Consumer SSD Battle

Want to talk about one of the articles in SPCR? Here's the forum for you.
Pierre
Posts: 156
Joined: Sun Jan 17, 2010 3:55 am
Location: Greece

Re: Consumer SSD Battle

Post by Pierre » Tue Nov 02, 2010 2:16 pm

MoJo, the bit of analysis Pierre fetched very much contradicts a naive reading what you said! Exponential usually refers to something like failures rates of 2% the first year, 4% the second, 8% the thrid, 16% the fourth and so on.
That's exactly what I was commenting on...thank you
I don't know what you call relevant but I went back to the paper and I see that remembered right: they've found strong correlations between SMART data and failures. The only SMART data which is unambiguously about the surface as far as I know is the reallocation count which had a fairly strong correlation with failures. But others parameters had as good (if not better) correlations.
The variables that the report mostly expands on -scan errors*, various reallocation types - seem to be related more to the drive's disk surface than the reading/writing, rotating mechanism...
Unless I read the report hastily, it seemed to me these were more clearly related to disk failure probability than other variables (seek errors, CRC erros, spin retries, power cycles)

* I am including scan errors in this category (of variables more related to disk surface issues) based on what the report says:

"Drives typically scan the disk surface in the background
and report errors as they discover them. Large scan error
counts can be indicative of surface defects,"

On the basis of these findings, I said that surface defects - not having to do with the moving mechanism - could be more related to failures than the wearing of the moving parts. Unless I am misunderstanding something, this should also reply in part to the following:
When the disk is on it is rotating all the time. The motor is running constantly. It doesn't matter if the head is moving or reading/writing data, the drive still spins. In Google's case the drives never stop so start/stop cycles were not included in the data they used. Equally surface defects are not affected by activity, beyond being accessed from time to time which for a drive that is nearly 100% full of randomly accessed data (like Google's) definitely fits that category.
I didn't say surface defects are affected by activity but that it might have more to do with disk failures.

It is true that Google's study has its limitations...I don't argue with that...but you brought up their report, and I pointed to some differences between what you said and what it said.
It is also quite weird that you referred me to the Google study to base your argument on hdd servos/motors, when this constant motor movement is not displayed as a prominent failure variable by that report. And then you criticise it for not taking into account the variable of start-stop counts.
(yet because the drives are in server environment it does not mean starts/stops do not occur, but that power cycles do not occur)

It does address the issue of power cycles - that MikeC touched upon - somewhat by saying:

"Power Cycles. The power cycles indicator counts the
number of times a drive is powered up and down. In
a server-class deployment, in which drives are powered
continuously, we do not expect to reach high enough
power cycle counts to see any effects on failure rates.
Our results find that for drives aged up to two years, this
is true, there is no significant correlation between failures
and high power cycles count. But for drives 3 years
and older, higher power cycle counts can increase the
absolute failure rate by over 2%. We believe this is due
more to our population mix than to aging effects. Moreover,
this correlation could be the effect (not the cause)
of troubled machines that require many repair iterations
and thus many power cycles to be fixed."

SMART keeps track of how many surface defects it sees. What constitutes a SMART failure is defined by the vendor and tends to be somewhat conservative, but if you look at the raw numbers for things like remapped blocks the pattern is fairly clear. The reason SMART typically fails to warn you before a drive fails is simply because its counters do not reach the limits set by the manufacturer. At work we use PC Check which is a bit more realistic with the numbers and which also runs the SMART long test. Most manufacturer's tools can run the long test and it often fails due to scanning the entire surface of the drive and picking up new defects which then push the remapped block count over the limit.
I have HDD Sentinel running all the time, and I always check my hdds for SMART events not portrayed by the manufacturers' tools and test them arduously from time to time...
Moreover, when I first connect a new drive, I always use Sentinel's tools for testing for hardware issues and within the first two weeks I run a full "hdd regeneration" test (reading, writing, re-reading, rewriting) to make sure there are no weak sectors...
...so I'm not arguing -nor did I argue- with you there...

MoJo
Posts: 773
Joined: Mon Jan 13, 2003 9:20 am
Location: UK

Re: Consumer SSD Battle

Post by MoJo » Tue Nov 02, 2010 4:55 pm

HFat wrote:The Google study says that only 9% of their drives report any reallocations which goes to show the 99% figure is indeed on the high side...
Okay, I should qualify that. 99% of drives will have re-allocated sectors, they just might not report it. Most do report some though, but you have to examine the SMART logs to see it. For some reason the counters don't always increment but if you look at the log you can see failures that would have resulted in re-allocation. Maybe it is because the SMART data only seems to use one byte for the reallocation count, i.e. a maximum of 255, and most modern drives have more than 255 spare sectors... I don't know, in all honesty.

Manufacturers also fiddle with SMART data at the factory. Most drives begin life with some sectors mapped as bad during initial testing at the factory. It just isn't economical to try to make every drive 100% perfect, much like how CPUs often have failed cache which is disabled at the factory. The difference is that CPUs with good cache will be sold at a higher price, but HDDs are all just branded "500GB" as long as they meet the minimum good sector count. That makes me wonder if some drives have more spare sectors than others.

In other words: even brand new virgin drives sometimes (often?) have re-allocated sectors which are not included in SMART data. Don't misunderstand me here: I am not scaremongering. It is simply a fact and does not make HDDs any less reliable (see later on). My GPU has lots of broken stream processors but I don't think that makes it less reliable (if anything the opposite is true due to reduced heat output.) All I am saying is that re-allocated sectors are common and a normal part of HDD operation these days because you can't economically build something that can store trillions of bits without a single error.

