阿里通信如何查看流量:Intel X25-M SSD: Intel Delivers One of the World's Fastest Drives

Full View
Intel X25-M SSD: Intel Delivers One of the World's Fastest Drives
by on 9/8/2008 4:00:00 PM
Posted in
Intel finally lifted the NDA on the performance of its first MLCSSD. Let's just say that after today, a Core 2 isn't the only Intelcomponent you want in your system.
Page 1
Be sure to read our latest SSD article:The SSD Relapse for an updated look at the SSD market.
Noone really paid much attention to Intel getting into the SSD (SolidState Disk) business. We all heard the announcements, we heard theclaims of amazing performance, but I didn't really believe it. Afterall, it was just a matter of hooking up a bunch of flash chips to acontroller and putting them in a drive enclosure, right?

Thecloser we got to release and the more time I spent with competingproducts, the more I realized that Intel's biggest launch of 2008 wasn'tgoing to be Nehalem - it was going to be its SSDs. If Intel couldprice them right, and if Intel could deliver on the performance, thebiggest upgrade you could do for your PC - whether desktop or notebook,wouldn't be to toss in a faster CPU, it would be to migrate to one ofthese SSDs. Combine Nehalem and one of these mythical SSDs and you werein for a treat. But that was a big if...Intel still had to deliver.
Wealready talked about the drives back at IDF. The Intel X25-M and theX18-M, available in 80GB capacities, 2.5" and 1.8" form factors(respectively) with 160GB versions on the way. Today we are allowed toshare performance data and pricing information, one of which is moreimpressive than the other. Intel will be selling the X25-M at $595 MSRPthrough OEMs and channel vendors, although I hear the street price maybe lower.

Bothof the -M models are based on Intel's MLC flash, while a X25-E usingSLC flash will be due out by the end of this year. I'll detail thedifferences in a bit.
The pricing is rough, that puts Intel'sX25-M at cheaper than SLC drives on the market but more expensive thanMLC drives. Your options are effectively to get a 128GB MLC drive, an80GB Intel X25-M or a 64GB SLC drive. But as you can expect, I wouldn'tbe quite this excited if the decision were that easy. Over the nextseveral pages we're going to walk through the architecture of a NANDflash based SSD, investigate the problems with current MLC drives (andshow how the Intel drive isn't affected) and finally compare theperformance of the Intel drive to MLC, SLC and standard hard drives(both 2.5" and 3.5") in a slew of real world applications.

Ifyou want to know the ending first I won't make you wait. Intelabsolutely delivered with its first SSDs. After I completed my initialtesting of the drive I sent AnandTech Senior Editor, Gary Key a message:
"I think Intel just Conroe’d the HDD market."
Honestly,within 6 months I'd expect it to be just as important to have one ofthese drives in your system, as your boot/application drive, as it wasto have Conroe in your system back in 2006. The only issue here, theonly problem I have is the price. I was hoping for something much lowerfrom Intel and although the pricing is justified based on theperformance, it ensures that the X25-M like most high performance SSDs,remains a luxury item.
While the X25-M isn't the world's fasteststorage device across the board, it is among the fastest. And in theareas that it does dominate, it does so unbelievably well. The othergreat thing? You’ve got one of the world’s fastest hard drives, and itcan fit in your notebook.
Let's get to it.
Page 2
How SSDs Work
The building block of NAND flash is the N-channel MOSFET:

Each"cell" is made up of one of these transistors. In a single-level cell(SLC) flash device, one of these transistors can hold 1-bit of data.You write data to the cell by electron tunneling; apply a high enoughvoltage to the gate, create a powerful enough electric field, andelectrons will tunnel through the oxide and into the floating gate.Remove the voltage and the electrons will remain in the floating gate.Apply the voltage across the channel instead of the gate, reverse thebias and the electrons will go in the other direction. Simply put,that's how flash works - you've got two states, 0 and 1, and the stateis preserved even if the cell has no power, making it ideal for astorage device.
Programming flash is an iterative process. Thecontroller will apply voltage to the gate (or the channel), allow someelectrons to tunnel and check the threshold voltage of the cell. Whenthe threshold voltage has reached some predetermined value, it’s nowprogrammed and your data is stored.
MLC vs. SLC
There aretwo forms of NAND flash used in SSDs today: Single-Level Cell (SLC) andMulti-Level Cell (MLC). The difference between the two is the amountof data stored per cell, with SLC it's 1-bit per cell and with MLC it's2-bits per cell. The key here is that both SLC and MLC take up the sameamount of die area, so MLC effectively doubles your capacity at thesame price.
Intel actually uses the same transistors for its SLCand MLC flash, the difference is how you read/write the two. With SLCthere are only two voltages to worry about, since there are two states(0 or 1). With MLC, there are four states (00, 01, 10, 11) and thus ittakes longer to access since you don't want to accidentally write thewrong bit of data; you've got the same min and max voltage, you simplyhave more graduations in between the two now:

SLC (left) vs. MLC (right)
Below is a table of some basic stats on SLC vs. MLC performance:
SLC NAND flash MLC NAND flash
Random Read 25 μs 50 μs
Erase 2ms per block 2ms per block
Programming 250 μs 900 μs
Erasingperformance is the same between the two, read performance takes twiceas long on MLC flash and write performance can take almost four times aslong. If you've ever heard people complain about MLC write speedbefore, this partly why. Do keep in mind though, the numbers we'retalking about here are ridiculously low - even 900 μs to write to MLCflash is much faster than writing to a mechanical hard disk.
Thebiggest advantage of SLC ends up not being performance, but lifespan.To understand how flash wears, we first need to look at how it'sorganized in a storage device.
Page 3
The Flash Hierarchy & Data Loss
We've alreadyestablished that a flash cell can either store one or two bits dependingon whether it's a SLC or MLC device. Group a bunch of cells togetherand you've got a page. A page is the smallest structure you can program(write to) in a NAND flash device. In the case of most MLC NAND flasheach page is 4KB. A block consists of a number of pages, in the IntelMLC SSD a block is 128 pages (128 pages x 4KB per page = 512KB per block= 0.5MB). A block is the smallest structure you can erase. So whenyou write to a SSD you can write 4KB at a time, but when you erase from aSSD you have to erase 512KB at a time. I'll explore that a bit furtherin a moment, but let's look at what happens when you erase data from aSSD.

