Intel halves SSD power draw with 20nm chip shift Intel has updated its line of low-end solid-state drives by introducing the 335, a 240GB 6Gb/s Sata unit based on 20nm Flash chippery. The chip giant claimed the 335 delivers up to 500MB/s sequential read and 450MB/s sequential write speeds. For random read and writes, it can do 42k IOPS and 52k IOPS, respectively, Intel said …
Because IOPS is about as good a statistics as BogoMIPS.
Operations per second means nothing. Is the operation a 1-byte delete, a 1Mb-read or a 1Gb write? Without further context, the IOPS number means nothing.
Even stating bandwidth - i.e. MB/s - can be deceptive and wrong and vary enormously over time for a storage device.
Same when you state the same for databases - transactions per second? What kind of transactions? Without further context (which is never provided and/or never consistent across manufacturers) you might as well be stating random numbers.
Rule #1 - never buy anything based on statistics without TESTING your workload first.
With respect to Lee's Rule #1 and other points (which I completely, wholeheartedly agree with)... those IOPS numbers are not completely ridiculous for an SSD. PCIE SSDs routinely clock in the 6 figures for IOPS and HDD-replacement SSDs usually clock in the 5 figures. To the contrary, an SSD that only has 3k IOPS is very unusual. I wouldn't be surprised if those EMC drives have been intentionally throttled for some reason.
While shady performance metrics are not unusual in the storage arena, Intel is typically considered to be conservative with their benchmarks. Put another way: I am not aware of any reviews where they have tested Intel SSDs and saw actual performance below the benchmarks.
Warning, crappy car analogy: I can take an Evo, soup it up and get 500+ HP in it and have great fun in it... but I'm not going to throw a trailer behind it and use it to haul freight 24x7.
Consumer SSDs *do* have great performance but there are other factors in the enterprise space that must be considered... like write endurance (SLC vs. eMLC vs. MLC vs. TLC), reliability, form factor and power consumption. There are very good reasons why our servers aren't stuffed full of OCZ Vertex Max IOPS Edition SSDs ; )
I don't want it faster.
I don't want it smaller.
I don't want it with prettier software.
Gimme some damn data capacity on the "old" speeds / form factors / software / prices and I'll buy one TODAY. In fact, I'll have two.
Seriously, SSD is a huge enough leap without poncing about with a 0.1% improvement here and there. But what stops me upgrading is that I'd have to have FOUR of them in my laptop to not lose space. I *do* happen to have a laptop with 2 drive bays, but I'm very fortunate to have that and not everyone has that option. But still both drives I use are 1Tb, so whatever I do I sacrifice huge amounts of storage space, at great additional cost, for a speed increase of the resulting tiny storage space. So I'm stuck with spinning disks for the moment.
The consumer market for these would be HUGE at the right price because you can make a PC appear twice as fast to those inexperienced with computers just by reducing RAM / CPU and slapping one of these in (and those people barely use all their storage anyway). But few target that market with SSD.
The nerd/gamer market would be HUGE at the right price because you make the PC fly, but those people tend to have LOTS of storage that they would need to sacrifice. And few target that market with SSD.
In fact, I can't really work out what the market is for SSD at the current prices. Guinea pigs?
Gimme a 1Tb at something less than my laptop cost in the first place (and no more than 1.5 times the cost of the 500Gb, which should be no more than 1.5 times the cost of the 256Gb - meaning 1Tb = 2.25 * 256Gb) and I'll buy it. Maybe two if it's cheap enough.
People seem to forget that the SSD is a STORAGE device. If it lags behind other storage devices in, well, the storage stakes, then it's a bit pointless. I'd gladly sacrifice 50% of the speed increase of SSD over hard disk for double the capacity (and don't get me started on "hybrid" SSD disks - that's NOT what we're talking about) or even double the reliability.
Abso-bloody-lutely! You've pretty much found the only reason that prevents me from replacing all my apparently 'inferior' storage (according to the SSD zealots, that is) with solid state.
Capacity sucks. It's THAT which needs improvements, not minor increments on IOPS or overall transfers speeds. We need a proper reason to shift to the newer medium, and it's not going to happen until we have devices of equal or greater capacity of our existing drives for no more than 1.5x the cost. I keep hearing about the huge speed boosts and the like that SSD users apparently gain, but really... I'm a gamer mostly. My Steam drive alone is around 3/4 full, and that's a 1TB unit. Speed is meaningless if you barely have enough room for 1/3rd of your stuff to begin with.
So yes. Come on, you manuifacturers of such arcane electrickery, get your bloody acts together and start making devices with SIZE, we already have the speed.
SSD is significantly more expensive than rotational drives but it's also typically more reliable, far, far, faster for most applications so you cannot really compare it for mass storage (unless you are loaded).
A 3TB SATA drive can be had for about the same cost as 256Gb of SSD. Manufacturers could make larger capacity drives but it would be a very small market unless you reckon home users would pay £1200 for a 3Tb SSD drive when they could get a 3Tb rotational one for £100.
I think you miss the point.
If a 256Gb costs, say, £100 (ballpark), then a 1Tb drive with THE SAME CONTROLLERS AND THE SAME CHIPS but just more of them in the same space (which is exactly how SSD's are typically constructed - pop one open and see the blank "template" PCB with gaps for extra chips that would be used in the larger capacity drives), then it should cost NO MORE than 4 times the price. Assuming economies of scale and that a lot of the cost is in the PCB's, encasement and the drive controller itself and 2.25 times the price is about right. In fact, it costs nearer ten times the price.
