PCi-e Riser (Memory)

[Freezer]

Anyone know where I can purchase a PCi-e Riser exclusively to add more memory? I thought Logic supply had some. It's use is not for "system" memory, but for other means.

 

[dorsetknob]

something like this
http://www.amazon.com/dp/B000EPM9NC/?tag=tec06d-20
http://www.newegg.com/Product/Product.aspx?Item=N82E16815168001

 

[newtekie1]

There are cards that use memory sticks as storage, dorsetknob linked to a couple, but there is no way to use a PCI-e card to add system memory. The way system memory is addressed and controlled will not allow it.

However, with the advent of SSDs, there really hasn't been a need for PCI-e cards that take RAM and use it for storage space.

 

[Freezer]

Exactly, but PCIe.

They may have been phased out entirely, but just looking. I'd like something that could handle over 64GB (DDR2/DDR3) of ram or more.

 

[EarthDog]

Add more memory (ram) or storage (HDD)?

EDIT: You clarified when I posted, LOL....

Yeah, I never heard of PCIe ram that was actually not a ramDISK. Why do you need more than the 32GB you already have? Why not upgrade that to 64GB? What will this be used for???

 

[dorsetknob]

as far as i know there is nothing PCIx form factor available if its to use DDR then your have to go with what i posted
"" IF YOU CAN FIND ONE ""
sugest you read the wilki
https://en.wikipedia.org/wiki/I-RAM

then search ebay or craigslist ect

 

[Freezer]

Add more memory (ram) or storage (HDD)?

EDIT: You clarified when I posted, LOL....

Yeah, I never heard of PCIe ram that was actually not a ramDISK. Why do you need more than the 32GB you already have? Why not upgrade that to 64GB? What will this be used for???

Extending cache disk.

The availability would be great, but since there isn't any support for PCIe...

 

[EarthDog]

So it seems a simple ramdisk is all that is needed. A lot of boards can already do that by assigning some of your system ram to a ramdisk.

I guess I am not sure what you are doing that you are using all 32GB or why you cannot upgrade to 64GB and not use a ramdisk type thing... *throws more on the wall... seeing if it sticks, LOL*

 

[newtekie1]

Or a M.2 drive, or M.2 Drive and a PCI-E adapter.

 

[rtwjunkie]

Exactly, but PCIe.

They may have been phased out entirely, but just looking. I'd like something that could handle over 64GB (DDR2/DDR3) of ram or more.

They have been phased out. Gigabyte made the best one, but at the time it only took an additional 8GB of DDR-1 I believe. Also, it was only PCI.

 

[Freezer]

So it seems a simple ramdisk is all that is needed. A lot of boards can already do that by assigning some of your system ram to a ramdisk.

I guess I am not sure what you are doing that you are using all 32GB or why you cannot upgrade to 64GB and not use a ramdisk type thing... *throws more on the wall... seeing if it sticks, LOL*

Expensive. A riser would be more cost efficient, and if it were to support all chips from DDR/1/2/3/4, granted it was large enough.

It would be nice if manufacturers updated these.

 

[Aquinus]

You can't add more system memory beyond what the motherboard itself support. PCI-E doesn't extend system memory. Your only options are to buy high density DIMMs or to buy a server motherboard that uses registered memory which supports more DIMMs.

Extending cache disk.

Windows caches recently used data in memory and doesn't remove it unless you run out of space.

What exactly are you trying to achieve?

 

[MIRTAZAPINE]

Unfortunately those gigabyte Iram is limited to sata one speed. I have been trying to find out things like pcie ram drive to extend or upgrade from my 32GB z97 motherboard limit. So far. I found that it is just not worth the expense. There are ddr3 ram drive but they are expensive and meant for server customers. The ddr3 ram drive unlike the Iram does not require a battery as the content in the drive is being backup into an ssd drive as it is use. I got to search it up but the price is usd$1000 and above.

At that price it is better going with an x99 board or server board if you want more ram. Or the cheaper way is just to put 4 ssd in raid 0 if you want to do it for swap file.

Edit : ok so this ramdrive is 6k!
http://www.thessdreview.com/our-rev...ve-101-ramdisk-review-500k-iops-ddr3-storage/

 

[Freezer]

You can't add more system memory beyond what the motherboard itself support. PCI-E doesn't extend system memory. Your only options are to buy high density DIMMs or to buy a server motherboard that uses registered memory which supports more DIMMs.

