Do you use SSD's now or plan to buy?

[Topper]

If you look the velociraptor one year ago the price was the same but look at this SSDs the price is getting lower every 4-6 months so in one year the velociraptors will cost more than any SSD in the same capacity.

That's why I currently have a velociraptor but will upgrade mine with an SSD when the 320GB cost $200-$300.

That's the price range I am looking for but I thought that was unreasonably low.
Maybe in a year?
.

 

[Tesselator]

I don't doubt you.... but....

Found here. Thank you XKCD.

I just don't think that a linear model for SSD pricing is adequate at this moment in time - I think we may be hitting lower and lower price drops. Starting to see the exponential, for lack of a better term. Of course, my speculation could be wrong and I'm just being an ass.

Awesome projection dynamics! Math at work in the real world! I love it!

 

[nanofrog]

I don't doubt you.... but....

Found here. Thank you XKCD.

I just don't think that a linear model for SSD pricing is adequate at this moment in time - I think we may be hitting lower and lower price drops. Starting to see the exponential, for lack of a better term. Of course, my speculation could be wrong and I'm just being an ass.

Somehow, I doubt linear applies either.

That's the price range I am looking for but I thought that was unreasonably low.
Maybe in a year?
.

I think it will take a bit longer than that, considering the fact the 1K unit price = $440 for 160GB listed on the G2 version. It won't go 2x the capacity at the same price, let alone lower. Maybe in 2 - 3 years (3 being more likely IMO), but not a single year.

 

[VirtualRain]

That's the price range I am looking for but I thought that was unreasonably low.
Maybe in a year?
.

Does anyone recall when the first WD 74GB Raptors hit the market in 2004/05?... they were around $300. Then the 150GB drives came along... still at $300... then the 300GB VR's, again around $300. Now top-performing SSD's are around $300. The fastest desktop drives at any given time will likely always command a $300 price tag... but they will get larger and faster over time.

Traditionally this has been my observation...
The fastest desktop drives were always around $300
The largest desktop drives were always around $200
The best value drives were always around $100
The cheapest/smallest desktop drives were always around $50

Over time, I think SSD's will follow the same pattern. The newest fastest drives will always command a premium... which is $300-$400 now for a top performing SSD. My guess is that if your drive budget is $200, you can't expect to buy the top performer of the day... eventually you will get the capacity you want but I suspect it won't be the best performer .

 

[bajee]

I think it will take a bit longer than that, considering the fact the 1K unit price = $440 for 160GB listed on the G2 version. It won't go 2x the capacity at the same price, let alone lower. Maybe in 2 - 3 years (3 being more likely IMO), but not a single year.

I beg to differ, you do know that the intel x25-M (50nm) launched last year around september was around $595

http://news.cnet.com/8301-17938_105-10034775-1.html?part=rss&tag=feed&subj=Crave

and now the upgraded version with intel x25-m (34nm) launched around 225$, that's more than 50% off, and its not even September, that's less than a year

 

[wpc33]

that's more than 50% off, and its not even September, that's less than a year

My bootdisk's future's so bright

 

[nanofrog]

I beg to differ, you do know that the intel x25-M (50nm) launched last year around september was around $595

http://news.cnet.com/8301-17938_105-10034775-1.html?part=rss&tag=feed&subj=Crave

and now the upgraded version with intel x25-m (34nm) launched around 225$, that's more than 50% off, and its not even September, that's less than a year

That happens to be for the 80GB versions. A 320GB doesn't exist yet, and will be slower to fall, as you'd need multiple generations (2 min), and larger models to help push prices down.

From the article you linked, lets use the $7.43/GB breakdown. At 320GB, that would be a price tag of $2377.60USD. Rather pricey. So to keep the prices low enough to attract buyers, they'd wait until they can get the retail $/GB down. Given the fact Intel wants to pay for the fab shrink, and there's no 320GB model, there's no way it will fall that quickly in a year.

Even now, you can get the smaller drives for less $/GB, so making a stripe set is not only cheaper, but faster as well. So figure 4x 80GB models (G2), you get essentially 4x the speed, so near 1TB/s reads, for ~$900USD (higher at retail prices, but not drastically). Faster speeds, and less money.

Given this particular fact, which way would most people prefer to go, assuming they've the budget?

