LinkBack URL
About LinkBacksRead more.And it discussed its second gen Optane DC devices codenamed Barlow Pass.
Read more.And it discussed its second gen Optane DC devices codenamed Barlow Pass.
Stability, and Ultra fast large storage. Yes!
I take it, like Optane at the moment, this is going to have limited benefits for most of us?
I can't see how, for normal use, having Optane between RAM and SSD will make any major difference that can't be had from just adding more RAM at a substantially lower cost?
If there's a proper benefit to be had, I'd love to try it out but I don't want to waste money on stuff that is going to make zero difference for my rather mundane uses.
> in addition to this DIMM-based persistent memory hardware advancement,
> Intel is working with Microsoft to add support for the tech in client versions of Windows
I'd like to see what good system programmers would do with 4 DIMM slots in quad-channel mode, and 2 or 4 more DIMM slots populated with Optane DIMMs e.g. a bootable ramdisk appears feasible, with all the required changes to a BIOS/UEFI subsystem. With Optane DIMMs installed, a "Format RAM" option could be added to a BIOS menu, and chosen during a fresh OS install. Once a fresh OS install is finished, the BIOS would add that "ramdisk" to the list of bootable drives.
More layers drives down cost, so I'm sure consumer drives will see some benefit.Originally Posted by philehidiot
As for Optane, it is cheaper than RAM, but more expensive than flash.
OFC if Intel manage 5 bits per cell that is going to drive cost down but make performance way worse than current SSDs and probably decrease lifespan as well. That makes an Optane middle layer an easier sell, if it wasn't easier for SSD makers to just include a single level flash block in the SSD to give the boost there (which ISTR some already do).
But yes, for most of us the performance change won't be noticeable in the slightest.
You've said that before, but PCs just don't work like that. The BIOS doesn't have an option to partition a normal disk, let alone format it which would involve knowing *how* to format the drive which changes across Windows versions before you start thinking about Linux and BSD users etc. That is just not the job of the BIOS.
Besides, Optane requires wear levelling, so if you drive it like a ramdisk your expensive Optane sticks would be dead in hours. So a good system programmer would drive it like an Optane disk knowing the properties of the silicon to make the best of the good properties and tip-toe around the bad ones. I believe that is what Intel have already done, so you can already boot from Optane (perhaps not in Windows yet, I wouldn't know) just not involving the BIOS because that would be wrong.
Last edited by DanceswithUnix; 30-09-2019 at 07:43 AM.
> The BIOS doesn't have an option to partition a normal disk, let alone format it
Correct: no current BIOS has that functionality (that we know of).
That's why we proposed adding that feature to a BIOS, by way of an experiment.
Also, UEFI subsystems are intended to add functionality which a BIOS lacks.
FYI, here's the provisional patent application which we wrote several years ago:
http://supremelaw.org/patents/bios.e...lication.1.htm
> if you drive it like a ramdisk your expensive Optane sticks would be dead in hours.
No.
The suggestion is to "drive it like a Windows C: partition", otherwise transparent to the system.
The 4 x DDR4 DIMM slots in quad-channel model would also be available for a heavy-use ramdisk.
Because Optane is a third layer, tucked between DRAM and NAND, the system designer
has more choices for memory management, in general.
A good calculation is to compare the raw bandwidth of DDR4-3200 with x16 PCIe 4.0 in "4x4" mode.
You might be surprised with the results of that calculation.
The performance decrease and "lifespan decrease" of QLC and PLC don't concern me at all form a home user perspective. What concerns me is the "power off lifespan". With every bit that's added, the differences in voltages is halved. i.e. SLC is X volts difference between 1 and 0, MLC is 0.5X Volts between 00 and 01 etc, TLC is 0.25, QLC is 0.125 etc.
Depending on the temperature of the drive when the data was written, sometimes the data can begin degrading after just a few days!
Source: https://www.extremetech.com/computin...haracteristics.
"In a perfect world... spammers would get caught, go to jail, and share a cell with many men who have enlarged their penises, taken Viagra and are looking for a new relationship."
Intel Optane DC Persistent Memory Coming To Desktop Workstations
https://www.legitreviews.com/intel-optane-dc-persistent-memory-coming-to-desktop-workstations_214370
I'm not saying you're wrong, just that I don't see it happening.
What you are suggesting sounds trivially easy. It could of course just be that someone has taken out a patent on the idea so now it isn't worth even a few days work to implement if it leads to months of legal wrangling for the motherboard makers because there is no way anyone is going to add cost to the motherboard to implement this.
