SSD Caching w/Z68 Chipset: Is it worth it?

See:
Intel Z68 Express Chipset Preview: SSD Caching And Quick Sync : Z68 Express Makes Its Debut


After reading the above article a few hours ago, I was a little disappointed with the results... then, after thinking about it - I am even more positive that my next desktop (and hopefully notebook) build will be with an SSD + HDD combo powered by the new Intel Z68 chipset.

At first, a 2x to 4x improvement (not seen in boot/shutdown times!) over a standard HDD just didn't seem 'enough'. Then I re-read the article a second time and took notice of some important aspects of this SSD Caching implementation:

1) Not all SSD's are used for caching by default,
2) You choose how much space to allocate for caching up to 64GB,
3) The 64GB 'limit' is actually a positive, not a negative.


With this information and a few hours of letting my mind wander into other interests... I came up with the follow 'recipe' for speed with these tools:

First, get the biggest, baddest, fastest SSD you can afford - the Intel 510 250GB is an excellent choice in my opinion right now.

Second, get the biggest, fastest HDD(s) you need (Hitachi's all the way...).

Now add a 'K' series 2nd generation iCore, the Z68 chipset and the appropriate supporting peripherals (like 16/32GB RAM, for example ) and set-up the system as follows:

Install Win7x64U to the SSD and any other apps/utilities/programs you need/use daily. On the example Intel 510 250GB SSD, this should leave me with at least (232-100) 132GB free space on the SSD (I typically use 100GB for my 'C' drive partition, including about 20GB free space).

Install IRST 10.5 (not yet available) and allocate 18.6GB and up to 64GB of your SSD to be used for caching - (I would allocate all 64GB available).

With this scenario, you will have the benefits of fast bootup/shutdowns (O/S installed on the SSD) and the nice additional benefits of speeding up your HDD's up to 4x or more too.


Why 'or more'? Because in the article, they are using one of the slowest SSD's available: an Intel X25-V 40GB SSD - but that doesn't mean we have to.

So, while Intel is aiming for the performance enthusiast on a budget, that doesn't mean we have to (...stick to a budget, that is... ).

In the above scenario, I would also create a partition of 'only' 164GB initially and not 232.8 as using the full 250GB (nominal) capacity will allow. 100GB for my C: drive and 64GB for the 'cache'. This step will further enhance the dependability of the SSD/system by greatly increasing the spare area (these specific recommendations will put it to around 38% 'spare' from the default of ~9% on the Intel 510 SSD).


To summarize, I have been using (on and off - ) eBoostr since 2009 and it mostly mimics this 'caching' philosophy (and it worked too).

The Z68 / IRST 10.5 combo will bring ReadyBoost, eBoostr, SuperFetch and Intel Turbo Memory to modern levels with real and immediate results (okay, not immediate... HDD access' will be noticeable after the first run. ).


So, what do the great minds on NBR think? Have I fully exploited the new tools we'll have available to us soon (in about two months, so May/June)?

Of course, the above assumes that SSD's will not be 'affordable' nor available in TB configurations for some time to come; hence the need to find a way to speed up our 2TB x 'N' data drives (where 'N' is the number of additional 2TB drives our system's can house).

Thanks for reading and for contributing any additional and pertinent information.

 

For the unfortunate Windows users, there is a thing called ZFS by Sun Micro(now Oracle) that does all this at the file system level with other goodies.

 

Dangit. I knew about the features, but now with these results it looks like I'm gonna have to wait until May to put my computer together. I've got everything except the motherboard sitting in the closet (including a 128GB SSD)... it can wait another couple months

 

chimpanzee,

Sure, I've no doubt ZFS is superior - but... at what (complete) cost?

And, will it run my progs?


Pitabred,

I felt exactly the same way. lol..

May/June is the time for most of my upgrades anyways (in the last 6/7 yrs) - but this is excruciating reading about all these goodies and not being able to even get a taste.

I'm sure this new system will be the same money (~$2,500/workstation), but it should also last me admirably for the next 3 or 4 yrs too (at an estimated triple my current productivity).

Can't wait.

 

I suppose it would depend on what exactly you're putting on your HDDs. I mean, I'm running a 240 GB SSD right now with a 640 GB platter, and all I have on my platter right now are music/graphic/video files, as well as downloads. Oh, and pagefile. Even with that, right now I'm using a bit less than 70 GB of the SSD (still have games and things I could install... if I get around to it), and only 18 GB of my platter (still need to get around to putting all of my music on this machine... probably reencoding to lossless this time since I have a lot more space than I did before... 100 GB was my previous (single!) HDD. ). I think the second to last and last paragraphs really sum it up; if you're already getting such a large SSD, why not "use it all" and put the files/programs you definitely want to be fast on there?

Really, to me this just kind of feels like a Momentus XT writ large; it's the same idea, just with the possibility for a larger cache that you provide yourself. So just like the Momentus XT, it's better than "just a regular" hard drive, but it doesn't quite match up to a full SSD/HDD combination. After all, even with 64 GB, it's cache, so depending on the caching algorithms, things will get moved in and out at certain rates. So really, if it's something you know you'll be using a lot for a period of time, why not just copy it all to the SSD, and then remove it later? That would give you greater benefits, albeit at increased management demands on you.

 

Why I said 'unfortunate'. There is already port of ZFS in linux,BSD and at one point Mac OS X(but dropped).

 

Judicator,

When I was actively using eBoostr, it competed and excelled in opening/closing programs, general responsiveness and almost anything I did with my VAIO (P8400) system compared to ... an Dell M6400 with a Samsung SSD. The only thing noticeable was bootup and shutdown times which were superior with the SSD based system.