Also consider what Google's definition of "failing" is. In my line of work that means crashing, failure to boot, corrupt files etc. Google only class a drive as failed once it is dead and no longer working or having unrecoverable read/write errors.
Pierre wrote:I don't know what you call relevant but I went back to the paper and I see that remembered right: they've found strong correlations between SMART data and failures. The only SMART data which is unambiguously about the surface as far as I know is the reallocation count which had a fairly strong correlation with failures. But others parameters had as good (if not better) correlations.
I agree entirely, but would add one important point: it isn't just the level of the counters, it is the speed at which they are increasing which is important. Before failing the rate at which certain parameters increase accelerates. Graphing the parameters over time is a really good idea :)

Pierre: Drives do scan in the background, unless they are in a Google server in which case they are in constant use :) All I can say is that I see a lot of machines which the manufacturer's tools say are fine, the BIOS says is fine and most desktop tools seem to say is fine but which PC-Check says are failed. PC-Check reads the event logs as well as the raw counters and seems to fail drives more aggressively than other software. It is very reliable, as sure enough when you replace a drive it failed the problems go away. We do over 150 repairs a month of which about two thirds to one half need drives replacing. Again, I am not scaremongering - by definition any PC brought in for repair is broken and real life failure rates are a lot lower. I just wonder how many people suffer from crashing and other problems because their HDD is failing but the manufacturer's tools say they are fine.

Oh, and BTW, when we return drives that PC-Check says are broken to the manufacturer for warranty replacement and list "SMART failure" as the cause they always replace them. Clearly they don't just go by the raw counters either :)

HFat
Posts: 1753
Joined: Thu Jul 03, 2008 4:27 am
Location: Switzerland

Re: Consumer SSD Battle

Post by HFat » Tue Nov 02, 2010 6:30 pm

Pierre wrote:The variables that the report mostly expands on -scan errors*, various reallocation types - seem to be related more to the drive's disk surface than the reading/writing, rotating mechanism...
Unless I read the report hastily, it seemed to me these were more clearly related to disk failure probability than other variables (seek errors, CRC erros, spin retries, power cycles)
CRC, retries and cycles are irrelevant or non-existent in Google's experiment.
But it's interesting that they link seek errors to surface issues. Wikipedia talks about issues with "the mechanical positioning system". If these are surface errors, why are the sectors not reallocated anyway? Or are you saying it's a subset of reallocation events?
MoJo wrote: Most do report some though, but you have to examine the SMART logs to see it. For some reason the counters don't always increment but if you look at the log you can see failures that would have resulted in re-allocation.
Are you saying most of your drives have errors in their error logs? Mine don't.
MoJo wrote:I just wonder how many people suffer from crashing and other problems because their HDD is failing but the manufacturer's tools say they are fine.
When drives fail it tends to be pretty obvious. Very few people use the manufacturer's tools or even know they exist anyway.
Would you care to post SMART logs and test results of a drive which seems to be working but is really retuning bad data so often it's a worse problem than software issues? Like if you were to write 1M files all over the drive and read them back, how many would have errors? And would the result be the same if you read them a second time? I'm curious.

If you really have an way to predict failures of individual drives, you ought to work for Google or something. Either they're lying or they don't know how to do that. So Occam's personal hygiene tool says you're mistaken.

Pierre
Posts: 156
Joined: Sun Jan 17, 2010 3:55 am
Location: Greece

Re: Consumer SSD Battle

Post by Pierre » Wed Nov 03, 2010 12:08 am

But it's interesting that they link seek errors to surface issues. Wikipedia talks about issues with "the mechanical positioning system". If these are surface errors, why are the sectors not reallocated anyway? Or are you saying it's a subset of reallocation events?
Not seek errors, scan errors are attributed to surface defects by Google...
The report mostly expands on scan errors and the various reallocation types, which are the variables it finds more directly related to disk failures

I will check that PC Check software, although I am very very happy with HDD Sentinel...it also has different available algorithms to measure the health of a drive based on raw smart data, logs and history of the drive's performance...I always use the stricter analysis -suited for servers.

MoJo
Posts: 773
Joined: Mon Jan 13, 2003 9:20 am
Location: UK

Re: Consumer SSD Battle

Post by MoJo » Wed Nov 03, 2010 7:34 am

HFat wrote:Are you saying most of your drives have errors in their error logs?
No.
HFat wrote:When drives fail it tends to be pretty obvious. Very few people use the manufacturer's tools or even know they exist anyway.
Would you care to post SMART logs and test results of a drive which seems to be working but is really retuning bad data so often it's a worse problem than software issues? Like if you were to write 1M files all over the drive and read them back, how many would have errors? And would the result be the same if you read them a second time? I'm curious.
I can post the SMART logs if you like. I don't have time to test the rest of the stuff but I will save some logs from a drive which has failed but not yet passed the SMART counter failure levels. Sometimes these drives give read errors when you try to clone them, but often they don't. I will try to find one which does not give any errors but none the less has failed. Obviously you will have to take my word for it that replacing the drive fixed the problem but I can't see any way around that.

Often the signs of a failed drive are not that obvious. We get people who bring the machine in for virus removal but when we run the tests their HDD has failed as well. It is hard to convince them to replace the drive, but you can be sure that anyone who doesn't will probably be back within a few weeks. We have to do the tests and inform the customer because if they do come back we need to be able to show them that the problem isn't virus related and is in fact due to their HDD, i.e. not covered under the warranty on the work.

Typical non-obvious signs of a failing drive include drivers crashing (often video or IDE/SATA drivers), corrupt user profiles, failure to install drivers and for IDE drives dropping back to PIO mode in Windows.
If you really have an way to predict failures of individual drives, you ought to work for Google or something. Either they're lying or they don't know how to do that. So Occam's personal hygiene tool says you're mistaken.
I think you don't understand what I am saying, because what I am saying is that the results Google had support my argument. You seem to switch between supporting their conclusions and the disagreeing with them whenever I also agree with them. Are you perhaps getting me confused with Pierre?

HFat
Posts: 1753
Joined: Thu Jul 03, 2008 4:27 am
Location: Switzerland

Re: Consumer SSD Battle

Post by HFat » Wed Nov 03, 2010 9:06 am

So you're not claiming that you can determine that 99% of drives reallocate sectors based on SMART counters or logs. And you're not claiming that the reason SMART fails to predict failures is that its thresholds are not as agressive as those of the software you use to convince your customers to pay for a HD replacements when their OS is having issues. Good. Let's move on.

