Discussion Optane Client product current and future

[IntelUser2000]

I was never able to replicate that. This is 4 900P drive in VROC 0 in my own system.

That's a bit strange, but it could be that CDM isn't a good enough benchmark. Most sites use IOMeter when doing such tests. RAID setups always added latency. Even with NAND SSDs RAID setups were questionable. It was really with HDDs you wanted RAID.

 

[nosirrahx]

That's a bit strange, but it could be that CDM isn't a good enough benchmark. Most sites use IOMeter when doing such tests. RAID setups always added latency. Even with NAND SSDs RAID setups were questionable. It was really with HDDs you wanted RAID.

I ran through all of the basic disk tests like ATTO and the like, none of them showed bad results for VROC 0. I do remember some review site being critical of the bad results stating that they also could not replicate the poor performance.

I am away on vacation but I could do some more testing when I get back.

On a related note, I am not sure why I never though to look for 2280 -> 22110 adapters before but they are totally a thing. I am going to buy a few to test laptop clearance as they do appear to add some height. Obviously on desktops this could solve the problem outright as long as there is component clearance.

If I can get a 905P into this laptop the performance will be pretty crazy.

 

[IntelUser2000]

I ran through all of the basic disk tests like ATTO and the like, none of them showed bad results for VROC 0. I do remember some review site being critical of the bad results stating that they also could not replicate the poor performance.

TH's results may be particularly bad, but Optane with super low latency will definitely be impacted by it. Just run using IOMeter.

https://www.pcper.com/reviews/Stora...-Quad-Optane-vs-Quad-960-PRO/Performance-Focu

In practice, a drop from 86K QD1 IOPS to 55K won't matter. At 55K that's still 220MB/s.

 

[nosirrahx]

In practice, a drop from 86K QD1 IOPS to 55K won't matter. At 55K that's still 220MB/s.

Not only that but a lot of the files I transfer on and off of this array are huge so most of what you feel is the massive sequential speed.

 

[cbn]

32GB Optane is down to $41.99 free shipping at Newegg:

https://www.newegg.com/Product/Product.aspx?item=N82E16820167427

Wow! That is really cheap!

 

[cbn]

Still Interested in Carson Beach Optane.......

........which I reckon will be the best mainstream Optane to get before Optane goes mainstream on NVDIMM-P and SO-NVDIMM-P.

I wonder if Carson Beach will use the same Controller as the 905p (M.2)?

 

[IntelUser2000]

An Intel Innovation Day presentation says Stony Beach 2 and Carson Beach is coming second half of 2018. There might have been delays as we're into November already. However, at least we now know what the "Next Gen Optane Memory" meant on the 300 series chipset presentations.

 

[Dayman1225]

Available first on Google Cloud: Intel Optane DC Persistent Memory

Intel Optane DC Persistent Memory Readies for Widespread Deployment

Beta Programs now available for Optane DC Persistent Memory.

Intel® Optane DC Persistent Memory Operating Modes Explained

It also appears that Optane DC PM has two operating modes:

In the volatile Memory Mode, DRAM acts as a cache for the 3DXP (so DRAM doesn't count towards total memory). In the persistent App Direct Mode, latency-sensitive data is stored in DRAM, other data in 3DXP.

 

[cbn]

An Intel Innovation Day presentation says Stony Beach 2 and Carson Beach is coming second half of 2018. There might have been delays as we're into November already. However, at least we now know what the "Next Gen Optane Memory" meant on the 300 series chipset presentations.

I wonder if Stony Beach 2 ends up being Stony Beach but using up to eight Optane dies per package and M.2 2242 form factor? (To me this is more believable than seeing up to eight dies per package on M.2 2280)

And Carson Beach ends being available (with up to eight Optane dies per package) in both M.2 2280 and BGA form factor (re: BGA form factor SSD is even smaller than M.2 2230 by quite bit).

 

[cbn]

Looking at the 12 memory channels per socket specification of Cascade Lake-AP:

Wouldn't it now be very likely Optane NVDIMM-P will developed by Intel?

One reason I bring this up is because having 12 DIMM slots per socket seems a lot more manageable than having 24 DIMM slots per socket.

1 Optane NVDIMM-P replacing 1 DRAM DIMM and 1 Optane DIMM.

 

[IntelUser2000]

I wonder if Stony Beach 2 ends up being Stony Beach but using up to eight Optane dies per package and M.2 2242 form factor? (To me this is more believable than seeing up to eight dies per package on M.2 2280)

I cannot confirm, but Stony Beach 2 may already exist. Certain presentations say it's Stony Beach with the L1.2 power save feature. It may just be the M10 model. The availability of laptops with Optane Memory suggest that's certainly possible and within the H2 2018 timeline. Just like ARK page doesn't say Mansion Beach refresh for 905P but just Mansion Beach, it may be same for Stony Beach 2.

