I hope not enterprise-only price points. I was looking at one of the NVME Samsung M.2 drives this fall, but if this is early next year and will have something at similar capacities at, say, a 50% premium, I think I'd jump on it.
That would make it what, $300 for 256GB & $600 for 512GB? No clue if that's optimistic or not... There might be extra costs if this ends up requiring new motherboards etc, there's so little details to go on. Hopefully they have U.2 versions like the 750 tho...
For a boot M.2 NVMe drive, I'd gladly pay $200-300 for 128GB with that speed and RELIABILITY. Would be great in addition to a couple of Samsung 1TB 3D TLC SATA drives (on sale hopefully) in RAID 1 for storing the rest of my important files.
I'm seriously considering a Thinkpad P50/70 early next year. ^^These drives will be the icing on the delicious sweat cake.
Maybe we shouldn't be talking about pricing yet? They had said that these would be expensive in the beginning, and for enterprise use. I believe that. I've also read that they said that it wouldn't be priced anywhere near consumer drives for years.
This is a completely new technology, and we all know what that means in terms of pricing.
Probably, tho the X25-M fell dramatically in price from the first gen to the second, so there's a precedent for a quicker drop on their part. Even at $2-4/GB it'd be cheaper than the original X25-M too...
I think you're wildly optimistic. Because of relatively low volumes and specialist applications, you're more likely to see 2x or 4x that initially. Tech prices fall, but always from a high base.
Too few details ,add some more. BTW Micron had a presentation on friday and they showed a roadmap (page 36 in the available pdf). Seems that gen will arrive rather fast at the end of 2016/start of 2017. I imagine they'll just add 2 more layer and no shrink given how fast it arrives.
It was never supposed to be 1000x faster in this scenario. It will be a lot faster (more than 5-7x) when doing write workloads, though.
However THIS kind of performance is exactly what we need. High performance at *low queue depths* ! Pretty much all SSD's have given us lots of performance at high queue depths for a while now, but most real world client workloads are LOW queue depth.
Even 7x could be pretty impressive depending on price... That's a bigger jump than we've seen out of any single SSD since the X25 days. I was skeptical they'd price this in any sort of mainstream way out of the gate but if they're talking about Ultrabooks...
Could this threaten NAND based PCI-E SSD (at 2x the premium of SATA drives) before they even take hold of the market?
NAND based PCI-E SSDs will be available for only slightly more than SATA ones very soon. Why? There isn't a big cost difference to make them and competition will force it.
There'd have to be more of a demand for competition to matter... If Samsung hasn't bothered taking SM951 to retail demand can't be that great. Hope you're right tho, already bought one myself but lower prices would mean more pressure for 3D Xpoint etc etc.
You need a large enough market to justify retail costs. OEM distribution is a lot more efficient: less contacts, less support, single point of negociation, fixed prices, contracted volumes... Retail will come once enough products that can use it are available in the market and an upgrade product would make sense. PCIe NVMe drives will come. The whole market is bullish about it, multiple controller houses have chips for it. Lots of demos for it a Flash Memory Summit early August. Should be here by xmas hopefully or early next year.
Writes is supposed to be where the biggest gains are: - bit access: you do not need to erase a full page to modify one bit - faster writes Intel is revealing the pieces one at a time to keep people interested till 2016.
Did they say anything about data retention? I heard recently that SSD's have a data shelf life of two years (as in, if you unplug it and store it, in two years the data will be gone).
Retention is a different matter than maximum write cycles. Nevertheless, NAND flash data retention is dependent on maximum write cycles. It is also temperature dependent, which lead to a scare a while back. Under normal storage conditions (<= 30 C) Data retention for MLC NAND flash is around 10 years for a new drive, falling to 28 months for a drive that has reached its maximum write endurance. That falls to a matter of days when the storage temperature exceeds 55 C.
But does it matter? Who is going to use expensive SSDs for archival storage?
Intel has already hinted that these things won't be cheap. They are meant to be superior to current SSDs. You will likely see prices of $3-$5 per gig, and this product will most definitely be targeting the enterprise.
They'd probably sell some to enthusiasts/pros even at those prices, it'd make a 128GB about $400-650, 256GB $750-1,250. I'm pretty sure that's STILL cheaper per GB than the X25-M was originally.
Remember that the first SSD cost $3,600 for 32GB. So this will be pretty cheap when compared to that, and may give some idea as to how prices could drop over the years.
