Not for most, no.
Here, it is not cheering the 'rise of the midrange'.
You're missing the point, which is that it's sequential write. Even QD=1 should show similar results.
You raised the issue. I was merely supplying the evidence to support (most of) your point. Why are you mad at me? ...maybe you're a Samsung fanboy?
Not true. Try unzipping a big archive, for instance. ZStd compression is very fast, and you can get good enough compression ratios on some data for it to saturate the write bandwidth of these SSDs.
Because they were discontinued, I presume. Even if the were still included, they're at a disadvantage with their PCIe3 interfaces.
Huh? Why do you think NAND will ever reach the latency of Optane? As far as I understand, NAND isn't on track to achieve this, and nobody is saying otherwise.
Huh? It's an order of magnitude slower than DRAM, has vastly lower endurance, and isn't a whole lot cheaper. It fails the test as a DRAM substitute, which was a big part of Intel's strategy.
You presume it could be. We don't know that. As far as we know, an unlimited budget and development time wouldn't be adequate to make it commercially viable. It's a fundamentally different technology that you can't simply presume will follow the same cost & density curve as DRAM or NAND flash.
If you look at the density of the 2nd gen Optane dies, all the volume in the world wouldn't make it competitive.
It's not clear why you'd want a DRAM cache, since it can already sustain more than 75% of PCIe 4.0 bandwidth and the amount of DRAM you could fit in there would only last a few seconds. In a typical use case, the OS buffers writes using system memory, which has an order of magnitude more bandwidth and far greater capacity than what DRAM the drive could hold. Plus, using system memory to buffer writes is more flexible.If you gave a 5800x a similar sized cache of dram that the others have of dram+slc, what would the following chart look like?
That chart is Optane at it's weakest.
I'm not. I wish Optane would get cheaper and faster, as much as anyone here. However, I think Intel usually makes sound business decisions, and the fact that they saw no way forward and couldn't sell it or spin it off means the technology simply didn't have enough gas left in the tank. I also trust Gelsinger to make data-driven decisions and listen to what his technical folks tell him.
* sigh *
Then you probably already know that even ramdisks can only occasionally beat the 5800X in this practical bench:
Oh, I don't have Windows on that machine.
Optane is still the best thing out there, for a few niche use cases. For instance, to hold the journal of a high-end all-flash array.
Hmmm... then you might find this thread interesting:
Depends on how you're using it. My point was (and is) that it burns a heck of a lot of power, which was contrary to the wish expressed in the message I was replying to.
Source? I checked the listing on Ark and it says nothing about the PBW figure, for the PMem 200-series.
ark.intel.com
Intel debuted their first 3D NAND products before any specific Optane products were even announced. Again if you don't think they knew where that market was going I got nothing.
I don't think anyone following the tech would forget the awful 1000/1000 marketing slide when the first officially announced the 3D XPoint technology. The late shipments likely had a much bigger impact than anything else (I think the PMem was 2 years late) due to how quickly the storage market was moving the whole time. I believe Intel had already stopped making their own controllers for their client SSDs a few years before Optane products launched because it wasn't worth the cost as the margins were so low.
What evidence do you have of this again? Rhetorical question because there isn't any evidence. The only thing that we actually know is that the cost for Intel to spin up their own fabs for 3D XPoint didn't make financial sense for the company as they're trying to make cuts to increase profits.
It's not about doing stuff on a "whim" as you keep putting it. It's about targeting cuts where it makes sense to the company monetarily. He's not interested, and thinks they should be out of the business which means it's a division that will be high up on that list. In the vast majority of businesses the quality of a product is completely irrelevant when compared to the sum of money it is going to make you right now because that's how big businesses are run today.
Nobody would buy something that currently has a limited market and zero production facilities. Continued development of the technology would have to be done along with having a fabrication line before selling or spinning off would be possible.
You don't have a clue how business works if you actually believe that when it comes to publicly traded companies. Here's an example that leaps to mind: Verizon was forced to stop rolling out FIOS in the US due to the high amount of capital expenditure by angry investors. Their service has been profitable for years now, was always going to be profitable, but it wasn't going to happen soon enough so the rollout was killed. That may sound irrational to you or me, but that's the sad truth of how the game is played these days.
