Discussion Optane Client product current and future

[cbn]

16GB Optane is now $25 at Newegg:

https://www.newegg.com/Product/Product.aspx?item=N82E16820167426

(Would like to see either Gemini Lake Mini-ITX boards or NUCs with PCIe 2.0 x 2 M.2 NVMe slots to accommodate this)

 

[IntelUser2000]

I was a fan of the $50 CDN 32GB Optane. I see it as a bigger deal because not only GB/$ is greater, the whole thing about block and caching thing is an extra feature you can't get on the 16GB. Intel should use the 16GB only as a bundle and make the extra cost only $25, not equal to MSRP of $44. Most stores sell 16GB less than that.

Still need to see what the M10 drives feature before I decide. I believe M10 will have read/write speeds similar to 800P.

 

[cbn]

I was a fan of the $50 CDN 32GB Optane. I see it as a bigger deal because not only GB/$ is greater, the whole thing about block and caching thing is an extra feature you can't get on the 16GB.

Agreed....for Windows the 32GB is better value per GB for that reason.

For Linux I eventually want to see how 16GB Optane compares to the 60GB Patriot Flare with Gemini Lake. (Not sure when this will happen though)

 

[Dayman1225]

Technology and Cost Trends at Advanced Nodes

Thought the probably educated guess of wafer cost per bit of XPoint in the future would interest a few here from Scotten Jones, ICKnowledge

 

[Dayman1225]

Micron has put 2nd Gen 3DXP into Production!

Source: Investor Day PDF

 

[cbn]

Technology and Cost Trends at Advanced Nodes

Thought the probably educated guess of wafer cost per bit of XPoint in the future would interest a few here from Scotten Jones, ICKnowledge

How multi-bit (not sure when this will happen) is positioned will be interesting.

(My guess is that Intel-Micron will make the future single bit even more high performance (ie, lower latency) than today's single bit and the multi-bit will slot below the "new single bit".)

 

[IntelUser2000]

Brief thoughts on the M10: I've yet to fully conclude because there's only one review, and I think other reviews will give a complete picture(I'd like to see battery life tests from TH, or anyone else for that matter) but, it does not look too good.

The biggest issue is that its too expensive. And so far it looks like a differently branded 800P, and since they both work as caching drives, 800P is better. M10 may have 8th Gen core platform specific changes, but because they are meant for mobile and let's face it, the only battery life review we get is from TH with their 45W CPU 3 year old laptop, we'll never know anyway.

 

[cbn]

For the battery life on notebooks I hope two things happen:

1.) Optane gets DDR interface on mobile chips soon* (2019? 2020?). I reckon this is where Optane should really pull away from NAND (even if NAND also gets DDR interface) and I also reckon (because of the lower latency) this would also allow Intel to get more aggressive on power than NAND on DDR interface would allow. (Though it should be mentioned the top NAND drives do well on NVMe)

2.) 4200 (or lower) rpm 2.5" HDDs appear.

*Still crossing my fingers for a "Celeron 300A type" 20th Anniversary Celeron in 2018 with 16GB DDR interface Optane.

 

[IntelUser2000]

Regarding power consumption on notebooks, and Optane-based DIMMs, may be too early for laptops.

Here's when it'll happen.

"Windows 10 5th Anniversary Edition now includes support for Optane DIMMs. Select systems with Optane DIMMs will have an extra power mode in addition to the traditional Sleep/Shutdown/Restart modes. The mode is called Optane Sleep and will bring you the best of Sleep and Shutdown modes. System wakeup and sleep speeds of sleep mode with zero power mode of shutdown."

I reckon in a few years. This year, and maybe even the next is too early. Even if only the OS can be put into the Optane DIMMs, the benefits of a Suspend-to-RAM state with RAM that uses zero power to retain data is an amazing advantage. The OS will have to keep the traditional S states for the sake of compatibility, you can't abandon them. Just introduce an extra power state. StR replacement is best because StR skips BIOS/EFI load time. The system, when using Restart will also have to flush the contents of Optane DIMMs for the sake of security, and to get rid of errors that are only solved by a restart.

Of course, significant hardware/software changes are required to support DIMMs that don't need power to retain data, and for the rest of the system to do the same.

The laptop and phone manufacturers will advertise "infinite standby" on their ads. The world of computing will be great.

 

[cbn]

M10 may have 8th Gen core platform specific changes, but because they are meant for mobile and let's face it, the only battery life review we get is from TH with their 45W CPU 3 year old laptop, we'll never know anyway.

