Micron's 3D XPoint QuantX Product Slays Intel's Optane

[It's me!]

Micron Kicks Intels Arse pretty soundly in the 3D XPoint test results, but it seems that Intel and Micron are betting huge on 3D XPoint, and they aren't even done building it themselves. All of the test results they've shown are from single-deck materials, when the final version will have double decks. That means, basically, that they aren't even in the ballpark of being as far along as they claim. No wonder they are a year behind, and counting. 

Also, they've explained why their claims went from 1000X faster than NAND to only 4X faster, but still have no naswer on why they said it was 1000X more endurance, but now it is only as much endurance as most regular enterprise SSDs. 

I'm starting to think its gonna flop on em. 

Quote

 

The final material will consist of two layers, which Micron refers to as "decks," although... IMFT was still working with prototype single-deck materials. There were newer 3D XPoint revisions arriving every 60-90 days, so Micron employed daughtercards on its SSD prototype (next page) that allow it to fast swap new revisions. The material will expand to a second deck when it has matured; perhaps it already has.... All of Micron's (and probably Intel's) performance demos were conducted with single deck prototypes.

 

http://www.tomshardware.com/reviews/3d-xpoint-guide,4747.html

 

http://www.tomshardware.com/reviews/3d-xpoint-guide,4747.html

[alexyy]

I don't know what any of this means but it sounds awesome!

[DocSwag]

To be fair, part of the reason the Micron ssds are doing better is because they have like at least around 10 times more storage capacity...

[It's me!]

5 minutes ago, DocSwag said:

To be fair, part of the reason the Micron ssds are doing better is because they have like at least around 10 times more storage capacity...

3D XPoint doesnt scale by capacity, the performance is the same no matter what size the SSD is. There are slides that show Micron's SSDs at 120GB are just as fast as the 1.6TB models. 

[NumLock21]

3D Xpoint is designed together by both Intel and Micron since 2006. Micron calls theirs QuantX and Intel calls it Optane. The extra performance from Micron most likely comes from a 3rd party controller by CnexLabs.

 

[DocSwag]

31 minutes ago, It's me! said:

3D XPoint doesnt scale by capacity, the performance is the same no matter what size the SSD is. There are slides that show Micron's SSDs at 120GB are just as fast as the 1.6TB models. 

Really? I assumed it works the way NAND does. I guess I'm wrong. I do remember hearing that the main bottleneck in the 3D XPoint based SSDs is the controller, not the storage, so I guess that would make sense.

 

It looks like the cnex engineers must've done a better job designing the controller than Intel's engineers. ??

[Energycore]

750,000 IOPs at QD2

 

My RAM is slower than that.

[It's me!]

34 minutes ago, DocSwag said:

Really? I assumed it works the way NAND does. I guess I'm wrong. I do remember hearing that the main bottleneck in the 3D XPoint based SSDs is the controller, not the storage, so I guess that would make sense.

 

It looks like the cnex engineers must've done a better job designing the controller than Intel's engineers. ??

Yeah, According to the article Intel stuck with SMI, a normal SSD controller vendor, while Micron went with CNEX.

 

End result, Intel got pwnt. 

[patrickjp93]

1 hour ago, It's me! said:

3D XPoint doesnt scale by capacity, the performance is the same no matter what size the SSD is. There are slides that show Micron's SSDs at 120GB are just as fast as the 1.6TB models. 

Not true. Where did you come up with that idiocy? It scales by due count just the same as NAND does, only each individual die also has vastly higher performance.

[It's me!]

1 minute ago, patrickjp93 said:

Not true. Where did you come up with that idiocy? It scales by due count just the same as NAND does, only each individual die also has vastly higher performance.

Well, if you look at the bottom album of this page it shows VERY little differentiation between the different capacity points. Yes, the media itself may scale, theoretically, given a perfect interface, controller, firmware and whatnot. As those don't exist in the real world, for all intents and purposes, a 200GB QuantX is almost identical to its 1.6TB brethren when its behind the PCIe 3.0 x4 interface, which is the same connection the Intel uses. Therefore, the capacity shoudn't be the limiting factor, unless the Intel ASIC just sucks even more than thought. 

http://www.tomshardware.com/reviews/3d-xpoint-guide,4747-6.html

 

[patrickjp93]

1 hour ago, It's me! said:

Well, if you look at the bottom album of this page it shows VERY little differentiation between the different capacity points. Yes, the media itself may scale, theoretically, given a perfect interface, controller, firmware and whatnot. As those don't exist in the real world, for all intents and purposes, a 200GB QuantX is almost identical to its 1.6TB brethren when its behind the PCIe 3.0 x4 interface, which is the same connection the Intel uses. Therefore, the capacity shoudn't be the limiting factor, unless the Intel ASIC just sucks even more than thought. 

http://www.tomshardware.com/reviews/3d-xpoint-guide,4747-6.html

 

NAND SSDs also just scale up to the limit of the interconnect on which they sit, but they do scale by die count. 3DXP does as well.

[It's me!]

10 minutes ago, patrickjp93 said:

NAND SSDs also just scale up to the limit of the interconnect on which they sit, but they do scale by die count. 3DXP does as well.

The point is, the Intel SSD isn't slower due to its lower capacity. Perhaps I could've worded it more eloquently, but that basic point is true. 

[patrickjp93]

1 minute ago, It's me! said:

The point is, the Intel SSD isn't slower due to its lower capacity. Perhaps I could've worded it more eloquently, but that basic point is true. 

Except it is. A single-die solution vs. Micron's 24. Seriously, use your brains people.

[It's me!]

6 minutes ago, patrickjp93 said:

Except it is. A single-die solution vs. Micron's 24. Seriously, use your brains people.

The picture of the Intel card shows 4 packages, so there are at least four die. Also, the die density is 16GB, so math says theirs 9 or more die, dependent upon the amount of over provisioning. 

[It's me!]

That also doesn't explain why a 200GB Micron card SMOKES the 140GB Intel. 

[Sauron]

Will we just ignore the fact that the micron drives have over 10 times the capacity? 

3 hours ago, DocSwag said:

To be fair, part of the reason the Micron ssds are doing better is because they have like at least around 10 times more storage capacity...

whew, I thought I was the only one to notice.

[patrickjp93]

2 hours ago, It's me! said:

The picture of the Intel card shows 4 packages, so there are at least four die. Also, the die density is 16GB, so math says theirs 9 or more die, dependent upon the amount of over provisioning. 

No it doesn't. It shows 1 die on the Intel products that have actually been benchmarked so far.

[patrickjp93]

2 hours ago, It's me! said:

That also doesn't explain why a 200GB Micron card SMOKES the 140GB Intel. 

There is no 140GB Intel item to test yet. The only test we've seen is for a single-die 16GB or single-die 32GB product.

 

As for Micron's numbers, they're probably flat out wrong. Wait for 3rd party reviews.

[It's me!]

1 minute ago, patrickjp93 said:

No it doesn't. It shows 1 die on the Intel products that have actually been benchmarked so far.

Obviously you have not read the article that you are referring to, as it is the only explanation for the nonsensical reply. It is a 140GB AIC that they demo'd at IDF. 

 

Here, let me help you and take a picture from the article and show it to you. The big black thing there, that is a 140GB Optane Prototype SSD. PCIe 3.0 x4. There is no such thing as a single 3D XPoint die that is 140GB, mind you. 

 

[It's me!]

5 minutes ago, patrickjp93 said:

There is no 140GB Intel item to test yet. The only test we've seen is for a single-die 16GB or single-die 32GB product.

 

As for Micron's numbers, they're probably flat out wrong. Wait for 3rd party reviews.

Please see the previous reply. Obviously, you haven't been paying attention. Anandtech also covered this same 140GB Optane device. 

[It's me!]

2 hours ago, Sauron said:

Will we just ignore the fact that the micron drives have over 10 times the capacity? 

whew, I thought I was the only one to notice.

Read the remainder of the comments, and you will see that it does make perfect sense. Here, let me help. 

This picture shows that no matter the capacity, the performance of 3D XPoint is nearly the same. See the 200GB Micron? It is only a minor difference between it and the Micron 1.6TB. Now, with the knowledge that they are essentially the same performance across all capacity, the comparative Intel v Micron chart takes on new meaning, eh??

This is 3D XP, the old rules of capacity scaling do not apply. 

[It's me!]

12 minutes ago, patrickjp93 said:

No it doesn't. It shows 1 die on the Intel products that have actually been benchmarked so far.

This is a clearer picture from Anandtech's coverage. 

 

[It's me!]

On 11/26/2016 at 2:51 PM, NumLock21 said:

3D Xpoint is designed together by both Intel and Micron since 2006. Micron calls theirs QuantX and Intel calls it Optane. The extra performance from Micron most likely comes from a 3rd party controller by CnexLabs.

 

Yeah, and according to that article, Intel is still using SMI, which is an SSD controller manufacturer that makes low-rent 4 channel controllers. Lol. 

[Sniperfox47]

The other question to ask is just how much more expensive are Micron's drives?

 

Intel and Micron are, initially at least, targeting two *very* different markets with their drives. I wouldn't be surprised if they were *many* times more expensive than their Intel counterparts.

[SansVarnic]

I am not really surprised this is Micron's specialty.