More Intel Optane Memory Reviewed

[CSI_PC]

OK one important aspect that just clicked in with me.
NVMe SSD with Optane Memory Cache.

This will be bottlenecked on anything but HEDT CPUs and motherboard as with consumer CPU it must go through the DMI for both NVMe devices.
Really this needs to be retested once the Skylake-X is launched if context is using the high performance NVMe SSD and seeing how will Optane memory cache works in this configuration.

Cheers

 

[Miikun]

You're not going to have enough PCIe lanes for that. You'll need 4 for your NVMe drive, and another 4 for the Optane drive. On a Kaby Lake, that only leaves 12 left for your GPU.

Optane vs NVMe drive is going to be "either/or, not both", unless you're on an X299 board with a 28/40lane CPU.

12 lanes wouldn't bother me, or 8 even. I think I care more about reducing load stutters and significantly shorter level load times than an extra 5% frame rate at some exotic setting, nor do I buy top of the line GPUs (again, not the titan guy). Faster I/O has always made me happier than a slightly faster GPU (cause i'm impatient) so I spend accordingly.

 

[Bandalo]

12 lanes wouldn't bother me, or 8 even. I think I care more about reducing load stutters and significantly shorter level load times than an extra 5% frame rate at some exotic setting, nor do I buy top of the line GPUs (again, not the titan guy). Faster I/O has always made me happier than a slightly faster GPU (cause i'm impatient) so I spend accordingly.

Right now (until we see some more tests to the contrary), if you care about load stutters and load times, you're fine with a regular SATA SSD. All the testing I've seen that compares NVMe to SATA SSDs shows a nearly undetectable difference in game load and level load times between them. The bottleneck is elsewhere in that particular scenario. Faster I/O hits diminishing returns pretty quickly. The jump between a HDD and an SSD? HUGE and easily detectable by a typical user. The jump from a good SATA SSD to an NVMe drive? Most people probably would never notice it unless they were running benchmarks. (kinda like that 5% frame rate thing)

Dropping to 8x PCIe for your graphics card is probably not going to be a problem for most people, at least right now with current cards. That might change, it might not. I'd personally not want to handicap my build right out of the box by making that decision. I'd rather have a CPU with enough lanes to support a full-speed graphics card AND a couple NVMe / Optane drives.

 

[CSI_PC]

You're not going to have enough PCIe lanes for that. You'll need 4 for your NVMe drive, and another 4 for the Optane drive. On a Kaby Lake, that only leaves 12 left for your GPU.

Optane vs NVMe drive is going to be "either/or, not both", unless you're on an X299 board with a 28/40lane CPU.

It is even more restrictive than that.
From what I have read the Optane Memory Cache must go through the DMI and PCH 200-series chipset with Kaby Lake.
I have read this now from a few sources mentioning this as a limitation with NVMe (storage) on NVMe (Optane cache) in context of Kaby Lake.

Need to see how this is going to be handled by HEDT and Skylake-X.
Cheers

 

[CSI_PC]

Here is the requirement for Optane Memory Cache from Intel guide.

1. Supported 7th Gen Intel® Core™ processor and chipset.
2. M.2 type 2280 connector on a PCH Remapped PCIe* Controller and Lanes in a x2 or x4 configuration with B-M keys that meet NVMe* Spec 1.1
3. System BIOS that supports the Intel® Rapid Storage Technology (Intel® RST) 15.5 driver
4. Windows* 10 64bit or above installed on a supported storage device (Hard drive, SATA SSD or SSHD)

Supported Platforms and Chipsets Intel® Optane™ memory is designed to provide functionality for the following Intel® PCH Storage Controllers:
1. Intel® 200 Series Chipset Family SATA AHCI/RAID Controller SKUs:
 Q270
 Z270
 H270
 Q250
 B250
2. Intel® C230 Series Chipset Family SATA AHCI/RAID Controller SKUs: C236

So for now it does need to go through the 200-series chipset and so the PCH/DMI 3.0 (x4 PCIe).

Cheers

 

[Bandalo]

It is even more restrictive than that.
From what I have read the Optane Memory Cache must go through the DMI and PCH 200-series chipset with Kaby Lake.
I have read this now from a few sources mentioning this as a limitation with NVMe (storage) on NVMe (Optane cache) in context of Kaby Lake.