What the Google study shows is that you can justify replacing a drive after the first error depending on your drive replacement costs, drive failure costs and the age of the drive. You shouldn't wait wait for the second error unless the costs are balanced enough that it makes a difference. I had drives which have survived errors (including the dreaded offline error) and have been working fine for years but I wouldn't take the chance if I couldn't afford failures. Some of these dubious drives ended up failing of course.

Pierre,
The few drives I looked at with smartctl as well as Wikipedia's list lack "scan error" counters which is why I assumed these were seek errors but I see I was mistaken. Perhaps "scan errors" are a type of reallocation. In any case we don't know if these errors are caused by surface defects alone or if mechanical issues are contributing. Clearly aging increases the amount of errors anyway.

Zap
Posts: 21
Joined: Sun Apr 16, 2006 9:52 pm
Location: Kenosha, WI

Re: Consumer SSD Battle: WD vs. OCZ vs. Intel

Post by Zap » Fri Nov 05, 2010 6:46 am

Lawrence Lee wrote:A Windows 7 image loaded with our test suite was cloned to drive
What cloning software was used? Did you verify the SSD was properly aligned after the cloning?
porkchop wrote:i'm surprised the kingston fared so well since it uses a 'inferior' jmicron controller, i guess they've sorted out the old problems.
i believe the wd uses the same controller.
You don't think Jmicron could improve on their products in newer generations?
alain wrote:The sf2000 is an enterprise grade controller, will probably be a while before it shows up in consumer ssd's.
Actually I'd say it would show up pretty quickly in something like an OCZ Vertex 3 Pro Limited Edition or something like that. :lol:

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Re: Consumer SSD Battle

Post by dhanson865 » Mon Nov 08, 2010 3:41 pm

Zap wrote:
porkchop wrote:i'm surprised the kingston fared so well since it uses a 'inferior' jmicron controller, i guess they've sorted out the old problems.
You don't think Jmicron could improve on their products in newer generations?
Kind of hard for him to notice since they haven't advertised the JMicron name in relation to a newer generation SSD controller since 2009.

They print the name Toshiba on a big chunk of the chips to avoid the stigma of the JMicron name. JMF612 and JMF618 are basically the same controller but the 618 will have the Toshiba name on it.

Apparently JMicron quietly introduced the JMF616 this summer and it's found it's way into some drives but this isn't one of them.

http://www.pcper.com/article.php?aid=872&type=expert shows the drive porkchop is surprised about still uses the old JMF618 from 2009.

As far as I know no JMF616 drive has been benchmarked at anandtech.com yet but looking at the specs I don't expect it to fare much better than the 612/618.

The catch is they are selling this as a value drive while using the full 8 channels of the controller. So if you compare it vs the 40GB Intel product with only 5 channels instead of the 10 channels in the 80GB or 160GB model you see it doing OK. Make the comparison vs the faster Intel drives or say the C300 and the Kingston falls short.

You'll also see a boost for the Sandforce drives above 60GB but it's not as pronounced as the difference between the Intel 40GB and 80GB.

This is why benchmark charts need a control at each end. A modern HD like a 600GB raptor or one of the 500GB+ per platter drives from seagate/WD/samsung for one end and a 160GB Intel or C300 256GB for the other end so the value drives get plopped right in the middle and show you some context vs the overall market.

Of course if you do this you have to drop the HD out of the random write chart and possibly others as it won't show without ruining the scale but it's still worth putting it on real world benchmarks like boot times. None of the charts in this review have that problem.

SPCR did well with their choices of HDs but those charts are missing faster SSDs which would change the perceived value of the Kingston SSD in those charts.

And given the price range of the SSDs in the review they could easily justify discussing other drives

WD silicon edge blue 64GB is above $200 which pits it squarely vs the Intel 80GB drive in cost.

The Kingston 64GB and OCZ 60GB drives are both right around the cost of the C300 64GB. All three drives are in that $125-$150 segment.

porkchop
Posts: 496
Joined: Thu May 25, 2006 1:19 am
Location: Australia

Re: Consumer SSD Battle

Post by porkchop » Mon Nov 08, 2010 5:50 pm

zap wrote:You don't think Jmicron could improve on their products in newer generations?
i do and they clearly have, i wrote myself that they've sorted out some old problems.
the cheaper kingston beat the vertex 2 in some benchmarks which i found surprising, i honestly expected the vertex 2 to beat it soundly in every test.

colm
Posts: 409
Joined: Tue Jan 31, 2006 8:22 am
Location: maine

Re: Consumer SSD Battle

Post by colm » Sun Nov 21, 2010 5:33 pm

just got to reading article. great stuff.

I am slightly frowning at load times on regular hdd.. altho, what I compare to is my 80gb, (wd800jb - pata) @ 31000 hours right now.

I am way way under a minute loading win7. It has been my favorite late model pata since I began in the 90s.

the stuff about "TRIM" has got me waiting for smarter SSD. Like my pc upgrade ...
I was searching until this read. Saved me some time..

boe
Posts: 331
Joined: Sun Feb 01, 2004 9:47 am

I would recommend against the OCZ

Post by boe » Fri Nov 26, 2010 6:01 pm

I had the ocz revo card for a couple of weeks and it seemed fast - that was until in the second week it kept bluescreening and by the end of the second week it died completely. Any system I put it in wouldn't even get to POST on boot up so the motherboard couldn't even count off memory - nothing.

I don't think all SSD are defective. I have a RAID SSD on my Sony laptop and have used it nearly a year without a single hiccup.

My guess is that X2 version of the REVO is because they are trying to fix the issues they've had with the original models although I don't know if they actually are any more reliable.