Windows caches recently used data in memory and doesn't remove it unless you run out of space.

What exactly are you trying to achieve?

It would not be "system" memory, as I previously mentioned it would not be.

We all know, now, PCIe doesn't support the use of memory risers, atm, but someone could design something to support memory risers if they'd really wanted to. Nothing is impossible, just how probable, how complicated, and when. What do you think PCIe SSD drives are? Basically the same thing in some very basic form, but they're outrageously expensive. Yet a memory riser that would support all formats of memory, mix and match, would be very useful for some folks, not only help eliminate the huge pool of overstocked modules gathering dust. Though we'd have to rely not only on hardware, but whether the software will support the functionality of being persistent.

Plus if it were possible on a modern mainboard with PCIe a memory riser would be great as a dedicated cache or swap drive without taking away from other system resources.

 

[Aquinus]

What exactly are you trying to achieve?

@Freezer : Could you answer this question?

 

[rtwjunkie]

@Freezer, for it to be persistent, it would need a battery/batteries, since RAM is volatile.

 

[Freezer]

@Freezer : Could you answer this question?

I did

It's probably more or less a moot point with today's technology and SSD drives, however, if I had 20 memory modules laying around and doing nothing but collecting dust... I think I'd be able to install them into a riser and get some use out of them. It'd be nice as a scratch disk, temp directory, or paging, since neither of them need to be persistent. It'd just provide the ability to off-load resources that the system currently uses from system resources and manually force these resources to use mixed pairs of memory installed on the riser instead and be 2x - 4x faster, than having to allocate space on a drive or ramdrive for a swap, scratch disk, or page file.

@Freezer, for it to be persistent, it would need a battery/batteries, since RAM is volatile.

That's not the case using RAMDisk software, or is it? I'm unsure if it uses an alternate power source, not a programmer, but you can create a persistent RAMDrive.

Sooner or later we won't be using SSD/HDD to store data on. IIRC someone is working on a biological method by using H2O. Well, lets not get off-topic

 

[OneMoar]

what
I don't even....
go home freezer you are drunk

 

[tabascosauz]

I did

It's probably more or less a moot point with today's technology and SSD drives, however, if I had 20 memory modules laying around and doing nothing but collecting dust... I think I'd be able to install them into a riser and get some use out of them. It'd be nice as a scratch disk, temp directory, or paging, since neither of them need to be persistent. It'd just provide the ability to off-load resources that the system currently uses from system resources and manually force these resources to use mixed pairs of memory installed on the riser instead and be 2x - 4x faster, than having to allocate space on a drive or ramdrive for a swap, scratch disk, or page file.





That's not the case using RAMDisk software, or is it? I'm unsure if it uses an alternate power source, not a programmer, but you can create a persistent RAMDrive.

Sooner or later we won't be using SSD/HDD to store data on. IIRC someone is working on a biological method by using H2O. Well, lets not get off-topic

You don't need to pretend that you've come up with some innovative idea that no one in the industry has thought of. These things exist, and they're expensive at $6-12K. You want something like that? Be prepared to shell out $6-12K. Can't afford it? Sit back, and do as everyone else does by upping the RAM in your system to 64GB of DDR4, and enjoy the added capacity. Also, you can't expect to just be able to throw old RAM in there and have it work regardless of whether it's DDR2 or DDR3. You'll need to pick one, and that's why these expansion cards don't sell.

If speculation about ambitious, not-yet-existent forms of data storage isn't "off-topic", I don't know what is.

 

[Freezer]

Pretend? Not even pretending.... it's also nothing "new," it's OLD system peripherals that no one wanted and was costly in the day, not only that it was marketed for servers. I know how much those cards were worth in the 90's.... even in 2000's. Just wasn't sure if they were still manufactured. However, today, these components would be less costly than a PCIe SSD drive to manufacture. Though, I don't know exactly why, and no one said, why manufacturers stopped production for risers. So all I'm left is theories, and perhaps manufacturers just think a riser just looks aweful in a system in a day and age where people are more concerned with minimization and modding their systems to look "pretty" ... which I care less about. I care about functionality, power, and the ability to do my work and add peripherals that would allow me to expand beyond that. I understand what you folks are saying, but that's not "ideal" in my position, just trying to be economical.

Yep, I could max my memory to 64GB or even 128GB, but I'd like to ability to go beyond that without sacrificing "system" resources, or take away from available system resources. The point is... having a memory riser that would accept the addition of mixed matched memory and put them to use, rather than sit around collecting dust or tossed into the recycling bin. It'd be nice to have an additional 32GB, 64GB, 128GB, 256GB, etc... of memory to use ***NOT SYSTEM MEMORY*** for other purposes.

I know the probably is quite high and no one is going to manufacturer something for just one individual. Tho, the only reason I see, and read, why memory risers are no longer being manufactured is because there is a lack of demand, or quite probable the lack of awareness of how effective memory as storage can be. However, if enough people show an interest, perhaps it may be something manufacturers would consider providing again. Unless someone here is agitated that I was being optimistic and had an idea or suggestion to share.

Lol, not drunk, and yes someone is working on H2O as a data source. It's ok tho, people laughed at Einstein, Tesla, Newton, Pascal, Hawkings, Musk (Tesla Powerwall, is somewhat overpriced), etc... haha. I'm not even in their bracket. A wise man once said, "Great minds are often oppressed by the meek and toothless."

 

[Aquinus]

however, if I had 20 memory modules laying around and doing nothing but collecting dust... I think I'd be able to install them into a riser and get some use out of them.

You think incorrectly, and you over simply the difficulty of doing what you suggest.

It'd be nice as a scratch disk, temp directory, or paging, since neither of them need to be persistent.

Why do you think servers and registered memory are expensive? One doesn't simply "throw dram together and use 20 sticks."

It'd just provide the ability to off-load resources that the system currently uses from system resources and manually force these resources to use mixed pairs of memory installed on the riser instead and be 2x - 4x faster, than having to allocate space on a drive or ramdrive for a swap, scratch disk, or page file.

I think you missed the part where I said Windows caches recently used files and data and doesn't remove them from system memory unless you run out of free (uncached or unused,) memory. Believe it or not, there is a point of diminishing returns when it comes to memory capacity and it's exacerbated by things like SSDs.

go home freezer you are drunk

I think people need stop stop thinking that computers just "work" and can throw stuff together. Simple fact is that if it was so easy, it would have been done already. The world isn't only full of idiots which is why we have things like i7s, DDR4, SSDs and chicken salad sandwiches.

However, today, these components would be much less costly than a PCIe SSD drive to manufacture.

You're still wrong because production costs are still going to be similar. Tell me, how many traces would be required to run 20 DIMMs to any number of memory controllers, along with the interconnects from the PCI-E bus to the controller? You sir, are not smarter than an engineer and should not try to act like one.

No offense, but you're doing drunk people a disservice with this kind of thinking.

I'm going to unsub myself before I earn myself an infraction.

 

[Freezer]

LOL@Aquinus, clearly not a single shred of entrepreneurship or creativity in your mind. I take no offense. I can certainly visualize others rolling their eyes and scoffing, but you're only limiting yourselves to what you already know and clearly unwilling to expand on the possibilities. That's pretty much how innovation works.

1) Memory risers were for SERVERS and were expensive because they were catered specifically to that market... PLUS they were designed specifically for ECC modules and to run as SYSTEM resources.

2) The intention here, which you clearly seem to fail at understanding, is the ability to use mixed pairs of memory as a data source, etc. It's certainly doable, and no one said it would be a walk in the park, or cost efficient from the start.

3) I'm talking among the lines of home desktop usage. There is no need for ECC, no need for exact paired modules within a home setting, and some WS settings, with what I'm theorizing about. There is a way despite what you seem to allude to be delusional optimism. Besides, you haven't provided any credible evidence claiming that other manufactures already attempted this, or quotes AS TO WHY they decided not to further production of memory risers or why manufacturers have not shown an interest with the home consumer.

Yep, some said raspberry pie was impossible.