Need to see how this is going to be handled by HEDT and Skylake-X.
Cheers

Intel has said support for Optane will be added to new chipsets as well, so the X299 should support it in theory. And given you've got either 28 or 40 lanes with the HEDT CPUs, I'd expect to see motherboards with 2 or more m.2 slots when the X299 boards start rolling out. So you could do one Optane and one NVMe in theory, but the value of that configuration remains to be seen.

 

[CSI_PC]

Intel has said support for Optane will be added to new chipsets as well, so the X299 should support it in theory. And given you've got either 28 or 40 lanes with the HEDT CPUs, I'd expect to see motherboards with 2 or more m.2 slots when the X299 boards start rolling out. So you could do one Optane and one NVMe in theory, but the value of that configuration remains to be seen.

I agree,
but remember that even HEDT has a PCH/Southbridge going through DMI 3 and the RST is integral to it.
Optane Memory on Kaby Lake also requires the PCH with the integral RST.

So I am not positive this is going to be ideal for those who want a nice fast Memory Cache with their NVMe drive even with the X299, but lets see because I would love a great tech solution myself.

Cheers

 

[Miikun]

Right now (until we see some more tests to the contrary), if you care about load stutters and load times, you're fine with a regular SATA SSD. All the testing I've seen that compares NVMe to SATA SSDs shows a nearly undetectable difference in game load and level load times between them. The bottleneck is elsewhere in that particular scenario. Faster I/O hits diminishing returns pretty quickly. The jump between a HDD and an SSD? HUGE and easily detectable by a typical user. The jump from a good SATA SSD to an NVMe drive? Most people probably would never notice it unless they were running benchmarks. (kinda like that 5% frame rate thing)

Dropping to 8x PCIe for your graphics card is probably not going to be a problem for most people, at least right now with current cards. That might change, it might not. I'd personally not want to handicap my build right out of the box by making that decision. I'd rather have a CPU with enough lanes to support a full-speed graphics card AND a couple NVMe / Optane drives.

It really depends on how the assets are stored on the HDD, low latency access is a boon for small assets, make very little difference if the assets are large. So it's more of an advantage on some MMOs where they incrementally add content, and can't bundle them in a practical fashion since some assets can't be bundled by zone/level/region. If it's a game that loads faster off a ramdisk, it should be improved by optane. Otherwise, if it had no impact on small file access, you wouldn't see such significant improvements on OS boot times. But we'll have to wait for the actual test results. I haven't updated my desktop system yet, so I'm happy to look at the results before buying in. But i'm tired of putting reasonable sized SSDs on all 4 of my computers, i'd be much happier knowing I only have to worry about that on my laptops.

 

[Bandalo]

It really depends on how the assets are stored on the HDD, low latency access is a boon for small assets, make very little difference if the assets are large. So it's more of an advantage on some MMOs where they incrementally add content, and can't bundle them in a practical fashion since some assets can't be bundled by zone/level/region. If it's a game that loads faster off a ramdisk, it should be improved by optane. Otherwise, if it had no impact on small file access, you wouldn't see such significant improvements on OS boot times. But we'll have to wait for the actual test results. I haven't updated my desktop system yet, so I'm happy to look at the results before buying in. But i'm tired of putting reasonable sized SSDs on all 4 of my computers, i'd be much happier knowing I only have to worry about that on my laptops.

A 32GB Optane cache drive isn't going to help much with larger games that push 60GB each. It might help for some things, but you're still going to see high latency loads with Optane+HDD compared to a straight SSD setup.

 

[CSI_PC]

The Optane has incredible 4K read performance though, which ties in with what Miikun mentions.

The problem I have is that it needs to use the PCH/DMI 3 and not direct to CPU and it is associated only with the boot drive and so system as well.
Intel need to be able to make this more configurable and link to other drives where usually one only loads games from and also with more flexibility that it is not forced to use the PCH/DMI connection but can be direct to CPU (not going to happen anytime soon IMO).

Anyway one has to cope with the DMI contention if owning the Optane Memory Cache and also a fast NVMe SSD.
It may be better to own the 375GB Optane SSD (PCIe x4 slot) and use the Intel software to hook that into being part of the system memory pool, a freakingly expensive idea though for memory caching and losing the SSD as a storage device

Cheers

 

[Miikun]

A 32GB Optane cache drive isn't going to help much with larger games that push 60GB each. It might help for some things, but you're still going to see high latency loads with Optane+HDD compared to a straight SSD setup.

