Discussion Optane Client product current and future

Page 24 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

cbn

Lifer
Mar 27, 2009
12,968
221
106
For the Optane caching a separate NVMe NAND drive.....this also gets the additive Sequential write?
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
There's nothing on the RST release notes for this. It might be easier to implement the feature on one drive.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106

Pretty surprising results on Blender and Gaming. (Less impressive results with Premiere)

Screenshot_22.png


Screenshot_23.png


Screenshot_24.png


P.S. The storage being cached by the 32GB Optane is a 1 TB SATA 6 Gbps SSD.

Looking at those results and considering that H10 is available on lower end hardware maybe Microsoft will begin to allow the option to change "swappiness" in Windows?

P.S. Sure Optane cache is not that fast in Adobe Premiere, but for a Celeron or Pentium user this would be very low on the list of tasks performed. (Main tasks for Celeron user would be web browsing, Youtube, office and Main tasks for a Pentium user would be web browsing, Youtube, office and gaming).
 
Last edited:

VirtualLarry

No Lifer
Aug 25, 2001
56,383
10,070
126
Looking at those results and considering that H10 is available on lower end hardware maybe Microsoft will begin to allow the option to change "swappiness" in Windows?
That's a good point. Much like SuperFetch being mostly useless on an SSD system, much of the "pre-paging" behavior that is designed to prevent the worst-case latency scenarios from occurring while paging with a mechanical HDD, could be elided, with a system with an Optane cache drive being used for paging. (Move to JIT paging, rather than pre-paging.)
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Having more free DRAM to write to in a low DRAM system would be great.

One thing I noticed during my informal 58GB 800p browsing test (Windows installed on the 800p) was that as I opened more and more web pages my system got slower and slower due to lack of free dram to write to (However, at the same time the first 10 web page I went to could still open really fast). The system was holding on too much in DRAM believing I only had a hard drive to page to.
 
Last edited:

cbn

Lifer
Mar 27, 2009
12,968
221
106
Yep, that is NVDIMM-P.

P.S. Regarding NVDIMM-P I have been wondering if Intel is working on something the equivalent of that but on the GPU die? (Imagine perhaps a GPU using TSVs (ie, stacked GPU dice*) with DRAM right above that and Optane right above the DRAM.

*Think GPU designed with high performance per watt at low voltage....with the lack of absolute performance made up by stacking the dice.

GPUs would do well with internal storage for graphics data. You could eliminate load times and texture pop in as the primary storage feeds the on GPU storage.

Imagine a typical 12GB of frame buffer and then another 64GB of low latency storage right on the GPU.

Game code would need to change to take advantage of this but I bat that would not be that hard as all you are doing is changing where the data is, not the actual data.

The problem with that idea, like with all others is cost.

64GB Optane is quite pricey. If you want developers to code for it, then you'd want it widespread. You would want it not just on GTX 1060 class cards, but much lower. Literally every new discrete GPU out there needs to have it on board, so of course AMD's as well. Look at what's happening with RTX and DLSS.

In addition to the discrete cards, the future consoles would be a good place to have Optane that is pin compatible with HBM.

Maybe this will be achievable by the time third gen optane comes out? (Thinking node shrink to boost endurance sufficiently enough to cover for both increased switching speed and increased parallelism per GB)

P.S. If Optane that is pin compatible with HBM does come out I expect it will help Intel do very well in the data center (Xe, NNP and FPGA). Consumer Xe will also get a boost, but I don't think Intel will want consoles......AMD will though.
 
Last edited:

Dayman1225

Golden Member
Aug 14, 2017
1,152
974
146
1557699735027.png
Looks like Optane Persistent Memory will be coming to Client Workstations later this year, not HEDT but we are getting there
 
  • Like
Reactions: cbn

cbn

Lifer
Mar 27, 2009
12,968
221
106
View attachment 6259
Looks like Optane Persistent Memory will be coming to Client Workstations later this year, not HEDT but we are getting there

Very interested to see how DreamWorks will deploy those in their workstations:

https://forums.anandtech.com/thread...n-software-development.2529334/#post-39295413

3:56 Scott Miller (of Dreamworks) talks about moving persistent memory closer to the end user.

5:47 Scott Miller (of Dreamworks) wants to federate the 3DXpoint DIMMs across the enterprise which he describes as sort of a fabric vision but with end user workstations rather than data center gear.

19:37 A Question was posed to Scott (of Dreamworks) about how persistent memory could change (animated) movie workflows. His answer was he wanted to do something called "Sequence based or whole film based development" rather than working on individual shots (about 80 frames) within each workstation. This would take 200TB, 300TB or 500TB of federated persistent memory spread across many machines to hold all the assets (and drive work) to work on the whole movie at the same time.

36:47 Scott Miller (of Dreamworks) mentioned designing APIs around in memory data structures rather than for filesystems.

I do hope Dreamworks allows someone to do an article on that.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Looks like Optane Persistent Memory will be coming to Client Workstations later this year, not HEDT but we are getting there

I knew that back when this roadmap came out: https://www.extremetech.com/wp-content/uploads/2019/04/Intel-Client-CPU.jpg

The 18C version of Cascade Lake-W has the code-name "Apache Pass" next to it. Apache Pass is the code-name for the first generation of Optane DC Persistent Memory.

The good news is the possibility the earlier roadmap being wrong, because while the Apache Pass is only shown for what can only be described as a Cascade Lake refresh in 2020, Intel slide says it'll be this year. Perhaps the same, the client mobile roadmap will be wrong and we'll see more 10nm products earlier and more commonly.

