Gigabyte i-RAM 2 Information

[mikeblas]

Well,
At the current state of mobo/cpu development - the mobo/cpu is the reason that we can't just throw more RAM at our systems. Fix them then we'll be able to pass the 4GB barrier on our PCs.
That plus cost of the components.

Cheaper than what? The memory cost is about the same or slightly cheaper per stick; the i-RAM itself costs as much as a half-decent desktop motherboard.

Maybe you missed this:

I think it's been said before in this thread, but I'll say it again: there's nothing wrong with peripherals like the i-RAM if they're applied sensibly. But they're not cure-alls, and I don't think they make much sense for general-purpose computing until you've maxed-out the physical memory you can put into your system.

IOW, the point isn't building a machine that has 16 gigs of memory. The point is maxing-out the memory on the machine you already have. If you want to build an i-RAM-based machine to play with booting Windows from a RAM drive, that's great; be my guest, and I hope you have fun with it and report your results here.

On the other hand, if you're expecting general-purpose performance improvements by installing one in a system that doesn't have the maximum amount of physical memory already installed, I think you're approaching the problem the wrong way.

 

[rodsfree]

Cheaper than what? The memory cost is about the same or slightly cheaper per stick; the i-RAM itself costs as much as a half-decent desktop motherboard.

The "Cost of components" thing was about ECC regeristered DDR. Which is required by the current mobo's that can pass the 4GB limit.

Let's look a little closer at this. The cheapest server/workstation mobo I've seen that can go over 4GB is around $250-350. Each stick of ECC Reg DDR for it cost about 2.5 times what regular DDR does. And that's not counting the price premium for socket 940 Opterons or Xeons. Getting a little pricey to play in that sandbox.

Maybe you missed this:



IOW, the point isn't building a machine that has 16 gigs of memory. The point is maxing-out the memory on the machine you already have. If you want to build an i-RAM-based machine to play with booting Windows from a RAM drive, that's great; be my guest, and I hope you have fun with it and report your results here.

On the other hand, if you're expecting general-purpose performance improvements by installing one in a system that doesn't have the maximum amount of physical memory already installed, I think you're approaching the problem the wrong way.

Nope, I didn't miss it. And I agree that you should max out a system's RAM before you go for an iRAM. Unless the iRAM can meet certain conditions.

Maybe you missed this.

Interesting, concept though. I still think that the i-RAM with 4 - 8GB of really cheap failed testing DDR PC2100 ram would be the perfect pagefile drive. Up it to SATA 300 speeds and it'd be even better. Cause it would be the only drive on the market that could hit the rated speed of the interface. That would make it reasonable to do the standby to ram for instant on operation of windows.

But I couldn't see spending $1000.00 on one.

Newegg has 1GB Rosewill DDR 266 for $68.99 right now. If I could buy the chips that failed the binning process for $30.00. 4 X 30 = $120 + the cost of the iRAM, which really sould come down to like $75.00.

So, I could see $200.00 for a 4GB 'iRAM, already populated.
They could sell them at that point, especially if they were SATA 300.

So, if you missed that, my point is that an i-RAM built that way would be cheaper than maxing out the memory on a motherboard and offer real world benefits to the average consumer or enthusiest - can you say "instant on from standby". Then it would end up in the mainstream PC market. Economies of scale would kick in and they would end up being cheaper for us all.


And based on memory bandwidth requirements, just about any RAM that is currently being thrown away would work. Which would make it cheaper, still.

The biggest issue of the iRAM right now is that it isn't being effectively marketed or supported by the major memory producers.

If Micron, Infenion, or Samsung got behind this thing as a way to turn failed DDR into a revenue stream and got the production line in gear so that the cost of the PCB and the really cheap controller chip would come down, then we'd see these things turning up in Dell desktops as a cheaper alternative to maxing out the system memory with expensive high quality DDR.

A system with 512 MB of DDR and a really cheap 2GB SATA 300 solid state drive for a pagefile would operate almost as well as a system with 2GB of DDR on it. And would probably be cheaper. Especially for the Home PC user. Who ends up swapping everything back and forth to a pagefile anyway. Having an email app open, 10 browser screens, a stock ticker, and watching a movie all at the same time will cause a lot of paging.

That's just my take on it any way.

You are looking at this from a server/workstation perspective. I'm looking at it from a mainstream PC perspective and a "how do I get this out into the world" perspective.

Try to see it from the senario that I discribed. It would only benefit the market as a whole to be able to get a solid state drive like this into the retail channel.

 

[mikeblas]

The "Cost of components" thing was about ECC regeristered DDR. Which is required by the current mobo's that can pass the 4GB limit.

Registered memory is required. ECC isn't required, but it's not a bad idea even for boards that aren't being built to have lots of memory.

Nope, I didn't miss it. And I agree that you should max out a system's RAM before you go for an iRAM. Unless the iRAM can meet certain conditions.

Then why are you pricing boards that can go past the 4GB limit? Then why are you asking about Socket 939 boards that can address more than 4GB?

So, if you missed that, my point is that an i-RAM built that way would be cheaper than maxing out the memory on a motherboard and offer real world benefits to the average consumer or enthusiest -

Instant on is interesting for some applications -- but not for anything I do. Booting is a very small part of the performance equation.

can you say "instant on from standby".

I'm a little lost with this one. Do you think that I have a learning disability or a speech impediment just because I dare to disagree with you? Isn't it a little absurd that you'd try to convince me of your point by talking to me like you're Mr. Rodgers addressing a preschooler?

If Micron, Infenion, or Samsung got behind this thing as a way to turn failed DDR into a revenue stream and got the production line in gear so that the cost of the PCB and the really cheap controller chip would come down, then we'd see these things turning up in Dell desktops as a cheaper alternative to maxing out the system memory with expensive high quality DDR.

Why aim at DDR? Why not use PC100 or PC133 memory? They still offer almost a gigabyte of bandwidth; an order of magnitude more than the SATA bus can deliver. These parts are nearly free. Even given the density problems, I think you'll still end up with something cheaper for the low-end segment you're trying to address.

A system with 512 MB of DDR and a really cheap 2GB SATA 300 solid state drive for a pagefile would operate almost as well as a system with 2GB of DDR on it.

I'm not sure I'm ready to agree with that. Paging is bad; it's really slow. You're not getting rid of paging at all -- in fact, you're encouraging it by using so little main memory. Why do you think paging to a fast device will perform as well as not needing to page at all?

And would probably be cheaper. Especially for the Home PC user. Who ends up swapping everything back and forth to a pagefile anyway. Having an email app open, 10 browser screens, a stock ticker, and watching a movie all at the same time will cause a lot of paging.

I wouldn't expect it to cause any on a machine with, say, a gig of memory.

I'll go picking the cheapest prices at NewEgg: a gig of PC3200 DDR memory is $66. is 512 megs of DDR, plus the PC board, plus the 2 gigs of cheaper DDR less expensive, in total, than $66? The 512 megs is $33, so now you have a $33 budget for 2 gigs of memory, your PC board and all its components and IP.

You are looking at this from a server/workstation perspective. I'm looking at it from a mainstream PC perspective and a "how do I get this out into the world" perspective.

Perhaps, but I also deeply understand paging and the way Windows works. So I know someone interested in performance doesn't want paging, even to a very fast device; they want main memory, regardless of what application they're running. The net effect that paging has on the system makes is profoundly undesirable. Even if the semiconductor disk solution can be made cheaper (at $66, how much chaper could it be?) than the additional memory, I'd be happy to pay the price to get the performance boost.

 

[rodsfree]

Registered memory is required. ECC isn't required, but it's not a bad idea even for boards that aren't being built to have lots of memory.

I agree - but the price for ECC Registered DDR means that it will never be mainstream.

Then why are you pricing boards that can go past the 4GB limit? Then why are you asking about Socket 939 boards that can address more than 4GB?

Again, Consumer Mainstream motherboard. The masses drive the market. And the OS and software that the average consumer uses are memory hogs.

Instant on is interesting for some applications -- but not for anything I do. Booting is a very small part of the performance equation.

The biggest complaint of the averager computer user ( i.e. not you ) is the length of time it takes their computer to boot or to switch between application windows.

I'm a little lost with this one. Do you think that I have a learning disability or a speech impediment just because I dare to disagree with you? Isn't it a little absurd that you'd try to convince me of your point by talking to me like you're Mr. Rodgers addressing a preschooler?

I'm sorry if the way I talk offends you. I'm a Southerner. We talk like this. I had no intention of offending you. And no, I'm not trying to talk down to you. But I do think that you unrealisticly take any comment in a discussion as a personal attack against you. Especially when that person takes an opposing view.

Why aim at DDR? Why not use PC100 or PC133 memory? They still offer almost a gigabyte of bandwidth; an order of magnitude more than the SATA bus can deliver. These parts are nearly free. Even given the density problems, I think you'll still end up with something cheaper for the low-end segment you're trying to address.

Great Idea! Even better than using DDR.

The only problem is that the production level of PC100 and PC133 is at about 5% of the current production level of DDR. Therefore economies of scale dictate that it wouldn't be economical to use this type of memory. You have a similiar situation in the printer market.
The cost if a 9 pin or 24 pin dot matrix printer is greater now than at any time in history. In fact I can buy a color laser cheaper than I can buy an industrial grade Epson dot matrix printer. The same one that used to sell for $75.00 after inkjets went mainstream. And the reason why is because there is an entire office niche market that uses impact printer forms. Like the IRS, banks, and mass marketers. ( I resisted the urge to say "Can you say" )

I'm not sure I'm ready to agree with that. Paging is bad; it's really slow. You're not getting rid of paging at all -- in fact, you're encouraging it by using so little main memory. Why do you think paging to a fast device will perform as well as not needing to page at all?

I wouldn't expect it to cause any on a machine with, say, a gig of memory.

Paging isn't bad. And I am encouraging it. In fact it was created to allow consumer level computers to deal with an operating system and applications that needed more resources than the consumer could afford to buy. And if you notice I said almost as well as.

And that's the way it's always going to be. Software needs have always out stripped the capabilities of the hardware that it runs on.

Vista is coming out soon. And from all reports it NEEDS more resources than the average $400 consumer PC. But it will be sold on a Dell running a Celeron D with maybe 512MB of ram. Or an eMachine Sempron. So how is Vista going to operate on that kind of hardware.
It's going to page. Just like when Windows 2000 first came out on a socket 423 1.3Ghz P4 with 32MB of RAMBUS on it. It paged like all hell.

I'll go picking the cheapest prices at NewEgg: a gig of PC3200 DDR memory is $66. is 512 megs of DDR, plus the PC board, plus the 2 gigs of cheaper DDR less expensive, in total, than $66? The 512 megs is $33, so now you have a $33 budget for 2 gigs of memory, your PC board and all its components and IP.

It is if you get the memory manufacturers to start providing the memory that fails the normal testing and binning process for this application. Then instead of simply throwing away the failed chips they can turn them into a revenue stream. They could sell 1 GB of failed DDR for $10.00 and make money off of it. Or at the very least change losing money on scrap to a no loss situation. Which is the same as making money on it.

Perhaps, but I also deeply understand paging and the way Windows works. So I know someone interested in performance doesn't want paging, even to a very fast device; they want main memory, regardless of what application they're running. The net effect that paging has on the system makes is profoundly undesirable. Even if the semiconductor disk solution can be made cheaper (at $66, how much chaper could it be?) than the additional memory, I'd be happy to pay the price to get the performance boost.

Mike, you know a lot about Windows and a lot about software design. I know a lot about the market and consumers in general.

Paging is not BAD. It's the way that our OS's and applications deal with the type of hardware that the average consumer is going to purchase. The average consumer is who BIll Gates and Micheal Dell are selling their products to, we are just a slight blip on their agenda.

And an iRAM set up the way that I've outlined would save Micheal Dell millions of dollars a year in the amount of RAM he has to buy and would make his customers extremely happy with his product. So, it would be a win-win situation for both the consumer and for Dell.

But you've got to get everything in the production chain created just so for something like this to happen. Until you get the right companies on board and willing to make this product a reality it's going to just be a niche item. A conversation piece or a way for an enthusist to increase his ePenius.

In some ways we agree and others we disagree. And it's because the perspective that we are seeing this item from. You are concentrating on if it would help you specifically. I am looking at how this could be used to benefit consumers, memory producers, and system builders all at the same time.

 

[unhappy_mage]

Paging isn't bad. And I am encouraging it. In fact it was created to allow consumer level computers to deal with an operating system and applications that needed more resources than the consumer could afford to buy. And if you notice I said almost as well as.

Paging is not BAD. It's the way that our OS's and applications deal with the type of hardware that the average consumer is going to purchase. The average consumer is who BIll Gates and Micheal Dell are selling their products to, we are just a slight blip on their agenda.

ECC registered memory is the wave of the future, and I'll be glad to see its arrival. It's time to make reliable hardware the standard. I want to know when something's broken or breaking, and which something to blame and/or replace. The price difference of ECC/reg will sink to about 20% higher than nonecc/nonreg as demand for it comes up and production gets higher, and then suddenly (say around 2010*) AMD and Intel will stop making chipsets that deal with nonecc/nonreg. I won't cry.

Paging is BAD. If you want to have your application run fast (or if frame rate matters), you can't be paging. To a ramdisk or not, doesn't matter. And at 60 fps, there's about enough time with a fast disk (e.g. ramdisk) to pull 1MB worth before you have to redraw. From memory, you can probably pull about 50 times that. But of course I'm doing math in my head again.

*Nostradamus said so. Look it up.

 

[mikeblas]

I'm sorry if the way I talk offends you. I'm a Southerner. We talk like this. I had no intention of offending you.

That's a relief. Then, what did you mean? I'm sorry you think I'm thin-skinned, but I don't spend much time in the south and can't bind "can you say?" to anything other than Mr. Rodgers explaining something to his target audience.

The only problem is that the production level of PC100 and PC133 is at about 5% of the current production level of DDR.

I had thought you were talking about enthusiasts -- the only people I'd see as willing to go through the gyrations to get workflow arranged so that the i-DRIVE was usable for high-speed storage (as opposed to simply using it for paging). I'd see that group buying PC133/PC100 parts off eBay, used, and building their drives with them.

But, sure, if this is to go into production, then you want something that's -- well, that's in production.

biggest complaint of the averager computer user ( i.e. not you ) is the length of time it takes their computer to boot or to switch between application windows.

Not an ergonomics issue? (Carpal tunnel syndrome or other RSI issues, or screen problems.) Not stability? Or complexity and troubleshooting issues? Bad error messages?

Do you have a source for that assertion? Not to call you out at all, and as an aside from the thread -- I want to know and read the research so I can better focus my efforts. If I could fix one thing about the computer industry, what should it be? Is it really boot time?

Time to switch between application windows? Isn't that dominated by paging? (Sometimes, it's just a shitty application -- Dreamweaver 8, for example.)

Paging is not BAD.

Paging is terrible for performance. Sure, as a stop gap, it allows more applications to run (or, the same applications with more data requirements) and that can be enabling. But if I can avoid it for $33 on an incremental, I will.

And I can't see why anyone wouldn't, even if you reach super-mega-turbo volume on the parts involved. Say you get to a point where, for $15, I can get a gig of i-DRIVE. I'd rather not set that up just to use as a pagefile -- I'd rather spend the $33 and reduce my pagefile dependency.

In fact it was created to allow consumer level computers to deal with

I think that paging predates the idea of a consumer buying a computer.

In some ways we agree and others we disagree. And it's because the perspective that we are seeing this item from. You are concentrating on if it would help you specifically. I am looking at how this could be used to benefit consumers, memory producers, and system builders all at the same time.

Actually, I really am thinking broader than you're giving me credit for. Regardless of the application, I don't see why someone wouldn't spend their money on more physical memory instead of using the i-DRIVE as a swap file because right now, in reality, it's cheaper than the i-DRIVE solution to the point of the max memory supported by the motherboard.

But let's say you do successfully engage this volume, and now a certain type of memory is cheaper. Why not just use that cheaper memory as system memory, even to the point of choosing the processor and motherboard to accept the cheaper memory? It's slower than the best memory you can buy, but I don't think memory bandwidth is limiting the "mainstream consumer market" that you're targeting. And wouldn't the processor manufacturers be pressured to come up with chipsets or processors that used this memor, if it was cheaper in a meaningful way?

 

[rodsfree]

ECC registered memory is the wave of the future, and I'll be glad to see its arrival. It's time to make reliable hardware the standard. I want to know when something's broken or breaking, and which something to blame and/or replace. The price difference of ECC/reg will sink to about 20% higher than nonecc/nonreg as demand for it comes up and production gets higher, and then suddenly (say around 2010*) AMD and Intel will stop making chipsets that deal with nonecc/nonreg. I won't cry.

Paging is BAD. If you want to have your application run fast (or if frame rate matters), you can't be paging. To a ramdisk or not, doesn't matter. And at 60 fps, there's about enough time with a fast disk (e.g. ramdisk) to pull 1MB worth before you have to redraw. From memory, you can probably pull about 50 times that. But of course I'm doing math in my head again.

*Nostradamus said so. Look it up.

u_m

Ok, let's do a little empirical study.
How many applications are there that require high speed operation or fast frame rates?
Now how many are there that don't?

Lots more spread sheets, web browsers, media players and stuff out there that would benefit from a faster cheap paging solution than would benefit from larger amounts of more expensive ram.

And I personally have as much RAM on my computers as I can afford.

My argument is that for the iRAM to be a commerically viable product it will have to fill a niche that the guy who purchases a computer from Dell would want to have and that would be economical for him to buy.

As an enthusist product - it's just not enough bang for the buck. IMHO

And ECC reg. DDR or DDR2 will not become mainstream.
Same reason we are not using the more efficient and faster RAMBUS modules - they were ECC reg IIRC.
It doesn't benefit Intel and the other mfgs to do so.

 

[unhappy_mage]

Lots more spread sheets, web browsers, media players and stuff out there that would benefit from a faster cheap paging solution than would benefit from larger amounts of more expensive ram.

Ah, your web browser uses more than 4GB of ram? You use firefox, too? I'm joking, of course, but the idea remains - what media player are you using that needs that kind of memory? Right now I'm running Winamp, Media Player Classic, PuTTY, Cream (an editor), Firefox (150 MB for 9 tabs?! Earlier it was at 700!), uTorrent. "PF usage" in Taskman" is at 776 MB, 500 mb are free (1GB physical memory). For the sake of argument, let me open Acrobat reader, Gimp, Ethereal, Nero, Word, Excel, Access, and Visual Studio. Now I'm up to 916 MB used and 390 free. Switching through all these apps takes as long as it does to click on them. The hard drive is barely touched during this. So why would I need more ram for office apps? What database are you using?

And if it doesn't require high FPS or fast operation, most people are content to let it page'n'wait. I just don't see the need for a "medium-speed" pagefile device.

My argument is that for the iRAM to be a commerically viable product it will have to fill a niche that the guy who purchases a computer from Dell would want to have and that would be economical for him to buy.

As an enthusist product - it's just not enough bang for the buck. IMHO

So as a consumer product, it is? How's that work?