Wheneveryou write data to flash we go through the same iterative programmingprocess again. Create an electric field, electrons tunnel through theoxide and the charge is stored. Erasing the data causes the same thingto happen but in the reverse direction. The problem is that the moretimes you tunnel through that oxide, the weaker it becomes, eventuallyreaching a point where it will no longer prevent the electrons fromdoing whatever they want to do.
On MLC flash that point isreached after about 10,000 erase/program cycles. With SLC it's 100,000thanks to the simplicity of the SLC design. With a finite lifespan,SSDs have to be very careful in how and when they choose toerase/program each cell. Note that you can read from a cell as manytimes as you want to, that doesn't reduce the cell's ability to storedata. It's only the erase/program cycle that reduces life. I refer toit as a cycle because an SSD has no concept of just erasing a block, theonly time it erases a block is to write new data. If you delete a filein Windows but don't create a new one, the SSD doesn't actually removethe data from flash until you're ready to write new data.
Nowgoing back to the disparity between how you program and how you erasedata on a SSD, you program in pages and you erase in blocks. Say yousave an 8KB file and later decide that you want to delete it, it couldjust be a simple note you wrote for yourself that you no longer need.When you saved the file, it'd be saved as two pages in the flash memory.When you go to delete it however, the SSD mark the pages as invalidbut it won't actually erase the block. The SSD will wait until acertain percentage of pages within a block are marked as invalid beforecopying any valid data to new pages and erasing the block. The SSD doesthis to limit the number of times an individual block is erased, andthus prolong the life of your drive.

Notall SSDs handle deletion requests the same way, how and when you decideto erase a block with invalid pages determines the write amplificationof your device. In the case of a poorly made SSD, if you simply wantedto change a 16KB file the controller could conceivably read the entireblock into main memory, change the four pages, erase the block from theSSD and then write the new block with the four changed pages. Using thepage/block sizes from the Intel SSD, this would mean that a 16KB writewould actually result in 512KB of writes to the SSD - a writeamplification factor of 32x.

Atthis point we don't have any data from any of the other SSD controllermakers on how they handle situations like this, but Intel states thattraditional SSD controllers suffer from write amplification in the 20 -40x range, which reduces the longevity of their drives. Intel statesthat on typical client workloads its write amplification factor is lessthan 1.1x, in other words you're writing less than 10% more data thanyou need to. The write amplification factor itself doesn't mean much,what matters is the longevity of the drive and there's one more factorthat contributes there.
We've already established that with flashthere are a finite number of times you can write to a block before itloses its ability to store data. SSDs are pretty intelligent and willuse wear leveling algorithms to spread out block usage across theentirety of the drive. Remember that unlike mechanical disks, itdoesn't matter where on a SSD you write to, the performance will alwaysbe the same. SSDs will thus attempt to write data to all blocks of thedrive equally. For example, let's say you download a 2MB file to yourband new, never been used SSD, which gets saved to blocks 10, 11, 12 and13. You realize you downloaded the wrong file and delete it, then gooff to download the right file. Rather than write the new file toblocks 10, 11, 12 and 13, the flash controller will write to blocks 14,15, 16 and 17. In fact, those four blocks won't get used again untilevery other block on the drive has been written to once. So while yourMLC SSD may only have a lifespan of 10,000 cycles, it's going to lastquite a while thanks to intelligent wear leveling algorithms.

Intel's wear leveling efficiency, all blocks get used nearly the same amount

Bad wear leveling, presumably on existing SSDs, some blocks get used more than others
Intel'sSSDs carry about a 4% wear leveling inefficiency, meaning that 4% ofthe blocks on an Intel SSD will be worn at a rate higher than the rest.
Page 4
How Long Will Intel's SSDs Last?
SSD lifespans areusually quantified in the number of erase/program cycles a block can gothrough before it is unusable, as I mentioned earlier it's generally10,000 cycles for MLC flash and 100,000 cycles for SLC. Neither ofthese numbers are particularly user friendly since only the SSD itselfis aware of how many blocks it has programmed. Intel wanted torepresent its SSD lifespan as a function of the amount of data writtenper day, so Intel met with a number of OEMs and collectively they cameup with a target figure: 20GB per day. OEMs wanted assurances that auser could write 20GB of data per day to these drives and still havethem last, guaranteed, for five years. Intel had no problems with that.
Intelwent one step further and delivered 5x what the OEMs requested. ThusIntel will guarantee that you can write 100GB of data to one of its MLCSSDs every day, for the next five years, and your data will remainintact. The drives only ship with a 3 year warranty but I suspect thatthere'd be some recourse if you could prove that Intel's 100GB/daypromise was false.

Justlike Intel's CPUs can run much higher than their rated clock speed,Intel's NAND should be able to last much longer than its rated lifespan
It'salso possible for a flash cell to lose its charge over time (albeit avery long time). Intel adheres to the JEDEC spec on how long your datais supposed to last on its SSDs. The spec states that if you've onlyused 10% of the lifespan of your device (cycles or GB written), thenyour data needs to remain intact for 10 years. If you've used 100% ofavailable cycles, then your data needs to remain intact for 1 year.Intel certifies its drives in accordance with the JEDEC specs from 0 -70C; at optimal temperatures your data will last even longer (these SSDsshould operate at below 40C in normal conditions).
Intel andMicron have four joint fabs manufactured under the IMFT partnership, andthese are the fabs that produce the flash going into Intel's SSDs. The50nm flash used in the launch drives are rated at 10,000erase/programming but like many of Intel's products there's a lot ofbuilt in margin. Apparently it shouldn't be unexpected to see 2, 3 or4x the rated lifespan out of these things, depending on temperature andusage model obviously.
Given the 100GB per day x 5 year lifespanof Intel's MLC SSDs, there's no cause for concern from a datareliability perspective for the desktop/notebook usage case. High loadtransactional database servers could easily outlast the lifespan of MLCflash and that's where SLC is really aimed at. These days the MLC vs.SLC debate is more about performance, but as you'll soon see - Intel hasredefined what to expect from an MLC drive.
Other Wear and Tear
Withno moving parts in a SSD, the types of failures are pretty unique.While erasing/programming blocks is the most likely cause of failurewith NAND flash, a secondary cause of data corruption is something knownas program disturb. When programming a cell there's a chance that youcould corrupt the data in an adjacent cell. This is mostly a functionof the quality of your flash, and obviously being an expert insemiconductor manufacturing the implication here is that Intel's flashis pretty decent quality.
Intel actually includes additional spaceon the drive, on the order of 7.5 - 8% more (6 - 6.4GB on an 80GBdrive) specifically for reliability purposes. If you start running outof good blocks to write to (nearing the end of your drive's lifespan),the SSD will write to this additional space on the drive. Oneinteresting sidenote, you can actually increase the amount of reservedspace on your drive to increase its lifespan. First secure erase thedrive and using the ATA SetMaxAddress command just shrink the usercapacity, giving you more spare area.
Page 5
What Happens When Your SSD Fails?
When your hard drivedies we all know what happens. You go to turn on your machine one dayand your OS doesn't boot, or your drive stops getting detected. Butwith SSDs their lifespan is far more predictable, so what does happen asthey near the end of their life? A well designed SSD will have a goodenough wear leveling algorithm to make sure that all blocks in thedevice get equal usage, so that when they fail, they do so at the sametime.
Intel's SSDs are designed so that when they fail, theyattempt to fail on the next erase - so you don't lose data. If thedrive can't fail on the next erase, it'll fail on the next program -again, so you don't lose existing data. You'll try and save a file andyou'll get an error from the OS saying that the write couldn't becompleted.
The beauty here is that the SSD knows exactly when itcan't erase/program a block, and if the drive knows, then you can usesoftware to ask the drive what it knows. In the near future Intel willbe releasing its own SSD tool that will let you query two SMARTattributes on the drive: one telling you how close you are to the ratedcycling limit, and one telling you when you've run out of reallocatingblocks. The latter is the most important because Intel fully expectsthese drives to outlast their rated limits. As bad blocks develop, theSSD will mark them as such and write to new ones - by telling you whenit has run out of bad blocks (or nearly run out of bad blocks), you'llknow exactly when you need a new hard drive.
This is hugelyimportant. While Intel's SSDs aren't exactly cheap, the beauty of flashis that it follows the same Moore's Law that CPUs do. In the next ~18months you'll be able to get a 160GB drive for the price of the 80GB, inanother couple of years we'll be at 320GB for the same price (mostlikely lower as SSD demands increase). Within the next five years we'llbe in a situation where the fans in your system are more likely to failthan your hard drive, and if your drive does happen to fail it'll tellyou well in advance. How nice of it.
Page 6
The Generic SSD
Intel isn't the first manufacturer toget into the SSD market, in fact quite a few of the companies weregularly deal with have SSDs. OCZ Technology, SuperTalent, Patriot,G.Skill, Mtron, and Silicon Power are just the names that contributedhardware for this review, but many more actually make SSDs.
{gallery 201}
Silicon Power drives provided byDV Nation
You'llnotice that most of these companies are memory companies, not chipcompanies. Cracking open any of their SSDs reveals the true nature oftheir SSD manufacturing business: most of them simply rebrand someoneelse's SSD.
Let's take the MLC drives for example, here we have MLC drives from OCZ (the infamous Core), SuperTalent and Silicon Power:




Every last one of the drives has the exact same layout, same flash devices and same Jmicron controller (the JMF602).
The same is true for SLC designs, here we have SLC SSDs from G.Skill and OCZ:




They are the same, you could give them ice crea...nevermind
Note that these are the same drive, in this case purchased from Samsung and placed in new housing.
Only Mtron offers a slight amount of innovation by using its own FPGA as a controller:


But we know very little about any of these controllers so it's tough to say if Mtron's FPGA is done well.
Forthe most part however, all existing MLC drives on the market are builtout of the same parts, as are all existing SLC drives. The MLC drivesall use the JMF602 and the SLC drives use the Samsung S3C49RBX01 hostcontroller with 32MB of on-board DRAM. The SLC drives are all pricedabove $640 for 64GB while the MLC drives are all under $300 for 64GB,Intel's 80GB MLC is priced in between the two (but closer to the SLCdrives) at $595.
We'll soon see that Intel's price is justified inthe market, but before doing that it's important to address an issuewith these very tempting MLC drives...
Page 7
Enter the Poorly Designed MLC
The great thing abouteveryone making MLC drives based on the same design is it helps drivecost down, which gives us a very affordable product. After rebate youcan buy a 64GB OCZ Core SSD, an MLC drive, for $240 from Newegg.Compared to the $1000+ that 64GB SSDs were selling for a year ago, thisis good cost savings. The bad thing about everyone using the samedesign however is if there's a problem that affects one of the drives,it affects all of them. And indeed, there is a problem.
Thesymptoms are pretty obvious: horrible stuttering/pausing/lagging duringthe use of the drive. The drive still works, it's just that certainaccesses can take a long time to complete. It's a lot like using a slowlaptop hard drive and trying to multitask, everything just comes to ahalt.
I first discovered this problem a couple of months ago whenI started work on an article looking at the performance of a SSD in aMac Pro as a boot/application drive. Super Talent sent me one of its3.5” drives, which I had assumed was a SLC drive. Application launcheswere ridiculously fast, but I noticed something very strange when I wasusing my machine. Starting to type in a document, or sending an IM, oreven opening a new tab in Safari would sometimes be accompanied by asecond-long pause. At first I assumed it was a problem with my drive orwith the controller, or perhaps a combination of the drive, the SATAcontroller on the Mac Pro’s motherboard and OS X itself. I later foundout it was an MLC drive and thus began my investigation.
SuperTalenthad received a lot of attention for its SSDs, and rightfully so - theywere starting to be affordable. OCZ however quickly took the spotlightwith its Core SSD, finally bringing the price of a 64GB MLC SSD to below$300. Users flocked to the Core and other similarly priced drives,because if you looked at the marketed specs of the drive you werebasically getting greater than SLC performance, at a fraction of thecost:
Advertised Specs OCZ Core (MLC) OCZ (SLC)
Read Up to 143MB/s Up to 100MB/s
Write Up to 93MB/s Up to 80MB/s
Seek < 0.35ms unlisted
Price < $300 > $600
However the real world performance didn't match up.
Let'sstart with the types of benchmarks that we usually see run in SSDreviews, here's a quick run of PCMark Vantage's HDD. Vantage paints theCore as a screamer:
PCMark Vantage HDD Test
OCZ Core (JMicron JMF602, MLC) 8117
OCZ (Samsung, SLC) 12143
Western Digital VelociRaptor (10,000 RPM SATA) 6325
Digginga bit deeper we only see one indication of a problem, performance inthe Media Center test is significantly slower than the VelociRaptor -but overall it's much faster, what could one test actually mean?
Windows Defender Gaming Picture Import Vista Startup Windows Movie Maker Media Center WMP App Loading
OCZ Core (JMicron JMF602, MLC) 48.1MB/s 72.5MB/s 90.4MB/s 47.9MB/s 23.2MB/s 33MB/s 17.8MB/s 20.3MB/s
OCZ (Samsung, SLC) 69.3MB/s 71.8MB/s 86.9MB/s 63MB/s 43.7MB/s 65.6MB/s 33.8MB/s 39.9MB/s
Western Digital VelociRaptor (10,000 RPM SATA) 27.5MB/s 20.1MB/s 59.0MB/s 22.9MB/s 58.5MB/s 113.3MB/s 15.2MB/s 7.6MB/s
Ifwe turn to SYSMark however, the picture quickly changes. The OCZ SLCdrive is now 30% faster than the MLC drive, and performance in the VideoCreation suite is literally half on the MLC drive. Something is amiss.
SYSMark 2007 Overall E-Learning Video Creation Productivity 3D
OCZ Core (JMicron JMF602, MLC) 138 143 111 134 168
OCZ (Samsung, SLC) 177 161 200 178 172
Western Digital VelociRaptor (10,000 RPM SATA) 179 155 222 177 169
Page 8
Delving Deeper
I had suspicions as to the nature of theproblem based on my experience with it in my Mac Pro. The SuperTalentMLC drive in my machine would pause, most noticeably, randomly when I'dwant to send an IM. What happens when you send an IM? Your logfilegets updated; a very small, random write to the disk. I turned toIometer to simulate this behavior.
Iometer is a great tool forsimulating disk accesses, you just need to know what sort of behavioryou want to simulate. In my case I wanted to write tons of small filesto the drive and look at latency, so I told Iometer to write 4KB filesto the disk in a completely random pattern (100% random). I left thequeue depth at 1 outstanding IO since I wanted to at least somewhatsimulate a light desktop workload.
Iometer reports four resultsof importance: the number of IOs per second, the average MB/s, theaverage write latency and the maximum write latency. I looked atperformance of four drives, the OCZ Core (Jmicron controller MLC), OCZSLC (Samsung controller), Intel MLC (Intel controller) and the SeagateMomentus 7200.2 (a 7200RPM 2.5" notebook drive).
Thoughthe OCZ core drive is our example, but please remember that this isn'tan OCZ specific issue: the performance problems we see with this driveare apparent on all current MLC drives in the market that use a Jmicroncontroller with Samsung flash.
4KB, 100% random writes, IO queue depth 1 IOs per Second MB/s Average Write Latency Max Write Latency
OCZ Core (JMicron, MLC) 4.06 0.016MB/s 244ms 991ms
OCZ (Samsung, SLC) 109 0.43MB/s 9.17ms 83.2ms
Intel X25-M (Intel, MLC) 11171 43.6MB/s 0.089ms 94.2ms
Seagate Momentus 7200.2 106.9 0.42MB/s 9.4ms 76.5ms
Curiouserand curiouser...see a problem? Ignore the absolute ridiculousperformance advantage of the Intel drive for a moment and look at theaverage latency column. The OCZ MLC drive has an average latency of 244ms, that's over 26x the latency of the OCZ SLC drive and 25.9x thelatency of a quick notebook drive. This isn't an MLC problem however,because the Intel MLC drive boasts an average latency of 0.09ms - theOCZ MLC drive has a 2700x higher latency!
Now look at the maxlatency column, the worst case scenario latency for the OCZ Core is991ms! That's nearly a full second! This means that it takes anaverage of a quarter second to write a 4KB file to the drive and worstcase scenario, a full second. We complain about the ~100 nanosecondtrip a CPU has to take to main memory and here we have a drive that'lltake nearly a full second to complete a task - totally unacceptable.
Inorder to find out if the latency is at all tied to the size of thewrite I varied the write size from 4KB all the way up to 128KB, but keptthe writes 100% random. I'm only reporting latencies here:
100% random writes, IO queue depth 1 4KB 16KB 32KB 64KB 128KB
OCZ Core (JMicron, MLC) 244ms 243ms 241ms 243ms 247ms
OCZ (Samsung, SLC) 9.17ms 14.5ms 21.2ms 28ms 28.5ms
Intel X25-M (Intel, MLC) 0.089ms 0.23ms 0.44ms 0.84ms 1.73ms
Seagate Momentus 7200.2 9.4ms 8.95ms 9.14ms 9.82ms 12.1ms
Allthe way up to 128KB the latency is the same, 0.25s on average andnearly a second worst case for the OCZ Core and other similar MLCdrives. If it's not the file size, perhaps it's the random nature ofthe writes?
For this next test I varied the nature of the writes, Iran the 4KB write test with a 100% sequential workload, 90% sequential(10% random) and 50% sequential (50% random):
4KB writes, IO queue depth 1 100% Sequential/0% Random 90% Sequential/10% Random 50% Sequential/50% Random 0% Sequential/100% Random
OCZ Core (JMicron, MLC) 0.36ms 25.8ms 130ms 244ms
OCZ (Samsung, SLC) 0.16ms 1.97ms 5.19ms 9.17ms
Intel X25-M (Intel, MLC) 0.09ms 0.09ms 0.09ms 0.089ms
Seagate Momentus 7200.2 0.16ms 0.94ms 4.35ms 9.4ms
Theaverage latency was higher on the OCZ Core (MLC) than the rest of thedrives, but still manageable at 0.36ms when I ran the 100% sequentialtest, but look at what happened in the 90% sequential test. With just10% random writes the average latency jumped to 25.8ms, that's 13x thelatency of the OCZ SLC drive. Again, this isn't an MLC issue as theIntel drive does just fine. Although I left it out of the table to keepthings simpler, the max latency in the 90/10 test was 983ms for the OCZCore drive once again. The 90/10 test is particularly useful becauseit closely mimics a desktop write pattern, most writes are sequential innature but a small percentage (10% or less) are random in nature. Whatthis test shows us is that even 10% of random writes is all it takes tobring the OCZ Core to its knees.
The problem gets worse as youincrease the load on the drive. Most desktop systems have less than 1outstanding IO during normal operation, but under heavy multitasking youcan see the IO queue depth hit 4 or 5 IOs for writes. Going much abovethat and you pretty much have to be in a multi-user environment, eitherby running your machine as a file server or by actually running ahighly trafficked server. I ran the same 100% random, 4KB write testbut varied the number of outstanding IOs from 1 all the way up to 64.Honestly, I just wanted to see how bad it would get:

Thisis just ridiculous. Average write latency climbs up to fifteenseconds, while max latency peaked at over thirty seconds for the JMicronbased MLC drives. All this graph tells you is that you shouldn't dareuse one of these drives in a server, but even at a queue depth of fourthe max latency is over two seconds which is completely attainable in adesktop scenario under heavy usage. I've seen this sort of behaviorfirst hand under OS X with the SuperTalent MLC drive, the system willjust freeze for anywhere from a fraction of a second to over a fullsecond while a write completes in the background. The write that willset it off will often times be something as simple as writing to my webbrowser's cache or sending an IM, it's horribly frustrating.
I didlook at read performance, and while max latency was a problem (peakingat 250ms) it was a fairly rare case, average latency was more thanrespectable and comparable to the SLC drives. This seems to be a writeissue. Let's see if we can make it manifest itself in some real worldtests.
Page 9
The Generic MLC SSD Problem in the Real World
Based onthe Iometer results I knew for sure that there was an issue with randomwrite performance on these SSDs, the only common thread between thembeing the type of controller (JMicron JMF602) and the MLC flash devicesbeing used (Samsung). But I wanted to see if I could get the highlatency writes to appear in a real-world benchmark-able way.

The Samsung SLC controller

Thefirst indication that something is wrong actually comes from runningthe Windows Vista install itself, the MLC drive takes 25% longer tocomplete the install (let’s just ignore the part about the full installnot being fully functional upon completion). Clearly there’s an issuewith write speed. I ran into something similar with OS X, but I didn'tput it together until now.
The problems are far worse in Vista.While OS X will just pause until the data is written, Vista doesn’t seemto respond well to unusually long file write delays. I haven’t beenable to get the Vista install to complete without errors on the OCZ Coredrive. One install completed but I was greeted with this error as soonas I hit the desktop:

Trying to reinstall gave me this error even before I booted into Vista:

Itlooks like the Vista install doesn’t do well with significant delayswhen writing files to the disk. The only way I could actually get areliable Vista install on the Core drive was by cloning another drivewith a working Vista image on it.
For the next test I tried creating a 200MB archive of pictures:

Sofar, so good. The OCZ Core performs no differently than the rest ofthe pack. Now let's try creating the same archive, but also extractingone at the same time:

Ah-ha!Now we're on to something, the SLC and Intel MLC drives are botharound 30% faster than the OCZ Core. Let's try creating the samearchive but extracting a much larger one:
Creating 200MB Archive Extracting 5GB Archive Number of Pauses
SuperTalent (JMicron, MLC) 83 seconds 573 seconds lots
Silicon Power (JMicron, MLC) 128 seconds 632 seconds tons
OCZ Core (JMicron, MLC) 60 seconds 222.7 seconds 20
OCZ (Samsung, SLC) 42 seconds 94.7 seconds 0
Intel X25-M (Intel, MLC) 42 seconds 113.7 seconds 0
Seagate Momentus 7200.2 72 seconds 260.6 seconds 0
Western Digital VelociRaptor 46 seconds 90.9 seconds 0
You'llnotice a new column called number of pauses; this column is the numberof times all disk activity ceased on the system, causing the wholemachine to stutter for a moment. You'll also notice that there arezeros in this column, unless the drive uses the JMicron controller.Also note the randomness of the problem, the OCZ, SuperTalent andSilicon Power drives all use the same hardware yet I saw tremendousvariations between runs. This is a manually timed test but the rest ofthe drives didn't vary nearly as much.
It's also important tonote that while the Seagate notebook drive performed similarly to theCore, it didn't suffer from the pauses. What this helps illustrate isthe nature of the problem, it's very bursty - you get a period of veryhigh performance followed by an abrupt stop. The abrupt stops, as wenow know, are these 0 - 2 second write latencies where everything in thesystem is completely starved of data until the write is complete.

Poor hungry CPU, it just wants to eat. Comic by Laura ofwww.laurascomics.com
Fromthe CPU's perspective, it expects new data on a nanosecond scale,waiting a full second for anything is deadly for performance.
Anotherway of quantifying the impact is looking at how long it takes to launchan application when we're in this high-latency write period. I triedextracting the same 5GB archive and launching PowerPoint 2007 orPhotoshop CS3 (not at the same time).
Launching PowerPoint 2007 While Extracting 5GB Launching Photoshop CS3 While Extracting 5GB
OCZ Core (JMicron, MLC) 8.5 seconds 24.3 seconds
OCZ (Samsung, SLC) 2.8 seconds 9.3 seconds
Intel X25-M (Intel, MLC) 3.85 seconds 10.5 seconds
Seagate Momentus 7200.2 21.3 seconds 46.5 seconds
Western Digital VelociRaptor 8 seconds 23.5 seconds
Allof the drives took longer to launch the applications, but while the SLCand Intel MLC drives still performed in a league of their own, the MLCdrives behaved like conventional hard drives. Try running anapplication while your disk is busy doing something else, or better yet,try running a couple - they take forever. SSDs fix that problem, or atleast they're supposed to. These MLC drives don't, at least notalways; thankfully the SLC drives and more importantly, the Intel MLCdrive don't exhibit this problem.
Page 10
A Lack of DRAM or a Lack of Cache?
The high latencyrandom write issue is annoying enough that it'd force me not torecommend any of the non-Intel MLC drives on the market today,regardless of how attractive their pricing may be. High performancewith caveats isn't exactly high performance to me.

Intel's controller (left) + DRAM (right)
Thatbeing said, the root of the problem is still unknown. My first thoughtwas that it was because the MLC drives had no DRAM buffer, and ifyou'll notice, Intel's MLC drive does have a DRAM buffer. I asked Intelabout this and it turns out that the DRAM on the Intel drive isn't usedfor user data because of the risk of data loss, instead it is used asmemory by the Intel SATA/flash controller for deciding exactly where towrite data (I'm assuming for the wear leveling/reliability algorithms).Despite the presence of the external DRAM, both the Intel controllerand the JMicron rely on internal buffers to cache accesses to the SSD.

Findinggood data on the JMicron JMF602 controller is nearly impossible, butfrom what I've heard it's got 16KB of on-chip memory for read/writerequests. By comparison, Intel's controller has a 256KB SRAM on-die.And I'm going to go out on a limb and assume that given Intel'sexperience with CPU caches, that its SRAM implementation is probablyvery well done.
With the JMicron based solutions, if you try andwrite too much to the drive (and trust me, it won’t take a lot) and thebuffers get full, the controller tells the system that it’s not ready towrite more data and you get a pause.
When you cause the JM602’sinternal buffer to overflow, your system runs in bullet-time.Applications take much longer to launch and close, windows take longerto appear, and there are distinct pauses in anything you want to do thatinvolves the disk. Want to send an IM? Well, that writes to an IM log- you can expect a pause before you can send your IM. Loading webpagesis the worst, reading from and writing to the cache wreaks havoc onthese cacheless MLC drives. Just for kicks I tried loading AnandTechwhile I was extracting a 5GB file on the SuperTalent 60GB MLC, it tookover 10 seconds for the website to load. Once the JM602 was free tofulfill the read request, the website just popped up - but until then itwas like my DNS was failing. It’s a lot like what happens to yournotebook if you try and do too much, the disk quickly becomes abottleneck.
Thankfully, as we've already seen, this problem isonly limited to JMF602 based MLC drives. The SLC drives and the IntelMLC are totally fine, so while I'll include these problematic MLC drivesin today's comparison, let me state now that I would not purchase one.
JMicron'sroadmap shows a new controller next year with an integrated ARM core aswell as support for external DRAM, which could alleviate theseproblems, but until now the controller, and drives based on it, aren'tworth it. You get a much better overall experience out of aconventional mechanical disk drive, and much better performance from theIntel SSD or any of the SLC solutions on the market.
The Test
CPU: Intel Core 2 Quad Q9450 (2.66GHz)
Motherboard: Intel DG45ID
Chipset: Intel G45
Chipset Drivers: Intel 8.1.1.1010 (Intel)
Memory: Corsair XMS2-8500 1066
Video Card: VisionTek Radeon HD 4850
OS: Windows Vista Ultimate 32-bit
Page 11
Overall System Performance with SYSMark 2007
Our firsttest is the full SYSMark 2007 benchmark suite. It's a system-levelperformance test that measures performance based on average responsetime, making it a great candidate for our SSDs, which should providepretty low response times, at least compared to standard mechanicaldisks.





TheIntel X25-M does very well here, just edging past the fastest desktopdrive on the market and outperforming all other mechanical disks in itsown 2.5" form factor. We saw a 15% increase in overall performanceunder SYSMark compared to the Seagate Momentus 7200.2; to get that sortof speed boost on a notebook would take Nehalem otherwise.
You'llnotice that I threw in the 150GB Raptor in the mix as well, that's ourcurrent CPU testbed HDD. Our Q9450 saw performance go up by 9% thanksto the X25-M, that's more of a performance boost than you'd get byupgrading to a QX9650 and again, this is a system-level test, not diskspecific.
Here you also see the problems with the JMicronequipped MLC drives, the Intel X25-M manages to outperform these drivesby nearly 30%. The Video Creation test is particularly bad with theIntel X25-M garnering a score nearly 2x of the JMicron MLC drives.We've already established that the X25-M is the only MLC drive that youshould consider, this simply helps provide more support.
Page 12
Real World Performance with PCMark Vantage
Next up isPCMark Vantage, another system-wide performance suite. I chose to runthe whole suite rather than just the HDD test to hopefully bettercharacterize real world performance of these drives.

If we look at the individual test subsets of PCMark Vantage we can see the drive's strengths at work.

The memories suite for example includes a test involving importingpictures into Windows Photo Gallery and editing them, a fairly benigntask that easily falls into the category of being very influenced bydisk performance. The end result is a 15% performance advantage over theVelociRaptor, a 16.6% advantage over the Samsung SLC based SSDs and a42% advantage over the 2.5" Seagate Momentus 7200.2 HDD - the X25-M isgreat for a desktop, but a miracle for a notebook.

TheTV and Movies suite shows that the X25-M won't always dominate. Herethe tests are focused on video transcoding which is mostly CPU bound,but one of the tests involves Windows Media Center which tends to bedisk bound. Despite the nature of the test, the X25-M competes at thetop of the chart but is bested by the VelociRaptor. It's performanceisn't bad, but not earth shattering. Again, compared to other notebookdrives it is a dream come true.

Thegaming tests are very well suited to SSDs since they spend a goodportion of their time focusing on reading textures and loading leveldata. All of the SSDs dominate here, but as you'll see later on in mygaming tests the benefits of an SSD really vary depend on the game.Take these results as a best case scenario of what can happen, not thenorm. You can also see how tempting it is to opt for one of thoseJMicron based MLC SSDs, they perform quite well here - the test simplydoesn't show the ugly side of living with them.

We'reback to utter domination in the Vantage Music test. Here the main testis a multitasking scenario, which SSDs do quite well in: the testsimulates surfing the web in IE7, transcoding an audio file and addingmusic to Windows Media Player (the most disk intensive portion of thetest). The X25-M is nearly 60% faster than the VelociRaptor, aroundtwice the speed of the Seagate Momentus 7200.2 and over 37% faster thanthe Samsung SLC based SSDs. When the X25-M is fast, it's very fast.

TheCommunications suite is made up of two tests, both involving lightmultitasking. The first test simulates data encryption/decryption whilerunning message rules in Windows Mail. The second test simulates websurfing (including opening/closing tabs) in IE7, data decryption andrunning Windows Defender.
Despite the inclusion of WindowsDefender, the X25-M's advantage over the VelociRaptor is only 18%. Iwould honestly expect more based on some of my other system scanningtests, but I believe the reason we're seeing general domination and notutter destruction is that the tasks being run alongside Windows Defenderare quite light on the disk. Yes, I am nitpicking an 18% victory -this drive is that good. The SLC drives do well here but are no matchfor the X25-M. It's in tests like this that the X25-M really earns itskeep, it delivers SLC performance at a much lower cost.
The Productivity test is awesome, let me explain:

Inthis test there are four tasks going on at once, searching throughWindows contacts, searching through Windows Mail, browsing multiplewebpages in IE7 and loading applications. This is as real world of ascenario as you get and it happens to be representative of one of themost frustrating HDD usage models - trying to do multiple things atonce. There's nothing more annoying than trying to launch a simpleapplication while you're doing other things in the background and havethe load take seemingly forever.
Note that the test itself isn'tvery write intensive, so even the JMF602 based MLC drives do well here.I can attest to this as one of the things that drove me to put a SSD inmy desktop was that I wanted my applications to pop up instantaneously,regardless of what I was doing. The pausing doesn't get a chance torear its head, so all of the SSDs rule the playing field here. TheX25-M delivers 2x the performance of the VelociRaptor here and is fasterthan every other drive. Enough said.
Page 13
Application Loading Performance
You know what I hateabout rebooting my machine? Starting all of my applications back up assoon as it boots, the OS is always crunching away at my disk long afterI've hit my desktop and it takes forever to launch my mail application,IM client and start a web browser. These aren't complex requests, butthey take forever on a conventional hard disk.
I ran a quick testto illustrate the point. I took my test image and booted it, as soonas I got a cursor in Vista I launched Adobe Photoshop CS3, PCMarkVantage, Google Chrome and Norton AntiVirus 2008. I timed how long ittook for all four applications to come up. Note that this is actually abest case scenario, since my testbed image is very clean with a minimalnumber of applications installed. As you fill up your disk with moreapplications and files the disparity grows. Remember that HDDperformance decreases as your drive fills up, SSD performance doesn't.
Launch: Photoshop, Vantage, Chrome & NAV
Intel X25-M (Intel, MLC) 7 seconds
Western Digital Green 1TB 15.6 seconds
Western Digital VelociRaptor 12 seconds
Thedifferences in time here aren't huge on paper, but they are tremendouswhen you're actually using the machine. The funny thing about an SSD isyou don't always appreciate the benefit when you've got it, but go backto a mechanical disk and it'll feel like something is wrong with yourmachine. The difference is noticeable.
Now let's look at individual application performance:

GoogleChrome launches quickly on pretty much anything, the difference betweenSSDs is negligible but there is a slightly noticeable differencebetween running a simple, quick loading app on a HDD vs. a SSD. Again,the differences are small, but noticeable. On an SSD the applicationlaunches are instantaneous, on a HDD you can "feel" the load.

PowerPointis similar to Chrome, but the application does take longer to load onmechanical disks. Again, the differences are noticeable - the X25-Mwill load PowerPoint in half the time of the VelociRaptor. You canargue that the price isn't worth it, but the point is that the loadcompletes much faster...let's look at what happens when we're loading anapplication that does take a long time to start.
Photoshop takes its sweet time loading on a notebook drive, over 10 seconds on the Momentus 7200.2:

Evenon the VelociRaptor it's a 6+ second ordeal. On the X25-M and all ofthe other SSDs, we're talking less than 3 seconds (the X25-M isnaturally the fastest at 2.3 seconds).
My rule of thumb hasalways been that a good SSD will cut application launch times in halfcompared to a desktop hard drive, so if you have a particularly beefyapplication that takes a good while to load, expect that time to go downconsiderably. Remember, these are best case scenarios, in a little bitI'll start multitasking and then we'll separate the men from the boys.
Page 14
Game Load Performance
I looked at three games for loadtime performance and I picked these titles because they run the wholespectrum of how well an SSD will work in a gaming machine. First up isSpore.
Here I'm timing how long it took to load the next stage ofevolution in the game, my character just grew legs and is about to walkon land for the first time:

You'llsee that the results are pretty much the same on all of these drives,we're reading from the disk but we're bound elsewhere in the system - anSSD won't do anything for you here and there are many games where youwon't see improvements in load times.
Next up, Oblivion - I'm simply loading a savegame:

Comparedto notebook drives, the SSDs do extremely well. Against the 3.5"competitors, the margins are smaller but there - you shave a couple ofseconds off of the VelociRaptor's time, more off of the 1TB WD drive (orany other 5400RPM drive). For some reason these sorts of performancegains are far less noticeable in real world usage than the improvedapplication launch times. Perhaps it's because when you're playing agame it's all you've got going on and you're more patient, whereas whenlaunching a web browser or an application you're trying to complete atask where time is more valuable. I'll refrain from any morepsychological analysis and stick to the drives though.
Finally weget to Crysis, and this is the other end of the spectrum because Crysisis a game that is very demanding on the I/O subsystem. We've got threeresults here: 1) How long it took to launch the game and load thebenchmark level, 2) Minimum FPS, 3) Average FPS. I'm only reporting theminimum and average frame rate of the first run, because in that run ofthe benchmark the textures and level data are still being streamed offthe disk and it actually impacts both minimum and average frame rate.

TheIntel X25-M manages to load the game and benchmark level in around 80%of the time of the VelociRaptor and about 90% of the time of the SamsungSLC SSDs. Compared to notebook drives the performance advantage isstaggering. Game load and level load performance can be improved andCrysis is a good example of that.
The minimum frame rate graphis very interesting because it shows the impact loading textures from aslow disk can have on minimum frame rates in a game. The Crysis testis a bit exaggerated since it's streaming data as fast as possible,which you don't always do in a game, but if you've ever felt your gamegrind to a halt and hear your disk thrash this is a good test of that.

TheX25-M has a 33% advantage here over the VelociRaptor, and I won't evenmention the utter destruction of conventional 2.5" HDDs. Compared toother SSDs, the Samsung SLCs come the closest but Intel still manages a25% advantage. The JMicron based MLC drives suffer terribly here,Intel's MLC drive is 63% faster. I will also point out that for acouple of runs the JMicron MLC drives managed a minimum frame rate of 3fps, several of those lovely pauses happened in the middle of thebenchmark which really changed things. The phenomenon was random enoughthat I reported the more common frame rate but it's worth pointing outthat the pausing issue can happen while gaming, which would be bad ifyou're playing any sort of multiplayer game.
Average frame rate isobviously affected, but you can see that the numbers are much closerindicating that the minimum frame rates are at least not sustained forlong periods of time.