From Tom's hardware:

https://www.tomshardware.com/reviews/notebook-battery-life-storage,5152.html

This type of testing is rare to see in hardware reviews. Running on battery power is expensive, because the test is very hard on batteries. We take the battery from full charge to empty. A battery will lose its ability to charge completely after 10-15 cycles. We have to replace the $100 battery after a 1% drop. During peak storage seasons, that's a new battery every month. Lenovo also used three ODMs to build Y700 batteries, so we have to purchase a specific part number. Not every seller is as detailed as we are with those part numbers.

Any other ideas on how to do the testing?

Anyone?

 

[cbn]

From the Anandtech M10 review:

https://www.anandtech.com/show/1274...ry-m10-64gb-review-optane-caching-refreshed/9

On the other hand, it is clear that no amount of fast storage can make up for a system crippled by too little RAM, which is a disappointment in a time when SSDs are getting cheaper but RAM prices are still climbing. Optane SSDs may be the fastest swap devices money can buy, but they're no substitute for having adequate RAM. The 4GB low-end configuration we tested is simply not enough anymore, and for future storage caching tests we will consider 8GB as the absolute minimum requirement before any storage performance upgrades should be considered.

For a system with 4GB RAM Windows allocates (by default) a page file of only 1.3GB.

That is not a lot space for page file.

Maybe the Optane software should have an option (at set-up time) to increase page file? (This when both the 16GB* and 32GB (and greater) Optanes are used)

*16GB mentioned because it is normally only block level cache.

 

[Billy Tallis]

Any other ideas on how to do the testing?

Anyone?

The obvious solution is to simply measure wall power, using a low-power desktop with mostly laptop components, such as a NUC or some all-in-one PCs. For regular desktops with 65+W processors it is hard to get all the platform-level power savings features enabled and working properly, so storage usually ends up being a very small portion of overall power consumption. Does anyone know of a NUC-like machine that uses LPDDR3 and has both SATA and M.2 PCIe ports?

 

[ksec]

Micron has put 2nd Gen 3DXP into Production!

Source: Investor Day PDF

No mention of Xpoint DIMM. And no mention of what were the improvement for Xpoint 2.

What is certain though, while you might not have hear as much about Machine Learning and Big Data as 6 months ago, things are still moving at rapid pace. And the need for much higher capacity as well as faster DRAM is ever increasing. Terabyte dataset are not uncommon, I wish in few years time we could just throw this into a giant machine with 10s of TB of Memory to do data processing.

 

[cbn]

Here is the battery life result on the 118GB 800p (as a standalone drive) from Tom's hardware:

https://www.tomshardware.com/reviews/intel-optane-ssd-800p,5497-2.html

Compared to the following older model 2.5" HDDs it actually does worse (with the exception of the Seagate Laptop HDD (2TB), a 3 platter 9.5mm platter design I believe) :

https://www.tomshardware.com/reviews/notebook-battery-life-storage,5152-4.html

But I wonder a few things:

1.) How would the smaller capacity M10 drives do on battery life (if hypothetically used as standalone) compared to the 118GB 800p?

2.) How would newer model 2.5" hard drives do in comparison to the older ones used in the above battery life test? (As reference points the Seagate Momentus Thin 500GB is from 2012 and the WD Slim 1TB is from 2013)

3.) How would the power consumption of a Optane M10 (or 58GB 800p) as cache compare to its use as standalone? My guess is that it would probably have lower consumption (under the same workload) since it is only handling the small files and sharing the load with an HDD (handling the large files)....and not handling both small files + large files.

(I bring this up because I wonder if M10 + 2.5" HDD might be better in battery life than the standalone 800p results suggest.)

 

[IntelUser2000]

Interesting part about the M10 in Intel support page:

https://www.intel.com/content/www/u...0/memory-and-storage/intel-optane-memory.html

8th Generation Intel® Core™ Mobile Processors now support system acceleration with Intel Optane memory M10 modules. The Intel Optane memory M10 modules are currently only available in mobile platforms. Consult your OEM Vendor for more details and product availability

 

[Edrick]

Anyone hear anything about the 900P/905P replacements?

 

[cbn]

There's also the question of how low power the M10s are. The 800P review on TH shows its acceptable, but nowhere near Chromebook level. The top level NVMe drives are behind SATA drives in power consumption, and 800P is middling in the field of NVMe drives.

https://www.tomshardware.com/reviews/intel-optane-ssd-800p,5497-2.html

Looking back on post #61 and #70, I wonder if the 16GB M10 (if used in a Chromebook or Ubuntu "Browser Build" Laptop) could be tuned for even lower power consumption while retaining enough performance to be used for a paging out during browsing? Is there such a thing as combining a PS2 read with a PS0 write?

Thinking about the BAPCO benchmark used by TH for battery life testing.... if hard drives (with their higher idle) are beating the 118GB 800p doesn't that imply the workload is not very bursty (like I am assuming Browsing would be?)