 

[VirtualRain]

... So figure 4x 80GB models (G2), you get essentially 4x the speed, so near 1TB/s reads, for ~$900USD (higher at retail prices, but not drastically). Faster speeds, and less money.

This raises an interesting subject on it's own. Intel and other SSD manufacturers are swiftly exposing new bottlenecks in I/O architecture.

Intel is going to need to focus on it's I/O Hub performance before we can really leverage the true power of multiple SSD's in RAID0.

Apparently, the ICH10 tops out at around 660MB/s of throughput, indicating that you can't even enjoy the full potential of 3 SSD's in RAID0.

So those seeking more storage performance would naturally be forced to turn to a high-end RAID card... However, the next bottleneck is the DMI bus between the IOH which terminates all the non PEG PCIe lanes and the X58 which is a 4x PCIe express bus that tops out at 1TB/s... so even if you ran four SSD's on a high-end RAID card on a PCIe 8x connector, you are still going to run into a bandwidth cap on the DMI link.

If SSD's continue to improve significantly, the entire I/O architecture of a PC is going to need to change.

I recall reading somewhere that Intel is actually planning a high-speed flash interface on the chipset. It's not surprising.

 

[nanofrog]

This raises an interesting subject on it's own. Intel and other SSD manufacturers are swiftly exposing new bottlenecks in I/O architecture.

Intel is going to need to focus on it's I/O Hub performance before we can really leverage the true power of multiple SSD's in RAID0.

Apparently, the ICH10 tops out at around 660MB/s of throughput, indicating that you can't even enjoy the full potential of 3 SSD's in RAID0.

Might I trouble you for a link?

So those seeking more storage performance would naturally be forced to turn to a high-end RAID card... However, the next bottleneck is the DMI bus between the IOH which terminates all the non PEG PCIe lanes and the X58 which is a 4x PCIe express bus that tops out at 1TB/s... so even if you ran four SSD's on a high-end RAID card on a PCIe 8x connector, you are still going to run into a bandwidth cap on the DMI link.

Actually, it's PCIe 2.0, so double the bandwidth. A meager 4x is capable of 2.0TB/s. So it's not quite that drastic.

I do want to spend some time looking at this though, as it seems odd that it would be this dismal an outlook (planning due to simultaneous development by the same company).

Most of it has to do with the fact systems are still centered around mechanical drives for cost reasons. So there's been no optimization yet for SSD's. This will change, but won't be complete until SSD completely replaces HDD technology. Personally, 5 years min. isn't too far off IMO, before $/unit capacity drops to mainstream affordability.

If SSD's continue to improve significantly, the entire I/O architecture of a PC is going to need to change.

Hopefully, this will be the case. Without something to push improvements, the technology will stagnate.

I recall reading somewhere that Intel is actually planning a high-speed flash interface on the chipset. It's not surprising.

Again, can I trouble you for a link?

Thanks.

 

[VirtualRain]

Might I trouble you for a link?

My, you are demanding! I'll try to dig up some more info this weekend.

In the mean-time a quick search discovered these...
http://www.ocztechnologyforum.com/forum/showthread.php?t=54207
http://www.xtremesystems.org/forums/showpost.php?p=3641646&postcount=29
http://www.ocztechnologyforum.com/forum/showpost.php?p=366570&postcount=15

Actually, it's PCIe 2.0, so double the bandwidth. A meager 4x is capable of 2.0TB/s. So it's not quite that drastic.

Actually, I think I was mixing up my T's with my G's...I'm pretty sure it's 1.1... only the PEG slots use PCIe 2.0. (I think).

Actually according to these links, DMI is 10Gb/s (1.25GB/s!!! ) http://en.wikipedia.org/wiki/Direct_Media_Interface
http://en.wikipedia.org/wiki/ICH10#ICH10

Either way, this link has to be shared with all the I/O except graphics (networking, storage, USB, etc.)

SSD's are quickly stressing the rest of the I/O subsystem that was previously never considered an issue.

 

[nanofrog]

My, you are demanding! I'll try to dig up some more info this weekend.

In the mean-time a quick search discovered these...
http://www.ocztechnologyforum.com/forum/showthread.php?t=54207
http://www.xtremesystems.org/forums/showpost.php?p=3641646&postcount=29
http://www.ocztechnologyforum.com/forum/showpost.php?p=366570&postcount=15

Actually, I think I was mixing up my T's with my G's...I'm pretty sure it's 1.1... only the PEG slots use PCIe 2.0. (I think).

