsung 960 Pro, it will probably be far more dollars for far less capacity, and it's likely that performance will be mixed rather than Optane being faster across the board. Regardless, once you get into that performance tier, adding more performance is irrelevant to consumer use.
Optane could still have a point if it were massively faster than NAND, as there are uses for very fast, non-volatile memory. The problem is that other than latency, Optane isn't much faster than higher end NAND-based SSDs. The original promise of 1000x faster than NAND ended up turning into maybe 2x as fast as NAND in some things and slower outright than NAND in others.
So Optane was supposed to have a point before physics decided to contradict Intel marketing. Physics tends to win those battles, which leaves Optane without much of a point.
In a laptop, Optane would have made it possible to have the laptop running, send all of the memory to an Optane drive in a fraction of a second, and then turn the laptop off. When it comes time to turn the laptop back on, you could recover to exactly where you had been within a fraction of a second and not need to reboot.
This isn't just sleep mode. This is a laptop that is really off. You can pull the battery, unplug it from the wall, move it across the house or the country, and everything. And then when you turn it back on, everything is right where it was within a fraction of a second.
I don't know if the software to do this exists, but in a desktop, it would be theoretically possible to copy the memory for a game to disk, then close the game. Rather than taking some tens of seconds to reload the game the next time, you could load the memory from disk and be exactly where you were last time in a small fraction of a second. This wouldn't be just closing and relaunching a game, but you could pick up a game that you hadn't played in a month and had many system reboots in between, and continue where you had been in a fraction of a second. Online games where you have to re-establish a network connection might take longer, but still vastly faster than relaunching a game.
Extremely fast and durable non-volatile memory would make it possible to do things like that. DRAM is fast enough, but can't because everything vanishes when you lose power. Neither NAND nor Optane offers either the performance or the write endurance to reasonably offer that sort of behavior.
Unifying non-volitle storage and system RAM would be significant. Optane isn't that, at least not yet in any sort of general case.
Durability:
Optane is rated at three "drive writes per day". That is, Intel promises that if you do enough writes to fill up the drive three times per day, it will last for at least the length of the warranty. For comparison, on some computers, DRAM can do three "drive writes" per second. So durability is too low by about a factor of 100000.
Latency:
Getting a few bytes from DRAM typically takes on the order of 50 ns. If you ignore various parts of the latency that Intel insists you should ignore, getting a few bytes from Optane takes on the order of 10000 ns. So latency is too high by about a factor of 200.
Throughput:
The enterprise-focused Optane SSD has sequential throughput of around 2 GB/s. For comparison, two channels of 2133 MHz DDR4 is good for about 34 GB/s of bandwidth. Some desktops have higher clocked or more channels than that. So throughput is too low by about a factor of 20 on the enterprise drive, and probably a lot more than that on the small cache drive.
Power:
The consumer Optane cache drive is rated at about 3.5 W under load, 1 W at idle. The enterprise SSD is 18 W load, 5 W idle. I'm not sure what DRAM uses, but from various sources online, it sounds like it's about on par with the Optane cache drive.
I mention the power numbers not because Optane is too high to use as DRAM, but because in many cases, there are trade-offs between performance and power consumption. You can ramp up performance at the expense of burning more power. Think of CPUs and GPUs, for example, where you can get a 15 W APU with not much performance, or you can get a 140 W CPU and 250 W GPU with a lot of performance. So even if Optane could increase performance dramatically, the power consumption would probably still make it not viable for use as DRAM.
Optane doesn't have to completely close the gap with DRAM on durability, latency, or throughput. Being half as fast as DRAM would be good enough to be interesting as a DRAM replacement for some purposes. But it sure needs to get a lot closer than it is in order to be interesting as a DRAM substitute in very many situations that NAND can't already handle.
Ok, I'm completely failing to see what the purpose of this product is? It's slower than an SSD, higher cost per GB? Lower max capacities? Useless as a cashe drive, useless as a storage/program drive? Slow than regular system RAM?
I mean honestly, what's the point?
"The surest way to corrupt a youth is to instruct him to hold in higher esteem those who think alike than those who think differently."
- Friedrich Nietzsche
LOL: http://www.tomshardware.com/news/intel-optane-memory-cache,33988.html
"Intel's Smart Response Technology (SRT) was so successful that we haven't talked about it since it was first introduced in 2011. We say that with a heavy dose of sarcasm, because Intel plans to bring the technology back with a faster media."
"The surest way to corrupt a youth is to instruct him to hold in higher esteem those who think alike than those who think differently."
- Friedrich Nietzsche
Some people will see the Intel label and a high price tag and buy the flashy box without question.