 

[rodsfree]

That's a relief. Then, what did you mean? I'm sorry you think I'm thin-skinned, but I don't spend much time in the south and can't bind "can you say?" to anything other than Mr. Rodgers explaining something to his target audience.

It's used kind of like a verbal exclamation point. Without having to raise your voice or anything. Just a different way of emphasising what you are saying.

I had thought you were talking about enthusiasts -- the only people I'd see as willing to go through the gyrations to get workflow arranged so that the i-DRIVE was usable for high-speed storage (as opposed to simply using it for paging). I'd see that group buying PC133/PC100 parts off eBay, used, and building their drives with them.

But, sure, if this is to go into production, then you want something that's -- well, that's in production.

This thing will never be used for true high speed storage. It is just not really big enough. I mean, even a game these days require 4+GB. and Windows. It just keeps getting bigger all of the time, with the updates and all.

Not an ergonomics issue? (Carpal tunnel syndrome or other RSI issues, or screen problems.) Not stability? Or complexity and troubleshooting issues? Bad error messages?

Do you have a source for that assertion? Not to call you out at all, and as an aside from the thread -- I want to know and read the research so I can better focus my efforts. If I could fix one thing about the computer industry, what should it be? Is it really boot time?

Time to switch between application windows? Isn't that dominated by paging? (Sometimes, it's just a shitty application -- Dreamweaver 8, for example.)

I can't give you published data.
I wish I could, but every computer that I work on for some one, they always ask me if I can get it to boot faster, or complain that it is just plain slow when they are running more than one application. Or a similar complaint.

Paging is terrible for performance. Sure, as a stop gap, it allows more applications to run (or, the same applications with more data requirements) and that can be enabling. But if I can avoid it for $33 on an incremental, I will.

And I can't see why anyone wouldn't, even if you reach super-mega-turbo volume on the parts involved. Say you get to a point where, for $15, I can get a gig of i-DRIVE. I'd rather not set that up just to use as a pagefile -- I'd rather spend the $33 and reduce my pagefile dependency.

Sure it is, but it's a lot cheaper for the average computer user to have a system that pages than for them to buy a system that doesn't.

I think that paging predates the idea of a consumer buying a computer.

In the Windows world - Windows 3.1 started it all.
C64 didn't page.
C128 didn't page.
Geos didn't page.
TI-99 didn't page
Zeos didn't page.
HeathKit didn't page.
IBM PC didn't page.
Apple's first computer didn't page.

Actually, I really am thinking broader than you're giving me credit for. Regardless of the application, I don't see why someone wouldn't spend their money on more physical memory instead of using the i-DRIVE as a swap file because right now, in reality, it's cheaper than the i-DRIVE solution to the point of the max memory supported by the motherboard.

I agree with this. My whole point is that for the i-DRIVE to become viable it has to become cheaper than investing in more onboard memory. And at some point the OSs and apps are going to start eating all of the ram that a motherboard can carry, anyway. **cough**BF2**cough** At that point we are going to need a cheap and efficient pagefile drive. To carry us through the several generations of computer design that will get us mainstream motherboards capable of having more than 4GB of ram.

But let's say you do successfully engage this volume, and now a certain type of memory is cheaper. Why not just use that cheaper memory as system memory, even to the point of choosing the processor and motherboard to accept the cheaper memory? It's slower than the best memory you can buy, but I don't think memory bandwidth is limiting the "mainstream consumer market" that you're targeting. And wouldn't the processor manufacturers be pressured to come up with chipsets or processors that used this memor, if it was cheaper in a meaningful way?

Well, the memory is already here. I'm sure that you could go to Samsung's factory and see piles of it that failed the binning process.

The current motherboards wouldn't make use of slower memory as system memory because the entire motherboard design culture has been to speed everything up. By everyone, including the mainstream market.

The issue here is going to be more "What is fast enough for THIS specific application?"
Because system memory needs to run at a certain speed - very fast compared to just a data storage device.

And we are in a "you can't get there from here" situation.

There is not going to be a cheap solid state drive of really usable size in the near future.
They are working on Hybrid drives that have huge caches to be the next step toward the true solid state drive. And the huge FLASH drives of course.

So, where do you think that the i-DRIVE could fit.
If it could be made cheap enough to benefit you.

As a pagefile / scratch drive is where I see it being a good niche product - if it could be produced cheaply enough.

Or, it'll end up dying and we'll just have to wait on a different solution.

 

[mikeblas]

This thing will never be used for true high speed storage. It is just not really big enough.

That's amusing to me, as it's the only viable application I see for the thing: as a scratch space for apps that need to spool, like Photoshop when editing a giant picture.

I wish I could, but every computer that I work on for some one, they always ask me if I can get it to boot faster, or complain that it is just plain slow when they are running more than one application. Or a similar complaint.

Bummer. I never would've chosen boot speed as the #1 problem. I would've picked complexity (someone has to show me how to do this), or fragilty (it worked yesterday), or bad tech support, or spam, maybe.

