The recently launched Crucial RealSSD C300 is (arguably) the fastest available SSD, when used with SATA600.
But Anand managed to brick his sample quickly and encountered an abnormally high max latency with this drive, obviously using early firmware.
It's known that both Intel drives and Indilinx based drives had issues with early firmware, but they were eventually fixed.
I wonder how long it's going to take this 'firmware stabilisation' for the C300 drives, anyone has some clues?
Link to Anandtech article: http://www.anandtech.com/storage/showdoc.aspx?i=3747

It's been out for what, 2-3 weeks ? And they are not even in stock in many stores here so you should at least wait until they have been in actual use before asking that question.
There is no magic here, to know if they are bug-free you need time and users.

Despite the chicken-and-egg syndrome at play, I recommend sticking to an established player unless price is no object. I suspect the entire market may be different in another year when the SATA 6 Gb/s support is widespread and makers have moved to smaller manufacturing processes.

Come back, one year.

The safe move is to buy an Indilinx controller based drive if you are an ATI/AMD fan or an Intel controller based drive if you are an Intel fan.

The Intel controller is better than the Indilinx but the faster drives seem to have more firmware bugs (including Intel).

If you want rock solid stability you'd get some crappy slow samsung or jmicron based drive.

I'd love to have a C300 or a 160GB Intel G2 but I'm not going to pay for either at these prices.

I'll either pop for a 64GB Indilinx or I'll continue to buy more 40GB Intel drive(s).

Oh and if I had all the money in the world I'd do a RAID 1 between a C300 and an Intel and do backups to a eSATA enclosure. I wouldn't trust any single drive controller to be the only key component between me and my data.

Well, that's not what I hoped to hear... Eunos, maybe your 'soup nazi' remark is exaggerated? I was thinking about waiting 3-4 months.
The drives are available here: http://www.overclockers.co.uk/showprodu ... ubcat=1427
I can afford one 256GB C300, for my new build, in 3-4 months, but two of them for Raid1 would be psychologically too expensive.
When I'll build the new system I also plan to get a USB3 enclosure with a cheaper 256GB (maybe Samsung) for backups.

If the problem is in the firmware/controller a RAID 1 of two C300 SSDs could be just as unreliable as a single C300. It is possible that both drives would brick at the exact same moment based on receiving the same write commands. I doubt that read commands are what bricked Anands SSDs.

If you use a 160GB intel + a 256 GB C300 to make a RAID1 of ~150GB. You will have unused space on the C300 but you'll increase your uptime probability dramatically.

Firmware updates would be individual instead of the entire array and a bricked drive during an update would not cause loss of data. The odds of two different controllers bricking themselves at the same time during normal use would be astronomical. You would however have to stop what you are doing and backup your data as soon as you lose redundancy in the array and replace the failed drive asap.

If you wanted the capacity you could RAID 1 between a 256GB C300 and a Crucial M225 256GB (Indilinx). You could then make a ~200GB RAID 1 partition. This would probably be measurably slower than the Intel + Marvel combo above but probably not noticeable outside of benchmarks and would be as reliable or more reliable with the added benefit of 50GB of additional usable space.

Of course you can just use a single drive and avoid the RAID concept but if you are going to raid an expensive SSD I'd suggest doing it with dissimilar drives.

Oh and the C300 is clearly the fastest 256GB drive.

Sandforce drives that beat it in some benchmarks top out at 200GB.

Intel drives that beat it in some benchmarks top out at 160GB (MLC) or 64GB (SLC).

Indilinx drives match the capacity of 256GB but the C300 wins any realistic bench vs Indilinx.

Jmicron JMF61x drives match the capacity of 256GB but are slower than all the above.

Samsung drives match the capacity but again are slower than all the above.

http://www.anandtech.com/storage/showdo ... i=3757&p=5 is from the article with the most recent benchmarks.

Thank you for the replies, dhanson865. The C300 seems very good value for (a lot of) money, since it sells at the same price point as an Indilinx drive.
But I do need a reliable solution, so early adoption is out of the question. I didn't think that two drives might brick at the same moment, thank you for pointing this out.
And using a second different drive for Raid1 would lower performance and increase cost per GB to unacceptable levels. So I'll stay away from Raid.

