Intel Spills the Beans on Optane: Kaby Lake Processor Required

"If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck"...

Is it just me or does anyone else (who's old enough to remember) kinda feel like history is repeating itself for poor old Intel (I say poor... they've been ripping us off for the last three generations of CPU).

Kaby Lake + Optane vs Ryzen + DDR4 looks and sounds awfully similar to..... The Pentium 4 2.4Ghz + Rambus memory vs AMD TBirds B + DDR2 quack, quack!

We all know how that fight ended up. The winner was the consumer :D
Show me some benchmarks before I care one way or another on this topic...

If this thing is really light-years better than a standard NVMe M.2 drive, and benchmarks and day-to-day real life operations proves it, then it'll be worth the upgrade.

If it's faster in benchmarks but not anything noticeable in real-world use, then it'll probably be some seriously slow adoption and may just go die the SATA Express route.
another middle finger at you straight from intel lol. Oh well, I won't have octane anytime soon so no biggie for me, but I'm looking forward to Ryzen if it doesn't suck and I'll dump my 2 skylake systems in a heartbeat just to give the finger back to Intel.
Same as 4K blu-ray. Guess my next few builds will be AMD. Gotta at least try to keep Intel inline.
Samsung thanks you for the extra business your handing to them.

Welcome back to the rambus days all!
Looks at Mac Pro with un-official operating system. Un-official cpu's. Un-official memory. Un-official video card. Un-official esata controller. Un-official harddrives.

Looks to heaven and screams "FU Apple you ungreatful fascists!"

It doesn't look like this Optane product really follows any established protocol. It uses an M2 SATA port, but it's not really SATA. And I'm sure there is some bus shenigans going on in the back ground. So I would really expect this to work on older chipsets/ports. If it's classified as memory, then there needs to be a memory controller.
You might be thinking of "Ready Boost" which wasn't for SSD's but any flash drive. The problem was that it wore out the flash memory too quickly by exceeding the finite reads and writes the flash memory could handle. This was in the infancy of the SSD and the earlier days of removable flash storage.
I think maybe he meant Turbo Memory?

It does seem like something they have tried a few times since then
Some of this non volatile memory will be in DIMM slots. It will be at speeds unlike any PCIe slot card. The memory controller probably is different for something like that. It would be nice if it could accommodate terabytes of data. SSDs would be obsolete. For now it probably is limited to the same address lines as DDR4. Some day there may be no storage, only RAM.
While I'm at it, I thought I'd point out some examples of what I mean. In the Northwood Pentium IV days, the Northwood core supported "Jackson Technology" which would later be known as Hyperthreading. Though it was only ever enabled on the Pentium IV 3.06 "B" CPU. The next revision of Northwood Pentium IV's known as "C" processors generally had Hyperthreading enabled. It was a technology that was built into all Northwood cores but not actually enabled. Similarly, Prescott Core "E" Pentium IV's had x86-64 / EM64T support which wasn't enabled until it made the transition from socket 478 to LGA775.

Sometimes support for better memory technologies, certain CPU instruction sets and features are built into the silicon for the purpose of bringing them out later. Sometimes Intel can enable these features via microcode updates and sometimes they can be enabled via BIOS updates. Other times these design elements simply aren't ready for prime time and can't be cut out of the design entirely and will remain unused until a new processor revision or stepping comes out. Sometimes the revision has to be a major one. If you read documentation concerning processor erratum, you'll find that processors can often behave in ways that are not intended under very specific circumstances. They are far more flawed than we generally realize. So, Intel may have wanted Skylake to support Optane technology but couldn't because it simply didn't behave as intended with the technology. This sort of thing often happens when a feature is designed to interface with something that's not released or available at the time the design was finished. In this case, Optane memory wasn't available for QVL testing during the earlier stages of Skylake's development.