So in the mean time, we have zero BIOS support, but the Linux community can boot just fine like that and they are the ones who really care about performance anyway so I think we're good
... and yeah DDR4 speeds are good, but there is already very little performance improvement between PCIe and SATA ssd systems despite a massive improvement in raw speed.
In short, there are better things to do with a small expensive pool of Optane than C:
> there is no way anyone is going to add cost to the motherboard to implement this
The cost of a motherboard with this feature does not necessarily need to increase
especially if it's "trivially easy".
That sounds like a "straw man" argument. Just rewind the clock enough years
and we hear:
"There is no way anyone is going to add cost to a motherboard to implement overclocking features."
> Linux community ... are the ones who really care about performance anyway
Tell that to Ryan Shrout, now at Intel.
https://www.pcgamer.com/pc-perspective-founder-ryan-shrout-joining-intel-as-chief-performance-strategist/
Patents are bought and sold routinely, so the existence of an applicable patent does not necessarily imply "months of legal wrangling". Another "straw man" argument.
Our provisional patent application expired, because we did not follow with a formal Utility Patent Application.
Again, see the link above to:
Intel Optane DC Persistent Memory Coming To Desktop Workstations
The arguments you are making discourage scientific research and reinforce bias against Windows users.
[QUOTE=DanceswithUnix;11281]No it won't. The DIMMs are rated for unlimited use at maximum speeds for 5 years. An Intel guy said it'll also last much longer than that.
We're not talking about having lifespans like a NAND cell with far less durability.
On a conventional SSD I expect that is true. Optane does require wear leveling, and last I heard the durability was only about a magnitude better than flash though with one big win feature up it's sleeve: it can be byte erased rather than flash working in blocks or pages. That should help quite a lot, but it isn't a magic bullet by any means. So wear leveling is there, and writes have to go through a driver that knows about optane, hence my comment that you can't just pretend it is a ramdisk.
Wear leveling of Optane has so far been a wash, the stuff is a magnitude more resilient (yay!) but costs a magnitude more (boo) which means for most uses you are better off just buying a NAND ssd ten times bigger than the Optane one as that gives the same real world performance but the flexibility of massively more storage for the same money.
The early M.2 Optanes used only x2 PCIe 3.0 lanes.
Yesterday, I found only one M.2 Optane that uses x4 PCIe 3.0 lanes.
Are there any others that I may have missed?
e.g. the 800P uses an x2 interface:
https://www.intel.com/content/www/us/en/products/memory-storage/solid-state-drives/consumer-ssds/optane-ssd-8-series/optane-ssd-800p-series.html
https://www.newegg.com/p/pl?d=Intel+Optane+800P
Yes, it requires wear levelling, but its all on the controller. On the DIMMs its all contained in the DIMMs.
And I'll repeat myself and say the DIMMs are rated at maximum write speeds 24/7/365 for 5 years, whatever that might be. At 2.3GB/s that's 360 Petabytes written for the 256GB module.
It doesn't matter whether you use it as RAMdisk or RAM, because you are basically limited by the write speeds regarding endurance. I think you just ignored my point about "maximum for 5 years".
Several months ago, I finally isolated a serious problem we were having with a backup storage computer.
Certain intermittent problems were driving me crazy e.g. the task writing a drive image file would
write a corrupted file. I would re-run that task, and the output file would be corrupt again.
Then, I would re-run that task a third time, and the output file would be OK.
There were certain video drivers that would also crash during install from a factory cd-rom.
Because that PC was built with quality DRAM with a lifetime factory warranty,
I simply did not suspect that a DIMM was failing after several years of reliable service.
We power all of our PCs with quality UPS / battery-backup units, so
all PCs are getting clean input power at all times.
One day, I tried updating the motherboard's BIOS, and one of the default settings
switched STARTUP mode to a full memory check: as I watched the BIOS count K bytes,
it quit after reaching a count that was about two-thirds of the total amount of RAM.
So, I pulled both dual-channel DIMMs and installed some spares I had in inventory.
Sure enough, at STARTUP the BIOS correctly enumerated the total amount of RAM.
The vendor of the failed DIMM honored their factory warranty, but they no longer
manufactured that particular model of DDR2, so they sent me a check instead.
Moral of the story: a lifetime factory warranty is no guarantee that quality DRAM will
last indefinitely during normal usage. It was certainly my fault for not suspecting
a failed DIMM much sooner, but it was my expectation that needed adjustment.
OK, if that's true then Intel have made great progress on their durability. That is basically the bit I was missing. The last time I read a spec sheet for an Optane SSD (which wasn't that long ago) it was quite easy to hit endurance limits despite a slower interface and larger capacities.
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
Reply With Quote