I don't understand why slower, cheaper drives like the Samsung or JMicron based ones would be more reliable than the Intel or Indilinx based ones.
If you know something specific, please share it in this thread. AFAIK the Intel drives only have a problem with the BIOS drive password.

Intel did have firmware dramas as recently as a few months ago, which appears to be fully sorted now. The performance of the MLC Intel is now middling in most of Anand's benchmarks, but I am confident recommending them because they have reached a stage of proven maturity in both the product and firmware. I believe the G2 came out as far back as late '08, so it's a real grandpa.

No harm in being an early adopter if the funds and enthusiasm are there, but one just needs to be very pedantic with a backup routine and willing to accept the higher probability of dramas. But if anyone asks my advice I continue to stick to the well-knowns.

We've covered a lot of topics in this thread so I'm going to separate my replies to hopefully reduce confusion. Reliability I'll save for later, for now I'm addressing RAID 1.

Keep in mind that adding a slower drive to the C300 in RAID 1 will not slow down the overall configuration for most usage patterns. It will increase the cost per usable GB that's for sure.

Read requests will be satisfied by the drive that replies first. If IO depth goes above 1 for any noticeable length of time both drives will respond independently increasing the read bandwidth higher than a C300 can do on its own.

For example (not real numbers from direct testing) If the C300 could do 330MB/s read by itself and a Indilinx drive could do 250MB/s read by itself the combination in RAID 1 would not be ~290MB/s but would instead be much higher something I don't have a proper number for but lets say >400MB/s easily. The request would have to be sufficiently huge to allow both drives to get up to speed on sequential reads at the same time but they would at times do so.

Again not real numbers, say you have random reads and the C300 can only do 75MB/s and the Indilinx will only do 30MB/s. The RAID 1 of the two drives will not be slower than the C300 alone. You will see read speed improve to >100 MB/s.

On the other side of the fence for writes there will be a small penalty in that all writes have to be satisfied on both drives (plus any wear leveling/garbage collection/trim activity) but so long as the slower of the two drives is responsive and your writes aren't bigger than the cache+flash write rate you won't notice that either.

Now if you know all that already and a RAID 1 primer wasn't needed I'm sorry but your phrasing of "using a second different drive for Raid1 would lower performance" makes me think you misunderstand how a RAID 1 array would behave in real world use.

Even at its absolute worst a RAID 1 array of two indilinx drives would be faster than a single Velociraptor and at it's best a RAID 1 of two indilinx drives would beat the read performance of a single C300. A RAID 1 of two C300 drives would be insanely fast and a RAID 1 of a C300 and a indilinx or intel drive is still going to beat the pants off of any single SSD in most usage patterns.

The nasty side of this situation is that I can't give you a URL to prove the point with public numbers. You can't just grab HDtune or HDtach and get raw sequential reads off of a software RAID1. The only way you could get an app like HDtune to benchmark a RAID1 array is if the array was created outside of the OS and looked like a single physical drive to HDtune (aka Hardware raid) which opens the possibility of getting crummy benchmark data from someone who doesn't have a raid controller that can truly keep up with modern SSDs. Most reviewers don't bother to try asymmetrical RAID setups because they don't give consistent easily reproduced benchmark data. The lack of easy testing doesn't mean it isn't a viable configuration though.

The performance you get out of a software RAID and modern SSDs or even a single modern SSD puts to shame small hardware RAID arrays with weak controllers that can barely keep up with 10,000 RPM drives.

I can tell you from experience that even a cheap $100 SSD in RAID 1 with a preexisting RAID 1 array of server grade rotating disks that the cheap SSD increased the performance of the server. I can tell you by the fact that the Read Queue length and Write Queue length as reported by windows performance monitor are now 0 when they used to be >0. I can't tell you what the Frankenstein array does in MB/s, I can tell you it works in a 24/7 always on server and hasn't caused any problems and has in fact made things better. I never did timed tests as the server handles multiuser load and the tests that would represent the user load I would need to model are more work than its worth when I can watch existing metrics and see the improvement in real world use.

OK, reliability. I don't work for a drive manufacturer so everything I can possibly say about this could be considered "hearsay" or a misunderstanding and to be sure I've read about and used every type of storage I could possibly get access to or afford but I don't always note external sources of my information. With that said I'll try to quote Anand at least to give someone else as a partial reference for what I'm thinking.



[quote="http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3755&p=5"]My RealSSD C300 died in the process of testing it with the 890GX. I don’t believe it was the motherboard or chipset, but right now it looks like I stumbled upon an untested usage case that put the drive in a state where it won’t even let a system POST anymore.

It’s because of situations like this that I’ve been very cautious in recommending any new SSDs. Hence my conclusion in the Vertex LE review:

“Go up another $100 and the recommendation is easily the Crucial RealSSD C300. Again, assuming that nothing horrible ever happens with the drive. I do have more faith in Crucial’s validation testing given that Micron is shipping the same drive to OEMs, but it’s still a brand new, unproven platform.â€

Wow, dhanson865, that's a lot of information, thank you very much for taking your time to write it.

Indeed, I have never used Raid, so please forgive my ignorance regarding the Raid1 performance with dissimilar drives.
Your explanation of the Raid1 performance in the first post is very good, and I'm sure other SPCR dwellers will appreciate it too.

If I understand correctly, if I would use a 256GB C300 drive and a 256GB Indilinx based drive, then there wouldn't be much of a capacity loss.
This is in contrast with using a 160GB Intel as the second drive, where the capacity loss would be at least 90GB, and I would feel sooo robbed.

There could be a problem with Raid, that the TRIM command wouldn't work? Maybe because TRIM requires the M$ AHCI drivers, and those don't support Raid?
The problem could get worse if the 'SSD Optimizer', or whatever the stand-alone trimmers are named, can't work in a Raid configuration with two different drives.
I may be wrong on this, as I said before I never used Raid, so please be gentle.

Regarding your second post, I had read those Anandtech articles before starting this thread, and I'm still in doubt about the reliability of older slower drives.
WD may be just talking BS when claiming their new SSD drives are slow because they are much more reliable.
I am using WD (spinning) drives in my main computer, backup computer and a WD passport, so I don't bash WD.

It's possible that the reduced reliability due to firmware bugs only applies to drives with cache? In which case is there a cache-less drive that's not rubbish speed-wise?

PS. After reading again the latest review at Anandtech, about the SiliconEdge Blue, and comparing the performance with other jMicron 618 drives, well... it seems WD engineers did a good job, because other implementations of JM 618 are a bit slower. But the decision to use JM instead of Barefoot or Marvell is quite strange if WD expect to sell those drives at a premium. Probably the bean-counters are to blame.

PS2. I noticed your hints: "I'd gladly take as many as you can give me..." and "Again I'll take every one you can send me...".
I do appreciate the time you spent writing the valuable information in your above posts, but I'm afraid I can't afford to reward you with a SSD drive.

Dhandson865

Please explain. Do the 30 or 32G Vertex SSDs have 64mb or 32mb?

How about the 64GB Vertx?

Software raid treats each drive as a single drive in windows disk management but neither drive shows up anywhere else. Instead an abstraction is created to represent the pair. I can't say if trim works with a software RAID 1 as I haven't tried it. It's only possible in Win 7 and Server 2008 R2. Regular Server 2008 doesn't support trim and R2 is 64 bit only. I haven't made the switch to 64 bit Server versions due to driver issues.

I do know that if you have an Intel controller in the SSD the toolkit can be scheduled to run daily to to a trim even if the RAID doesn't do it automatically in real time. The catch is your drive has to be visible to software like HDtune, speedfan, etc. If you can read the smart data you can manually trim the drive with a utility.

Unfortunately server grade hardware doesn't let you do that. SMB (system management bus) on the motherboard intercepts the traffic and prevents using manual trim on a SSD.

I'll probably fire up a Win 7 PC sometime this year and test the whole trim in RAID 1 concept just to prove a point but I haven't done so yet.

The last I saw manual trim on Indilinx was beta/less smooth than Intel/more risky.

The nice thing is with RAID 1 if you accidentally wipe one drive you can just undo the RAID and redo it after preparing the "blank" drive.

But honestly that is more work than it's worth. I'd just let the SSDs performance degrade for months at a time if automatic trim wasn't happening.



There is no such thing as a SSD without cache. The less cache the SSD has the worse the write performance will be. This is why Jmicron had such a bad reputation. Their earliest SSD controllers only had 8KB of cache and anything less than MB of cache wasn't enough to stop the whole PC from grinding to a halt while waiting for writes to complete on a SSD.



