• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Gigabyte IRAM

mikelz85

[H]ard|Gawd
Joined
Aug 3, 2004
Messages
1,342
Didn't see any IRAM threads on a quick search of the forum, so I figured I'd start one here.

http://www.anandtech.com/storage/showdoc.aspx?i=2480&p=1

Hard drives are pretty outdated IMO, and the disk i/o system is a large bottleneck of any system, thus an affordable solid state solution could be huge to everyone, offering load times that are way faster than hard drives. The technology should destroy even the fastest 15k Ultra 320 drives, if done right, with latency in NS, vs. the MS latency of hard disks.

I'm honestly very dissapointed with the performance of this thing. I was expecting to see windows booting in seconds, and a far larger difference between the Raptor and the IRAM. Further, if it was raptors in raid 0 used in the review, instead of a single drive, the difference would have been even smaller.

Why do you guys think the performance of this ram drive was so poor? Is it bottlenecked by the SATA bus to the point that it is seriously affecting performance, or is it something else that is going on? If it was the SATA bus bottleneck that is causing the performance hits (a typical solid state drive should be far faster IMO, the system should boot as if suspended to ram, and load apps far faster), then wouldn't running two in RAID 0 come close to doubling the numbers we saw on anandtech?

I'm one of those guys that initially was very impressed when I heard this thing was coming out, I've been waiting for a consumer ramdrive for years, and thought I was finally going to see great solid state performance make it to the market, however, these numbers are not that impressive.

So, why is it that the numbers are so low?

Is it a lack of optimizations on windows? (unlikely)
Is the a bottleneck on the SATA bus?
Is it a bottleneck somewhere else?
Will we not see a large boost until the ram drive is interfacing through a larger/faster bus, or even better, integrated into the actual motherboard?

What's the deal here? I was really expecting this to be the next big thing, a real must have piece of hardware, and to destroy even the best hard drives, but that just was not the case.

Anandtech also said that gigabyte told them RAID 0 on these had stability issues, I'm hoping this gets fixed, as the numbers might be far better if the bottleneck on the SATA bus is doing a lot of damage.

Also, the initial run is said only to be ~1000 units, and the initial cost running ~$150, that is far higher than ~$60 that was originally quoted by gigabyte at computex.

My original idea was to just have two of these running in RAID 0 for some serious performance, maybe with ~6gb of space total, but now, I don't think it is worth the money, the performance numbers just aren't there.


For those of you wanting to see exactly what a ramdrive can do, you can download software that will allow you to map out a "fake" physical drive from your system memory. It won't be much space, but the performance is insane, and should give you an idea as to why this tech is so exciting.
 
It is indeed exciting, but far to cost prohibitive to most people....including me. :(
 
I have been keeping an eye on cenatek's 'rocket drive' for a few years hoping they could get the costs down to something a bit more palatable for home use too. http://www.cenatek.com/product_rocketdrive.cfm

According to a gentleman I spoke to, their high r&d costs coupled with the target market meant that specific ramdrive would probably never see the price point required for widespread adoption. Its currently priced around $1600 for a 2Gb model, and $3000 for the 4Gb model.... Ouch!

Any way you slice it, gigabyte deserves major credit for trying to get a solid state storage device priced for mass consumption. My hope is that it sells enough for them to continue refining the technology and get the costs down even more.

It would be nice if they could leverage their buying power and ship a fully or partially populated version even if they went with proprietary sticks (possibly notebook sized dimms?). Seems like they could kill two birds with one stone that way the first being cheaper ram... say $60 per GB, and the second being a bit friendlier capacity... say 9Gb for a single slot and 18Gb for a double. That would put the pricing around $700 for a 9 or $1200 for an 18 assuming a $150 base card price. Nice to dream about at least.

While the performance numbers are not mind blowing, omfg, completely redefine storage kind of numbers, it’s not exactly a slouch :)
 
well, given what solid state number should be, I would have hoped for nearly instant boot times, and far better load times. Sure the cost is high, but if the numbers were what I feel they should have been, I would have gone for a 2xRAID 0 setup with 6x512mb and 2x1gb for 6gb of space, and it would have destroyed even a ultra 320 setup. However, this is just not the case. I am hoping that it peaks interest, and is not a dead product, however, the price increase from $50 to $150 and a initial product run of only 1000 units seems to be a bad sign, perhaps meaning low demand.

My hope is that one day this is properly implemented, either over a faster bus, with better design, or even better, integrated into the motherboard directly. I'd pay over $100 for a feature like that if it offered good performance.

Surely an onboard upgradeable ramdrive in the form of empty DDR slots would be far more valuable than RAID. It would also have more direct access.
 
I have a few points to make about the review, but first, please remember this is a beta card. It could possibly get better in the final implementation.

So with that said,

1. Yes Windows has a small part of the performance problem. The Windows OS is designed for Hard Drives and as such, hasn't been optimised for this kind of speed (but I believe that Windows Vista has optimisations for Ram+HDD combo units).

2. I do believe that it's a combination of the max speed of SATA (should be SATA2 for better performance), and the fact that the Memory speed really doesn't make a difference (so CAS3 or CAS2 doesn't make a difference on this card. This could be some kind of BIOS fix on the Card that Gigabyte could fix (hopefully).

3. I don't think there is a BUS problem with the card, I think its main issue is the actual chip on the Card that needs to be upgraded with some of the improvements listed above.


If the price is around $60 like was first advertised, I would jump on this in a heartbeat, but at $160, they really need to do some further tweaking to make it a better value to the consumer. I guess I will wait for the [H]ard to get a unit and review it before I make my final decision.
 
I can't remember where I saw this, but I think the RAM in iRAM only runs at 33 MHz..
 
neo86 said:
I can't remember where I saw this, but I think the RAM in iRAM only runs at 33 MHz..

with SATA being 1.5Gbps, even DDR RAM that runs at 33MHz should easily saturate the bus.
 
http://www.overclockers.com/tips00813/
http://www.overclockers.com/tips00806/

"This is my big fear about iRAM. When the comprehensive reviews come out, it will be found that it will be very good at cutting down the amount of time it takes to do certain things, but on the whole, it won't greatly increase performance because, like memory, increased speeds in one area has relatively little overall impact on total performance compared to, say, ramping up CPU speed.

I'm afraid a lot of people have impossible expectations for this device, and when the iRAM numbers from the reviews don't meet it, they'll just conclude, "This sucks." "

and

"My own experience with RAMDrives tells me that the benefits in real-time use are mostly going to be a split-second here, a split-second there. If you don't think that's worth paying for the card to get that for a few applications (if you're recycling a little RAM) or even for quite a few apps and Windows (if you buy the card and a lot of RAM), you're probably right. If cost is at all a serious factor in your buying decisions, and you have to prioritize, this is a pretty low priority."

==>Lazn
 
well, IMO, the numbers for solid state should be a lot better. If latency is in NS and not MS, it should make a world of difference if it is executed properly.
 
mikelz85 said:
well, IMO, the numbers for solid state should be a lot better. If latency is in NS and not MS, it should make a world of difference if it is executed properly.

So it makes a huge difference if your highschool yearbook is on your desk vs in your garage if you only read it once every 5 years?

The reason it doesn't make much difference is that drive storage is not a constant use item. If you hard drive light never shut off, then a solid state drive would make a difference, but because your computer only accesses the drive to read something off of it, then keeps it in memory till it is done with it, and then finally writes it back it really is not the big limiting factor people think it is.

Solid state drives have existed for mainframes and high volume databases for decades, they just cost several countries arms and legs. But for most users you don't need them, as drive speed is not that big of a factor.