The setup at that time was using an obscenely slow Lexar 16GB ExpressCard SSD (1+ hrs to copy ~14GB to it... that's slooooow) as the eBoostr's cache.

Of course, the read/access times is what made this work; but it was no 4x increase either (as the Z68 caching chipset offers).

While the numbers may look dismal (compared to simply using an top-of-the-line SSD), trust me, the responsiveness of the complete system (as described in the first post) should be through the roof.

Oh, and just to highlight how much eBoostr helped: the comparisons to the M6400 were with a 5400 RPM Scorpio Black installed (it only improved with the Hitachi 7K500 later on and lately the XT).


chimpanzee,

is there anything comparing ZFS more current than this:

See:
To timidly go where many have gone before: ZFS benchmarking


If it really is so superior, and with no downsides - all I can surmise is that a lot of money is changing hands to keep the laughingly slow NTFS file system around until SSD's become a 'must-have' component of any new system.

Is there a link where ZFS is compared objectively with NTFS and shows the good and the bad of both file systems? TIA.

 

unfortunately, ZFS is mainly used(and compared) in sever/SAN scenario.

No one is interested in comparing NTFS to these file systems.

On the linux scene, ZFS port seems to be much slower than modern linux file systems like ext4.

I mentioned ZFS only because these multi-level cache thing should really be done at the OS/file system level rather than what Intel(or seagate XT) is doing though for consumer usage, it may be good enough.

 

I still say that with your inital suggestion (buying a large SSD in the first place), you're better off just doing the regular SSD/HDD split. Again, this seems to me to be an extension of the Momentus XT, or the old Intel Turbo Memory, except instead of buying a module to attach to your motherboard, you buy a small SSD to do the same thing. And I do agree that eBoostr would be an excellent example of the kind of benefits you'd see using this sort of system.

Like I said, I'm mostly thinking that if you get a large SSD in the first place, you have enough room on it to put everything you want to be "fast" on it in the first place, which makes the whole cache thing less useful. If you were only buying a little 32 or 40 GB SSD, though, it makes more sense to use it as cache, because then you can get a greater variety of things cached onto the SSD, so to speak.

I'm not saying it's a bad idea in general, I'm just saying a lot of it probably depends on your usages.

 

I don't plan on updating my desktop until next summer (2012), so guess I can see how this all pans out. Love the idea in theory, just hope it works in reality.

 

The beauty of the Momentus XT is that it allows smaller laptops with a single HDD bay to have both speed and capacity without breaking the bank. The sum of the parts makes the Momentus XT an extremely well-rounded package, but none of the parts are particularly special by themselves.

Once you remove the single drive laptop limitation, hybrid solutions like the Momentus XT and SSD caching becomes much less interesting.

I'm perfectly happy with a SSD/HDD split - I don't really need super fast random access on my music and video files and with both Seagate and WD optimizing their storage drives for sequential reads/writes, moving data around isn't significantly slower than an SSD either.

 

If you do that, dunk it in oil. That one thermal paste thread has got me seriously considering an oil-submerged desktop, haha.


Anyway, on topic, I doubt I would do this supercache thing unless it's really, really good. I'd much prefer the simple SSD+HDD combo without the cache hassle.

 

htwingnut, Forge,

ha ha ha...

If it's good enough (for years) for enterprise usage, I think it will be good enough for us too.

There will be no hassle, it will just work and assuming a large SSD with not only extra capacity for the cache, but also more spare area for longetivity/dependability, there is no downside I can see at all.

Again, I agree that a single (large) SSD is ideal - but for me, all my data does not fit on a 250GB SSD right now, nor will it fit on a 512GB one either. This also doesn't take into consideration how filling it up to near capacity not only slows it down, but also makes it take a big hit in the longetivity dept. too.

~$400 for a 250GB SSD + 8TB or so of mechanical storage with this enterprise inspired 'caching' technique and the benefit/cost ratio far outweight running a single (much larger and exponentially more expensive) SSD and still having to waste precious work time simply moving/copying data back and forth to the 'faster' drive to complete it as fast as possible.

While a small SSD + large storage HDD may be the fastest setup if the HDD is used simply for archival storage - in an environment (such as mine) where a TB of new data can be generated in a day - an all SSD solution is still a pipe dream.

 

Well, again, I'm not entirely certain of how you "work", but would you be unable to fit all of your applications onto a 250 GB SSD? This would be the applications only, mind. The TB of new data you generate, do you really need fast access to them, or couldn't they just go onto the mechanical storage? You see, I'm agreeing with you in getting said large SSD and 8 TB of mechanical storage (as you put it), but I'm saying I'm not sure I see the point in "wasting" the space on the SSD to setup the cache, when I could just "use" it instead. Of course, I'm also manifestly not in your position (as regards work and all). I guess, in the end, you might just have to be a guinea pig and try it out for yourself. One of the nice things about the way we're talking about it is that from the sounds of it, you can easily do it both ways; you could set it up with the cache and see how it works out, and also try it out without the cache and see how much benefit you actually have. If the benefit isn't as great as you thought it might be, and you want the space that you were using as cache to put in a few other programs, then it's easy enough to get rid of the cache and do so.

 

Interesting thoughts. The reason I'm thinking about it is because of my current setup. My 160GB Intel is used for boot and some programs, while the Samsung holds media and games. (I'd love to put games on the SSD, but my Steam folder is a bit... bloated right now.) And so far I don't mind it at all. Most games load up just nicely for me, and for a 7200rpm HDD I tend to value space over speed.

But in any case I'll be watching closely. If it's as good as everyone is assuming it is, hey, who knows, I might do the supercache thing myself.