I probably wouldn't be so upset if the system didn't require me many many hours of system rebuild time to install the OS, apps, profiles, itunes and other syncs, e-mail downloads etc. I would be much more forgiving if Newegg would let you return the $1250 POS but they won't. I contacted tech support at OCZ via phone and they said you need to open a ticket online - I did. Then I called them and they said they couldn't help me if I had opened a ticket as only the person on the ticket could respond. I had called because the person assigned the ticket had disregarded all the tech details I had given and asked for them again and disregarded all my questions. Crap support might explain why Newegg won't let me return this item - I'm guessing others have had bad experiences and they are cutting their losses.
Last edited by boe on Sat Dec 11, 2010 10:03 am, edited 1 time in total.

Pierre
Posts: 156
Joined: Sun Jan 17, 2010 3:55 am
Location: Greece

Re: Consumer SSD Battle

Post by Pierre » Sat Dec 11, 2010 7:08 am

Would these new SSDs function properly in a motherboard that does not support AHCI?

victorhortalives
*Lifetime Patron*
Posts: 207
Joined: Wed Aug 08, 2007 1:50 am
Location: Xhystos

Re: I would recommend against the OCZ

Post by victorhortalives » Sat Dec 11, 2010 1:41 pm

boe wrote:I had the ocz revo card for a couple of weeks and it seemed fast - that was until in the second week it kept bluescreening and by the end of the second week it died completely. Any system I put it in wouldn't even get to POST on boot up so the motherboard couldn't even count off memory - nothing.

I'm guessing others have had bad experiences and they are cutting their losses.
Thanks for the warning - I was going to spend some spare funds on one but I'll buy a new Intel instead.

boe
Posts: 331
Joined: Sun Feb 01, 2004 9:47 am

Re: I would recommend against the OCZ

Post by boe » Sat Dec 11, 2010 1:55 pm

victorhortalives wrote:
boe wrote:I had the ocz revo card for a couple of weeks and it seemed fast - that was until in the second week it kept bluescreening and by the end of the second week it died completely. Any system I put it in wouldn't even get to POST on boot up so the motherboard couldn't even count off memory - nothing.

I'm guessing others have had bad experiences and they are cutting their losses.
Thanks for the warning - I was going to spend some spare funds on one but I'll buy a new Intel instead.
No problem. I sent the unit back to OCZ for a replacement a while ago - they still haven't sent anything back. The positive thing I took away from this is I'm going to image my hard drive far more often so I don't ever have to go through this much set up, tweaking and profile adjustment ever again.

ist.martin
Posts: 220
Joined: Fri Jul 18, 2003 11:59 am
Location: Vancouver, B.C.

Re: Consumer SSD Battle

Post by ist.martin » Thu Dec 16, 2010 9:11 am

Does anyone know anything about the Kingston SSDNow V100 series that are currently selling?

NCIX has the 64GB model for $99, which seems to be a great price. Is it a quality, reliable drive? Or is it worth spending another $30-$40 for another brand?

Thanks.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Re: Consumer SSD Battle

Post by dhanson865 » Thu Dec 16, 2010 1:52 pm

viewtopic.php?f=7&t=58422&start=90 has my details on that.

If you don't want to read the thread I'd say avoid the V100 drives and get either Intel Gen 2 (80GB/120GB/160GB) or Crucial/Micron C300 (64GB/128GB/256GB).

Anything other than Intel or C300 is either less reliable and/or significantly slower.

Oh, and you might want to look at the posts in the pricing thread as well. See viewtopic.php?f=7&t=58811&start=90 for those.

HFat
Posts: 1753
Joined: Thu Jul 03, 2008 4:27 am
Location: Switzerland

Re: Consumer SSD Battle

Post by HFat » Thu Dec 16, 2010 9:12 pm

I agree you listed the drives which look the best right now but you're overstating your case. Do you have any source of information you didn't write about so far?
All drives have their issues. Other drives are better than Intel's or Crucial's in some respects. So there are no simple answers as to which drive is best.

It's not straightforward to determine which drive is faster. Depending on what test you look, you'd say the V+100 is faster or slower than an Intel for instance.
And do we know that the V+100 is less reliable than Intels? Do we know that the latest Samsungs are less reliable for that matter? I would expect the V+100 at least to be less reliable but that's not a given.
What we know is that that the Intel drives are well-tested. They're not the fastest and they're more expensive than many alternatives but they have no surprises in store (for those who've done their homework anyway).

It's not clear that there's something that beats the V100 in its price range. I'd be more comfortable recommending a 40G Intel even though it's quite slow but that's not an option if you need 60G. If you can get an 80G Intel for $30-$40 more than a V100 it would be worth the extra money (the price difference is bigger here). A Crucial might be a good choice but I wouldn't recommend one to someone who gave no information about what they want to do with the drive. The Samsungs are in the same price range but I don't know enough about them to recommend them. Does anyone? The V+100 might also be a better choice than the V100 as well depending on the price difference. But I think it's too early to recommend them.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Re: Consumer SSD Battle

Post by dhanson865 » Fri Dec 17, 2010 8:48 am

HFat wrote:It's not straightforward to determine which drive is faster. Depending on what test you look, you'd say the V+100 is faster or slower than an Intel for instance.
If all you are going to do is use it in an external enclosure as a glorified thumb drive (USB 2 or USB 3) to move data around it's plenty fast enough for that.

But the result in the image below shows that it is woefully too slow to be used as a boot drive without noticeable slowdowns. I can't recommend any drive in good faith that makes less than 10MB/s on the random write test.

Image

And do we know that the V+100 is less reliable than Intels? Do we know that the latest Samsungs are less reliable for that matter? I would expect the V+100 at least to be less reliable but that's not a given.
The V100 and V100+ are likely reliable enough to take the chance on but they are definitely too slow to recommend. Samsung SSDs have always been reliable so again I would assume the Samsung 470 to be reliable. I haven't seen benchmarks on it nor have I seen it cheap enough to recommend over Intel/C300 choices.
It's not clear that there's something that beats the V100 in its price range.
I find it odd you consider the Intel drives expensive since they are cheaper than the Samsung 470 and C300 drives at some capacities the last time I checked. I suppose I should do another price check now.

