Robson

Is it here to stay or is it just a passing fad? Intel boasts that it will become a primary part of the computer but is it really necessary and will AMD come up with something to compare to it? or wait do they have something already? Oh and it has just recently been changed to "intel turbo memory" and no longer Robson. So I've heard.

 

Currently, I think it's a bit of a fad, you don't really see a big boost in performance unless you're using like 1 gig or less of RAM instead of the 2 gig that you should use.

Also unless Windows allows for you to use more than 1 ready-drive devices robson will be replaced by hybrid drives I think.

 

I think it will be replaced by solid state additions to harddrives, but the idea and concept is here to stay.

 

Isnt AMD also doing something similar in their next mobile plattform?

I think it is here to stay. SSD is just to slow as a primary disk, slow as in tranfer speed. Flash together with a normal disk have the greatest potential because you get the fast access time of flash together with the pure mb/s of a magnetic disk.

I dont understand how people say it is only for computers with 1GB of ram. That was readyboost with USB sticks and even those reviews didnt take into account alot of things when they tested. Readydrive and Turbo Memory (Robson) should be alot faster then Readyboost with USB sticks because 1. it can speed up start up (boot time). 2. The Robson is faster then a USB stick.

And dont bring up that month old anandtech review... they used a whitebook with a broken bios.

 

HP has decided Robson is not worth it. See this thread:

http://forum.notebookreview.com/showthread.php?t=127901

 

It's here to stay until SSDs I believe. 1. There's too much backing it and 2. I'm pretty sure there are some severe implementation issues with it currently. Slower performance makes absolutely no sense and reeks of implementation errors.

 

Can someone explain to me the advantages of robson over a usb flash drive with vista readyboost.

 

Robson (Turbo Memory) uses a faster flash technology and the programming can be more sophisticated because it'll always be available to the operating system.

 

And key is the fact that it can function before the OS is loaded.

 

RAM works as soon as the machine is turned on as well

At the end of the day IMHO nothing is as fast as RAM (except the CPU's cache and the CPU's registers). Robson, USB drives etc none are as fast as main memory to me this "turbo memory" etc thing is just a fad and will pass! This technology is just a way to make up for lack of memory and to me there is nothing like the original get 2GB if you are running vista and you will be fine...

 

Check232 is 100% right when saying that the main benefit of Robson (Readydrive) is that it functions before the OS is loaded. USB sticks (Readyboost) only start working once the OS has loaded. Why you compare Robson to Ram I dont understand...

Turbo Memory doesnt replace Ram... It was not made to replace ram... It is made to overcome the hard disk bottle neck. When booting or loading programs most of the time is wasted by the hard disk fetching all the files need from all over the hard disk loading it into ram. Robson speeds this up because files from Robson, especially spread out small files, can be loaded into ram faster.

 

It was probably done to avoid more complaints for vista not running smoothly. Probably partly subsidised by microsoft for now. Heheh....I bet the new mac OS probably doesnt need them despite having a better interface.

 

USB drives are also available at boot. Most, if not all, m/b's these days have the option of booting directly from removable storage such as a flash drive. With a few changes they could act exactly like Robson? Plus 266x flash drives are coming soon and I bet 300x isn't too far behind since they have CF cards running at 300x already.

So unless this Robson flash memory is much more than 300x, it feels like Intel is just trying to throw around some patented technology in hopes of earning more cash. Whether it's useful technology or not.

If it is a hell of a lot faster than the fastest flash drive and upgradable, I think it has a strong chance of succeeding. Maybe Microsoft will allow Windows XP/Vista to place the actual pagefile on the flash memory if it's fast enough.

 

Robson is started by the bios. The bios detects it and it is pretty much ready to use. For a usb stick Vista needs to detect the stick, install the right driver, check if it has ready boost data on it. Check if it is the right data, not old data or from another computer. I think all this would make the use of a USB stick limited at boot because the time you save is contrasted by the time it takes vista to detect a usb device and check it. Robson is constant and Vista can trust it so doesnt need to do checking or detecting.

 

wave, the question is what is it can you do significantly better with Robson than with RAM? In practical terms for somebody buying hardware is whether they should buy 1GB of Robson memory or add 1GB to their RAM?

