Print 30 comment(s) - last by AntDX316.. on Sep 2 at 3:57 AM

  (Source: MGM)
Unlockable ability will boost flash controller speed for faster reads and writes

Overclocking your DRAM memory remains one of the most popular aftermarket tweaks of enthusiasts.  The imperfect art of adjusting voltage and timings has long been able to offer snappier performance depending on your system's bottlenecks and the quality of memory sticks in the box.

While on the surface NAND flash storage resembles DRAM, it has thus far remained unavailable for overclocking ever since SSDs began to creep into the market in volume around 2003.

I. Ivy Bridge-E: Better Late Than Never

But that could soon change.  

Intel Corp. (INTC) quietly posted about a Sept. 10 technical presentation "AIOS001 - Overclocking Unlocked Intel® Core™ Processors for High Performance Gaming and Content Creation" at the 2013 Intel Developer Forum (IDF) in San Francisco, Calif.  At the presentation, it's expected to unveil the tardy 22 nm Ivy Bridge-E architecture Core i7-4000 and X79 platform, and with it SSD overclocking.  The presentation was first noted by EXP Review, a Chinese language hardware site.  A list of presentation topics comments:
  • Overclocking architecture of the Intel® Core™ i7-4xxx series processors on Intel® X79 Express Chipset based platforms
  • Overclocking architecture of the desktop and mobile 4th Generation Intel® Core processors
  • Live demonstrations and expert tips from an industry partner on how to push the limits of Intel® unlocked processors using software and hardware tools from the enthusiast ecosystem
  • One of the first public demonstrations of overclocking the Intel® Core™ i7 4xxx Extreme Edition processor
  • Demonstrating overclocking SSD technology
  • First public demonstration of AppTune beta based on Intel® Extreme Tuning Utility (Intel® XTU)
It's unclear whether the SSD overclock will be reserved for only the 22 nm Ivy Bridge-E series chips, or will be open to Haswell chips as well.  Intel typically uses its "E" series to introduce Xeon multicore editions of its architecture, and higher clocked desktop consume Core i7/Core i7 Extreme branded chips.

Ivy Bridge
The tardy Ivy Bridge-E platform is expected to be introduced at IDF, next month.

The Ivy Bridge-E launch is a bit late -- Intel launched Sandy Bridge-E in Nov. 2011, nearly eight months after the 32 nm Sandy Bridge's initial launch.  By contrast, the 22 nm Ivy Bridge launched in April 2012, but has seen no "extreme" update in nearly a year and a half (17 months).  At the head of the new consumer "Extreme" chips is rumored to be the Core i7 Extreme 4960X, a monster that packs 15 MB of L3 cache and is turbo-clocked to 4.0 GHz. 

II. SSD Overclocking -- SATA on Steroids or Hype?

When it comes to the SSD overclocking, the upcoming platform is expected to allow you to apply more voltage to your SSD controller to bump your read, write, and IOPS speed.  Performance will be capped at 4.8 Gbps (614 MB/s), though, due to the fundamental ceiling of the current SATA 3 standard implementation supported in Intel chipsets (The raw performance of SATA 3 is 6 Gbps).
Indilinx Barefoot 3 controllers
Indilinx Barefoot 3 SSD Controllers [Image Source: OCZ]

That still would seemingly suggest that overclocking SSDs could offer some impressive gains.

Currently the fastest SSD in terms of read speeds is Z-Drive R4 from OCZ Technology Group Inc. (OCZ), who boasts (under light load) a gaudy 547.3 MB/s (4.28 Gbps) read and 657.9 MB/s (5.14 Gbps) write; but there's a catch -- this pricey (~$2,800 USD) OCZ drive uses PCI Express.  High end SATA 3 drives from OCZ and Samsung Electronics Comp., Ltd. (KSC:005930) score a more sluggish 290-320 MB/s (2.27-2.50 Gbps) read speed and 370-400 MB/s (2.89-3.13 Gbps) write speed, according to the latest numbers from AnandTech's benchmarking (again, light load).

Samsung 840 SSD
While it's clear that the current real world performance numbers are far from the theoretical 4.8 Gbps spec limit, skepticism is warranted.  It remains unclear whether simply how far current drives can be pushed with higher voltage and faster clock speeds to the controller.  Regardless of the real gains, this does mean we will see upcoming enthusiast SSD packaging with a focus on controller cooling with flashy finned heat spreaders, like we've seen in the enthusiast RAM market (case constraints may limit creativity for now, though).

Intel is working on a new chipset, which is rumored to be dubbed the 9 Series (Haswell Z97 and H97), which will support the fresh SATA Express standard, bumping transfer speeds to 8 or even 16 Gbps (819.2 MB/s or even 1.55 GB/s), perhaps.  The bus improvement is expected to drop ahead of BroadwellIntel's planned 14 nm die shrink.  
OCZ Vector
With more breathing room on the bus and fresh overclocking abilities, enthusiasts and industry performance tuners may finally have the freedom to achieve something they long wished for -- overclocking their SSD.