So where does their profit go? To sourcing more chips en masse, assembling them cheaper, giving us larger drives? No. It goes into FASTER SSD's (which are already not selling as many as hard disks because people need storage in a disk, not speed) and where 5% improvement in speed is undetectable. We don't want new designs, new controllers, new boards, new chips, and all the costs of the R&D that goes into that. We want what you were producing 5 years ago for the same thousands-of-pounds it still is today, but at a sensible price. The speed is still stupidly fast over hard disk, the capacity goes past modern hard disks if you want, but the price is still ludicrously high despite being a "proven" technology that you've been making for nearly a decade. And the available sizes? Not changed that much since SSD's first became viable.
Your competition? Hard disks that cost one-tenth of the SSD and manage half-the-speed (or better with appropriate caching) anyway.
This is the point - consumers, like me, WON'T pay £1200 (actually nearer £5000) for a 3TB SSD - something that has the IDENTICAL chips from 6 £150-ish 512Gb drives in a single box. But I'd happily buy lots and lots of 1Tb drive for, say, £150 each, even if the technology was first-generation SSD.
This is my complaint - they are sold as storage devices, they replace storage device, yet in the critical area (storage!) they are lacklustre because of price. Whereas even the most basic and ancient SSD device has a speed that still kicks hard disks into the ground and stomps on them again. So what do they improve for generation upon generation of the device? The speed. Not the capacity.
Just when was the last time a consumer bothered to check the transfer speed of his hard disk? Now when was the last time he bought a new one because he'd filled it up, needed an external backup, or just needed to replace his existing disk? To make him compromise storage for speed is THE WORST pay-off when you have the technology just sitting there.
P.S. I wouldn't play the SSD reliability card in your first paragraph if I was you. In theory, yes, in practice, no.
Nobody ever said you have to replace all of your storage with solid state - I can't even begin to imagine what you're doing as a regular consumer that needs 3TB of SSD storage (which is to say: latency sensitive random IO.) If you're running some sort of massive database, sure, but then you're solidly into the enterprise market. SSDs and standard rotational magnetic storage have different strengths, and they're best used in some sort of combination. For most consumers, 160GB is plenty for the files which are actually read and written frequently, the rest can easily be stored on cheap magnetic disks.
SSDs give you the ability to pay for higher performance where you need it, and if you've evaluated your situation and decided that the performance isn't worth the price, then that's fine, but it's hardly true for everyone.
As for your claim that spinning drives manage half the speed of SSDs, that's only true for sequential IO. Once you start doing random IO (or merely more than one sequential operation per disk,) even the crappiest SSD will start to pull out a significant margin. Last time I bothered to benchmark spinning disks, year ago, the fastest 15krpm 2.5" SAS drive was more than 20 times slower than a then-current SSD under random IO. Since then SSDs have roughly quadrupled in speed, and spinning disks have gained maybe 25%.
First gen SSD controllers, by the way, are horrid (you're probably thinking of 3rd gen, at least,) and also can't even begin to operate with 1TB of flash. You can't just keep adding chips to a controller because: A) new, high-density chips use different interfaces, not compatible with old controllers, B) Controllers have a limited number of interface channels, each of which support a limited number of flash dies, C) Controllers are just now becoming able to address and manage that much NAND, due to hardware limitations in the controller. Old controllers (3rd gen for instance) probably can't handle more than 128GB (or 160GB for Intel, due to their odd architecture,) and there's no way around that, short of making a new controller.
And flash prices don't scale linearly with capacity because higher density chips cost more per Gb, and large drives require the use of multi-die flash packages, which also increase price per Gb.
Your argument is like I want my F1 car to have 7 sets and carry a family with luggage yet cost as much as a Ford Focus.
You can get 256Gb SSD for about £140 inc. these days - of course that is a lot compared to a rotational drive but they are for different markets. All the data I need to carry with me fits easily only 256Gb but you are never going to please everyone.
Those Seagate Momentus XT drives may be a 'compromise' for you - some of the speed of SSD with the capacity of a rotational drive - 750Gb for just less than £100 inc.
Why on Earth would you want or need 3tb of storage in a laptop, SSD at that?
I have an SSD in my desktop PC at home - runs the OS and the most commonly opened applications. I then have a hybrid drive which has all my Steam games on it and a further normal drive for things like video storage etc...
At work, I'm just installing a new storage solution for our network - it consists of a mix of enterprise SSD for virtual machines, SQL and other 'high IO' applications, SAS2 drives for ordinary storage, and then SATA3 for archival and D2D backup. Sure, we could have gone full SSD, but we would have paid roughly 10x as much and needed a *lot* more space to fit all the drives in.
You don't run a lawnmower on jet fuel, so why would you put everything on SSD?
By some rough calculations, it should extend the battery life of the T420 laptop (which can run for maybe 10 hours) by another hour, i.e. 10% improvement, which seems pretty good. And hopefully cooler means more reliable (I've used other manufacturer SSDs and they did not last 3 years. No idea if Intel is reliable or not).
I wonder why Intel is even bothering to mention this, as I doubt there's anyone that cares. It isn't as though previous SSDs were power hungry beasts causing laptops to become too hot to touch. While it sounds like a big drop in percentage terms, it's less than half a watt after all! If they shaved two grams off the weight would they think that's worthy of mention as well?
What market are these lower power drives for? The only one I can think of are laptops, where battery technology is a bit lame, so savings like this extends battery life. Maybe their ultra-book concept?
They're not much use for desktop PCs - no manufacturer or consumer cares about a half a watt difference.
Seems an odd move to be honest - unless its stage 1 in reducing chip sizes to be able to increase density in the form factor? But then, they're not exactly 'crammed' into those cases anyway.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