Crucial RealSSD C300 64GB $126.72
Samsung 470 64GB $129.99
Kingston SVP100S2/64G 64GB $160.59
Intel SSDSA2MH080G2K5 80GB $173.33

Intel SSDSA2MH120G2K5 120GB $227.99
Crucial RealSSD C300 128GB $248.99
Samsung 470 128GB $268.13
Kingston SVP100S2/128G $278.04

So if you want a 64 or 80GB drive the C300 and Samsung 470 are cheaper than the Kingston and if you are going to consider a drive as expensive as the Kingston why would you not pay $12 more for the Intel?

And if you want a 120 or 128GB drive the Intel is cheapest and the C300 is in the middle with the Kingston the most expensive. Samsung isn't cheap in this group like it is in the lower capacity.

I assume you are looking at prices in Euros? Are you factoring in rebates (I don't give rebates face value)? Are you looking at older Kingston model numbers and confusing them with the current V+100 series?

I'd like to see your prices shipped to your house including VAT and such for the drives I just checked.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Re: Consumer SSD Battle

Post by dhanson865 » Fri Dec 17, 2010 10:10 am

On the subject of part number confusion

SNV425-S2/64GB is the 64GB version of "SSDNow V Series Drive"
SNVP325-S2/64GB is the 64GB version of "SSDNow V+ Series Drive"
SV100S2/64GZ is the 64GB version of "SSDNow V100 Drive"
SVP100S2/64G is the 64GB version of "SSDNow V+100 Drive"

that's ignoring

SSDNow E-SERIES
SSDNow S100
SSDNow V+100E
SSDNow V+180
SSDNow M-SERIES

as most of those are more expensive and are less likely to be confused in low price discussions.

Of the 4 cheaper confusing 64GB models we have stats like

Code: Select all

           V      V+       V100   V+100
idle power 0.7W   0.065W   1.0W   0.05W
load power 5.2W   3.5W     6.4W   3.6W
but all 4 behave poorly when it comes to random writes. At least the V+100 would be of interest to laptop users for the low idle power draw but I'd feel sorry for the laptop user that is confused or misled by a retailer selling them the V100 that draws 20x more power at idle and 75% more power at load just because it's a few dollars/euros cheaper and the V100 is likely to be discontinued very soon.

HFat
Posts: 1753
Joined: Thu Jul 03, 2008 4:27 am
Location: Switzerland

Re: Consumer SSD Battle

Post by HFat » Fri Dec 17, 2010 12:32 pm

Intel SSDs are expensive compared to the cheapos as well as compared to the better drives when you want capacities like 64G where Intel has no offering. The price per gig of the 80G is good but, if you don't have a use for more than 60G, you're looking at a serious Intel premium.
I'm looking at prices in CHF. All drives are more expensive than in the US but there are some differences. While the relative price of the V+100 is better here (I thought it'd be cheaper for you), it's still among the most expensive drives.

The V+100 is potentially (it's too early to tell) a good drive for the average consumer who's running XP or OSX because its performance is consistenly good to outstanding. The price and inconvenience of switching to Win7 outweighs other considerations.
You're right to point out that some SSD packagers have misleading product names but your account of SSD performance is even more misleading. You fetched that graph from Anandtech so look at their benchmarks!
The random write performance of the V+100 is indeed horrible for its price but it's still fast compared to hard drives so it's not necessarily a deal-breaker depending on what the drive is going to be used for.
You can't sum up SSD performance in a single number because the controllers behave so differently. No SSD does everything best. While random writes are important for some uses, aside from the sometimes serious fragmentation issues which such short tests do not reveal, I think you should be looking mostly at random reads and sequential writes.
Random reads are probably what's going to impact regular usage the most (provided the system has enough RAM) for the average consumer (and for many other users as well). This is why Crucial drives deserve to be considered if you have TRIM.
When your random access performance is pretty much instantaneous (recall the typical user only access small amounts of data at a time randomly), what you're most likely to notice I think are differences between copying speeds because copying large files will easily push any drive to its limits. Sequential writes are the copying bottleneck which is why low-capacity SSDs can copy slower than hard drives. Then benchmarks which only look at the high-capacity models do not reveal this. The other types of SSD performance do not degrade so badly when comparing low-capacity models with the ones more commonly sent to reviewers. The 128G V+100 has an oustanding sequential write performance so there's hope for the low-capacity versions.
All in all, I think Intel's drives are the best choice. But the V+100 would be faster for many real-world applications because of its outstanding sequential write performance and its aggressive defragmentation. That's is why it might make sense not to pay a little more for an Intel with more capacity.

Blappo
Posts: 85
Joined: Thu May 27, 2004 10:21 am
Location: Waterloo, ON

Re: Consumer SSD Battle

Post by Blappo » Fri Dec 17, 2010 12:44 pm

@dhanson865
The 64GB version of SSDNow V100 Drive (SV100S2/64GZ) is constantly on sale @ NCIX.com (a Canadian online store) http://ncix.com/products/index.php?sku= ... omoid=1261 for $99 CDN. I'm also curious if it is a good value at this price. Where have you seen a review of the 64GB V100 that measures the performance and power usage?

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Re: Consumer SSD Battle

Post by dhanson865 » Fri Dec 17, 2010 4:00 pm

HFat wrote:Intel SSDs are expensive compared to the cheapos as well as compared to the better drives when you want capacities like 64G where Intel has no offering. The price per gig of the 80G is good but, if you don't have a use for more than 60G, you're looking at a serious Intel premium.
I'm looking at prices in CHF. All drives are more expensive than in the US but there are some differences. While the relative price of the V+100 is better here (I thought it'd be cheaper for you), it's still among the most expensive drives.
For those who like me have to google to figure out what CHF is http://en.wikipedia.org/wiki/Swiss_franc is probably more useful than Congestive Heart Failure. :)

OK, let me reexamine my dislike of low random write speed. In 2009 SSD anthology part 1 we have
I’d argue that if you’re a desktop user and you’re using an SSD as a boot/application drive, what will matter most is latency. After you’ve got your machine setup the way you want it, the majority of accesses are going to be sequential reads and random reads/writes of very small file sizes. Things like updating file tables, scanning individual files for viruses, writing your web browser cache. What influences these tasks is latency, not bandwidth.

...

Random write performance is merely one corner of the performance world. A drive needs good sequential read, sequential write, random read and random write performance. The fatal mistake is that most vendors ignore random write performance and simply try to post the best sequential read/write speeds; doing so simply produces a drive that's undesirable.
and in that article we see random write speeds of

X-25M Gen 1 23.1 MB/s
OCZ Vertex 2.41 MB/s
OCZ Summit 0.77 MB/s (presumably this paused/stuttered as well)
JMicron JMF602B 0.02 MB/s (obviously this causes several seconds pauses aka stuttering)

keep in mind even with several second pauses and stuttering the JMF602B out performed the Raptor and Momentus in the PCMark Vantage test.

Later in 2009 (SSD Relapse) we see new data

X-25M Gen 2 36.1 MB/s
OCZ Summit New 4.4 MB/s
OCZ Summit Used 1.2 MB/s

and we get a section of the article titled "What's Wrong with Samsung?" where he says "Holycrapwtfbbq? Terrible.

Now to be fair to Samsung, this isn’t JMicron-terrible performance. It’s just not worth the money performance."


So again we know that not only does 0.0x MB/s cause problems but even 4.x MB/s and 1.x MB/s are enough to make a drive not worth recommending.

Step forward a year to 2010 and all the Kingston SSDnow V drives are still in that Holycrapwtfbbq? Terrible speed range for random writes. But instead Anand puts some tact on it and says
Random write performance has always been the weak spot of Toshiba’s controllers, this latest combination of controller and firmware is no different. Compared to all other SSDs, the Toshiba based SSDNow V+ 100 doesn’t look very good. It’s even slower than the old Indilinx based Corsair Nova.
That quote is just over a month old. SSDnow drives just don't hang with the current higher speed drives. So what changed from OMG to doesn't look very good? Well back in the day Samsung 120 GB drive was $499 at release in March 2009 and now the Kingston drive is 3/5ths of that cost. Is the price dropping by 40% over 18 months enough for you to ignore poor performance? To me it isn't, there are still drives that perform way better for the money.

The quote is "Compared to all other SSDs, the Toshiba based SSDNow V+ 100 doesn’t look very good" not Compared to all other SSDs, the Toshiba based SSDNow V+ 100 doesn’t look very good but it sure is cheap.

Don't get me wrong if you gave me a free Kingston SSD I'd gladly put it to use. If it got to be half the cost per GB of the Intel drive I'd be a fool not to buy one. The hard part is an individual decision on how much cheaper it would have to be before you overlook the poor performance if you are using it as a OS/application drive and not just long term storage.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Re: Consumer SSD Battle

Post by dhanson865 » Fri Dec 17, 2010 5:01 pm

Blappo wrote:@dhanson865
The 64GB version of SSDNow V100 Drive (SV100S2/64GZ) is constantly on sale @ NCIX.com (a Canadian online store) http://ncix.com/products/index.php?sku= ... omoid=1261 for $99 CDN. I'm also curious if it is a good value at this price. Where have you seen a review of the 64GB V100 that measures the performance and power usage?
Not for a laptop, battery time would suffer, now for a desktop lets restrict our search to NCIX for a bit


Prices from NCIX.com This comparison would be more accurate if you gave me a postal code to use. I'm going to try "N2J 1B1" which seems to be around your area.

Crucial RealSSD C300 64GB $137.58 CAD free pickup at Markham Warehouse
Samsung 470 64GB not available at NCIX.com
Kingston SVP100S2/64G 64GB $179.90 CAD free pickup at Markham Warehouse
Intel SSDSA2MH080G2K5 80GB $184.99 CAD free pickup at Markham Warehouse and a you can try for the $30 rebate

Given those prices I'd be looking at the C300 but you want to consider the Kingston SSDNow V100 Series 64GB which is $99.99 with free pickup at the Markham Warehouse. Well it's 37% cheaper but the random write performance is much slower. How much slower?

Well it's based on the JMF618 controller which performs the same as the JMF612 controller (the JMF618's unique feature is its support for Toshiba NAND. If you use Samsung NAND that makes the controller a JMF612.). The closest you'll get to that on anandtech.com is the Western Digital SiliconEdge Blue 256GB review http://www.anandtech.com/show/2954/5 but performance drops from the 256GB version down the the 64GB version.

Anand 256 GB WD SSD 12.5 MB/s random write official spec for writes "Up to 170 MB/s"
Andnd 64 GB WD SSD not tested for random write official spec for writes "Up to 170 MB/s"

I expect that is either under-claiming for the 256GB or over-claiming for the 64GB.

Kingston V100 256GB not tested at Anandtech official spec for writes "230MB/sec"
Kingston V100 64GB not tested at Anandtech official spec for writes "145MB/sec"

OK, so it's not going to drop by a ton maybe 40% slower than the 256GB version on writes. So it'll be somewhere in the 5 to 10 MB/s range (40% off 12.5 would be 7.5 MB/s but I can't say how accurate that scaling is).

Assuming it is 7.5MB/s you are talking 37% cheaper than a C300 but 90% slower on random writes, PC Mark Vantage score would be in the low 14K range vs almost 16K for the 64GB C300 so it's about 15% slower on those sorts of benchmarks?

Totally up to you if that is a trade off you are willing to make. It's gotta be better than the Jmicron drives from 2008 that made bar81 so mad but I can't promise you that you won't have buyers remorse come spring. If it would make you mad to see that drive at $70 a month after you bought it don't do it. If you aren't the type to get mad when a product goes on clearance after you buy it at full price go for it. Me I'd rather get a long term bargain that I would be sure to be happy with.

Any review you can find that mentions a 64GB JMF612 or 64GB JMF618 would give you a reasonable expectation of performance for the $99 drive you are looking at.

HFat
Posts: 1753
Joined: Thu Jul 03, 2008 4:27 am
Location: Switzerland

Re: Consumer SSD Battle

Post by HFat » Fri Dec 17, 2010 9:00 pm

Now you're just giving bad advice. Instead of quote-mining Anand's reviews, read them! The numbers you're looking at have do not reflect real-world performance.
The C300 is not recommended if you don't have TRIM. So don't recommend it without qualification. Anand specifically says the V+100 would be better for most people in that case. Is the V100 similar in that respect? I don't know. Do you?
The C300 has a highish power consumption as well. If you're going to recommend against the V100 on that ground, you shouldn't turn around and recommend the C300 without mentioning it. Note that you can't compare power consumptions at load without taking into account the time needed to complete a task (as with CPUs).

Stuttering is a latency issue, not so much a throughput issue (yes, there is a relationship). I don't have latency figures for the V100 or the V+100 unfortunately. Do you?
4.4 MB/s of random writes is not that bad for a consumer drive, unlike 1.2 MB/s. Clearly the V+100 has a better controller when it comes to fragmentation. The V+100 seems to have been tweaked to have a low random write performance. It's not an accident like with the early SSDs when Intel was the only manufacturer who knew what they were doing. The drive seems to work hard to defragment when you write small chunks to make sure you'll get top speeds on large writes later. That would be bad for random write performance but they apparently figure their random write performance is good enough for the target market.
The bottom line is that drives with relatively low random write speeds such as the Kingson tested by SPCR perform well in the real world as long as the controller is able to maintain performance over time. Read the review this thread is about: SPCR recommends against the 40G Intel because of its poor sequential write speed. It mattered more than random write speed for the kind of stuff the review looked at (within reason: obviously you don't want a random write speed that's much slower than a hard drive). Drives with relatively low random write speeds also do well on Anand's consumer-oriented benchmarks.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Re: Consumer SSD Battle

Post by dhanson865 » Sun Dec 19, 2010 7:35 am

HFat wrote:The C300 has a highish power consumption as well.
I'm afraid that isn't true in comparison to the Kingston drives.

Code: Select all

power draw C300 64GB vs  V+100 64GB vs V100 64GB
Idle Power   0.092W      0.05W         1.0W
Active Read  1.7W
Active Write 2.4W        3.4W          6.4W
Unfortunately Kingston doesn't spec theirs per capacity nor do they spec read vs write power draw so we have to wonder how the power draw scales.

Still it is obvious that C300 in that capacity has lower active power draw than either of the Kingston's we mentioned and the V100 still uses 10x as much idle power draw. Yes the C300 uses more idle power than the V+100 but its such a small amount it's likely offset by the lower active power draw.

I'm sure that perception of the C300 being higher power draw is fueled by review sites focusing on higher capacity drives. To show the difference lets look at 128GB and 256GB.

Code: Select all

power draw   C300 64GB vs C300 128GB C300 256GB
Idle Power   0.092W       0.092W     0.094W
Active Read  1.7W         1.7W       2.1W
Active Write 2.4W         3.1W       4.3W
and that is assuming you get the 2.5" version, the 1.8" version draws even less power at all three capacities. If we assume the power spec by Kingston is accurate the V+100 might be lower power draw at 256GB than the C300 but at the lower two capacities I'd say it isn't.

Now you did mention C300 and non TRIM OSes and you are right that it is a concern to pay attention to. The quote below is from a thread on storagereview.com where I was discussing this when the Samsung 470 was released.

No one has corrected me on this table so I'd be glad if you or anyone else could double check for mistakes but here it is

Code: Select all

Controller       Firmware  GC/Trim/wiper Tool  Secure Erase Tool
Intel G1           Yes       No                  No
Jmicron            Yes       Yes*                Yes  
Indilinx barefoot  Yes       Yes                 Yes (assuming OCZ tool works on non OCZ drives?)
Intel G2           Yes       Yes                 Yes
Sandforce 1xxx     Yes       No                  Yes
C300               Yes       No                  No* (can use sanitary erase tool from OCZ?)
Samsung 470        Yes       Yes (beta)          No
I know there is a wiper.exe around and google searches make it seem that it is used for Indilinx and Jmicron (or are there two executables sharing a name)? I also remember that older Jmicron drives could be pre TRIM and G1 Intel drives are pre TRIM.

When I try to boil it down to Yes/No answers it looks nicer than it really is for some of these. The Intel Toolbox really feels so much friendlier/more useful than the bare bones tools for a jmicron or indilinx drives. I haven't used the Samsung SSD Magician yet but from the screenshots I'm expecting it to be impressive as well.

HFat
Posts: 1753
Joined: Thu Jul 03, 2008 4:27 am
Location: Switzerland

Re: Consumer SSD Battle

Post by HFat » Sun Dec 19, 2010 8:50 am

Anand took the trouble of measuring power consumption and got an idle power consumption an order of magnitude above the spec you quote for the C300. Going by his numbers, the C300 is going to consume more than the V+100 unless you do a lot of random writing (the C300 is less efficient at sequential writing actually) and potentially a lot more if it's idle most of the time. Going by the specs, the V100 might well consume more than the C300 in many cases but I wouldn't be surprised if that 1W idle power consumption was wrong.
If you want low power consumption, the V+100 is probably the best choice for a mostly idle drive while Sandforce drives would have the lowest power consumption at load.