Actually according to these links, DMI is 10Gb/s (1.25GB/s!!! ) http://en.wikipedia.org/wiki/Direct_Media_Interface
http://en.wikipedia.org/wiki/ICH10#ICH10

Either way, this link has to be shared with all the I/O except graphics (networking, storage, USB, etc.)

SSD's are quickly stressing the rest of the I/O subsystem that was previously never considered an issue.

Of course. Thanks for the links BTW.

The limit isn't the DMI, but the ICH10/R's limit at 660MB/s. As it stands, DMI is capable of more (1.25TB/s as you found out ), and future versions will get wider to handle the additional bandwidth.

I still want to look deeper into the ICH10R though, as to the cause of the cap. So I'll be searching around as well. So far, what I've come across, concerns ASUS boards, one of which I have (P6T6 WS Revolution). So I'm certainly interested. Though my gut tells me either an ARC-1231ML or ARC-1680ix8/12/16/24 would eliminate this issue. Not a cheap "solution", when users would expect the ICH10R to handle SSD's (5 anyway, before hitting the wall).

I'm thinking either firmware or chipset drivers (especially if it's limited to a single board maker), but I need more time. Not much is popping up, but the search parameters I've entered may be too specific.

Anything you can find out, would be helpful, especially if you stumble across information pertaining to other boards, not just ASUS.

 

[VirtualRain]

Of course. Thanks for the links BTW.

The limit isn't the DMI, but the ICH10/R's limit at 660MB/s. As it stands, DMI is capable of more (1.25TB/s as you found out ), and future versions will get wider to handle the additional bandwidth.

Now you are mixing your T's and G's... the DMI bandwidth is stated as being 10Gbps on Wikipedia... or 1.25GB/s... It would only take 5 SSD's on a decent RAID controller to max that out.

I still want to look deeper into the ICH10R though, as to the cause of the cap. So I'll be searching around as well. So far, what I've come across, concerns ASUS boards, one of which I have (P6T6 WS Revolution). So I'm certainly interested. Though my gut tells me either an ARC-1231ML or ARC-1680ix8/12/16/24 would eliminate this issue. Not a cheap "solution", when users would expect the ICH10R to handle SSD's (5 anyway, before hitting the wall).

I'm thinking either firmware or chipset drivers (especially if it's limited to a single board maker), but I need more time. Not much is popping up, but the search parameters I've entered may be too specific.

Anything you can find out, would be helpful, especially if you stumble across information pertaining to other boards, not just ASUS.

The OCZ SSD forums are a good place to find hardware enthusiasts that have railed the limits of their ICH with multiple Vertex drives in RAID0.

EDIT: Here's a link to the technology I mentioned where Intel is incorporating some kind of flash technology on the chipset. It's called Braidwood, and it seems their intent is to have some NAND flash cache on the north bridge. Access should be super fast as it's only a QPI hop away from the CPU.

http://news.cnet.com/8301-13924_3-10258748-64.html

 

[Andrew Henry]

Already have em and I'd NEVER go back!

 

[nanofrog]

Now you are mixing your T's and G's... the DMI bandwidth is stated as being 10Gbps on Wikipedia... or 1.25GB/s... It would only take 5 SSD's on a decent RAID controller to max that out.

DOH...Rather easy enough to do isn't it.

The OCZ SSD forums are a good place to find hardware enthusiasts that have railed the limits of their ICH with multiple Vertex drives in RAID0.

I haven't put much time into this one yet, but I'll take a look. So far, all the references I've located still seem to be with the ASUS boards. I'll give a good reading. Thanks for the link.

EDIT: Here's a link to the technology I mentioned where Intel is incorporating some kind of flash technology on the chipset. It's called Braidwood, and it seems their intent is to have some NAND flash cache on the north bridge. Access should be super fast as it's only a QPI hop away from the CPU.

http://news.cnet.com/8301-13924_3-10258748-64.html

Thanks.

It's interesting, but not a fundamental fix for SSD (or others) exceeding the various bus limits. Other companies, such as board makers, have done things that remind me of it. ASUS's Express Gate OS comes to mind (a Linux install in flash soldered directly to the board). Boot times claimed at 10 seconds. I've not actually tried it, but maybe I should give it a go out of curiosity.

 

[cube]

Too expensive, too small, and none of those in disk format have both good random and sequential write throughput.

 

[bajee]

From the article you linked, lets use the $7.43/GB breakdown. At 320GB, that would be a price tag of $2377.60USD. Rather pricey. So to keep the prices low enough to attract buyers, they'd wait until they can get the retail $/GB down. Given the fact Intel wants to pay for the fab shrink, and there's no 320GB model, there's no way it will fall that quickly in a year.

you totally don't get my point, that link was last year 2008. That was when prices of SSD are so expensive.

So imagine this, last year was 7.43/GB, and at the present intel is offering $2.81, that's more than half the price in less than a year! It also means, you get the 160gb version in much lower price than that of the 80gig, september last year.

Now isn't the 160gb twice the capacity of 80gig and Intel is offering it at much lower price than last year?

you're telling us it will be 3 years, but not a single year, and I'm presenting you facts, in less than a year, the price was cut down more than half.

Now you do know that with the new 34nm technology, Intel can fit twice as much on the same SSD?

See this, the back of the new Intel X25-M 160GB

Notice the whole back, is intentionally left blank?

from anandtech :

Intel is now using 16GB flash packages instead of 8GB packages from the original drive

So Intel can definitely produce 320gb if they wanted to. As for why they aren't releasing it yet, most people can infer that, Intel will not only just slap memory modules on that side, but also utilize internal raid technology which is prevalent on the 2nd generation SSD nowadays, like the one found in OZC. Making the 320gb faster than their 160gb and 80gb siblings

 

[nanofrog]

that link was last year 2008

I didn't pay attention to the date, and based the calculations on the cost/GB listed in the article.

...and at the present intel is offering $2.81

It's definitely better than last year. But you're assuming the retail will follow the same cost per gig on a new model that's never been offered.

But keep in mind, even though it scales that way on the manufacturing cost end of the equation, it doesn't always translate to the retail prices. ~$900 for 320GB would be really nice, if they actually hit the street at that. But this is dependant on capacity as well. It's a new line, and the production quantities may not be able to support such a drive just yet. This might explain why it wasn't released at the same time as the 80 & 160GB versions. And yes, I'd expect them to have planned a 320GB model from the beginning of the G2 series, so the PCB does have the room for the necessary components.

An increased markup isn't out of the realm of possibilties either, particularly due to limited supply (at least initially, when they do hit). I guess I've just seen situations like this too often, and have reservations about the potential/promises of initial low cost. At least the performance side isn't a disappointment. So I'll wait to see what they go for on the street when they finally arrive.

Hopefully, it won't be an issue for those willing to part with $900.

 

[brendon2020]

price is still to high for such low capacity drives, but they're getting close.

 

[theimacdude]

I don't see myself getting an SSD for my MacBook Pro anytime soon. The capacity just wont do for me. When they make then cheaper and higher capacity, then I would maybe give them a chance.
I'm looking to upgrade my 250GB In my MacBook Pro to a 500GB, so as you can see I don't think I'll be buying an SSD Anytime soon - unless they do ultra cheap, 500GB SSD's?
I can pick up a 500GB Hard Drive for £65, or get a 128GB One for 4 times the price, TBH, the speed is fine on my MacBook Pro, and 128GB is not enough. Maybe in a machine like a MacBook Air, but for a MacBook Pro, I don't see why anyone would want an SSD. Bare in mind this is a PRO machine, meant for things like, Photoshop, Final Cut, etc I guess the speed would be great the size just is not.... right for a pro machine.

 

[VirtualRain]

I don't see myself getting an SSD for my MacBook Pro anytime soon. The capacity just wont do for me. When they make then cheaper and higher capacity, then I would maybe give them a chance.
I'm looking to upgrade my 250GB In my MacBook Pro to a 500GB, so as you can see I don't think I'll be buying an SSD Anytime soon - unless they do ultra cheap, 500GB SSD's?
I can pick up a 500GB Hard Drive for £65, or get a 128GB One for 4 times the price, TBH, the speed is fine on my MacBook Pro, and 128GB is not enough. Maybe in a machine like a MacBook Air, but for a MacBook Pro, I don't see why anyone would want an SSD. Bare in mind this is a PRO machine, meant for things like, Photoshop, Final Cut, etc I guess the speed would be great the size just is not.... right for a pro machine.