In the Windows world - Windows 3.1 started it all.

Most of the machines you list didn't hvae a true operating system; they just sat in a BASIC interpreter after you started them. Windows is not the first OS to offer virtual memory. The first OS I experienced with VM was VAX/VMS, but I'd figure that some old Burroughs or IBM rig would have been the first to implement the idea. No consumer would go buy one of these; these old mainframes filled rooms and took real maintenance to keep running.

The current motherboards wouldn't make use of slower memory as system memory because the entire motherboard design culture has been to speed everything up. By everyone, including the mainstream market.

When you buy Dell's current sub-$500 rig, you're not getting the fastest motherboard around. While everyone in this forum was buying fast Athlon chips and 800 MHz Intel rigs, they were selling 400 MHz FSB machines to that crowd.

As I mention above (and as you've gathered, I'm sure), I think using it as a device for hold a page file is absurd. For spooled storage, sure—that seems very appropriate, and exactly what such a high-speed storage device is for.

 

[rodsfree]

As I mention above (and as you've gathered, I'm sure), I think using it as a device for hold a page file is absurd. For spooled storage, sure—that seems very appropriate, and exactly what such a high-speed storage device is for.

And that is just where it will stay.
In a very limited market niche. With a premium price tag on it.

Because the only other application for it would be as a pagefile drive.

It won't be an OS drive - Windows has gotten bigger than 4GB and nobody wants to reload their OS just because a battery failed.

It might make it as a Gamers drive. Except the latest games are bigger than 4GB and you'd still have the re-install issue. Unless you did some kind of autocopy script from you HD. Might have registery issues though.

There might be another use for it.
I just can't think of one.

 

[mjz_5]

And that is just where it will stay.
In a very limited market niche. With a premium price tag on it.

Because the only other application for it would be as a pagefile drive.

It won't be an OS drive - Windows has gotten bigger than 4GB and nobody wants to reload their OS just because a battery failed.

It might make it as a Gamers drive. Except the latest games are bigger than 4GB and you'd still have the re-install issue. Unless you did some kind of autocopy script from you HD. Might have registery issues though.

There might be another use for it.
I just can't think of one.

well, when they release the next version that hopefully supports 16 GB (8 2GB slots).. It will be an amazing OS/Appz drive. It will be the biggest jump in speed since they invented a hard drive or something

 

[SatinSpiral]

well, when they release the next version that hopefully supports 16 GB (8 2GB slots).. It will be an amazing OS/Appz drive. It will be the biggest jump in speed since they invented a hard drive or something

something like that. i'm sick and tired of ppl crapping on this thread about the limited capacity/speed/applicability of I-RAM 1. This is an I-RAM 2 thread, about a much more capable and useful product coming out in 07. NOT ABOUT A PRODUCT ALREADY RELEASED.

 

[Jason711]

how about running a website off a i-ram style product?

 

[SatinSpiral]

i'm not too knowledgeable about how servers work. does anyone know if a gaming server would benefit at all from I-RAM 2?

 

[Jason711]

i'm not too knowledgeable about how servers work. does anyone know if a gaming server would benefit at all from I-RAM 2?

nice idea.

 

[Ockie]

Yes, game servers or any server for that matter will benifit greatley from it, however, servers drives aren't usually small. As for a game server, this would work superb and since you are in a guranteed uptime enviroment (server co-location comes to mind) it would be great to host it there

 

[mikeblas]

Yes, game servers or any server for that matter will benifit greatley from it, however, servers drives aren't usually small.

Buy what mechanism would any server benefit?

 

[Ockie]

Buy what mechanism would any server benefit?

I think your taking the "any" too literally. Game servers, web servers, caching servers, all those would benifit in terms of accessing the drive quicker and responding quicker.

The biggest issue with game servers are that you can't load too many games per system due to the fact that when a map changes in one server it would lag the other servers... .right there alone is a great potential to eliminate that bottleneck.

 

[mikeblas]

I think your taking the "any" too literally.

Perhaps, but I wouldn't feel insane for thinking "any" means "any" instead of "some" or "most", or "certain kinds".

 

[mjz_5]

something like that. i'm sick and tired of ppl crapping on this thread about the limited capacity/speed/applicability of I-RAM 1. This is an I-RAM 2 thread, about a much more capable and useful product coming out in 07. NOT ABOUT A PRODUCT ALREADY RELEASED.

THANK YOU

 

[mjz_5]

I think your taking the "any" too literally. Game servers, web servers, caching servers, all those would benifit in terms of accessing the drive quicker and responding quicker.

The biggest issue with game servers are that you can't load too many games per system due to the fact that when a map changes in one server it would lag the other servers... .right there alone is a great potential to eliminate that bottleneck.

database serers will probably love this product

 

[mikeblas]

database serers will probably love this product

It's too small for anything substantial, though I guess you could put a temp workspace on it and hope for the best.