This somehow makes sense to me. My personal new board Asus 2011 and 3930k doesn't legacy support PCIE 3.0 every time I update the GPU driver I have to run a small program from Nvidia to enable 3.0?? But at least I can :) I'm sure I'm wrong but the info I found was Intel wasn't supporting PCIE 3.0 on 3930k, that came in the 49x series, but it's on the chip (at least my 3930k) and the P9x79 Deluxe 1st gen (blue board).

Maybe its something that can be bypassed by software or a MB vendor design or with a bios update?...if they were inclined to do so. Just a thought :)

(edit: I mean high end Z170 boards (maybe even older) might already have what's required. Who knows;)
Last edited:
I went from IvyBridge to Kaby Lake like a month ago. I'm good to go, but odds are, I won't go. I'm sure Optane is going to be expensive and not worthwhile for my home usage (some video encoding and gaming). Guess will just have to see when it actually releases.
This is without a doubt the biggest bullshit statement I've read today. It's still early though. :) Microsoft never set such a precedent. That's Apple you are thinking of. Optane support and Skylake almost certainly has little or even nothing to do with Microsoft. Even if it did, Microsoft is one of the many companies that holds onto legacy features and supports legacy hardware far longer than they probably should. The PC platform has been held back by legacy software and hardware compatibility since the very beginning. Microsoft and even Intel often provided alternatives to legacy platforms which were ignored by the public. The inability to support 32bit code through emulation is why IA64 never took off. Microsoft brought about a true 32bit only OS years before the public would accept it. Unfortunately, people wouldn't accept Windows NT for a long time because it had poor DOS program compatibility. Apple is the only company that gets away with telling it's users that you have no choice but to move on.


Hopefully other companies will come out with a similar product that has support for more than just the newest and greatest chips.
RAMBUS - never again!!!

Thought I would have learned my lesson but fell for tripple channel too.
RAMBUS - never again!!!

This is an enterprise technology that will revolutionize the enterprise storage market. This will be great for raid caches, high performance databases and even applications that used to require TBs of RAM. It may be used in the enthusiast consumer market but it is an enterprise product.
Last edited:
Pure Flat Memory Model = CPU Caches <> Optane CPU Caches (as big as current RAM) <> Optane DIMMs <> Optane NVMe <> Volume Storage (all cache-coherent).

Optane flat memory model = Ram > Optane DIMM > Optane NVMe > Volume Storage = shift in storage paradigm, everything is addressable in the same format it just lives in different places.

Its basically ZFS on steroids, unfortunately at the present time the software eco-system simply does exist for storage that's also live memory via any OS. Machine Learning, SETI, Anything embarrassingly parallel etc is what it's good for (and designed for) but the eco-system doesn't exist for that either.

Optane is currently:-

1. Very Late.
2. More than x100 times slower than performance promises made by intel multiple times.
3. Not as resilient as promised.
4. Not as compatible as expected (how hyper-threading was dragged into existence is a perfect analogy).
5. Promised capacity targets missed (that's why Optane is on M.2, It was supposed to start at DIMM level).

So much initial promise, but I ain't buying into their poxy SSHD half-way house Optane bullshit. Currently its the emperor's new clothes.

When a FMM Computing software/OS ecosystem exists I'll buy (if they ever fix it). What I won't do is buy 'ReadyBoost 2.0' (licensed from Microsoft :p).
This is an enterprise technology that will revolutionize the enterprise storage market. This will be great for raid caches, high performance databases and even applications that used to require TBs of RAM. It may be used in the enthusiast consumer market but it is an enterprise product.

At least in the short-to-midterm. It has the potential in a number of applications to simplify the memory/storage architecture. But as far as 2017-2018 are concerned, I agree 100%. Past that it may get fuzzy.
Don't bet on it. At least not for a year or two.

I am not sure why any home user would care anyway. From benchmarks it isn't significantly faster than Good M.2 SSD like a Samsung 960 Pro and likely much more expensive.
Thank for remembering things that i saw before. optane is RAMBUS memory and U2 slot all over again: trying to make proprietary something that could be standard. As it stands, m2 RAIDs are the way to go for [H]ard users.
I had to look up what optane was since I never heard about it.