 

[cbn]

The following results are from this video.

Would like to see how the same hard drive compares with Carson Beach Optane.

Does Carson Beach Optane allow the HDD to consistently beat the Intel 760p 3D NAND TLC NVMe SSD? (I'm thinking it probably could because a PCIe 3.0 x 4 Optane controller (combined with an increase in the amount of Optane to 64GB or greater) should allow for a substantial boost in QD1 Sequential Read.)

 

[cbn]

Looking at the lack of NVMe NAND scaling for consumer workloads mentioned in this post and this post (with this post showing an exception) I got to wondering about the future of PCIe 4.0 x 4 or 5.0 x 4 NVMe SSDs.

Could it be that PCIe 4.0 x 2 or even PCIe 5.0 x 1 (or PCIe 5.0 x2) becomes more popular that expected? This until the consoles get a hardware overhaul that promotes a greater need for high Sequential?

Maybe the next hardware overhaul that will boost the need for high Sequential will involve an advanced generation of Optane stacked on top of a Radeon or Intel GPU die?

Or maybe a new low latency Optane that boosts CPU single thread and efficiency?

Or maybe booth?

 

[Insert_Nickname]

Could it be that PCIe 4.0 x 2 or even PCIe 5.0 x 1 (or PCIe 5.0 x2) becomes more popular that expected? This until the consoles get a hardware overhaul that promotes a greater need for high Sequential?

We'll see. I wouldn't mind going for PCIe 4.0/5.0 x1 or x2 drives. Most lower end NVMe drives can barely saturate a PCIe 3.0 x2 link, so cutting down on link width seems a good idea. It'll reduce latency, and possibly power consumption for mobile use. Desktop can retain a x4 link, since power doesn't matter too much there. Even PCIe 4.0 x4 offers a whopping 7.88GB/s, with double that for 5.0. So it should be enough for the foreseeable future.

 

[nosirrahx]

380GB 905P M.2 delayed until 11/23

I just got the email from newegg.

 

[Billy Tallis]

380GB 905P M.2 delayed until 11/23

I just got the email from newegg.

Not too surprising. A Sunday release never made sense to begin with.

 

[IntelUser2000]

At least the ARK page is up for the 380GB M.2.

Same read/write specs as the bigger brothers, noticeably lower power use.

2600MB/s Read
2200MB/s Write
Active power - 9.35W
Idle power - 2.52W

 

[nosirrahx]

At least the ARK page is up for the 380GB M.2.

Same read/write specs as the bigger brothers, noticeably lower power use.

2600MB/s Read
2200MB/s Write
Active power - 9.35W
Idle power - 2.52W

I wonder why they did not offer a smaller capacity 2280 form factor model? Are they worried that the 800P would be completely obsolete in terms of both speed and capacity?

They sell lots of laptops that come with a 256GB NVMe drive, seems odd to me that they would literally lock this entire market segment out intentionally.

That new 7.68TB QLC SATA SSD cached to a 905P in a laptop would be far ahead of anything we will see out of a single drive for at least a year or 2.

 

[IntelUser2000]

I wonder why they did not offer a smaller capacity 2280 form factor model? Are they worried that the 800P would be completely obsolete in terms of both speed and capacity?

I don't think so. You can't fit 7 chips in the 2280 form factor, it needs to be 22110. 7 chips for the 7-channels.

 

[cbn]

I don't think so. You can't fit 7 chips in the 2280 form factor, it needs to be 22110. 7 chips for the 7-channels.

I do wonder if Carson Beach could make use of that controller in M.2 2280 by having some packages with lower amounts of chips than others. For example, M.2 2280 using four packages (both sides of PCB)......three packages with two die and one package with one die, etc. (7 total dies).

P.S. For 905p M.2 I wonder if we will eventually see a 760GB version using eight dies per package.

 

[cbn]

Desktop can retain a x4 link, since power doesn't matter too much there. Even PCIe 4.0 x4 offers a whopping 7.88GB/s, with double that for 5.0. So it should be enough for the foreseeable future.

Found out Radeon Pro SSG has PCIe 3.0 x 8 (via two PCI-E 3.0 x4 M.2 slots ) for its 2TB NAND SSD. (I thought it was only PCIe 3.0 x4)

More info on how Radeon Pro SSG works:

https://www.gamersnexus.net/guides/2537-how-the-radeon-pro-ssg-works-basics

"The extended framebuffer uses NAND Flash (we explain this here) storage. It's effectively an SSD bolted onto a GPU, and allows production users to locally store large files that are being actively crunched by the GPU. Whenever the GPU needs a large texture, big data, or composite objects, it'll talk to the CPU (via PCIe bus), which then talks to the system storage device or memory, depending on if that data happens to already be loaded in volatile memory. The SSG bypasses this process by storing files for entire projects locally. Those files remain resident until cleared by the application or user, and only need to be migrated to the SSG once (unless cleared regularly) when working on a project. In this way, it acts as proper solid-state storage; this is not your standard volatile VRAM.

Let's take an example of an architectural CAD model. You might be marking a city or using ray-tracing to determine where light will hit with your building – in Europe, there are laws restricting buildings to prevent light blockage to other buildings, which can be modeled with ray tracing. To load massive CAD models can take huge amounts of time as the GPU ping-pongs with system storage, from minutes to an hour, in some of the use cases that AMD has detailed to GN. An SSG means we eliminate the communication over PCIe to the CPU, then from the CPU to storage. That's a lot of latency and physical distance – we've asked for a specific number on latency, but haven't gotten one yet – and it's all cut by transacting locally. The GPU can now communicate directly with local storage when its own VRAM capacity has been exceeded or is tapped for other implementations. AMD's rep told us that the card would have 4GB “or more” of VRAM."

So maybe we will see Specialized Workstation Video cards sporting more lanes? Or wider buses?

 

[nosirrahx]

I don't think so. You can't fit 7 chips in the 2280 form factor, it needs to be 22110. 7 chips for the 7-channels.

The whole thing just feels short sighted. PCIe, U.2 and 22110, all form factors with little demand when it comes to storage.

They also don't seem to be interested at all in use cases where you cache a SATA SSD to a 800P which results in crazy capacity and performance (in fact they tell you not to do this).

All of this feels kind of rushed, like Optane was supposed to be a 2019/2020 product but they forced it out early and tried to force it into a market with no demand.

I like the product but only because I can use it in ways its not designed for. If Optane cached locked out the 800P I probably would not touch it anymore.

 

[cbn]

They also don't seem to be interested at all in use cases where you cache a SATA SSD to a 800P which results in crazy capacity and performance (in fact they tell you not to do this).

Yep, they don't recommend 800p for cache:

https://communities.intel.com/thread/127754

From Anandtech:

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

"The 800P and the new Optane Memory M10 are based on the same hardware and an updated revision of the original Optane Memory M.2 modules. The M10 and the 800P use the same controller and the same firmware. The 800P is usable as a cache device with the Optane Memory software, and the Optane Memory M10 and its predecessor are usable as plain NVMe SSDs without caching software. The 800P and the M10 differ only in branding and intended use; the drive branded as the 58GB 800P is functionally identical to the 64GB M10 and both have the exact same usable capacity of 58,977,157,120 bytes."

P.S. Have been wondering if Optane caching can actually reduce wear on a 3D QLC SSD (particularly a dram-less one). Also normally a 58GB 800p is required to beat a SATA SSD sequential write but I actually wonder if the 32GB Optane would work since 3D QLC SATA SSD writes slower? (EDIT: Will have to consider if SLC cache on certain 3D QLC SSDs is larger than the write buffer in the 32GB Optane)

 

[IntelUser2000]

Yep, they don't recommend 800p for cache:

https://communities.intel.com/thread/127754

The comments by the Intel guy suggest while there might not be an official support line for it, they won't block it either.

.....The M10 and the 800P use the same controller and the same firmware.

I wouldn't call them identical, not in practice. Laptops are notoriously finicky devices, just because of necessity. Simple driver based distinction can easily enable optimization that's specific to the M10. While the 800P is widely available to buy, the M10 is mostly offered through systems, more specifically, all configured with laptops.

ARK page also shows the 64GB M10 has lower active power than the 58GB 800P. Some Intel data shows M10 enables extra power management optimizations when used on the 8th Gen Core mobile platforms. You can see from reviews that battery life is quite decent on the M10-enabled laptops which is different from tests shown by AT and Toms.

The real details are likely hidden behind the curtains of laptop manufacturers that make proprietary hardware and driver modifications.

 

[cbn]

Does anyone happen to know what the write buffer size is for the 32GB Optane when used as cache? The 58GB Optane when used as cache?

I haven't been able to find this info yet.