DailyTech will be reporting live from the Intel Developer Forum in two weeks, so hopefully we'll have more details for you shortly.

Sources: Intel, EXP Review

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

Sounds like a bad idea
By ImmortalSamurai on 8/27/2013 1:18:53 PM , Rating: 5
I don't know how I feel about overclocking my SSD. If you do it incorrectly on a CPU and it blue screens down the road you can go back in and change the settings and try again till everything is fine. But, if you overclock your SSD and it fails down the road you are looking at data corruption and possibly losing everything on your C: drive.

And then possibly having to reload everything each attempt if the data corruption is bad enough. No thanks.

RE: Sounds like a bad idea
By Mr. Texel on 8/27/2013 1:28:54 PM , Rating: 2
Not if you do a backup image before overclocking.

RE: Sounds like a bad idea
By DanNeely on 8/27/2013 1:58:56 PM , Rating: 2
A backup from before you turned on SSD-OC will be of limited value if you take several months to discover you're suffering from low grade bitrot.

RE: Sounds like a bad idea
By Mitch101 on 8/27/2013 2:34:48 PM , Rating: 2
I'll pass on this idea of overclocking an SSD.

The idea of going from 1 second to .9 seconds doesn't interest me at the expense of potential data loss. If I need more speed Ill look at RAIDing 2-3 SSD's. Plenty happy with one SSD and Im 3 generations behind the latest Samsung Pro's.

RE: Sounds like a bad idea
By tayb on 8/27/2013 6:54:05 PM , Rating: 1
You're uncomfortable with the data loss risks associated with overclocking an SSD but mention the possibility of a 2 disk RAID instead?

RE: Sounds like a bad idea
By Monkey's Uncle on 8/27/2013 7:32:49 PM , Rating: 2
It's safer than overclocking a single SSD and will net you better performance with less risk of bitrot.

Remember that overclocking an ssd means pushing more voltage into its controller and memoory. Pushing more voltage into an SSD = a hot SSD controller. Hot SSD controllers = less reliability. Less reliability = daily backups highly recommended. Remember too that mlc has a finite write cycle life. That is calculated at standard voltages and frequencies. Try to write it faster or at higher voltages for extended periods and that duty cycle reduces by large amounts.

So ya, please do overclock your single SSDs. I'll raid mine and we'll see which is faster and lasts longer.

Note: My 2-drive QNAP NAS runs its drives in raid-0. Has been running steadily that way for 3 years. I have automated backups that run 3 times/week and have yet to rebuild my array in that entire time. Somehow I don't see an overclocked SSD having that level of longevity or reliability.

RE: Sounds like a bad idea
By B3an on 8/27/2013 9:37:24 PM , Rating: 2
All assumptions. No one has OC'd an SSD yet, or even knows exactly how this will work.

RE: Sounds like a bad idea
By Monkey's Uncle on 8/28/2013 11:50:50 AM , Rating: 2
The article is conjecture and my initial response plays along with this conjecture (you apparently didn't seem to get that). If Intel is intending to allow overclocking an SSD via the motherboard chipset, simply bumping past the current limit won't do it. The SSD's onboard controller itself (ie Indilinx, Sandforce, Micron controllers) needs to accommodate higher clock speeds and voltage it as well. I don't see any SSD manufacturers stepping up claiming to be working with Intel to make that happen, do you?

Until they do, this is all conjecture and based on that conjecture my response is perfectly valid. Overclock an SSD at your risk and assume your hardware life span will shorten drastically if you do (as ANY overclocker knows well). While manufacturers may build their hardware to accommodate overclocking, they will not accept liability for damages if you do - it is all on you if you leverage it.

RE: Sounds like a bad idea
By jRaskell on 8/28/2013 12:46:22 PM , Rating: 2
While it wasn't explicit, I think it's safe to say that Intel will step up and support Overclocking with their own SSDs. From that point on it becomes a selling point that many other manufacturers are going to want to be able to add to their checklists as well, otherwise they're looking at losing market share, something no serious business wants to deal with.

I also think it's rather overkill to claim that it will drastically shorten the lifespan of your hardware. Overclocking your CPU or graphics card also shortens their lifespans as well, but it's only drastic in extreme cases of overclocking. There's really no technical reason to believe SSDs will be any different.

RE: Sounds like a bad idea
By Samus on 8/28/2013 3:39:43 AM , Rating: 2
You know, I've never felt SSD's were too slow. It never occurred to me there was a need to overclock them, since they are limited by SATA anyway.

RE: Sounds like a bad idea
By AntDX316 on 9/2/2013 3:57:52 AM , Rating: 2
I don't think people understand what is being overclocked. I doubt the SSD overclocking is in the same method boat as CPUs/GPUs/RAM. It probably has to do with an algorithmic boost instead of driving more volts and turning the MHz up to make it run faster.