But I'm not sure that I'd trust it to do transactional writes -- if the power goes out when a write is half-done, does the battery finish the write, or just hold the half-written garbage in the store memory so that I have to recover and clean the database when I restart?

Data warehousing, on the other hand, where you need lots of workspace and the data is more ethereal, seems like it might be a good match.

 

[mjz_5]

It's too small for anything substantial, though I guess you could put a temp workspace on it and hope for the best.

But I'm not sure that I'd trust it to do transactional writes -- if the power goes out when a write is half-done, does the battery finish the write, or just hold the half-written garbage in the store memory so that I have to recover and clean the database when I restart?

Data warehousing, on the other hand, where you need lots of workspace and the data is more ethereal, seems like it might be a good match.

a server will have a UPS, so that power argument is a non issue. Second, this is regarding the i-RAM 2 device, it may be up to 16GB, which should be good enough for most databases.

 

[mikeblas]

I'd certainly not call a 16-gig database "substantial".

 

[mjz_5]

I'd certainly not call a 16-gig database "substantial".

thats pretty large.. And plus, this is just the begining. Give it a couple of years, the size of the ram drive will be large enough

 

[mikeblas]

Many of the customers I've worked with would consider 16 gigabytes a starting point for the system memory on their production servers.

16 gigs is only a single table of 100 million rows of 160 bytes each, with no space left over for any indexing or any other objects.

For truly large, see the Wintercorp survey; there isn't anything in the top ten that's less than 16 terabytes.

 

[hokatichenci]

Many of the customers I've worked with would consider 16 gigabytes a starting point for the system memory on their production servers.

16 gigs is only a single table of 100 million rows of 160 bytes each, with no space left over for any indexing or any other objects.

For truly large, see the Wintercorp survey; there isn't anything in the top ten that's less than 16 terabytes.

Yeah but obviously this device wouldn't be used by any big corporation, they'd have clusters of databases which are probably 4+ way and support 64-128GB. However, for your mom and pop group that is probably storing 100k to millions of data this might be useful. Not to mention that using this as a system drive would quickly speed things up, which would increase availability assuming the system crashes. I'm still looking cautiously at this for my workstation as an an alternative to digging into 3gb of my ram, though if I go quad-core I'd probably just buy the extra ram since my machine boots fast enough.

 

[mikeblas]

Yeah but obviously this device wouldn't be used by any big corporation, they'd have clusters of databases which are probably 4+ way and support 64-128GB.

Right. So, can we agree that 16 gigs isn't enough for anything substantial?

I don't see how using it as a system drive would "speed things up" in the application of a database server. What would it speed up?

 

[Jason711]

i think it would be great for a handfull of small servers.

 

[Ockie]

@Mikeblas

Are you seriously trying to contradict everything everyone is saying? It's rather quite annoying.



Anyways, as for databases, this would be a great product as you can raid these devices, so 16GB's might not seem that large, but you can raid as many of these as you have physical space. So 10 of them alone would be 160Gigs, which is pretty large.

 

[hokatichenci]

Right. So, can we agree that 16 gigs isn't enough for anything substantial?

Substantial to whom is really the question. No, not to mega-corporations that can blow 50 grand on a server, but to companies who've got maybe 4-5 grand, it might make a difference.

I don't see how using it as a system drive would "speed things up" in the application of a database server. What would it speed up?

Even if somethings written to memory it still has to get written to disk at some point. This is why databases love the huge/fast raid arrays. Using it as a system drive would decrease the time it takes to boot, therefore increasing availability.Maybe these days that isn't a big deal with clustering but I remember a few years back where every company was fighting over how many 9's they could fit into their availability, which basically came down to how fast the machines could reboot.

 

[mikeblas]

@Mikeblas

Are you seriously trying to contradict everything everyone is saying? It's rather quite annoying.

Nope. Only the stuff that I don't understand. Which is how people have conversations: it's a great way to learn something.

If you reread my notes carefully, I think you'll find that I'm not "contradicting everything everyone is saying". I'm asking questions; if someone asserts something is true and I don't agree with it, I've tried to respond by asking about it. How would that work? What would be the net benefit? And so on. For a comparison, I'll also share my own insight and experience and add that to the discussion.

For example, since there's no ECC on the memory used in the I-RAM drive, I would be hesitant to use it for database storage, or a few other server applications. Someone who asserts this is a great product for database applications must have thought that through, so I would want to ask them why they think the lack of error checking is acceptable for such applications. Would they absorb some of the perf gain from using this by turning on strict torn page checking in the database? Does that work out well? Is there another way to work around the data integrity issues that I don't know about? And so on.

Using the devices in RAID0 seems like an interesting idea. I wonder how much latency a RAID controller would add to to stack. It would be better than physical disks, sure -- but it would be a larger percent of the toal request time since the request time is so much smaller. That means the controller's own performance has more influence, since it's a larger part of the pie.

Would the controller have the bandwidth to handle it? If the I/Os are happening in a 1000th of the time they used to take, can the controller keep up?