I mostly put those comments in there just for the same reason you said you weren't trashing Western Digital. If some vendor saw my comments or even a random SPCR reader I don't want them to misunderstand. If Intel, Marvel, Crucial, Indilinx, OCZ, Kingston, etectera wants to send me review samples or just decides to reward me with a pallete of overstock one day I'll gladly take them but I didn't expect them to come from you.

30GB and 60GB shows the Vertex with 32MB cache but Crucial M225 has the same controller and 64MB of cache. Pick your source, if you google "indilinx 32mb cache 64mb cache" you'll see any number of stories from the anouncement of the OCZ drives mentioning that OCZ only used 64MB on the 128 GB and larger drives.

Crucial decided to use 64MB on the 64GB drive and as far as I know is the only company to do so with the 64GB indilinx SSD.

For some reason the reviewers are so focused on the 128GB and larger drives so no one has paid attention to the fact that the Vertex at 64GB is overpriced compared to the Crucial with more cache.

I think there has been coverage on how the 32GB drives are slower than the 128GB drives.

New firmware for the C300 fixes a ton of issues but as usual updating the firmware on your SSD may brick it so backup your data first and have a spare drive around to use if you brick the SSD. Oh, and besides even if it isn't bricked the firmware update wipes the drive.

http://forum.crucial.com/t5/Solid-State ... td-p/12363

Apparently the new firmware bricks the drives, at least under certain circumstances, according to this Anandtech article:
http://www.anandtech.com/show/3694/cruc ... rmware-fix
I think there should be a rule for SSD firmware updating: don't do it sooner than a month (or two?) after release.

Holy circular references batman! Or should I say thanks Gilligan? I'm not sure which reference would seem funnier to you, don't know how popular those 1960s era shows were in your country.

Anyway the anandtech link you posted is where I got the URL I posted and WHY I posted it.

Bricking SSDs is common, it's happened with OCZ, Intel, Crucial, etcetera.

The good news is if you do it right and/or you are lucky enough to buy the drive after the firmware is updated it's a big plus that this firmware fixes most any issue you could have cared about for the C300.

Just another week until the firmware updater is fixed. The v0002 firmware should be stable. Then another month or two for people to start trusting it.

but are Tzupy and I the only ones on SPCR closely watching this drive? Is price the reason?

I guess at $350+ there aren't a lot of SPCR's willing to test this baby but I honestly think it'll be the drive to beat come summertime when the firmware brouhaha has died down.

If they came out with a 64GB version of the C300 that was cheaper than the 80GB Intel G2 it'd sure be nice.

Diminishing returns. Assume that upgrading from a hard drive to an Intel SSD will make some previously slow things awesomely fast. Now upgrading to the fastest available SSD will make these things even a bit faster. But is that little bit more anymore worth the loss in trust?

Here's a couple of quotes from Recommended Hard Drives that sum up my thoughts.

* we recommend using [SSDs] if you can afford them, but keep to the models that have the best reliability. Not surprisingly, given the company's resources, Intel SSDs seem tops in this regard.

* Small differences in [hard drive] performance are almost impossible to appreciate in actual use because there are umpteen bottlenecks and overheads in the PC that obscure such differences.

I'm still far more interested in the GB/$ equasion being improved rather than another few MB/s being squeezed out of the SATA interface. For enthusiasts who are not interested in RAID 0-ing common or garden SSDs, though, the C300 is a performance king.

Crucial quietly started selling the C300 in a 64GB version at $150. This is a significant price drop considering its the only 6GB/s sata SSD on the market right now.

Note it's just as fast as the 128GB and 256GB versions on reads but is noticeably slower on writes (just as all the lower capacity drives are).

64GB Crucial RealSSD C300 2.5-inch SATA 6GB/s
CTFDDAC064MAG-1G1

http://www.crucial.com/pdf/Datasheets-l ... online.pdf

http://www.crucial.com/store/partspecs. ... 064MAG-1G1

Andand has been hinting of a coming C300 review that is coming soon now that the firmware has been fixed. Hopefully he'll actually test all three capacities.