The cache is used only while writing data, as I use the drives a lot more in reading than writing it is OK. The difference in reading is mostly because of the interface so a dramless M2 drive is fine for my usage.
Intel & Micron were probably developing 3DXpoint for at least 5 years before they ever publicly mentioned it. At that time, the exact trajectory of NAND couldn't be known with much certainty.
If this is true, then it's not a "Gelsinger" thing, as you keep claiming. If it's a sound business decision, then anyone reasonable would come to the same conclusion.
Intel has a long history of making big investments that don't pay off for years (a hallmark of semiconductor manufacturing). If they thought Optane would be viable in several years hence, I'm sure they'd have either found the finances to keep it going, or found an interested buyer who would.
But there were production facilities that could've been bought. And the big NAND makers all have their own fabs that they could retool to make 3D XPoint, if it were viable.
Verizon is also a different sort of business. Their revenue is mostly service-oriented, which is probably why their investors don't like seeing them make big capital investments. Perhaps they're also worried that regulation could make the broadband market less profitable - who knows?
That's mostly true. From what I remember of the HDD era, there are a few cases where the in-drive cache is also useful for speeding up reads.
I wrote that for M2 PCI Express drives, reading is in general fast enough, their writing speed is slower, and if they have DRAM, some do not, DRAM cache is used while writing.
SSDs don't only use DRAM to buffer writes or fill reads from cache. They also use it to cache the block map, which they do to implement wear-leveling and to exclude failing blocks from circulation.
I wasn't talking about hybrid HDDs, but rather pure HDDs. Back before SSDs became dominant, a fairly common question was how much impact the amount of a drive's cache had on performance.
No source was provided, and I didn't find any PBW or DWPD rating in Intel's published specs on their PMem 200 modules. Are you going to simply take an anon poster's word for it?
No source was provided, and I didn't find any PBW or DWPD rating in Intel's published specs on their PMem 200 modules. Are you going to simply take an anon poster's word for it?
I find the claim that they warranty continuous writes @ max-rate rather incredible. Not least, because it's not how anyone else warranties their storage devices (AFAIK).
Here's a claim that the 800 GB P5800X can endure 146 PBW, in its life:
I feel the need – the need for speed (and endurance): Intel Optane edition | Dell Technologies Info Hub
It has only been three short months since we launched VxRail on 15th Generation PowerEdge, but we're already expanding the selection of configuration offerings. So far we've added 18 additional processors to power your workloads, including some high frequency and low core count options. This is...infohub.delltechnologies.com
...but that's a SSD that can potentially do wear-leveling. Do the PMem DIMMs do wear-leveling?
Also, I wonder about over-provisioning - do we know the raw capacity of the individual Optane chips? Why are the Optane SSDs offered in different size increments than the PMem DIMMs?
And with the max PMem 200-series listed on Intel's website being only 512 GB, it seems like the PBW probably isn't over 3x as much as the larger P5800X.
www.intel.com
Thanks! I did look, but not super hard. I don't know why they didn't just put this stuff in their ark.intel.com database, as some of their other Optane products have it.
Intel is a big company with big revenues and profits. Investors don't like it when those numbers shrink. If he can produce an absolute increase in profits, they would accept a reduction in revenues. Otherwise, the only acceptable way for Intel to shrink is by spinoffs and selloffs.
Corporations do a relatively small number of things, while Government does a mind-bogglingly large amount of them. It's not an overstatement to say that government is the underpinning of modern society. Everything from industrial and technological standards (via NIST, ANSI, ISO) to administration of air travel (FAA), food safety, transportation, pipelines, trade & commerce, communications, financial oversight... a vast array of things most people don't see or interact with on a daily basis, yet are critical to many things we take for granted.
It's about cost. That's the elephant in the room. And even with endurance as high as you quoted, it's still about 1/10th as fast as DRAM, yet it's approaching DRAM pricing. So, that makes in an unappealing alternative to DRAM. And as persistent storage, NAND is vastly cheaper and good enough for most. There's simply not a good business case for it.
Putting your PC to sleep is effectively "instant on". And you still need to have some way of rebooting after installing updates or when your PC gets flaky.
Don't compare Optane with consumer SSDs. Compare it with datacenter SSDs, and then it doesn't look so special.
Facebook
Twitter
Instagram