In SSDs' case, 50x more IOPS than HDDs bought about 3x single-operation performance. By the same ratio, 7x buys +42%, and if anything it'd be a lot lower since other system components will become the bottleneck. I know these are extremely rough numbers but you get the idea.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
50 Comments
Back to Article
Memo.Ray - Tuesday, August 18, 2015 - link
How much do we pay to "obtain" one of these?!icrf - Tuesday, August 18, 2015 - link
I hope not enterprise-only price points. I was looking at one of the NVME Samsung M.2 drives this fall, but if this is early next year and will have something at similar capacities at, say, a 50% premium, I think I'd jump on it.Impulses - Tuesday, August 18, 2015 - link
That would make it what, $300 for 256GB & $600 for 512GB? No clue if that's optimistic or not... There might be extra costs if this ends up requiring new motherboards etc, there's so little details to go on. Hopefully they have U.2 versions like the 750 tho...lilmoe - Tuesday, August 18, 2015 - link
For a boot M.2 NVMe drive, I'd gladly pay $200-300 for 128GB with that speed and RELIABILITY. Would be great in addition to a couple of Samsung 1TB 3D TLC SATA drives (on sale hopefully) in RAID 1 for storing the rest of my important files.I'm seriously considering a Thinkpad P50/70 early next year. ^^These drives will be the icing on the delicious sweat cake.
Amoro - Tuesday, August 18, 2015 - link
Sweat cake? Gross.lilmoe - Tuesday, August 18, 2015 - link
victim of the edit buttonaakash_sin - Wednesday, August 19, 2015 - link
Haha!Wardrop - Tuesday, August 18, 2015 - link
Hahamelgross - Wednesday, August 19, 2015 - link
Maybe we shouldn't be talking about pricing yet? They had said that these would be expensive in the beginning, and for enterprise use. I believe that. I've also read that they said that it wouldn't be priced anywhere near consumer drives for years.This is a completely new technology, and we all know what that means in terms of pricing.
nandnandnand - Tuesday, August 18, 2015 - link
A good starting point for XPoint is a 16/32 GB M.2 drive for laptops to hold OS and applications.thetuna - Wednesday, August 19, 2015 - link
Yeah... no.We already tried that with dinky SSD caches.
jjj - Tuesday, August 18, 2015 - link
Maybe in 3-4 years. They got no reason to sell it that cheap when they can sell all they make for a lot more.Impulses - Wednesday, August 19, 2015 - link
Probably, tho the X25-M fell dramatically in price from the first gen to the second, so there's a precedent for a quicker drop on their part. Even at $2-4/GB it'd be cheaper than the original X25-M too...prisonerX - Tuesday, August 18, 2015 - link
I think you're wildly optimistic. Because of relatively low volumes and specialist applications, you're more likely to see 2x or 4x that initially. Tech prices fall, but always from a high base.cnxsoft - Wednesday, August 19, 2015 - link
I've just checked the price for a P3700 SSD on Amazon... A 400 GB SSD already costs $900, so the 3D XPoint version should be even more.vred - Tuesday, August 18, 2015 - link
It is unoptaneable.AnnonymousCoward - Wednesday, August 19, 2015 - link
Hahaddriver - Tuesday, August 18, 2015 - link
A kidney would suffice.Timbrelaine - Tuesday, August 18, 2015 - link
Maybe in America. With VAT I wouldn't be surprised to see prices hit two kidneys in Europe.Pietje Bel - Tuesday, November 17, 2015 - link
It wil about have the price of DRAM per GByte.source: https://www.youtube.com/watch?v=tkMXZJ1FwKw
So that's roughly 10 times more expansive as NAND flash memory per GByte!
Pietje Bel - Tuesday, November 17, 2015 - link
I mean: It wil about HALF (1/2) the price of DRAM per GByte!!!!!!Shadowmaster625 - Tuesday, August 18, 2015 - link
Wow we're seriously running out of names....abhaxus - Tuesday, August 18, 2015 - link
Could be worse... The improved version next year could be the iOptane, for maximum punography.kyuu - Tuesday, August 18, 2015 - link
Excited for the technology, but "Optane"? Hard to believe that some asshat in a marketing department actually gets paid to come up with this crap.nandnandnand - Tuesday, August 18, 2015 - link
It could be called Asspain and if the price is right I'll still buy it.lilmoe - Tuesday, August 18, 2015 - link
lol +1AndrewJacksonZA - Wednesday, August 19, 2015 - link
What's wrong with the name AsSpain? Is it some kind of Europe-envy? "We do it AsSpain does it!"Shadow7037932 - Tuesday, August 18, 2015 - link
Still better than the Agilent -> Keysight name change.jimjamjamie - Wednesday, August 19, 2015 - link
I don't see why they don't just go with something like Xpoint i7-8635E, at least it would be familiar bullshitjjj - Tuesday, August 18, 2015 - link
Too few details ,add some more.BTW Micron had a presentation on friday and they showed a roadmap (page 36 in the available pdf). Seems that gen will arrive rather fast at the end of 2016/start of 2017. I imagine they'll just add 2 more layer and no shrink given how fast it arrives.