Actually it should, if it's properly designed, it will only cache small files, as larger files don't benefit from latency as much as it spends significantly more time in transfer. If your interface is taking 10ms to transfer your 1MB file, the difference between 1us and 10us access latency doesn't matter. When your file is 512 bytes, it's huge.

 

[Bandalo]

Actually it should, if it's properly designed, it will only cache small files, as larger files don't benefit from latency as much as it spends significantly more time in transfer. If your interface is taking 10ms to transfer your 1MB file, the difference between 1us and 10us access latency doesn't matter. When your file is 512 bytes, it's huge.

That assumes that Optane caches on a per-file rather than per-block basis, which I don't think it really can at the hardware level.

Even then, I'd still question how much you're going to see this difference at the keyboard and not in a benchmark.

 

[Miikun]

That assumes that Optane caches on a per-file rather than per-block basis, which I don't think it really can at the hardware level.

Even then, I'd still question how much you're going to see this difference at the keyboard and not in a benchmark.

It can, sequential access at low queue depths can be analyzed and statistically modeled with reasonable accuracy, but you're right there's no guarantee there's any smart logic behind it (but considering this is likely windows only product like SRT, it may be not only FS aware for NTFS, but OS and init state aware so it can aggressively optimize boot performance without post-boot access patterns gradually pushing out boot time optimizations through platform agnostic (ignorant) design patterns). My expectations is that for games like MMOs, and other use cases where disk access patterns are less deterministic than in say benchmarkable single player games, this will be have it's advantages since the application devs can't necessarily optimize around it. Well that's my hope at least. My second hope is that this technology is covered in AMD's cross licensing deal with Intel so this won't be an intel-only product forever, we don't want another vendor lock of a good technology, and I'd imagine Intel would be happy to sell optane as a companion to AMD PCs. Please no gsync/freesync nonsense. I'm actually quite excited for this to show up on AWS, I'd be happy to get an R4 box with those PCI-E optane cards / dimms than have to pay for an X1 in heavily distributed applications where the data has to traverse ther network eventually.

 

[zalazin]

It sounds to me just a glorified version of SSD caching. I bought an HP laptop came with 24gb ssd and 1 TB hd. Put in a 240 SSD Msata and set cache to 64gb Not bad. You know what is better? An actual 480 GB ssd in the hard drive bay in place of mechanical. And just using the 240 and 480 ssd as just drives. No comparison ssds are wayyyy faster than caching any mechcanical hd.....

 

[Miikun]

It sounds to me just a glorified version of SSD caching. I bought an HP laptop came with 24gb ssd and 1 TB hd. Put in a 240 SSD Msata and set cache to 64gb Not bad. You know what is better? An actual 480 GB ssd in the hard drive bay in place of mechanical. And just using the 240 and 480 ssd as just drives. No comparison ssds are wayyyy faster than caching any mechcanical hd.....

Don't knock SSD caching, the 512GB l2arc SSD cache I have in front of my spinning rust on my media server really performs about as well as reading off a native SSD and 512GB provides me a pretty good cache hit rate. Optane is closer to reading off a system RAM cache of the contents of your HDD, which is pretty darn fast. Just because SRT cache drives were low performance monstrosities and capped at such a small volume size doesn't mean that SSD caching itself is a bad idea, after all storage has always been tiered cachng layers for cost reasons. If cost were no issue, we'd just have 8TB of dual ported SRAM in every PC. This seems like a reasonable solution if you want to use big disks and get good performance on your active set. Look how much of a difference the L3 cache makes relative to the size of your system RAM, it's great stuff! This can be too.

 

[Bandalo]

It can, sequential access at low queue depths can be analyzed and statistically modeled with reasonable accuracy, but you're right there's no guarantee there's any smart logic behind it (but considering this is likely windows only product like SRT, it may be not only FS aware for NTFS, but OS and init state aware so it can aggressively optimize boot performance without post-boot access patterns gradually pushing out boot time optimizations through platform agnostic (ignorant) design patterns). My expectations is that for games like MMOs, and other use cases where disk access patterns are less deterministic than in say benchmarkable single player games, this will be have it's advantages since the application devs can't necessarily optimize around it. Well that's my hope at least. My second hope is that this technology is covered in AMD's cross licensing deal with Intel so this won't be an intel-only product forever, we don't want another vendor lock of a good technology, and I'd imagine Intel would be happy to sell optane as a companion to AMD PCs. Please no gsync/freesync nonsense. I'm actually quite excited for this to show up on AWS, I'd be happy to get an R4 box with those PCI-E optane cards / dimms than have to pay for an X1 in heavily distributed applications where the data has to traverse ther network eventually.