First of all Robson works only with Vista right now. It does not work with XP. It does not work with Linux (whether it ever will is unclear). So unless you're planing to run Vista, it gives you nothing. (I, for one, am not planing to run Vista until I see a serious advantage other than eye candy.)

Let's deal with the two ends of the memory spectrum because those are special cases:

1. If the machine has all the RAM it can take, then yes, adding Robson memory will give you something that RAM can't but just because you can't add more RAM.

2. What if a machine is RAM-starved? In this case there is no question that the first thing to do would be to add RAM before even thinking of Robson. What RAM-starved means really depends on which OS you're using and which applications you're running.

For all the cases in between, where the machine does not have its RAM maxed out and it is not RAM-starved. I still think RAM is the best option for performance... although price-wise it really depends on how many RAM modules needs to be bought to bring an additional 1GB into the machine. Let's go over the advantages of Robson:

1. Booting rapidly. Hmm... ok maybe. Except that I usually suspend-to-ram rather than shutdown my machine. It takes maybe a second to come back. Granted, once in a while, I'll have to boot up and then there's an advantage that Robson will (probably) give me that RAM can't. But what about battery life? If the machine has been shutdown, it does not use power but a suspended machine does use power. What about that? Not a practical issue. I've left my laptop suspended-to-ram for several hours without experiencing any significant power drain. (There was a power drain, I'm sure, but nothing that made any difference for me.)

2. Intelligent disk cache. Again, RAM is your friend. The most idiotic alternative scenario would be to add 1GB of RAM and ask the OS to treat it (for HD accesses) in the same way it would treat a Robson cache. I mean "idiotic" in the sense that this alternative requires pretty much no thinking on the part of the engineers implementing it. I'm sure there would be ways to do this better by modifying the concept of Robson and bringing it more in line with traditional RAM caching. For instance, do you need to have 1GB dedicated to the disk cache at all times? Or what if you need 1.5GB temporarily. With Robson the size is fixed. In RAM you can allocate space as you need it.

As far as caching goes, the one thing that Robson can *possibly* do better than RAM is caching writes to the disk. RAM can cache writes just as well and faster. The problem is what happens if the OS is ceases to function in an uncontrolled way (e.g. no power, crash, freeze). In theory it would be possible to recover the last write commands that were in the Robson cache when the OS comes back up. With RAM, that's not possible. It's not clear however that this would be a significant advantage *in practice*.

As usual, there's a difference between the theoretical advantages of a technology and the advantages in practice. My impression is that an additional 1GB of RAM in practice would be more advantageous than Robson. Of course you won't find Intel or Microsoft saying that because they are both invested in that technology. There's more to say but that's about as much as I want to write right now.

 

Ofcourse having more ram will have a greater benefit then Robson. I never disputed that, but it doesnt mean that Robson is bad.

Robson has one purpuse: speeding up the time it takes normal magnetic hard disk to load data into ram. The speed up is fastest when there are many small files because flash has a seek time close to 0 and higher random access speed then a hard disk. It does not replace ram. And no matter how much ram you have you need to put data into the ram somehow. Atleast if you reboot the ram will be empty and the data has to come from somewhere. Ofcourse if you only suspend to Ram all the time and got 4GB already I dont see much benefit to Robson for you either. I have 2Gb of ram and use a 2GB SD card with Readyboost. The applications that I use every day do load alot faster now. The SD card I use benchmarked at about 12mb/s. Intel says Robson can do about 35mb/s so it should make the effect even more noticable. But it will always be hard to benchmark because it will be a few milliseconds here and half a second there but never 10seconds faster to start Photoshop. But all the small gains do add up and make using a computer nicer.

As a side effect Robson uses less power then having to load from hard disk, giving you better battery life because the hard disk can stay off longer.

Also Robson is alot cheaper then ram. Lenovo charges 50$ for 1GB Robson and 100$ for upgrading from 1GB to 2GB of ram (I know Ram cost a lot less in the open market but just to give an idea because Lenovo is over charging for both). The offical retail price for Germany for 1gb Turbo Memory seems to be €29,99, 1GB of Ram costs between 35-40. My guess is that it will cost less then €20 in europe and less then $20 in the USA once it is widely available.

Robson will be a cheap way to speed up low and mid level laptops. For high end systems it might be nice to have but not much speed gained.

 

wave, I've read your post but my opinion has not changed. Giving the reasons why would take more time than I care to spend right at this moment. I'll try to spend some time later on this. In the meantime, I found that AnandTech tried Robson and found it deficient. Here's the link:

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=2985&p=1

I find it also revealing that HP is not planing to implement that technology and that there are manufacturers who, although they are not saying anything about the value of Robson explicitly, are selling Santa Rosa laptops without the possibility for Robson.

 

As far as I know that the whitebook anandtech used had a defective bios. I read reviews in two magazins (printed ones so no link sorry) both testing the same notebook and both said that they cant comment about Robson because the bios they got was broken.

 

I think some of you are missing the point of Robson - its not to supplement Ram, but its to reduce the seek and load time from the hard drive. Overall you'll get a bigger system boost by increasing ram. Robson is supposed to give an incremental boost to disk intensive activities - booting the OS, loading common programs, etc.

I think the jury is still out on Robson. Like other new technologies it may take a few software revisions until we can accurately assess its value. I chose to bet for it and included the module for $40. I can understand others betting against it and not including it in a purchase at this time.

 

Currently, it's nothing to brag about.There might be some good improvements both on the hardware and the software sides.

 

It is the next logical step in combating the large discrepancy between CPU speed and data access speed.

We are progressing towards the day that computers will do away with hard drives of any kind all together. And I don't mean the embedded systems types with no storage and 256Kb of RAM. I'm talking about on-die storage, as much as 100GB directly next to the CPU. The whole hierarchy of cache-RAM-HDD is designed to combat costs and technological limitations. When those two obstacles are overcome in time, the hierarchy will become obsolete in all but principle.

 

From the article:

Let me get this straight. Intel themselves sent this unit to Anandtech and when Anandtech gets statistics that Intel does not like Intel blames it on the BIOS??? Am I the only one finding it is just an excuse to get out of embarrassment? ("The dog ate my homework!") Note that the article was written on May 10th and it appears that Intel still has not explained the details of why Turbo Memory did not work right on that system ("the BIOS was faulty" is a vague answer) and nor have they sent another system to Anandtech that would demonstrate the benefits they claim.

 

Most motherboard bios' can be configured to boot from a USB flash drive without an existing O/S. With a few bios tweaks couldn't it be used identically the same way that Robson is being implemented. The only problem I see is that USB 2.0 has a transfer speed limit.

 

Does anyone know whether there are any plans for the Robson use in Linux.
If they can use it for a part of the swap memory I think there will be a good performance gain.

 

I've searched the Linux kernel mailing list (LKML) and found no discussion *at all* about Robson. (I also searched for Turbo Memory, found nothing). I'm not surprised given that Robson is a bad idea.

This being said, I've heard that Intel may be providing the drivers for it since they've been pretty good in the past at supporting Linux. For instance, they are going to have drivers for their latest wireless chipset (the one that supports n-draft) this summer. Could be faster but that's pretty fast. (Much better than ATI for instance.)

Another thing to point out. I've never tried it but you can use a flash drive as swap space. I have a fairly good idea of how I would do it manually. I do not know of an automatic way to do it.

 

That's precisely the point. People should be educated about the benefits of suspending to RAM instead of being sold a solution for a problem that does not exist.

My 5400RPM hard disk which I currently use through a USB interface does 28.67MB/s. I'm using it through a SATA-to-USB interface because I'm transitioning from PATA to SATA and my old laptop only has a PATA interface. Once I put it somewhere where it can be directly plugged into a SATA interface, the speed should be even faster. Robson does not look so great with stats like these.

There's also a difference between the scenario you are talking about and the choice between Robson or RAM. What you are talking talking about is the opportunity to put to use a flash card that would otherwise probably go unused. It's there. You've paid for it already. Well, ok, why not put it to use? You are making a choice between using or not using a facility you already own. (By the way, by using your card, you are using ReadyBoost, etc. but this is *not* Robson.)

The situation is different with Robson. It is a mini PCIe module. The only way to get it right now is when buying a new laptop. (I'm disregarding the choice of adding it later because I think this is a rare situation right now since most laptop owners do not have that option at this moment.) It is *in that situation* that it makes sense to think about additional RAM vs adding Robson. And what I'm saying is that *in that situation* if performance is the main concern, then adding more RAM is a better choice overall. Somebody who is planing to buy a laptop with 1GB or less should put money on memory to bring it to 2GB instead of buying the Robson module. The price differential is so small as to be insignificant. Somebody who is planing to buy a laptop with 2GB should think about adding an additional 1GB of RAM instead of buying Robson. Why? Because RAM can do everything Robson does and do it more efficiently than Robson *except* for persistence between reboots and the fact that RAM draws more power than a flash drive (but we are talking about very little power here).

Additional RAM gives you that benefit albeit to a lesser degree because RAM needs refreshing. Here's the deal: whatever caching can be done with Robson, the *very same caching* can be done in RAM which is faster than Robson.

This is about the only criterion regarding which adding 1GB of RAM is not as good as adding 1GB of Robson. However the prices of RAM are falling down quickly right now. 2GB of RAM (2 DDR2 SODIMS 667) is about $70 right now on the open market although I've seen it for about $45 on sale. 1GB of Robson costs about $40 at Powernotebooks.com. Note: I don't think it is possible to buy Robson on the open market right now. I've priced a jump from 1GB to 3GB to about $120-150, ok that's a significant difference. Still, if performance is key, there is nothing like RAM for quick caching.

 

That is when you test it with large files and write or read in sequance. Randam reads and writes will be much slower because you need to position the hard disk head many times. Try copying many 1-2kb pictuer files (like maybe your web browers cache) from one disk to another. You will see it is much slower. This random read/write speed is much faster with flash and ram. That is the only advantige that flash has compared to the harddisk. Well plus that it uses less power.

The only advantige flash has over RAM is that it doesnt lose its data when there is no power. And not everybody can suspend to ram all the time. I run on battery alot and I need to shut down or atleast suspend to disk. If you use your laptop as always being pluged in then you are fine with suspend to ram.

edit:** You can buy Intel Turbo Memory in germany in online shops. I am sure you can buy it in the USA soon too. And even if not other companies will build something similar for mini pci-e or maybe even the express card slot.

 

I think you are right. The software I used was doing sequential reads and writes. Still, it does not change my contention that an additional 1GB of RAM gives better performance than an additional 1GB of Robson.

Ok, here's the scenario. I'm on the bus traveling between two cities. My battery is old and crappy and won't keep a charge more than 1.5 hours. So I work and work until the battery light on my laptop starts flashing RED. The OS tells me to plug in or shut down right away if I don't want to lose my work. What do I do? Do I plug in? I can't because I'm on a bus. Do I shut down? Hell no! I don't want to have to go through a shutdown and reboot. I suspend to RAM. I'm still running on battery and the laptop won't be plugged back in until at least 2 hours later. When I'm able to plug in, 2 hours or more later, the power is still present. The light is still flashing RED. Everything is fine. I have not lost anything. I do this *very* frequently.

Suspending to RAM takes very very little power. So little power that it makes no sense to worry about it.

 

Lemur, I know you are putting alot of fact's forward.. but the big picture of telling people to just "get another gig of ram" is that no matter how much ram you give to an MS operating system, there must be a paging file/swap space allocation, and windows will write to it even if you "disable" it. The only thing disabling the swap file does is make some applications very unstable since they rely on the swap being there, and the OS will cache things like DLL's to it anyways.

The point of Robson is to essentially remove the need for the swap/paging file for user applications, while at the same time they have come up with other uses such as suspending to it. The other nice thing about robson is that through software it learns what needs to stay going to help your system out. (The key point is software controls this, which means applications like Photoshop can be creative with their clip-boards and whatnot and store to the Robson flash leaving your system memory free to shuttle data to and from the CPU for calculations, leave more of the faster resources available to keep the user experience snappy. The swap file doesn't offer this level of control).

Ram isn't a solution that just sits around waiting for you to call a bit of code again, once the CPU no longer has need for it the registers are flagged "blank" so more can write to it.

Ram and Robson are not the same technology, they are not meant to be used the same, and their use doesn't overlap. Robson is a hardware upgrade for Swap and Paging files if anything, which is why it is important that it can suspend your information state. Since this is the case, why not let users suspend to it anyways, and free up your system ram so you can wake up the computer faster with more available memory?

 

Sorry, I can't parse what you are saying here. The only thing that comes to mind is that RAM is in fact used by the operating system to buffer files so that accessing files multiple times do not result in multiple accesses to the hard disk.

Ok, the reply "RAM and Robson are not the same technology" has gotta be the weakest argument that can be made. First of all, both are solid state memory. The difference between the two is that RAM must be refreshed because when we say RAM in this context we're actually talking about DRAM. The corollary to refresh is that RAM requires constant power in order to avoid forgetting its contents. On the other hand, RAM is a much faster memory than flash. But both are solid state memory. When we're talking difference between RAM and flash, that's a difference in degree, not in kind.

Now consider this. There is a fundamental difference between RAM and disks (floppy, HD, CDROM, whatever). It's a difference between a memory made of integrated circuits vs a memory made of spinning platters and moving heads. We're talking about an essential difference in operation: a difference in kind rather than degree. But is this fundamental difference an obstacle to use RAM as a disk? Not at all. Ever heard of RAM-disks? With RAM-disks, we have RAM used to perform the function of a entirely different kind of technology.

Given that the fundamental chasm between RAM and disks has been bridged with RAM-disks, I'm not impressed by argument that "RAM and Robson are not the same technology". Whatever caching can be done in Robson can be done in RAM. To a great extent, this caching is already being done by Windows. Heck, caching is really at the level of "OS Design 101" and I'm sure Microsoft is not so stupid as having overlooked that. If there is any algorithm used in Robson that Windows does not already use to manage its RAM, then the pressure should be put on Microsoft to implement it because there is no inherent technological hurdle that prevents caching things in RAM the same way it would be cached in Robson.

The only areas in which RAM cannot compete are persistence across reboots and power consumption. That's it, that's all.

 

Yes Ram is used to buffer files for the CPU to process, however things like DLL's and some system drivers etc which are called regularily do not stay in ram, they get shuttled to the swap file to make more system memory available for current tasks. The will get refreshed into memory as needed.

Your argument isn't about the same things I'm saying, just like you did in the other Robson thread. You are talking about physical implimentations, and I'm talking about the way the operating system literally uses the technology, making most of your reply meaningless.

Did you skip my entire section on the flexibility for software developers to use Robson for "needed but less important" short term storage? You can not do this with system memory unless you develop in lower level languages (As in unmanaged C++), and the current trend is for run-time languages (.NET, Java, Python/Ruby/Other dynamic interpreted languages). Again, using my example for photoshop. Imagine your clipboard has 400mb of "undo" data... This can be pushed out of the main system memory which will keep the application more responsive (as well as the rest of the operating system), while still allowing nearly instant recollection of the data since your data is in flash rather then a paging file. Currently Photoshop uses the swap for these sort of tasks for just this reason. If you leave the OS to decide as you use your application the system memory will bloat up with your undo lists until you overflow the entire system into your swap file.

Short answer is Robson is faster then a paging system, and Robson puts the developers a bit more in charge of their memory model on higher level languages. The same languages that make up 80% of enterprise level applications, and 60% desktop applications, and the number is growing.

Vista's implementation is only ONE way to use it. It will take time for more software to show up on the market.

The other side of this discussion that you seem to avoid completely is the fact that Robson is an enabling technology. It may not stay, but it sure is required in order to get more companies developing smaller die manufacturing plants, which enable larger capacity at a cheaper price which is needed to get to the real goal.. Flash drive based HDD's. Intel just recently backed a new huge startup company in Israel with the sole job of bringing this technology mainstream. They are ~50% owners, and provided nearly 2000 of the employee's as well as a fabrication plant.

Edit: As for your comment about RAMDISK's, yes, I have tested several from many companies and they just don't pan out for consumer use. I've used/sold both drives from LaCie as well as Gigabyte and Asus. They are at the same hurdle as the flash based HDD's. The difference being, Intel is not a memory company and doesn't want into that rat nest, going this way they can drive the technology without requiring the industry since they are already a leader in flash technology.

 

pyro9219, please tell me specifically what RAM-disk you have used.

 

I have used "soft" implementations of this with drivers for windows 2000 and various other OS's. But I was referring to physical products, some of which I sold years ago.

Gigabyte i-Ram is the only one I can report on.
Review : http://techreport.com/reviews/2006q1/gigabyte-iram/index.x?pg=1
Sale : http://www.amazon.com/Gigabyte-GC-R...7?ie=UTF8&s=electronics&qid=1182552178&sr=8-1

LaCie never made it retail (was back in 2001-2002) but I sold them to test sites for evaluation, and they were an external solution that connected via firewire. I think this is where they failed, since at the time firewire was basically only a Mac feature.

The Asus product is probably vapor ware since gigabyte released their i-Ram.

 

Ok, when I'm talking about RAM-disks, I'm only talking about software solutions. My point was that you can use RAM on a SODIM module as a disk. The gap between the two has been bridged.

I find nothing in your last substantial reply that has any bearing on the issue at hand.

 

I fail to see your point, other then perhaps there is a question being asked - Why Robson when we have had software based RamDisk's since the old days of Mac's(in the mid 90's) as well as windows 98 through XP/2003 (no idea if the drivers exist for vista)?

All I can think is that back in the early days when this was exciting technology ram wasn't a nickel per mb like it is today and the desire to go that way was lost.

Or perhaps what I said earlier about Intel, being a technology leader, didn't want to join the DRAM market, but the new wave of devices using flash did seem good so they went that way. Now we are looking at today where Intel is a flash market leader, and they are leverage where their own research and abilities exist to provide this as a solution.

It is something to think about though... perhaps as a software developer I could license that technology and have my software install a ramdisk, then using that as my own personal cache system to avoid some of the crap the O/S does with the memory stack. Any language could also right to it since it can be mapped as either a folder or a drive letter... Very cool!

The only flop I see with this is there are a number of know bugs where the OS's will fight over the memory addresses and can corrupt data in the system memory... Maybe more research should be done in this area?

EDIT: Just remembered I've done projects that use SQLITE http://www.sqlite.org/ and this can be implemented as a memory resident database if wanted... could have some fun with both idea's combined.

 

Well, seeing as it doesn't work, I wouldn't even call it a fad. I say wait for Robson 2.0, if anything. With HDDs with a 3GB/s transfer rate, a working Robson module's help would be minimal.

 

Anandtech is only one place that has showcased the technology. there are video's online of the technology working just fine.

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3009 this article show's functionality working.

 

The link is broken.

 

No, they were down for maintenance.

 

They need to give better error messages. When I followed the link, IT got a 404.

 

Anyway, it is time to point out what I've pointed out already here. The inner quotes in what follows are from the very same Anandtech article:

 

So, I haven't searched the entire thread here, but what I understand from Lemur's post is that you can use USB flash drives EXACTLY LIKE turboboost? I.e. they have the functionality of RAM?

 

Romanian, I don't understand what you're trying to say. I'm saying that RAM can perform all the functions that Turbo Memory can perform except for two things: lesser power consumption and persistence across reboots.

 

Wow. I did read the whole thing.

So what's the consensus? I'm seeing that Turbo Boost really doesn't make noticable difference.

If I'm getting Vista Ultimate, 2.0GHz processor, 2GB of RAM, is not getting turbo memory gonna be noticably slower?

It also makes a difference because without Turbo Memory, I'll just get the 512 and buy 2 GB 3rd party, saving money over buying the 2GB from Lenovo PLUS $50 for the Turbo Memory.

I was leaning towards getting it. Now I'm wondering if it is just a waste. My computer stays on almost constantly, or I put it in sleep/hibernate. I rarely reboot. And I do pretty basic stuff with it (no games, not much heavy video or photo editing).

This is really my last hurdle to determine. Seems there is at least SOME benefit. The question is will I really notice it?