ac's comments are entirely correct. This article is probably the exact opposite of what Intel wants to see. Still, this whole Optane project is a pile of broken promises. Every potential customer should be aware of that fact before he/she makes the decision.

Simply put, another failed Intel project. They realized this way back before launch but pushed it anyway because they were desperate for the revenue.

Another potential use for these is for professionals that already have fastest NVMe drive but still find their daily apps too slow. How much benefit would be in that use would entirely depend on the app. One such test would be trying to test various popular compilers/engines like Intel, MSVC++, C#/Unity, Unreal Engine, CryEngine etc, during repeat compilations with minor changes. You could even compare that to having all the project files in ram drive - something easy to talk about but rarely implemented. I see this Optane aiming eventually toward a "trustable, reliable ramdrive" and it's interesting to see how much this first iteration lags behind that goal. ( - laptops already have battery but in desktop use it would be interesting to see which "ramdrive" preserves your work if power is cut abruptly: Optane or some software ramdrive)

These benchmarks are assuming the device is meant for storage.

Caching is not same as storage, so different benchmarks should be used.

A properly done benchmark for this type of device should use user-interface automation and responsiveness profiling. eg. How many ms does it take to open up windows explorer or browser or restore things from pagefile. And how well does the cache retain things.

Obv. a very fast nvme drive gets things to "good enough" level for most things but what if you wanted to have say 2x 8TB drives in a mirror as a system drive and have this cache in front of those? To get same with SSD's would cost a great,great deal more.

A very legit question however is, in such 2x8TB + current optane module config, are these caches large enough and do they work as good as one would expect (ie. they def. should not cache anything running in WMP but they def. should cache say first and last xMB of a video so that it starts up quick even while the HDD mirror setup is still in process of spinning up if it was spun down due to idle time)...