Great idea. Not at Newegg yet, but the 64 GB "CT64M225" is $190 and sold out.

http://www.newegg.com/Product/Product.a ... 6820148318

Not sure why you'd want that drive over the C300 but if you do just look at http://silentpcreview.pricegrabber.com/ ... &viewmod=2 you'll see that you can easily order the CT64M225 for $180 no tax and free shipping.

Pricegrabber doesn't show the C300 64GB yet but http://www.google.com/products?q=CTFDDA ... &scoring=p has it at $142 + $9 shipping.

wait, I see you are in Australia. Does newegg ship to AU where other US retailers don't?

On storagereview I was discussing this drive with someone who wasn't impressed with the sequential write numbers. The following is an excerpt from my reply to him (though SR automatically shortens URLs so I'm having to chop some here on SPCR):
-------------
anandtech page 5 and anandtech page 2

2MB sequential READ on 3Gbps SATA


anandtech page 6 and anandtech page 2

4KB random READ on 3Gbps SATA


Now we know the Read speeds are the same for the 64GB, 128GB, and 256GB versions of the C300. So the benchmarks above should remain valid for the 64GB version. However the Write speeds are not the same.

Manufacturer specs on read speeds 3Gbps SATA:


Manufacturer specs on write speeds:

We know the manufacturers specs don't equal the benchmark results but they do at least honestly proclaim lower capacity drives will have lower performance.

You seem to have a pessimistic view of the situation if you think 70MB/s is slow. By that reasoning the Intel 80GB and Intel 40GB drives are slow as well.

btw while we have the benches out anandtech page 6 also gives us

4KB aligned random WRITE on 3Gbps SATA


Which heavily favors the C300. We just don't know short of testing the 128GB and 64GB C300 and the 80GB and 40GB X25 how much they would be worse than the top end drives in their families.

We do however have sequential read tests of one more model from anandtech page 5 and anandtech page 2

2MB sequential WRITE on 3Gbps SATA


Which not only shows that the C300 is faster in this test it also shows how the 80GB and 40GB Intel drives are slower than the 160GB Intel drive.

If you go one step further and compare the spec to the sequential write speed Anand got you have this table
Write Speeds Manufacturer Spec vs 2MB sequential WRITE test


Which still looks pretty favorable to the C300 line.

-------------------------

Now let me be clear the C300 isn't faster than the M25 in all cases. But Until Anand benchmarks the 64GB version I'd be hesitant to suggest that it is slower than an Intel SSD.

Reason I posted the CT64M225 link is because I am confused about the line-up - looks like you get a lesser model for more money.

Newegg don't ship to Australia, the one time I bought from there I had to ask a friend in the States to act as a middleman. However, I keep an eye on the site for reference.

The C300 should be much faster than the Intel in almost all situations, though whether the end user would notice a difference is another matter. As Anandtech said regarding OCZ's internal-RAID PCIe card:

Most desktop users would find it difficult to realize a measurable performance difference between the RevoDrive and a single Vertex 2. While the jump from a HDD to SSD is significant enough in most day to day tasks to tell the difference, application launch times and most conventional desktop uses won’t be affected by the RevoDrive.

The CT64M225 is a leftover from the time before the C300 64GB part was created (it just went on sale a few days ago), it'll take time to either stop selling ones that have already been produced or to lower the price well below the C300.

I'm expecting the production of the M225 series to end very soon knowing that the C300 has already taken its place and the Intel G3 and Indilinx Jetstream will be coming soon.

Hi, on overclockers.co.uk someone has left a review of 64GB C300 drive with performance numbers. To be taken with pinch of salt etc etc but may help place this drive's performance. here.

Looks to me like this is the drive of the moment to some extent. The 40GB intel drive is £100 and this 64GB drive is £120 so better value and, possibly, faster. It's also cheaper per GB than the 128 and 256GB C300 drives...
Regards, Seb

Hexus review of the 64GB C300

What I want to know is, will I see the same performance on my old 965P chipset (Windows 7)?

Is there anything new chipsets added for SSDs that my old chipset won't do well?

You might want to read http://www.anandtech.com/show/2973/6gbp ... -x58-p55/1 which discusses motherboard bottlenecks and the C300.

I doubt it's that big of an issue though. If you have a traditional hard drive the C300 will be faster in 95% or better of the situations you put it through.