RE: Sounds like a bad idea
By XZerg on 8/27/2013 1:36:37 PM , Rating: 2
or brick the drive.

RE: Sounds like a bad idea
By Flunk on 8/27/2013 2:03:52 PM , Rating: 2
It really doesn't make sense for desktop loads, most SSDs already far exceed the raw storage bandwidth that current software is designed to use.

What I could see it used for is on SSD caches on high-volume server systems. For those, longevity isn't as important and raw throughput is key.

RE: Sounds like a bad idea
By Makaveli on 8/27/2013 2:32:46 PM , Rating: 2
"SSDs began to creep into the market in volume around 2003"

did I miss this in 2003?

When was there SSD's in the consumer market in 2003?

"Overclocking architecture of the desktop and mobile 4th Generation Intel® Core processors"

Isn't haswell intel's 4th Gen i7?

ivy-E which is based on Ivy bridge was the 3rd Gen!


RE: Sounds like a bad idea
By DanNeely on 8/27/2013 3:03:38 PM , Rating: 2
My only guess for the 03 date, assuming it's not in error and is for PCs, would be that it was an inflection point in the industrial sector. There were DIY solutions with IDE to SDCard/Compact Flash adapters around then but they were really niche.

Alternately 03 could be early thumb drives; I don't recall when they first started coming out in useful sizes without breaking the bank. The first flash based ipod was released in Jan05; but there were a few older flash based 3rd party entries.

As for you second comment, the list of the presentations bullet points clearly indicate that Intel is talking about more subjects than just the two headlined here. One of them is overclocking haswell systems.

RE: Sounds like a bad idea
By Monkey's Uncle on 8/27/2013 7:45:21 PM , Rating: 3
Actually m-data introduced flash-based diskless drives way back in in 1995. The whole idea of diskless drives has been around since the 70's (supercomputers) & 80's (high-end PCs).

In 1999, BiTMICRO made some presentations and announcements about flash-based SSDs. 2007 saw the first PCIe-based SSDs.

The first commercial SSDs on a SATA interface didn't show up until 2009 (OCZ Cebit demo, and brought to market by Micron). They were obscenely expensive (and still not anyone's idea of cheap today).

Short answer - there were no commercial SSDs as we know of them today available in 2003, though there were devices that could have been considered solid state drives dating all the way back into the 70s.

RE: Sounds like a bad idea
By Reclaimer77 on 8/27/2013 2:39:38 PM , Rating: 2
Luckily there's no point in using this feature until we get our hands on actual 2GBps SATA Express mobo's and SSD's. Intel is just getting ahead of the game here, as usual.

It's kind of a cool option I guess? But already the cheapest SSD you can buy will blow you away. I'm not sure if there's really a lot of performance to gain here.

RE: Sounds like a bad idea
By Moishe on 8/29/2013 4:42:59 PM , Rating: 2
Yep. If we're bumping up against the SATA limit, we need a new one.

But I guess they're not focusing on people who need safe data. They're focusing on a potentially new, cool enthusiast market.

RE: Sounds like a bad idea
By Treckin on 8/27/2013 3:54:11 PM , Rating: 2
I believe you're treating it like a mechanical drive, ie. if it fails at all it's total data loss.

The reality is the opposite and one of the prime reasons to use SSDs - the parralellism offers inherent data security. If you lose a few blocks or even a whole NAND array, you still have access to the rest of the data.

Yes critical OS files might get corrupted, but you wont lose your pron or music videos.

Even if the controller dies, it should be easier and safer to replace then opening a mechanical drive to replace a motor, head, or circuit board.

RE: Sounds like a bad idea
By retrospooty on 8/27/2013 5:08:38 PM , Rating: 2
I would be more worried about the few odd bits here and there that got corrupted and you weren't aware of it... 6 months later after you may have gone through your backups cycle and overwritten older data, now you find some files that are corrupt and your backup is also corrupt.

Overclocking SSD just seems dumb to me. They are fast as hell already. If yours isnt, its either old, or cheap. Anything mainstream today is crazy fast.

RE: Sounds like a bad idea
By Monkey's Uncle on 8/27/2013 7:50:43 PM , Rating: 3
What you are referring to is bit-rot. It is insidious and really hard to find until it bites you in the ass. Most of the time the CRC functions of the disk controller fixes any bit-rot that slips in, but if you are overclocking the controller, you have no idea if you are borking the CRC checksum handling.

Agreed that overclocking mass storage is dumbass as hell. If you need fast go wit raid 0. If you need fast and safe there is raid 5 (sorry but the safety net is gonna cost you 25% of your space).

RE: Sounds like a bad idea
By Moishe on 8/29/2013 4:37:12 PM , Rating: 2
Ditto... Especially when SSDs are faster than regular hard drives already. I'd think it'd be safer to just pick up a large amount of ram and use a ramdrive.

"If you can find a PS3 anywhere in North America that's been on shelves for more than five minutes, I'll give you 1,200 bucks for it." -- SCEA President Jack Tretton

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki