NAND Flash Faces Off - Synchronous vs. Asynchronous @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,534
NAND Flash Faces Off - Synchronous vs. Asynchronous - News flash! All flash NAND is not created equal! Sure, you know about multi-level and single-level NAND when it comes to speed, but what about synchronous and asynchronous NAND inside your shiny new SSD? We have answers and tell you where your money is best spent for real data speed.
 
Thank you very much for this article. It was very informative, articles like this are why I love this site.
 
Thank you very much for this article. It was very informative, articles like this are why I love this site.

Yes indeed. I haven't made the move to ssd yet but, I will definitely keep this info in mind when I do.
 
Does this have any relevance to those SSDs not running on a sandforce controller?
 
Thank you for the very informative article. It really goes to show that real-world tests can bring out the good (and the bad) in the actual performance levels of these drives compared to synthetic benchmarks. Also, Kyle posted this thread 2 days ago? lol.
 
Just want to echo others in saying this is a very informative article and well appreciated. I learned something today that will help me with an upcoming system upgrade purchase.

From what I learned today, OCZ has sunk even lower in my opinion with how they handled the Vertex 2 debacle. They have spread a lot of FUD in their "moderated" forum regarding this issue. I am looking forward to other manufactures of SSD to catch up to OCZ's performance so that I can give them the money instead to OCZ.
 
Just want to echo others in saying this is a very informative article and well appreciated. I learned something today that will help me with an upcoming system upgrade purchase.

From what I learned today, OCZ has sunk even lower in my opinion with how they handled the Vertex 2 debacle. They have spread a lot of FUD in their "moderated" forum regarding this issue. I am looking forward to other manufactures of SSD to catch up to OCZ's performance so that I can give them the money instead to OCZ.

you may want to just look at the patriot wildfire in that case. it uses the same type of nand as the max iops edition.
 
Nice article, but screw some of those graphs. There is absolutely no consistency in them. I appreciate articles that shed light onto subjects, however I despise disingenuous graphs that pretend to start at 0%, but really start at 80%, to make a point that is capable of making itself. It makes it look as though your argument is cannot stand on its own and that you have to resort to visual trickery to convince the reader. Patronizing bullshit. Newspeak is detestable whether its coming from politicians, from grandma, from vendor benchmark comparison PR, or from tech site articles.

At the very least you could have labeled the starting points of the graphs.
 
thank you for the great article! pay no attention to the goofball above
 
It was an interesting read, but what I came away from it with was the fact that when deciding between sync and async, you get proportionally better performance from a proportional increase in cost. So, this can't be about value for the dollar. It's also not about any need is only satisfied by specific performance scores.

It seems to be about what technology has the performance crown, by any margin. Some of the bar charts are give a misleading impression of the actual difference in scores because they don't measure from zero - this magnifies the difference and makes a basically hyperbolic argument.
 
I didn't intend any graph trickery but will look into the starting point for next time. The numbers are accurate so look at those and not the distance from the left I guess.
 
Nice article, but screw some of those graphs.

Thanks for your kind and insightful posting. Roger that, all graphs from 0. We have done it this way for over a decade, we are not going to start something different now.

The issue with the graph has been answered and will be corrected going forward. We will not spend more time discussing it. Please let's stay on topic to the article, rather than format of the graphs.
 
Interesting article, but for those of us who are using X58 mobo's will we really notice a difference? The point made in the test setup paragraph just says they don't hit full speeds, yet there is no mention of a speed difference when using the Marvell controller or SATA II with a synchronous vs. asynchronous SSD.
 
Interesting article, but for those of us who are using X58 mobo's will we really notice a difference? The point made in the test setup paragraph just says they don't hit full speeds, yet there is no mention of a speed difference when using the Marvell controller or SATA II with a synchronous vs. asynchronous SSD.

well technically this does show you that it will improve performance on sata 2 because the 4k speeds are higher on sync vs async. you'll never saturate your controller with 4k speeds, so it does actually make a difference, no matter what controller you're using.

personally for me, I've never touch an async drive, but I'd also never touch a 120gb drive. anything above 60gb is too much for me because I raid them. I ended up going with 60gb force gts on blind faith and they seem to do really well, being the only sync 60gb sandforce drive I could find.

I wish corsair/patriot would have released a 34nm or 32nm toggle on the new controller at 60-64gb size, but that's okay, force gt's will do for now.
 
Finally what I have been looking for. I have not jumped in to the SSD world yet because of the "lack" of data. After reading this, I went to the Corsair website, only one out of the four Force series indicate that they are Synchronous. The other 3 do not. Maybe as consumers we should assume that the other 3 are Asynchronous, but after dealing with the Federal government over the years, IT HAS TO BE WRITTEN or else it can be whatever it wants.