You will get more performance increase putting that same RAM into system memory than using it as a ramdrive. (esp with 64 bit CPU's that allow for more than 4GB of RAM)

==>Lazn
 
yeah, but executed correctly, a ram drive should boot far quicker than the IRAM does (of course anandtech did mention windows boot process not being optimized for low latency access), and the game load times should be far quicker. Really, with the speed of ram, there's no reason not to be able to boot in seconds, just as if the PC had been suspended.

My point is that the IRAM offers nowhere near the performance of server solid state solutions, and that a quality solid state solution really shouldn't cost all that much. I wouldn't be at all suprised to see 15k SCSI beating out the IRAM, and properly implemented, solid state shouldn't be losing to moving parts.

IMO drive speed is still a large factor. The technology hasn't moved very far in the last 10 years or so. Drives aren't really that much faster now. We haven't see the same rates of progress with processors/ram/gpus etc. with storage devices. Sure the size has increased, but speed is still lacking. Having used a ram drive mapped from my system ram, I can say that the HD is not anywhere in the same league.

I feel that the disk IO system still represents a pretty large bottleneck, given the power of everything else in our systems. Maybe it doesn't matter unless you are doing larger read/writes, but I wouldn't mind cutting a 20-30-40second operation down to 2-3-4 seconds.

Also, I wasn't really aware of basic desktop use that would benefit from adding more than 1gb of ram. XP doesn't seem to be that big of a hog, and I'm not sure how much I could tweak it to use 2gb instead of 1gb. That said, past 2gb I don't think that adding more RAM will do anything, if you see a way to employ it, let me know.
 
mikelz85 said:
and the game load times should be far quicker.
Game loads is mostly CPU as it has to uncompress the data, this is why RAID 0 is not much of a benifit for game load times.
My point is that the IRAM offers nowhere near the performance of server solid state solutions, and that a quality solid state solution really shouldn't cost all that much. I wouldn't be at all suprised to see 15k SCSI beating out the IRAM, and properly implemented, solid state shouldn't be losing to moving parts.
True the IRam is not that good a implemenation but it is cheap. (heck the "CPU" it uses is a Xilinx Spartan FPGA)
IMO drive speed is still a large factor. The technology hasn't moved very far in the last 10 years or so. Drives aren't really that much faster now. We haven't see the same rates of progress with processors/ram/gpus etc. with storage devices. Sure the size has increased, but speed is still lacking. Having used a ram drive mapped from my system ram, I can say that the HD is not anywhere in the same league.
Put in a 10 year old drive and benchmark it compared to a modern 16MB buffer SATA drive. It has drastically improved, it is still the "slowest" part of the system, but the whole point of the HD has always been cheap - large, if that were not the main purpose of the drive we would have developed it a different direction.. Now that we will soon be reaching drives that are large enough for everyone, I expect the emphasis will change to speed.
I feel that the disk IO system still represents a pretty large bottleneck, given the power of everything else in our systems. Maybe it doesn't matter unless you are doing larger read/writes, but I wouldn't mind cutting a 20-30-40second operation down to 2-3-4 seconds.
As I said, the thing that drove drive research was larger, the market shaped it.. If speed were king we would have 1GB 300,000 RPM drives instead of 500GB 7200RPM drives, or we would have solid state, but we don't. But I hope we soon will have faster drives.. I would not mind a faster drive either.

==>Lazn
 
I feel that we have reached a point where it makes a lot of sense to have 2 forms of storage. Many already do this in the form of raptors, having 1 drive or 2 raptors in raid for things needing fast access, and then another HD for all the "storage". To me, having 6gb or so of incredibly fast storage, and then 300+gb of slower storage (long term parking) makes sense. If only the ram drive put up better numbers, because done right, it should completely eliminate the disk I/O bottleneck, which just isn't the case.

Again, my hope is that someday a mobo manufacturer will throw one of these onboard, with a direct pipeline, or at least more direct than the SATA bus. It really wouldn't be all that expensive to do, all we really need is a battery backup, and DRAM can be used for storage. I've been hoping for this ever since I saw RAID blow up, and solid state would be a hell of a lot bigger to users than RAID.
 