Your table is somewhat wrong or outdated. I assume all reputable drives have some form of GC (which might or might not rely on TRIM). I don't think there's much point in getting it right anyway because what matters is how the drives perform in the real world and Anand's test have revealed rather complex behaviours under load in some cases. Determining whether a drive supports TRIM doesn't tell you how it's going to perform.
So far as I can recall he said G1 Intels don't use TRIM and perform OK while G2s don't perform as well without TRIM. The C300 performs fairly poorly without TRIM. The V+100's performance seems rock solid without TRIM. The old Samsungs stuff used by OCZ didn't perform well but I don't think he looked at the Samsung-branded drives in detail (that's a shame). I'm not much interested in Indilinx and Sandforce drives so you'll have to check his reviews for yourself. I don't know any other source for solid information about how the performance of a drive degrades under use.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Re: Consumer SSD Battle

Post by dhanson865 » Sun Dec 19, 2010 11:33 am

HFat wrote: Your table is somewhat wrong or outdated. I assume all reputable drives have some form of GC (which might or might not rely on TRIM).
Maybe English isn't your native language or maybe I'm just not being clear enough labeling that column. That column isn't about native GC or native TRIM support.

The purpose of that column is to show whether or not a executable program exists to allow for example Windows XP users to do something to restore performance on a drive that has been used in an OS like Windows XP that does not support TRIM.

ALL SSDs have some form of GC, All modern SSDs have TRIM support. The question is what does each controller manufacturer and SSD manufacturer do to provide support for users that use an OS that does not support TRIM.

Sandforce gets a No in that column because I haven't seen a download on the sandforce/OCZ/corsair etcetera web sites for windows XP users to manually trigger GC or TRIM functions. Sandforce drives fully support TRIM when the OS supports it.

Indilinx provides a tool called wiper.exe that does that. Intel provides a tool that does it. Samsung provides a tool for the new 470 drive that does it.

Does that make more sense now?

HFat
Posts: 1753
Joined: Thu Jul 03, 2008 4:27 am
Location: Switzerland

Re: Consumer SSD Battle

Post by HFat » Sun Dec 19, 2010 1:45 pm

Yeah, it makes a bit more sense. Except that "wipe" usually means the same thing as "secure erase" but you have them in different columns. And that there are of course wipers/erasers that work with all drives. And TRIM and GC are not the same thing. A tool that replaces the OS support for TRIM shouldn't overwrite what's on the drive. Assuming that's what your second column means, I still don't get what the third one means: a way to overwrite all the flash on the device (as opposed to what you can overwrite by filling the free space on the drive with a huge file or something)?
You talk about triggering the GC. But no special tool should be needed for that. If it needs a trigger (idle GC shouldn't), presumably it can be triggered simply by overwriting free space.
In any case the table doesn't tell you about performance. Maybe the V+100 supports TRIM as you say but it doesn't look like it relies on it for performance. And TRIM support doesn't guarantee performance (especially not delayed TRIM which is the best these tools for XP might give you). It's just that the C300 relies on it to get good performance. You say SandForce gets a "no" for isntance but the tool might not be needed. If you don't use encryption (which would slow down your SandForce drive anyway), compression should obviate the need for TRIM.
The main point of TRIM seems to be to increase the useful capacity and longevity of drives by giving the wear-leveling more flash to work with.
Overwriting free space (no special tool required) might help performance and/or wear-leveling if a TRIM tool isn't available. There are reports claiming it helps but testing would be required to know how much...

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Re: Consumer SSD Battle

Post by dhanson865 » Sun Dec 19, 2010 2:51 pm

HFat wrote:Yeah, it makes a bit more sense. Except that "wipe" usually means the same thing as "secure erase" but you have them in different columns. And that there are of course wipers/erasers that work with all drives. And TRIM and GC are not the same thing. A tool that replaces the OS support for TRIM shouldn't overwrite what's on the drive. Assuming that's what your second column means, I still don't get what the third one means: a way to overwrite all the flash on the device (as opposed to what you can overwrite by filling the free space on the drive with a huge file or something)?
You talk about triggering the GC. But no special tool should be needed for that. If it needs a trigger (idle GC shouldn't), presumably it can be triggered simply by overwriting free space.

...

Overwriting free space (no special tool required) might help performance and/or wear-leveling if a TRIM tool isn't available. There are reports claiming it helps but testing would be required to know how much...
uggh, no. Overwriting free space on a SSD can only be undone by a Secure erase and is a sure way to slow down a SSD.

the fact that they named it wiper.exe may confuse you but it is not working at the level you think of when you say "format" a partition in windows or "delete" a file in windows.

There are not programs like what wiper.exe does for all drives. Blame Indilinx for using the exe name that overlaps with tons of other utilities from the past. http://www.ocztechnologyforum.com/wiki/ ... l_speed.3F describes the tool or see http://www.anandtech.com/show/2829/13 for mention of what wiper.exe does.

Wiper.exe, the Intel Toolbox, and the Samsung SSD Magician tool allow for manually initiating TRIM on a drive. The process is different on Intel drives and different again on Samsung 470 drives.

Secure Erase resets the drive (other than SMART data) back to factory fresh removing all partitions and all traces of data. There is no way short of a proper secure erase to get a drive to that state. Old school hard drive programs that write all 1s or all 0s will not have the same effect. Secure erase doesn't just affect the user data it also affects the SSD controller data and all spare area in the drive. No command other than the ATA secure erase command can touch the non addressable portions of the SSD.

According to http://www.windowsitpro.com/article/fil ... tate-.aspx you can use HDDErase on any SSD if your SSD is in legacy mode in the BIOS. Assuming that is true then the No in the third column just means that the support/download/FAQs for a particular drive don't tell you how to do a secure erase on that drive. As important as knowing how to do a secure erase is I consider that significant.

Intel's Toolbox v2 provides a secure erase function in the GUI so you don't have to do it from a dos boot which makes it much easier to do if you are working on SSDs frequently or are new enough that a little hand holding is needed.

Please understand just because this one dos tool HDDErase will do a secure erase it's not the same as zeroing out a drive. https://ata.wiki.kernel.org/index.php/ATA_Secure_Erase shows doing this with HDparm. Traditional rotating disks will implement a secure erase but it is handled differently with that sort of media and takes much longer.

I'm not even going to begin discussing GC in any detail. I've already spent too much time on this thread.

Post Reply