Intel touts user-defined app cache Vista speed boost tech
Turbo Memory revamped
Will Intel's revamped Turbo Memory software - the code that comes with its Centrino-oriented Flash cache modules - make good on the technology's promise to accelerate Windows Vista application load times?
The chip giant certainly hopes so, but its solution - to let users, rather than Windows, choose which apps are cached doesn't look too friendly for the mass of non-technical users.
Intel's new Turbo Memory Dashboard
Intel has been offerings Turbo Memory for 18 months now. Before then, the technology was better known by its codename, 'Robson'. The modules sit inside a PC and tie into Vista's ReadyBoost and ReadyDrive features, which use Flash stores to cache frequently uses apps and data, even when that storage is connected to a PC's USB chain.
The idea is that since accessing data on a Flash chip is almost as fast as getting it from memory, Flash can be used as an ad hoc addition to a Vista PC's RAM. Yes, USB's bus throughput is much slower than a hard drive's interconnect speed, but this is outweighed by much tinier time it takes to find the data the system wants.
Or so the theory goes.
The jury's still out. HP last year said it didn't see much value in the technology, and reviews of ReadyBoost have shown that while its underlying data management code - aka SuperFecth - delivers a benefit, it's small one.
Vista's SuperFetch system monitors users' behaviour so it can cache the right apps in ReadyBoost memory, if it's present.
Intel's new Turbo Cache Dashboard app, part of the Centrino 2 era Robson package, lets users themselves choose which apps they want to cache, a technique it calls "user pinning".
'Pin' your favourite apps
Pinning an app essentially tells Dashboard to put it in the Turbo Cache memory and to notify Windows it's there. The result is that you can be sure the apps you really do use regualarly are cached for quick(er) access.
And it's not just apps - the Dashboard will let you cache folders full of files too.
We look forward to trying this out to see if it does indeed make a difference. But the question remains, how many ordinary users will bother to drill this far down into their system to see if there is any appreciable improvement? And is it likely to encourage notebook vendors to begin including Turbo Memory - many of them don't - beyond their top-of-the-range machines?
COMMENTS
Write back?
Wouldn't it be better to use the Flash as a write-back cache to keep applications from waiting for writes to commit to disk?
Waste of time
Agree with the above comments, this is a pointless bodge - and an incredibly ugly-skinned one, judging by that screenshot.
ReadyBoost hasn't increased speeds much, and no-one wants the extra management overhead of having to assign applications to a separate drive manually. If I wanted to do that I'd just copy the program files over myself.
Flash-cached HDs are more of the same. These features are an exercise in overcomplication, from engineers who have nothing better to do with their time and desperately need to give their product 'features' even if they're of questionable usefulness.
In the long run, the entire system disc of a typical PC is going to be an SSD anyway.
Feels more like a bodge than a true solution.
This falls down to the whole ancient argument of a Disk Based OS vs ROM based OS. To be as easily maintainable and updateable, the MS OS is loaded and launched from disk. As a direct result, the OS fires up, loads drivers, loads services, launches the (heavy) Vista GUI, allows users to log in, then performs a ton of log-in activities involved in setting up the environment and pre-caching things.
I say this whole method can go out of the Window(s) in favour of a whole new paradigm. I couldn't tell you what that is, but as a start, you could perform the equivalent of a base heavily configured image, like how a hibernated PC comes out of hibernation, why not simply allow Windows to configure itself a fully configured memory 'image' after Windows has installed, then every following reboot, load it straight into memory? Doing so would, at the least, save Windows from needing to continually address and load tens of thousands of individual files from disk upon reboot, whilst referring to the registry and other files for the current configuration info. I have no doubt this would improve reboot speeds. Furthermore a similar scheme could take place for Application initialisation to speed that up.
Maintenance of Windows itself would suffer as a result of this, as this 'base reboot image' would need updating every so often as Windows Updates occur, or as users install apps, or as Windows indexes your disks. But efficiencies in imaging techniques could alleviate this.
I have no doubt Windows performs perceivably slowly specifically because the OS has grown out of this whole 10,000 x 30Kb file mindset that Windows is made up of - take a quick ganders in your Windows directory (System32 is a killer, my sys has over 2300 files in system32 alone).
.
If this idea is worth anything, I offer it for free just because one day I'd like a Windows system to fire up Office Outlook within the time it takes to sneeze. Somebody do it for me!?
Re: Wrong question
...or even "Why not spend the same money on more RAM to begin with?". This will make all sorts of things go faster, with no configuration at all.
RAM isn't that pricey these days, and your OS ought to know already how to use it to best effect. To be cost effective, this Turbo stuff is going to have to be almost free and even then it is only of interest to users who have a regular habit of discarding large applications just before they need to use them.
On the other hand, if Intel *can* make a slightly slower but significantly larger form of volatile storage, there the possibility of rejigging the whole cache hierarchy. Instead of several gigs of ordinary RAM, you'd have a few hundred megs of very fast stuff and quite a lot of gigs of this rather slow stuff.
The technology probably has a future, but this particular productisation looks like a flop.
Wrong question
"But the question remains, how many ordinary users will bother to drill this far down into their system to see if there is any appreciable improvement?"
The real question is: Who is going to upgrade to an operating system where this kind of crap is necessary to make your system's speed anywhere near usable?