Gentlemen, I am the owner of two PCI based Platypus brand RAM drive products, an 8 GB card and a 2 GB card. The Platypus products are 33mHz 32-bit PCI cards that use custom a ASIC coupled with a basic SCSI chipset to interface the RAM to the PCI bus. I must say, they are rock stable, an the drivers work seamlessly with WinXP. Unfortunately, since they do ot interface to the motherboard chipset drive connectors, no real provision to boot from them exists, which is one huge plus for the iRAM. I can tell you from experience, they will not dramatically boost any ROUTINE task that non power users are accustomed to. I was at first dissapointed by the apparent lack of "speed boost", but soon came to realize the truth. YES, they have a sustained transfer rate on my system of 110 Mb/sec; YES, they have virtually no access time latencies. The problem is, as stated by an earlier reply, Windows and the nature of where real speed lies. If you use them for their targeted applications, they will rock. Anyone using Photoshop or a 3D rendering program that gobbles up swap file space and does a lot of "swapping" will benefit enormously. Databases that can have their indices and tmp/temp areas remapped to the RAM drive will also benefit. The nice thing about the iRAM is the price, I would buy one just for my swap file if nothing else.
 
I look at the Gigabyte IRAM as

A good foot in the door. and a step in the right direction.

Gigabyte brought the ability for an entry solid state drive that is flexable with size to the consumer desktop for an affordable price.

It will get better with time. This is the first revision... you know...

Its like getting a brand new Windows OS the DAY its released... shutters... talk about bugs and room for improvment you know?

Things they should do on a next version...
Make the Ram Dual Channel for more bandwidth internally.
4 Slots on top, 4 slots on bottom, ability for 2GB dimms, for a total max space of 16GB if the user wanted.
Make the Memory controller a little more flexable... Higher speeds if the user wants but will work with slow ram too.
A Faster Interface such as a SATA II or by default utilizing 2 SATA II ports for Automatic Raid 0 or Even SAS / SCSI

Just some thoughts. I would like to get a solid state drive. I hate Hard drives... and think that its horrible that we have all this advancement in CPUs and Video cards.... but so little in hard drives.... comparativly.
 
FYI.....
Samsung just released a 32GB flash chip....NAND variety I think...google for the reviews if you need specifics. And they said that 2 of them could be ran in a 64GB setup.

here's the google....link

Now that is getting into the replace a hard drive zone!
Reguardless of access speeds or any preformance issuses....1 of these could make a tinnie-tiny MP3 player/flash drive that could hold insane amounts of data/music.

If the access speeds come up then this could have great possibilities for a permanent solid state drive or iDrive setup.

And the main problem with the speeds of the iDrive is that it is being accessed via the SATA 150 controller.
And you are not going to go any faster that the controller will allow.
Maybe a true SATA 300 controller and the iDrive could be impressive - and maybe not.

The things that I'm currently reading suggest that the SATA 300 / SCSI u320 are bumping the physical limits of the mechanical HD design....i.e. you just can't move reader heads around any faster, even if you can spin the platters faster.

The point of all this is: that to surpass the current max speed systems we'll need new controller technology, probably a new bus design, and faster, high capacity solid state chips. Or some smart cookie will come up with a faster mechanical HD design.

Until all of that comes together you're not going to see the super fast access speeds from a true storage device that you see from an in-memory logical ram drive.

Just my thoughts....I really want it to happen too - but I don't see it for a long time.
But these Samsung chips and the iDrive style of technology are the first steps in that direction.

 
rodsfree said:
FYI.....
Samsung just released a 32GB flash chip....NAND variety I think...google for the reviews if you need specifics. And they said that 2 of them could be ran in a 64GB setup.

here's the google....link

[...]
No they didn't. Samsung has released, in limited quantities, a 16 Gbit NAND IC, which stores 2 GB apiece. Up to 16 chips are supported, for a total of 32 GB. 16 chips can't be practically executed in a "tiny" device any smaller than a 2.5" hard drive form factor.
 