 

[EarthDog]

What market would it serve that existing products, like a ramdisk, do not already take care of?
(Understand the 90% of PC owners don't have 'old memory sticks and stuff' laying around so you are looking into the jaws of a niche market out of the gate)

What tangible/quantifiable benefits would a consumer actually have with this over existing technologies?
(I've seen you mention resume old ram sticks)

How do you forsee mixing memory, which is already considered a 'no no' for stability reasons, of different generations is going to perform and be stable?(I would imagine as fast as its slowest part. I would also imagine the overhead of trying to splice data together different speed/timings/generations/brands whatever you are thinking, has to be massive?)

I think you are a dreamer, there is nothing wrong with that, I just haven't heard anything that makes me feel it's a solid idea at this time.

 

[Freezer]

What market would it serve that existing products, like a ramdisk, do not already take care of?
(Understand the 90% of PC owners don't have 'old memory sticks and stuff' laying around so you are looking into the jaws of a niche market out of the gate)

Agreed, total niche market. Though just theoretical as I've been on about. Again, purpose would serve the same as what ramdisk already offers, but eliminating the factor of using system resources which ramdrive requires. If a system has 64GB of memory installed on the MB for system usage and a user allocates 12GB of that to ramdrive, they're down to 52GB of system memory, plus what the system allocated to itself and the running applications. That may sound ridiculous, since there's plenty of resources available... but what if they were working on a large CAD, or IMAGE file and need a large scratch disk? That memory could be used up fairly quickly. So, lets say memory card risers were available for NON-ECC and PCIe, had 32 or 64GB of memory installed, then the user could easily use the entire riser for the purpose of a scratch disk. Using memory as a scratch disk is much faster than an SSD partition, even in raid0.

What tangible/quantifiable benefits would a consumer actually have with this over existing technologies?
(I've seen you mention resume old ram sticks)

Per above, maybe other uses as well... page file, etc.. running a system within memory only, which has no hard drive and requires a larger than typical memory cache. For example, a user or company has a MB that only supports up to 16GB of memory, the require 64... they have extra memory in bins not used and can purchase a memory riser to expand their needs at a fairly reasonable cost. They add the riser and move whichever application to memory allocated on the riser. I'm sure there are needs for this, but very minimal market. Recycling memory, efficiency (maybe). Clearly something like this would be software controlled, just as RAMDrive.

How do you forsee mixing memory, which is already considered a 'no no' for stability reasons, of different generations is going to perform and be stable?(I would imagine as fast as its slowest part. I would also imagine the overhead of trying to splice data together different speed/timings/generations/brands whatever you are thinking, has to be massive?)

You said it... slowest memory. I'm sure there is a way to stabilize memory with mixed modules, though, you're not going to see DDR4 speeds out of a DDR2 module. You could try. Since the riser isn't needed as "permanent" storage, fault tolerance and reliability isn't of much concern, but ought to be designed to provide a user with as much reliability as possible. As previously stated, if the user sets the riser to act as a scratch disk, the data drive ought to be capable to work under heavy utilization and provide the amount of memory that it was initially setup with. It shouldn't tank from 64GB to 22GB due to a configuration error or due to mixed pairing. The only drawback is a user adding a DDR4 modules in the riser and expecting the speed of DDR4 when they may have also used a couple DDR2 modules. DDR2 modules R/W access times are faster than SATAIII SSD drives.

I think you are a dreamer, there is nothing wrong with that, I just haven't heard anything that makes me feel it's a solid idea at this time.

Quite true. It's really a niche market and I don't believe it would be very popular at the moment... tho today it would have application in certain areas in retail and perhaps GOV (where security is of high importance) centers.

In all fairness, we will eventually no longer use HDD/SSD drives... look at the M.2 drives now... 10 , 20 years from now we could be running a system purely off memory. When HDD/SSD/M.2 drives would be a thing of the past.

 

[xvi]

I think people are getting excited over using the wrong terminology.
If I'm not mistaken, you're looking for a kind of PCIe ramdisk adapter. The reason why you'd prefer this over a SSD is that you have memory laying around that you'd like to put to some kind of use.

Personally, I'd recommend just selling any extra memory you have and using that to fund a PCIe SSD such as this one by Intel. If you have enough memory to be useful as a ramdisk, you'll almost certainly get better value trading it for a PCIe SSD.

we will eventually no longer use HDD/SSD drives... look at the M.2 drives now...

M.2 drives are just SSDs in a much more convenient form factor. SSDs were put in the shape of a 2.5" hard drive to help with adoption/compatibility.