Would you not benefit from a smaller faster boot drive in your laptop and then a larger external or NAS setup for whatever is consuming most of that space?

I have iWork, iLife, CS4, FCS, and Office 2008 on my boot/apps drive and it's only consuming 40GB. So even a single 80GB SSD would be plenty for your OS/apps and working files. All my FCS content is on a separate 1TB drive. Could you not do the same?

 

[lord patton]

I too think SSD drives will continue to fall rapidly in price. That ~128 GB drive may not be too far from $100. It's certainly not 24 months away. Three factors:

1) Controller Tech
2) Firmware and support margins
3) Flash pricing

1) While there are a lot of SSD brands now, most all of them are using the Indilinx controller. So there's not much price pressure on the controller, and I figure that will change. Surely there will be other controllers soon that don't stutter and lag, and the price of all the controllers will drop.

2) The RAM costs what it costs, the controller costs what it costs, but the cost of developing the firmware and support tools for the SSD is spread out over all the units that a manufacturer can sell. As adoption rates rise and they sell more, they'll be able thin the margins a little bit, especially as there are so many brands competing in this space.

3) As I understand it, a lot of factories in Asia are shuttered right now due to the global recession. Prices haven't dropped a lot because the supply tightened up. As things start humming, those factories are going to start producing product, and that might give a significant, "non-linear" drop in prices.

These are mostly back-of-the-envelope considerations—I would welcome anyone with a more robust knowledge of the details to chime in. Still, I expect SSD prices to fall quite a bit, probably halve their price every 8-10 months.

 

[bajee]

1) While there are a lot of SSD brands now, most all of them are using the Indilinx controller. So there's not much price pressure on the controller, and I figure that will change. Surely there will be other controllers soon that don't stutter and lag, and the price of all the controllers will drop.

I think you are mistaken, you are referring to the JMicron controller, that was prevalent on generation 1 SSDs. Only a handful are using the Indilinx controller, which is what we call generation 2 SSDs. The JMicron controller, is the one producing lag and not the Indilinx.

Basically Indilinx was founded by ex-Samsung engineers who formed their own company. So its basically an improved version of the samsung controller. Moreover, the generation 2 SSDs used Indilinx and Samsung controllers, so none of the controllers are exclusive, since there is an option. The only problem with the samsung ang indilinx controllers are they slow with aging, and also their random write performance is miles away of that of Intel's.

It's definitely better than last year. But you're assuming the retail will follow the same cost per gig on a new model that's never been offered.

Well, most of us assumed that SSD prices won't go down much within a year. About 95% of people are predicting 3 to 4 years . Now Intel just slap their face with their new pricing, while offering a price that's 1/3 of its original launching price. So being pessimistic doesn't make technology advancement to go slower

2) The RAM costs what it costs, the controller costs what it costs, but the cost of developing the firmware and support tools for the SSD is spread out over all the units that a manufacturer can sell. As adoption rates rise and they sell more, they'll be able thin the margins a little bit, especially as there are so many brands competing in this space.

3) As I understand it, a lot of factories in Asia are shuttered right now due to the global recession. Prices haven't dropped a lot because the supply tightened up. As things start humming, those factories are going to start producing product, and that might give a significant, "non-linear" drop in prices.

These are mostly back-of-the-envelope considerations—I would welcome anyone with a more robust knowledge of the details to chime in. Still, I expect SSD prices to fall quite a bit, probably halve their price every 8-10 months.

You do know that the answer to global recession is lowering the prices? Much like what is Apple is doing now by lowering their macbook pro lines. All of the their other lines will soon follow.

 

[nanofrog]

Well, most of us assumed that SSD prices won't go down much within a year. About 95% of people are predicting 3 to 4 years . Now Intel just slap their face with their new pricing, while offering a price that's 1/3 of its original launching price. So being pessimistic doesn't make technology advancement to go slower

It's realism to me. As I said, I've seen too many unfulfilled promises based on a substantial dose of blind optimism.