vlado08 - Tuesday, August 18, 2015 - link
So instead of 1000x it actually is 7x faster?extide - Tuesday, August 18, 2015 - link
It was never supposed to be 1000x faster in this scenario. It will be a lot faster (more than 5-7x) when doing write workloads, though.However THIS kind of performance is exactly what we need. High performance at *low queue depths* ! Pretty much all SSD's have given us lots of performance at high queue depths for a while now, but most real world client workloads are LOW queue depth.
witeken - Tuesday, August 18, 2015 - link
1000x faster is literally faster; lower latency.Impulses - Tuesday, August 18, 2015 - link
Even 7x could be pretty impressive depending on price... That's a bigger jump than we've seen out of any single SSD since the X25 days. I was skeptical they'd price this in any sort of mainstream way out of the gate but if they're talking about Ultrabooks...Could this threaten NAND based PCI-E SSD (at 2x the premium of SATA drives) before they even take hold of the market?
Flunk - Tuesday, August 18, 2015 - link
NAND based PCI-E SSDs will be available for only slightly more than SATA ones very soon. Why? There isn't a big cost difference to make them and competition will force it.Impulses - Wednesday, August 19, 2015 - link
There'd have to be more of a demand for competition to matter... If Samsung hasn't bothered taking SM951 to retail demand can't be that great. Hope you're right tho, already bought one myself but lower prices would mean more pressure for 3D Xpoint etc etc.frenchy_2001 - Monday, August 24, 2015 - link
You need a large enough market to justify retail costs.OEM distribution is a lot more efficient: less contacts, less support, single point of negociation, fixed prices, contracted volumes...
Retail will come once enough products that can use it are available in the market and an upgrade product would make sense.
PCIe NVMe drives will come. The whole market is bullish about it, multiple controller houses have chips for it. Lots of demos for it a Flash Memory Summit early August. Should be here by xmas hopefully or early next year.
lilmoe - Tuesday, August 18, 2015 - link
This is probably without any sort of parallelism.jjj - Tuesday, August 18, 2015 - link
They are insisting on read (and a bit of mixed) so got to wonder how slow write is for now.frenchy_2001 - Monday, August 24, 2015 - link
Writes is supposed to be where the biggest gains are:- bit access: you do not need to erase a full page to modify one bit
- faster writes
Intel is revealing the pieces one at a time to keep people interested till 2016.
prime2515103 - Tuesday, August 18, 2015 - link
Did they say anything about data retention? I heard recently that SSD's have a data shelf life of two years (as in, if you unplug it and store it, in two years the data will be gone).melgross - Wednesday, August 19, 2015 - link
That's been disproven. This is supposed to have 1,000 times the write lifetime, per cell, of NAND. Possibly more.Jaybus - Friday, August 21, 2015 - link
Retention is a different matter than maximum write cycles. Nevertheless, NAND flash data retention is dependent on maximum write cycles. It is also temperature dependent, which lead to a scare a while back. Under normal storage conditions (<= 30 C) Data retention for MLC NAND flash is around 10 years for a new drive, falling to 28 months for a drive that has reached its maximum write endurance. That falls to a matter of days when the storage temperature exceeds 55 C.But does it matter? Who is going to use expensive SSDs for archival storage?
eek2121 - Tuesday, August 18, 2015 - link
Intel has already hinted that these things won't be cheap. They are meant to be superior to current SSDs. You will likely see prices of $3-$5 per gig, and this product will most definitely be targeting the enterprise.Impulses - Wednesday, August 19, 2015 - link
They'd probably sell some to enthusiasts/pros even at those prices, it'd make a 128GB about $400-650, 256GB $750-1,250. I'm pretty sure that's STILL cheaper per GB than the X25-M was originally.gospadin - Wednesday, August 19, 2015 - link
The original Intel X25-M 80GB was $595, which is about $7.50/GB.http://www.anandtech.com/show/2614
valnar - Wednesday, August 19, 2015 - link
Take my money now!melgross - Wednesday, August 19, 2015 - link
Remember that the first SSD cost $3,600 for 32GB. So this will be pretty cheap when compared to that, and may give some idea as to how prices could drop over the years.AnnonymousCoward - Wednesday, August 19, 2015 - link
So how do you convert IOPS to actual performance?In SSDs' case, 50x more IOPS than HDDs bought about 3x single-operation performance. By the same ratio, 7x buys +42%, and if anything it'd be a lot lower since other system components will become the bottleneck. I know these are extremely rough numbers but you get the idea.
TelstarTOS - Friday, August 21, 2015 - link
Shut up and take my money!