And if Browsing is a bursty workload I wonder how would M10 (or 58GB 800p) compare to eMMC? (I know eMMC has a low absolute power number....but how different would the idle between M10 Optane and eMMC be? How much difference in race to idle?)

 

[cbn]

Interesting part about the M10 in Intel support page:

https://www.intel.com/content/www/u...0/memory-and-storage/intel-optane-memory.html

Here is a laptop with eighth generation Intel "H" processor, GTX 1050 dGPU and 16GB Optane memory:

http://www.dell.com/en-us/shop/cty/...&acd=1230923830920600&VEN3=112304412697571341

EDIT: Here is a laptop with eighth generation Intel "U" (15W) processor (no dGPU) and 16GB Optane:

http://www.dell.com/en-us/shop/cty/...&acd=1230980731501410&VEN3=110304414036615029

(Both laptops above are 15.6")

 

[IntelUser2000]

doesn't that imply the workload is not very bursty (like I am assuming Browsing would be?)

It's literally the only Optane battery life test, it could simply mean its power management features are not working. It is after all, a 45W CPU platform.

 

[cbn]

Deleted.

 

[cbn]

Below is a patent from Micron:

http://appft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PG01&p=1&u=/netahtml/PTO/srchnum.html&r=1&f=G&l=50&s1="20160276022".PGNR.&OS=DN/20160276022&RS=DN/20160276022

Some embodiments include architectures in which two or more memory array decks are vertically stacked. One or more of the stacked decks is configured to have different operational characteristics relative to others of the stacked decks. For instance, one or more of the decks may be configured to have rapid access times suitable for utilization in XIP (execute in place) applications and/or dynamic random access memory (DRAM) emulation applications, and one or more others of the decks may be configured to have stabile, possibly slower access, storage suitable for utilization in long-term storage applications. Further, one or more of the decks may be configured to have more endurance than others of the decks. For instance, one or more of the decks may be suitable for a lifetime of approximately 100,000 cycles, whereas one or more others of the decks may be suitable for about 1,000,000 cycles (in other words, at least one of the decks may have a durability of at least about 10-fold more cycling times than another of the decks). The difference between the endurance of the decks may result from structural differences between the decks. For instance, a deck with higher endurance may have reduced thermal disturb and/or other memory-loss mechanisms as compared to a deck with less endurance. However, the deck with less endurance may have other advantages (for instance, faster access times, etc.) as compared to the deck with higher endurance. Accordingly, each memory array deck may be tailored for applicability relative to specific memory functions.

According to that reduced latency (access time) can be achieved by sacrificing endurance of the 3DXpoint layers.

Endurance can be regained by having slower writes. (I am sure there are some exceptions though)

Increasing parallelism regains the speed lost by the slower writes.

They may or may not do this. Patents are very unreliable sources of information on future products though.

Maybe we see TSVs (for speed and capacity) first show up on the DDR5 level Optane DIMMs? Then after seeing what effect that can bring Intel and Micron get even more aggressive on TSVs (which leads to more tuning of latency).

 

[IntelUser2000]

(cbn, you could have used the space for the deleted post instead of creating another one)

Sacrificing endurance for latency is not a good idea especially on a part that displaces most of DRAM. Especially technology that's already significantly behind.

Anyway I believe Intel will continue to aim Optane products for high performance, endurance, and high price. This becomes their way of segmenting the memory division so 3D NAND can become the volume penetrator while 3D XPoint ends up the one bringing the dough. For Intel DIMM that's ok, because you have the performance to go with it. I think they are misguided on making very expensive storage though.

 

[cbn]

Sacrificing endurance for latency is not a good idea especially on a part that displaces most of DRAM.

Not endurance, just latency (I edited).

See the quoted area from post #103:

According to that reduced latency (access time) can be achieved by sacrificing endurance of the 3DXpoint layers.

Endurance can be regained by having slower writes. (I am sure there are some exceptions though)

Increasing parallelism regains the speed lost by the slower writes.

 

[IntelUser2000]

Not endurance, just latency (I edited).

That's interesting. So the same stack would have a slow section and a fast one. It's like SSDs that use circuitry to enable TLC on one end but SLC/MLC for performance on the other.

 

[cbn]

That's interesting. So the same stack would have a slow section and a fast one. It's like SSDs that use circuitry to enable TLC on one end but SLC/MLC for performance on the other.

The patent does mention different types of layers, but I was thinking of a die with all low latency, low endurance layers.....but with an increasing amount of TSVs.

(With the TSVs, the high endurance layers shouldn't be needed because each die can be more energy efficient at X write speed level. Less energy put into the die means less wear and tear on the die.