For most, the current pricing is still too high for mainstream (replaces HDD in most, if not all systems). People, and particularly system vendors want inexpensive, large capacity drives. SSD is neither ATM. This will change, but it's not there yet. Perhaps when it gets to ~$0.20/GB, proliferation will start to seriously displace HDD's (currently going for ~$0.10/GB on consumer models; 1TB units @ 7200rpm).

You do know that the answer to global recession is lowering the prices? Much like what is Apple is doing now by lowering their macbook pro lines. All of the their other lines will soon follow.

To an extent, but it still has to cover costs (direct & indirect), combined with a profit. Shareholders tend to be greedy, and Intel's are no exception. There is real issue of having to pay for the 34nm facility, so it counters the lower production costs to some extent. New lines also tend to need some kinks worked out, and until sorted, yields tend to be lower than the fab's capacity. This is likely the biggest reason for the lack of the 320GB model. Just not enough supply yet, to manage placing it in production. Yet.

I've seen component costs go crazy just after release over supply/demand (it's a very volatile market). Intel does sell parts to other vendors, but they're not selling the controllers. The 34nm flash NAND, almost certainly yes when the fab is running at full capacity, as it's a large market. Last I saw though, it's only running at 50%.

What I'm hoping, is Samsung's 30nm 64Gb MLC chips (also a 32Gb SLC as well) ship in drives, so competitive products will also ship. Good old fashioned competition. They're already out, but like IM Flash Technologies (actual company name making the 34nm 32Gb MLC component), it's also new, and will have the same spin up issues.

I'm still holding some reservations on a lowish MSRP until I see the 320GB actually release.

 

[lord patton]

I think you are mistaken, you are referring to the JMicron controller, that was prevalent on generation 1 SSDs. Only a handful are using the Indilinx controller, which is what we call generation 2 SSDs. The JMicron controller, is the one producing lag and not the Indilinx.

Basically Indilinx was founded by ex-Samsung engineers who formed their own company. So its basically an improved version of the samsung controller. Moreover, the generation 2 SSDs used Indilinx and Samsung controllers, so none of the controllers are exclusive, since there is an option. The only problem with the samsung ang indilinx controllers are they slow with aging, and also their random write performance is miles away of that of Intel's.

I should have written more clearly. Yes, Indilinx are the "good" ones, and the the previous JMicron controllers produced the lag. I know that. I guess I was talking from the point of view that Intel, Samsung, and Indilinx-based SSDs are the only ones worth considering. In that group, there are a lot of Indilinx based brands, and not too many other choices. Do either Samsung or Intel sell their controllers to third paries?

So when it comes to controllers with acceptable performance, n=2. There's room for another "option" or two. When there are more, prices will drop.

Interesting info regarding Indilinx's history. Do you know anything regarding their plans for a next generation controller? Are there other vendors with products in the pipe?

You do know that the answer to global recession is lowering the prices? Much like what is Apple is doing now by lowering their macbook pro lines. All of the their other lines will soon follow.

It's not that simple. Some vendors, like Apple, will have margins that can support price cuts, and may even make up for a price cut with increased volume.

But that's not the situation with commodity silicon. Often the price falls so much that factories can't sell their product above their cost. That was the case before the crash of late '08. There had been so much increase in production facilities during the boom years that some companies could barely stay profitable, even when running at full capacity.

Then the recession hit, and a lot of these fabs went under. I assume they defaulted on their loans and their facilities are owned by banks. They're not producing product and that's restraining supply. That's keeping prices of wholesale Flash from dropping much over the past 10 months.

As the economic outlook improves, some investors will buy those factories from the banks, for less than they originally cost (gotta love foreclosure firesales). Production will ramp up and prices will drop.

Just my prediction. I'm no expert "analyst".

 

[VirtualRain]

I think SSD pricing will plummet fairly quickly... it's not going to be a protracted affair that takes years for SSD's to become mainstream.

The reason is that there are simply too many tangible benefits. It's not like some marginal technology improvements of the past.

There's a significant pent-up demand... looking at the results of this poll (65% are waiting for prices to drop)... the first manufacturer to get prices to mainstream levels will capture a huge opportunity. Also, as prices drop, more people will buy which will accelerate the cycle. There's a race going on to win this market. It will happen sooner than later IMHO.

What strikes me as particularly odd is that neither WD or Seagate seem interested.