Would have been nice to see support for Optane PMM on Cascade Lake X though. Otherwise its a maybe 200Mhz clock bump. I think enthusiasts would have been very much on board with even the datacenter PMM prices. The 128GB is $600, which is not bad at all for the advantages it brings. Some gamers spend multiples of that on auxiliaries, like chairs, LED bling, and decals.
 
  • Like
Reactions: Dayman1225

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
In an interesting comparison between implementation and power use, the dual-port Optane SSD DC D4800X has a slightly lower 2400MB/s Read but higher 2400MB/s Write compared to the single-port versions. But the single port P4800X is set at 18W while the D4800X is set between 20-25W depending on capacity.

D4800X also has slightly higher than double the latency at 20-30us.

It looks like if Intel were to set power limits higher, we could also get enthusiast drives with 2.4-2.5GB/s bandwidth. But getting that 10% improvement results in something like 30% increase in TDP.

I am highly inclined to believe now Carson Beach and Bombay Beach probably uses a controller with new design and/or new process to lower power. Achieving 1GB/s write at 3.5W power levels require substantial work. The 100GB version of P4801X uses 7W to do so. Assuming the client version can reduce that to 5.5W, its still a reduction of 37%.
 
  • Like
Reactions: cbn

cbn

Lifer
Mar 27, 2009
12,968
221
106
Now that Icelake-U (with hardware mitigations for Speculation based attacks) is out I do hope we see retesting of Optane H10 (along with Intel 760p).

Reason: The hardware fix should result in a substantially better 4K QD1 read compared to the firmware fix used by Whiskey Lake.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Reason: The hardware fix should result in a substantially better 4K QD1 read compared to the firmware fix used by Whiskey Lake.

This is a good point. I hadn't thought of that. They probably won't retest it.

But, we should see laptops coming with H10. Computex is showing laptops with H10 coming. Then we'll see it in a relative sense.

https://www.notebookcheck.net/Dell-...for-a-starting-price-of-800-USD.421438.0.html
https://www.notebookcheck.net/Alien...h-even-thinner-Alienware-m17-R2.422290.0.html

Only Dell so far.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
This is a good point. I hadn't thought of that. They probably won't retest it.

But, we should see laptops coming with H10. Computex is showing laptops with H10 coming. Then we'll see it in a relative sense.

https://www.notebookcheck.net/Dell-...for-a-starting-price-of-800-USD.421438.0.html
https://www.notebookcheck.net/Alien...h-even-thinner-Alienware-m17-R2.422290.0.html

Only Dell so far.

I do hope the higher 4KQD1 Read encourages Intel to release a consumer version of memory drive.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
I doubt Intel will be sending me an Ice Lake system for storage testing until they have another new storage product to use with it.

Haha I'm glad at least you are considering it. Your test was unique that the benchmarks weren't just their testing recommendation.
 

nosirrahx

Senior member
Mar 24, 2018
304
75
101
I am starting to wonder if the 815P is a myth. Perhaps they are holding it back so the hybrid drives can get the spotlight?
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
With Icelake-U it would be nice to see H10 compared alongside M15/815p.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
I am starting to wonder if the 815P is a myth. Perhaps they are holding it back so the hybrid drives can get the spotlight?

I hope that is not the case.

I see H10 (with higher 4K QD1 Read due to PCIe x2 and Icelake hardware fix) being stronger in laptops. This particularly if Intel released a consumer Memory drive to compensate for slow 4K QD1 write.

I see 815p (with faster 4K write and Sequential write) being stronger in desktop. This particularly if using 815p to cache a RAID-5 volume of the new URE spec improved hard drives.
 

nosirrahx

Senior member
Mar 24, 2018
304
75
101
BTW, I think Intel may have broken compatibility with greater than 58/64GB modules for caching in a recent update.

On one of my systems I had an unrecoverable crash after some updates so I reinstalled with all current drivers. All 3 of my 118GB modules are invisible to Optane caching now but my 58GB module works fine.

I will do some testing on earlier Optane builds to see if I can confirm this. If they did this intentionally its a pretty low move to break existing installs.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Optane M15 announced:

https://www.tomshardware.com/news/intel-optane-memory-m15-ssd,39509.html

Sequential Read up to 2000 MB/s, Sequential Write up to 900 MB/s.

Capacities: 16GB, 32GB and 64GB.

Big question I have:

How fast is the 32GB M15?

900 MB/s write or 450 MB/s write? (450MB/s is sweet, but 900 MB/s would be so much better. Furthermore, the 4K QD1 write would be faster too)

(Waiting to see specs show up on Intel Ark)
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
900 MB/s write or 450 MB/s write? (450MB/s is sweet, but 900 MB/s would be so much better. Furthermore, the 4K QD1 write would faster too)

Nice! ARK specs are up. Available in Q3.

Optane Memory M15 "Carson Beach"
64GB: 1900MB/s, 900MB/s, 5W active
32GB: 1650MB/s, 550MB/s, 3.8W active
16GB: 850MB/s, 300MB/s, 3W active

Read and write latencies are down on all of them. The write latency reduction is significant.

The active power increases with M15 compared to M10. I was right about the write performance being related to TDP, and why Intel can't get super fast sequentials out of them. I'm having doubts on the controller being on the new process, and they are using higher TDP to achieve better write performance. Perhaps its a redesigned controller rather than one on a new process.

Double the write performance on the 16GB and nearly double on the 32GB version. That's good. PCIe x4 also increase sequential read substantially.
 
Last edited:
  • Like
Reactions: cbn

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
They changed the layout on the M15. I wonder if they are going to put something on that space there below the Optane branding?