Lazn_Work said:
If you hard drive light never shut off
You called? :p

My hard drives get a workout. Part of this is because I only have 512mb of ram and am regularly 500mb into swap. Part of this is because I'm sharing a lot of files. And part of it is because I'm a Power User. I want one of these, even after I upgrade to 4gb of ram.

 
unhappy_mage said:
You called? :p

My hard drives get a workout. Part of this is because I only have 512mb of ram and am regularly 500mb into swap. Part of this is because I'm sharing a lot of files. And part of it is because I'm a Power User. I want one of these, even after I upgrade to 4gb of ram.


1GB is the minimum these days, I am even getting work PC's fitted with it now. 2GB is better, and a noticable improvement, even for gaming. I have not bothered with more than that except for some servers at work.

==>Lazn
 
rodsfree said:
FYI.....
Samsung just released a 32GB flash chip....NAND variety I think...google for the reviews if you need specifics. And they said that 2 of them could be ran in a 64GB setup.

15MB/sec(ave) for NAND flash memory... Whats the average transfer rate of modern disk drives nowadays?
 
60-75MB/sec in the outer zones, 30-45 in the inner zones for most current drives.
 
J-Mag said:
15MB/sec(ave) for NAND flash memory... Whats the average transfer rate of modern disk drives nowadays?

Flash drives are getting faster... I'm pretty sure that next gen. flash gets around the same speed as modern hard drive... 'cept the speed doesn't degrade... It's always the same anywhere on the drive. :cool:
 
And you can easily stripe the flash chips inside the device. So if you had 4 of the Samsung chips for a total of 8gb, you could get 60 MB/s to/from the device, which isn't bad at all. I understand the seek times are lowered as well (correct me if I'm wrong?). This would make a nice boot drive, and maybe to load games with lots of small files (ut2k4) off of.

 
So taking this from u_m...


unhappy_mage said:
And you can easily stripe the flash chips inside the device. So if you had 4 of the Samsung chips for a total of 8gb, you could get 60 MB/s to/from the device, which isn't bad at all. I understand the seek times are lowered as well (correct me if I'm wrong?). This would make a nice boot drive, and maybe to load games with lots of small files (ut2k4) off of.

And this.....

xonik said:
Up to 16 chips are supported, for a total of 32 GB.

And striping across all 16 chips would give you ....what? 240MB/s
Or is my figuring screwed up?
You'd have to be able to get at least 120MB/s, right?

And that is very respectable, and would give you a desent size drive too.



 
And even more if you load mods. My brother just noticed he was out of disk space (again :rolleyes: ) and looked around for something to delete. He happened on the ut2k4 folder, and checked - it was 12gb.

In any case, my point is that it shouldn't be difficult to stripe the chips inside the device. And the reliability of Flash is higher (let's see you get a head crash, or break the motor) so it shouldn't be as much of a problem as raid 0 is in reliability terms.

Me, I'll wait on higher-density, lower-priced SRAM. That stuff'll be fun to play with.

 
Unfortunately, flash has a limited # of writes you can do before it fails completely. I wouldn't put anything that requires a lot of writes on a flash drive. I'm looking forward to MRAM though. That'll take a while though since density is only like... 128Kbit per chip or something like that.
 
That maybe so, but compared to a regular hard drive it's not that impressive. ;)
 
When the market is ready for flash storage, either the flash technology will be good to go, or the software engineers will design with these limitations in mind. I'm not overly concerned.

In the meantime, there have been talks of hybrid hard drives that incorporate large flash caches into the system, in an attempt to simultaneously minimize slow hard drive accesses and extraneous flash read/writes.
 
I like the hybrid idea. Flash still isn't there density wise, but having a 2GB low latency "Flash Cache" (I had better trademark that, as it has a great ring to it ;)) allows for the best of both worlds - the high density of magnetic storage with the performance and power saving potential of flash. 2GB is plenty of space to boot an OS or perform some other I/O intensive task.
 
Back
Top