I think the problem is you're trying to optimize and cache an inherently slow media. And you can never design a perfect cache, meaning you're always going to have that cache miss penalty. Switching from one game to another is likely to see it, since your cache size is fairly small by today's standards. You might also see it switching between large levels in any game. Right when you really NEED it.

I'd rather see vendors sink money into bringing the price of large SSDs down to more reasonable level. If you could get me a good 1TB SSD for $200-300, you'd kill this particular use of Optane overnight. HDDs lost the wars, except for bulk media storage, it's time to move on.

 

[Miikun]

I think the problem is you're trying to optimize and cache an inherently slow media. And you can never design a perfect cache, meaning you're always going to have that cache miss penalty. Switching from one game to another is likely to see it, since your cache size is fairly small by today's standards. You might also see it switching between large levels in any game. Right when you really NEED it.

I'd rather see vendors sink money into bringing the price of large SSDs down to more reasonable level. If you could get me a good 1TB SSD for $200-300, you'd kill this particular use of Optane overnight. HDDs lost the wars, except for bulk media storage, it's time to move on.

All flash is a waste at the current price points since the remaining HDD manufacturers will fight a price war against flash and I don't see flash winning (pricing parity) in the next 5-6 years. In the meantime technology like Optane is basically just faster flash. Saying we should ignore optane and focus on flash is like saying we should have ignored flash and focused on HDDs back when SSD controllers didn't have TRIM. Optane style technology is destined be more expensive than flash for a decade or more, but I'll gladly choose the fastest caching layer for my bulk storage as long as it fits my active data set. If RAM was as cheap as Optane, and windows caching wasn't so inflexible, I'd just buy 64GB of DDR4, never reboot my desktop PC, and call it a day. All flash storage is nice, but it's just too expensive currently, and most of it is wasted since I only need access to about 100GB of data on any given day, and a terabyte of data on any given week. Everything else can be on the cheapest storage I can find, and i shouldn't need to manually waste time shuffling things between, since algorithms can do that in less time. I'll continue to stick to caching layers because it works great when properly designed, and gets you most of the performance with 5x~10x the space for the price. Optane just brings another nonvolatile faster-than-flash / cheaper-than-ram caching option, and I'd like to see it develop a bit.

 

[Bandalo]

All flash is a waste at the current price points since the remaining HDD manufacturers will fight a price war against flash and I don't see flash winning (pricing parity) in the next 5-6 years. In the meantime technology like Optane is basically just faster flash. Saying we should ignore optane and focus on flash is like saying we should have ignored flash and focused on HDDs back when SSD controllers didn't have TRIM. Optane style technology is destined be more expensive than flash for a decade or more, but I'll gladly choose the fastest caching layer for my bulk storage as long as it fits my active data set. If RAM was as cheap as Optane, and windows caching wasn't so inflexible, I'd just buy 64GB of DDR4, never reboot my desktop PC, and call it a day. All flash storage is nice, but it's just too expensive currently, and most of it is wasted since I only need access to about 100GB of data on any given day, and a terabyte of data on any given week. Everything else can be on the cheapest storage I can find, and i shouldn't need to manually waste time shuffling things between, since algorithms can do that in less time. I'll continue to stick to caching layers because it works great when properly designed, and gets you most of the performance with 5x~10x the space for the price. Optane just brings another nonvolatile faster-than-flash / cheaper-than-ram caching option, and I'd like to see it develop a bit.

I didn't say ignore Optane. I said it would kill this particular use of Optane (i.e., as a cache for a slower disk). The only reason I use cache in the first place is because it's a high-speed (but high-cost) method to access a low-speed (but low-cost) dataset. I'd rather have a 1-2TB Optane drive for $300 over a 32GB Optane + 2TB HD for $150. But, and this is a big factor right now, there's diminishing returns on drive speed. If you look at the "user experience", rebooting, opening programs, switching levels in games, etc, there's very little difference between a decent SSD and a REALLY fast NVMe drive (or Optane). On paper and in benchmarks, that extra performance Optane provides is awesome. But is a single user on a single machine going to see any difference between a full Optane-speed SSD and a good NVMe drive?

I agree you'll probably never get SSDs as cheap as HDDs, but you don't need to. You just need to get 'em a bit closer. The benefits of flash far, far outweigh those of HDDs, the only thing keeping HDDs in business is cost, and they can only go so low. Just due to manufacturing costs, you'll never get HDDs too far below their current pricing, but there's lots of room to drop flash storage prices.

 

[CSI_PC]

I didn't say ignore Optane. I said it would kill this particular use of Optane (i.e., as a cache for a slower disk). The only reason I use cache in the first place is because it's a high-speed (but high-cost) method to access a low-speed (but low-cost) dataset. I'd rather have a 1-2TB Optane drive for $300 over a 32GB Optane + 2TB HD for $150. But, and this is a big factor right now, there's diminishing returns on drive speed. If you look at the "user experience", rebooting, opening programs, switching levels in games, etc, there's very little difference between a decent SSD and a REALLY fast NVMe drive (or Optane). On paper and in benchmarks, that extra performance Optane provides is awesome. But is a single user on a single machine going to see any difference between a full Optane-speed SSD and a good NVMe drive?

I agree you'll probably never get SSDs as cheap as HDDs, but you don't need to. You just need to get 'em a bit closer. The benefits of flash far, far outweigh those of HDDs, the only thing keeping HDDs in business is cost, and they can only go so low. Just due to manufacturing costs, you'll never get HDDs too far below their current pricing, but there's lots of room to drop flash storage prices.

The Optane memory cache is a fraction of the price though when compared to the Optane SSD currently at 375GB, imagine the cost of a 1TB version of it and it will be scary as a new tech for most.
NVMe SSD still cannot compare to Optane in terms of performance.

The downside with the 32-64GB memory is that it uses the shared-contended DMI on Intel platform while also solely boot drive linked, and also technically 2x PCIe rather than the full 4x like the SSD Optane that also can connect direct to CPU PCI related lanes.
Intel needs to evolve their chipset-CPU interfaces-communication and software as it is now possibly a limiting factor with regards to Optane unless used in its more basic SSD on a PCIe x4 connection.

Personally I would want SSD with the performance characteristics of the Optane SSD.
Optane does have a future when one also considers what AMD is pushing with their Vega software-driver solution and unified/virtual memory pool, with Optane having the chance to supercede most of this if Intel evolved it correctly.
Cheers

 

[Bandalo]

The Optane memory cache is a fraction of the price though when compared to the Optane SSD currently at 375GB, imagine the cost of a 1TB version of it and it will be scary as a new tech for most.
NVMe SSD still cannot compare to Optane in terms of performance.

The downside with the 32-64GB memory is that it uses the shared-contended DMI on Intel platform while also solely boot drive linked, and also technically 2x PCIe rather than the full 4x like the SSD Optane that also can connect direct to CPU PCI related lanes.
Intel needs to evolve their chipset-CPU interfaces-communication and software as it is now possibly a limiting factor with regards to Optane unless used in its more basic SSD on a PCIe x4 connection.

Personally I would want SSD with the performance characteristics of the Optane SSD.
Optane does have a future when one also considers what AMD is pushing with their Vega software-driver solution and unified/virtual memory pool, with Optane having the chance to supercede most of this if Intel evolved it correctly.
Cheers

The cost right now isn't great. But this is also the first release, first revision products. SSDs were 10 times the price when they first were available, so this is a huge step in the right direction already, but I'd bet Intel can get the cost down pretty quickly to parity with regular SSDs.

Taking full advantage of Optane as the new SSD will require new chipsets, and possibly new CPUs and some big OS revisions to take full advantage of all the capability. I'd love to see SATA removed from the chipset and things reorganized to take more advantage of NVMe/Optane. If people still want it for bulk storage, leave a couple lanes free, and MB manufacturers can add in a few SATA ports via an external controller.

Optane has a lot of potential in a lot of areas IMO. I don't know how much that will flow down to the typical desktop user, but it'll be interesting to see.

 

[Miikun]

I didn't say ignore Optane. I said it would kill this particular use of Optane (i.e., as a cache for a slower disk). The only reason I use cache in the first place is because it's a high-speed (but high-cost) method to access a low-speed (but low-cost) dataset. I'd rather have a 1-2TB Optane drive for $300 over a 32GB Optane + 2TB HD for $150. But, and this is a big factor right now, there's diminishing returns on drive speed. If you look at the "user experience", rebooting, opening programs, switching levels in games, etc, there's very little difference between a decent SSD and a REALLY fast NVMe drive (or Optane). On paper and in benchmarks, that extra performance Optane provides is awesome. But is a single user on a single machine going to see any difference between a full Optane-speed SSD and a good NVMe drive?

I agree you'll probably never get SSDs as cheap as HDDs, but you don't need to. You just need to get 'em a bit closer. The benefits of flash far, far outweigh those of HDDs, the only thing keeping HDDs in business is cost, and they can only go so low. Just due to manufacturing costs, you'll never get HDDs too far below their current pricing, but there's lots of room to drop flash storage prices.

Well, I think we'll agree that you're not early adopter and that's fine, i've been on the bleeding edge of SSDs, having owned 1st gen SLC, pre-trim SSDs that slowed to floppy speeds before trim, barefoot 1-2, a couple gens of intel, sandforce 3 generations, a slew of micron and samsungs. You'd rather just wait for tech to hit the mature state before adoption (1tb optane for $300) and that's your perogative, but it's early adopters that fund the technology and let it survive to maturity. And I encourage ppl to vote for technology with their wallet, even the kind of crappy implementations like SSDs in the early commercial days because doubters can totally kill products before they mature, just look at the amount of mudslinging against VR, and news articles telling people to skip a generation or two while HTC relies on a profitable outlook to even consider a refresh. That and I don't think this 1st gen of optane is that crappy at all, or that these m2 sticks wlll stay at this size forever in their first incarnation (as in you'll be able to stick a much bigger m2 optane drive in that same kaby bridge motherboard in 2018 for 30% less per GB) . We can't even be sure the prices being discussed will be street prices, those are often baked in with retail margins. After all, the optane crossbar structure is a stacked 3d design, they can scale the density once they get their yields up and stay in the m.2 format. Technology matures gradually, but only if people keep buying it. When people just plan to buy the next version, you end up killing the product (like the Osborne).

 

[Bandalo]

Well, I think we'll agree that you're not early adopter and that's fine, i've been on the bleeding edge of SSDs, having owned 1st gen SLC, pre-trim SSDs that slowed to floppy speeds before trim, barefoot 1-2, a couple gens of intel, sandforce 3 generations, a slew of micron and samsungs. You'd rather just wait for tech to hit the mature state before adoption (1tb optane for $300) and that's your perogative, but it's early adopters that fund the technology and let it survive to maturity. And I encourage ppl to vote for technology with their wallet, even the kind of crappy implementations like SSDs in the early commercial days because doubters can totally kill products before they mature, just look at the amount of mudslinging against VR, and news articles telling people to skip a generation or two while HTC relies on a profitable outlook to even consider a refresh. That and I don't think this 1st gen of optane is that crappy at all, or that these m2 sticks wlll stay at this size forever in their first incarnation (as in you'll be able to stick a much bigger m2 optane drive in that same kaby bridge motherboard in 2018 for 30% less per GB) . We can't even be sure the prices being discussed will be street prices, those are often baked in with retail margins. After all, the optane crossbar structure is a stacked 3d design, they can scale the density once they get their yields up and stay in the m.2 format. Technology matures gradually, but only if people keep buying it. When people just plan to buy the next version, you end up killing the product (like the Osborne).

I think Intel is going to make money on Optane for server purposes, not an odd-ball SSD/HDD cache arrangement.

I'm all for early adoption to a point, but this current implementation is going backwards IMO. The HDD with a built in SSD cache has been done, and has lots of limitations, and was mostly killed as SSD prices came down. Lets not resurrect it just to demo another possible use of Intel's new memory tech.

I honestly think Intel is a bit afraid. Optane/Xpoint is pretty damn good in the server market. I'll bet a full SSD implementation will probably end up pretty comparable to a top-notch Samsung SSD and the only difference will be that it'll cost a lot more. I can't think of any other reason for them to roll this particular solution out to the desktop market.

 

[westrock2000]

I still think Apple has done it best with Fusion, and I don't understand why others haven't followed suit.

Fusion adds an SSD of any size with a HDD of any size for a single storage device of total combined space and simply prioritizes storage to the SSD based on file usage.

So stuff you don't need much, or are large sequential files get moved physically to the HDD and smaller files get put on the SSD. And thats only after you have used the SSD....initially everything goes to the SSD. So write speeds stay good too.

I don't know if there are patent issues or what, it just seems strange Windows hasn't tried to replicate this yet.

I used a Sandisk ReadyCache for a while on a work computer and it did help (it was noticable). but it had sync issues where if the SSD became out of sync with what was on the HDD it would dump everything and start over. You do not get that with Fusion since the data on the SSD is actual data.

And no, you do not need official Apple drives to do this.