HEXUS.tv - guides :: Intel Centrino Pro - Robson demonstration

[Steve]

So what exactly can a gig of built-in Flash memory do for your notebook? Intel gave us a sneak peek - and you will be surprised how much extra performance Intel Turbo Memory (aka Robson) can provide.

http://www.hexus.tv/show.php?show=78

[funnelhead]

I would like to see a third laptop in that presentation, just to the right of the machine with "Turbo" Memory - a machine with a real 2gb of ram.

Obviously this is a way to bodge around not having enough ram in your system using cheap Flash memory as a paging drive. But if you consider yourself "pro" enough to buy centrino "pro" - shouldnt you know how much ram you need in your machine?

It's very clever marketing - sell the "pro" goods to people who don't know how computers work, but they really should rename it "centrino cheapskate oem"

f

[DR]

I would like to see a third laptop in that presentation, just to the right of the machine with "Turbo" Memory - a machine with a real 2gb of ram.

Obviously this is a way to bodge around not having enough ram in your system using cheap Flash memory as a paging drive. But if you consider yourself "pro" enough to buy centrino "pro" - shouldnt you know how much ram you need in your machine?

It's very clever marketing - sell the "pro" goods to people who don't know how computers work, but they really should rename it "centrino cheapskate oem"

f

Funnelhead,

I don't think you understand everything the Pro has in it - Mobile vPro and such too.

This is one feature, which is pretty negligable cost to the system but adds performance. There is a load of things which will make the Pro offering a pretty sweet setup

The other advantage with this is power saving - using flash instead of a HDD for getting data to the ram makes a big difference

Robson is designed to augment RAM not the HDD

It will increase performance even if you have 2GB ram

Do not confuse this with TurboCache which was a cheap bodge in essence

So a meaningful comparison in the way that you suggest would means would be to have a 3rd laptop there with a multi-way RAID0 array of 15Krpm/16MB cache drives... good luck with finding such a laptop

[this_is_gav]

Good marketing, but it's like ReadyBoost - it's not a replacement for memory, it's a means of allowing Vista to play with things more efficiently.

It would make a difference to a machine with 2GB RAM, and quite probably 4GB, though the effect will become much more diminished the more RAM you have.

My only worry is that it will lead to companies such as Adobe, famed for their bunging half of their Reader into memory on startup, just abusing it and loading pre-caching all-sorts. It's a measure to alleviate that problem, not to encourage more companies to do so. I'd like the choice please, OpenOffice like.

Sorry, got a bit side-tracked there, and moved onto RB rather than Turbo Memory.

[funnelhead]

ok i can see how having a flash drive to load common apps into at startup would reduce subsequent hard drive access and prolong battery life. That is cool, though i do worry about that being abused as gav does by lazy or unscrupulous companies (cough, adobe)

As for using this to increase performance - well i thought the machine was just using the flashmem to hold the paging file rather than the HD - if this is not the case, then what does it do?

Always happy to learn more!

f

[Steve]

As for using this to increase performance - well i thought the machine was just using the flashmem to hold the paging file rather than the HD - if this is not the case, then what does it do?

That's ReadyBoost, not Robson

[funnelhead]

OK I have educated myself somewhat. Here are some quotes from the intel website as to what robson is:

Here is the non technical blurb that tells you what robson does

"Over the past 10 years, there have been significant improvements in HDD system interface speeds—now available at up to 3.0 GB/s with SATA—and these speeds allow large files stored on sequential sectors on the hard drive platters to be transferred very quickly. However, the vast majority of real-world HDD data transfers is actually relatively small and tends to be scattered all over the HDD platters. Getting these small bits of data off the HDD and up to the system is limited by the time required for the HDD platter to spin to the right position and the read head to move over that position. Testing has shown that these mechanical latencies make up approximately 95 percent of the time required to read small files in real-world situations. For example, Intel testing has shown that within a small transfer requiring 5.4 milliseconds, about 5.13 milliseconds are consumed by mechanical latency. Essentially, systems are forced to wait for HDDs to catch up to processor performance.

Additionally, all the mechanical work required to find the requested small files on HDDs is a major source of power drain, particularly when multitasking"


here is the more informative bit that tells you what robson IS

Robson Technology Architecture
Robson technology consists of four components:

* Robson driver. This is the heart of Robson technology—the Robson driver interfaces to Microsoft Vista’s ReadyDrive and ReadyBoost technologies, interfaces to the Intel® Matrix Storage Manager driver, and provides all the control functions necessary for managing NAND flash in a caching role.
* Intel® NAND flash memory. Robson cards are powered by Intel NAND memory that offers 40 Mb/s reads and 12-15 Mb/s writes.
* Diamond Lake ASIC controller. This controller translates the system’s PCI Express signals to the Intel NAND memory and manages the retirement of Flash memory blocks as needed.
* OROM BIOS. Robson technology includes BIOS-level option ROM (OROM) code which manages NAND access before the OS and drivers load. The OROM also controls separation handling, critical to maintaining data integrity.

----------------------------------------------------------

So basically robson doesnt just take the swapfile, it acts as an intelligent cache redirecting hard drive saves onto itself instead.....which although slower than HD for large files, is actually quicker when dealing with small files..

am i close?

f

[James Morris]

Yes, that's not a bad summation. The access time for Flash is obviously many times faster than for a hard disk, although raw throughput of contiguous data will probably be slower.

[robg1701]

I think the idea is great, helping fix the HD problem with cheap flash...the implementation obviously works wonderfully...but as metioned already, they did sort of skew this demonstration by usign a laptop which didnt have enough memory to begin with. Vista+Photoshop with 1GB ? Please

Id like to see a 1GB laptop with a 1GB Readyboost pen drive in that comparison (i expect itd be far slower than Robson, but still faster than the poor crippled laptop)....but more importantly id also like to to see a 2GB with and without robson comparison please, not jsut the 1GB with and without, to see how much better it is really going to be for me (i still expect it would be a very sizable lead, since HD performance sucks on desktops let alone laptops), as a user who actualyl configures his machiens with enough memory for what i do

[James Morris]

I bet a 2GB memory notebook with the top 4GB Robson implementation would still rock. I'm sure you wouldn't see anywhere near the performance boost, but the hard disk could spin down for a lot of the time. Indeed, with the OS cached there it might not need to spin up at all to boot the system. That's bound to save you loads of battery minutes. I'd still like to see this proven myself, though...

[Steve]

Indeed, with the OS cached there it might not need to spin up at all to boot the system. That's bound to save you loads of battery minutes.

Of course, device initialisation would probably necessitate a spin-up.

[James Morris]

Yes, that's true. Unless there's some way round that. Vista does have a rather different structure for device drivers, after all.

[Steve]

If Vista itself can avoid it, then there's the BIOS and drive's firmware to contend with.

Supposing the BIOS tries to detect the drive but the drive's clever enough to respond without spinning up... that might work, but I think for SMART purposes it'd have to undergo a spin-up cycle.