Whatcan we conclude here? SSDs can be good for gaming, but they aren'tguaranteed to offer more performance than a good HDD. And where SSDs dooffer an impact on gaming performance, Intel's X25-M continues todominate the charts.
Page 15
The Killer App: Multitasking Performance
Here's wherethings get interesting. I ran two tests, in one I extracted a 5GBarchive and tried to run Photoshop after 30 seconds of extraction and inthe second test, I extracted a 5GB archive and tried to scan my systemfor viruses using NAV 2008. Simple enough, right? I'm reporting thetimes of each task individually.
The extraction task took the longest to complete but the standings speak for themselves:

Here'sone area where the Samsung SLC based devices actually come out ahead,by a good margin. The Samsung SLC SSD finished the extraction in 102seconds, compared to 161 seconds for the X25-M. Even the VelociRaptordid better here at 116 seconds, but remember you need to look at bothtasks for a complete picture:

LaunchingPhotoshop took 5.2 seconds for the Samsung SLC SSD, it was like weweren't even running another test in the background. The X25-M did fineat 12.2 seconds and the VelociRaptor was much slower at 27.3 seconds.The JMicron based MLC drives didn't do too bad here either, althoughthey were a little slower than Intel's MLC offering.
The realstress test was this next multitasking scenario. Quite possibly one ofthe most annoying thing about viruses is having to run real timescanning and protection software all the time, especially with atraditional HDD in your system. The extraction task is the same asbefore, but the other task is a full system scan in Norton Anti-Virus2008. I timed both:

TheWD 1TB drive would always complete the extraction task quicker than allof the other drives, but paid the penalty in the scan test (which iswhy you have to look at both charts for a full analysis). The SamsungSLC drive is still the overall winner here, followed closely by Intel'sX25-M. The JMicron based MLC drives do horribly here, taking over twiceas long to complete as Intel's MLC.
The mechanical disks howeverdo a lot worse. While the Intel X25-M took 3.5 minutes to extract the5GB archive, the VelociRaptor took over 17 minutes. The SeagateMomentus 7200.2 took over 23 minutes!
Let's look at the NAV results:

TheX25-M took around 5.3 minutes, the Raptor needed more than 23 minutesand Seagate Momentus made me wait over 40 minutes. It's these sorts ofusage scenarios that really make SSDs worthwhile, and they are the mostrealworld you can get. When we started looking at real worldperformance the PCMark Vantage numbers may have looked a bit ridiculous,but by now you should see that they are more of the middle ground whenlooking at performance of these drives.
Page 16
Power Consumption & Battery Life
Intel lists theidle power consumption of the X25-M at 0.06W and the "typical workload"power consumption as 150mW. This is far lower than any conventionalhard drive and relatively comparable to other SSDs.
In a desktopsystem using the Intel X25-M will shave off around 3 - 5W of totalsystem power at the outlet under idle or load conditions. The savingsare less compared to a 2.5" drive, but the real question there is howmuch more battery life do you get if you switch to the X25-M?
Wedidn't have enough time to run through a full suite of battery lifetests but using MobileMark 2007's Productivity test I looked at theIntel X25-M, the OCZ Core (JMicron based MLC), the OCZ SLC SSD (Samsungbased SLC) and the Seagate Momentus 7200.2.

Overallperformance was highest on the OCZ SLC SSD, but only marginallycompared to the Intel X25-M. The OCZ Core managed to come in thirdplace, still faster than the mechanical disk.

Thebattery life of the system was actually lower with the OCZ Core,presumably thanks to the strange write performance issues. Butgenerally the SLC, MLC and HDD all last around the same amount of time.Intel's X25-M boosted battery life by a bit over 6% or 27 minutes, notan insignificant increase in battery life.
Page 17
Final Words
At the high level, SSDs are still the key totruly solid performance and this is where the issues with the JMicronbased MLC drives are really unfortunate, because it means that the mostaccessible SSDs on the market can actually deliver a pretty bad userexperience. But if you look at what Intel's X25-M and the Samsung SLCdrives can deliver, it's really quite good.
As I've mentionedbefore, the random write issues with JMicron JMF602 based MLC SSDs aresimply unacceptable and in my opinion they make the drives unusable foruse in any desktop or notebook that you actually care about. Next yearwe may see a JMicron controller that fixes the problem but until then,I'd consider those drives off limits.
This thing is fast, and Iwant one in my system...actually, two. It's the only SSD that I wouldactually go out, buy and stick in my desktop machine at this point. Ithink that's the first time I've ever said something like that in areview, but I'm absolutely convinced. I've been using SSDs in mysystems for a few months now and I'm hooked.
What Intel did withthe X25-M is show the world what is possible with MLC flash. You getbetter than SLC performance, at lower than SLC prices. Despite that, theabsolute only thing that bothers me about Intel's X25-M is the price.Although Intel is totally justified in pricing the X25-M at $595, I washoping for pricing inline with the JMicron based MLC SSDs. At $300 -$400 this would be a no brainer for any enthusiast, and honestly even at$595 it's worth considering if you have other drives for data storage.
The other complaint is obviously capacity; at 80GB you can getby with this being the only drive in a corporate notebook or even yourpersonal notebook if you've got external storage, but in a desktopmachine 80GB is a bit shallow. Thankfully with better reliability thanconventional hard disks you should be able to put two of these in RAID0, doubling capacity without any fear of reduced reliability. Then weget back to the pricing problem unfortunately.
If Intel can getcapacities over 100GB at reasonable prices in the near future, I'd saythat the X25-M would be the best upgrade you could possibly do to yoursystem. I'm curious to see what pricing and availability will be likefor the 160GB drives, but Intel is being pretty tight lipped about them.
The implications of the X25-M are tremendous. I mentionedbefore that it is the great equalizer between the notebook and desktop,you can finally have a desktop usage experience on your notebook ifyou've got one of these in there (or a SLC SSD). If capacities growquickly enough, these SSDs could mean that desktops would startaccepting 2.5" drives, allowing for smaller form factors. Apple isalready doing this on the Mac mini, and we've seen 2.5" drives used insystems like the ASUS Eee Box, but now you don't have to give upperformance.
Intel didn't start the SSD revolution but it suredid kick it into high gear. Companies like Samsung and JMicron arereally going to have to step up their game if they want to compete withIntel going forward. While Intel was light on details about the tricksthey implemented in their controller, it's clearly enough to completelychange the way we look at MLC SSD performance. And if this is the sortof performance we can expect out of its MLC drive, I'm wondering whatwill happen when we look at its SLC drives.
To me this is biggerthan Nehalem, but then I look at the price tag and think that Nehalemwill probably be a cheaper upgrade. Intel redefined the performance ofthe MLC SSD, I only wish they also redefined the price...