 

[lukek]

I can't wait to see the concrete specs and maybe an eval. of the I-RAM 2, about 2 weeks ago we switched over one of our databases at work to I-RAM 1.3's. I have a eBay business that me and several friends run. We recieve about 1500-2000 emails a day from that and 2 of our websites. We run outlook and shipworks on our server and our email database alone is about 7gigs a month. We just moved everything over to a pair of I-RAM's and its amazing, we have one 8 gig stripe for outlook express and another 8 gig stripe (4 I-rams) for our shipworks although that one is only around 3 gigs of data a month. Anyways, processing anything and searching is instant, we use to index everything overnight but this is soo much better. We of course do nightly backups to 2 smaller (500gb) scsi data server's. But I find that this item would def benifit high traffic smaller databases that many home business and small businesses have. I currently ordered a pair of these for home use and will probably setup my HTPC OS/ and some apps on it, since its silent. Just thought you guys would like to hear of a reallife use of these puppies.( sorry its not I-RAM 2 related, although a 16gb setup would be sweet, also if anyone is interested we did setup the 4 I-RAM's on a PCI-E raid card to see if they would work, its not really supported but we did see read write speeds of about 440-450

 

[Jason711]

Nope. Only the stuff that I don't understand. Which is how people have conversations: it's a great way to learn something.

If you reread my notes carefully, I think you'll find that I'm not "contradicting everything everyone is saying". I'm asking questions; if someone asserts something is true and I don't agree with it, I've tried to respond by asking about it. How would that work? What would be the net benefit? And so on. For a comparison, I'll also share my own insight and experience and add that to the discussion.

For example, since there's no ECC on the memory used in the I-RAM drive, I would be hesitant to use it for database storage, or a few other server applications. Someone who asserts this is a great product for database applications must have thought that through, so I would want to ask them why they think the lack of error checking is acceptable for such applications. Would they absorb some of the perf gain from using this by turning on strict torn page checking in the database? Does that work out well? Is there another way to work around the data integrity issues that I don't know about? And so on.

Using the devices in RAID0 seems like an interesting idea. I wonder how much latency a RAID controller would add to to stack. It would be better than physical disks, sure -- but it would be a larger percent of the toal request time since the request time is so much smaller. That means the controller's own performance has more influence, since it's a larger part of the pie.

Would the controller have the bandwidth to handle it? If the I/Os are happening in a 1000th of the time they used to take, can the controller keep up?

i understand where you are coming from. however you have to understand that this is a very new "consumer level" product. we are simply coming up with ideas to make use of what we might have. and you have to consider that this technology will improve. it doesnt do much good to totally scrutanize it unless you can actually do something about it.

 

[mikeblas]

Samsung has announced a 32gig FLASH-based drive.

 

[seto]

I got one as a lark. Yeah, it was expensive, but computing is my hobby. I even bought 4 sticks of low end OCZ 1G DDR400 for it. So yeah, I tossed my money away.

I am using it as my scratch disk. MSIE temporary folder, system and user temp directories are now moved onto it. Nothing else.

I did copy over the occational app, but you are still running head on into the <4G size (remember there is still format overhead) as well as the SATA 1 speed of 150 (max).

OK, you can get your mad chuckles over the derfag speed, but all in all, I just don't think about it. IE response has dramatically increased, but other than that, no other changes.

There is no point in moving my page file to the I-Ram. I already have 4G installed on my on my motherboard (of which, only 3+ are usable. Damn you Win XP 32 and lack of BIOS device driver relocation!).

As a side note, yes I plan on buying an Agiea PPU card as soon as they are available too.
Long live my disposable income! (Hmm, 10 lap dances or a PPU.)

 

[mikeblas]

Damn you Win XP 32 and lack of BIOS device driver relocation!).

Huh?

 

[seto]

If you decide to fully populate your motherboard with max ram, you wuill be limited with howq much you can use.

The BIOS takes up adress space for its devices and crams them after the 3G. Until recently, no consumer board could have more than 3 G so no loss. Now, with 4G available, the BIOS has to move its device addressing after 4G or you will not be able to access that memory.

XP is also limiting factor. The biggest reason for me to install Vista is to be able to fully use the 4G or main system ram that I have.

But its still dependant on BIOS support.

 

[mikeblas]

The BIOS takes up adress space for its devices and crams them after the 3G. Until recently, no consumer board could have more than 3 G so no loss. Now, with 4G available, the BIOS has to move its device addressing after 4G or you will not be able to access that memory.

You may find memory-mapped regions there, like the AGP aperture. But I don't think you're going to find any "BIOS device drivers" up there—though I'm not sure I know what you precisely mean by that phrase.

Check out page 417 of "Windows Internals", Fourth Edition; it's got a memory map of the system address space, and a discussion of what ends up where.

XP is also limiting factor. The biggest reason for me to install Vista is to be able to fully use the 4G or main system ram that I have.

What is it that you expect Vista to change?