Great article and good data, now ALL the vendors need to identify this on ALL websites
 
Any word on crucial M4's, are synchronous or asynchronous?
 
How does this affect video encoding if at all? Great article though. Wish it had came out a few days sooner, just made a purchase of 2 60gig ssd drives.
 
For those of you that called us insincere in our attempts to share information with our readers for using "disingenuous graphs," I have gone back through the graphs this morning and checked each and every one. We had one graph that was not scaled from Zero. The HD Tune Pro Read graph is the only graph that was not scaled to our normal standards. The rest are scaled from Zero, as we usually do. Going forward we will put the Zero on the axis in order to make sure you are not led astray. Calling us "disingenuous" is not appreciated and quite frankly it offends me personally. But if you think that HardOCP is "disingenuous" then I would suggest moving your reading and forum usage elsewhere.
 
Andy idea why there seem to be so many problems with the SF-2281 SSDs being reported? Especially on the OCZ Forums?

Specifics and links always help, otherwise this is a loaded question.
 
The m4 isn't affected, an m4 is an m4.

When i get home I'll post the Blackmagic video tests. This is a new benchmark that I'm still validating.
 
Chris, thanks for the great read. I think it's important to keep the results in perspective, though, as you've noted in a previous article. My personal experience with a handful drives with different performance characteristics is that there is very little subjective difference between them. I've had an Agility 3, Intel 510, Force GT, and Performance 3 all in the same machine, and they all feel exactly the same when booting, launching anything, shutting down, etc. The only difference between them has been the consistency of performance, which this article seems to hint at. The A3 and (to a lesser degree) the Force GT were pretty variable in how "snappy" they felt from day to day, while the 510 and P3 seem rock steady.

Just my .02.
 
Wow, I thought this was a site for people looking for performance. If Nvidia/AMD released a new graphics card that was 1% faster everyone would be freaking out that they wanted it. But I guess that doesn't apply to SSDs. Anyway, great article as usual guys.
 
Wow, I thought this was a site for people looking for performance. If Nvidia/AMD released a new graphics card that was 1% faster everyone would be freaking out that they wanted it. But I guess that doesn't apply to SSDs. Anyway, great article as usual guys.


Please explain the object of your post, because I do not understand what exactly your point is.
 
I have an original Intel 80GB SSD and have been looking at the new ones.

Great article, I feel you really pointed out the differences in speed for most daily users. I was lost as to what ones to buy. This helped alot!

Also thanks for the Great Coupon! I'm getting the CSSD-F120GBGT-BK today!
 
How does this affect video encoding if at all? Great article though. Wish it had came out a few days sooner, just made a purchase of 2 60gig ssd drives.

video encoding doesn't gain much of a performance boost from faster hd access because of how much the cpu bottlenecks it. if you can download an application to watch hard drive activity, you can see that your drive is never pulling in that much data at a time.

personally I've encoded off my hd, off an ssd, and even purely in ram, and the difference is way too small, like margin of error difference, especially when I was doing 1080p encoding.
 
Thanks for that article. I wish the manufacturers would actually list this type of imformation.

Although an SSD or two would help performance of my system, I am still not convinced it is the best way to go for the price as I can get 4x 320GB drives and a Dell PERC/5i RAID controller that can use up to 1GB DDR2 for less than two of those SSDs with the performance (besides seek) being most likely higher. (4x of those 320GB drives in RAID 0+1 gives about 300MB/s.. if all 4 were in RAID 0, I would expect around 400-475MB/s).
 
@gjs278

Thanks for the reply. Good to know that I shouldnt be to affected by this async issue on that front at least. Got a 120gig sync for my os/apps. Usually I do a lot more research on this stuff before I buy but I had some money burning a hole in my pocket and fired from the hip. It should all work out in the end though. :)
 
The reason why this matters, is many things are already compressed.
Nearly all media is already compressed. Video files, sound files, pictures, movies, TV shows, etc. Most games have some form of compression going on to fit onto fewer disks to save costs. Text files aren't compressed, and neither are things like DLL files, many OS files aren't compressed.

If someone wants to test if something is compressed already, use a program that can do zip or rar files and zip or rar up the file or folder in question. If the original folder size is 5 gb, and the new zip file size is 4.9gb, almost everything there is already compressed.

What I'd like to see is the faster sandforce controller and faster flash combined with the compression option. Then we could get the best of both worlds.

Thanks [H] for doing good informative articles on subjects like this!
 
Thanks for that article. I wish the manufacturers would actually list this type of imformation.

Although an SSD or two would help performance of my system, I am still not convinced it is the best way to go for the price as I can get 4x 320GB drives and a Dell PERC/5i RAID controller that can use up to 1GB DDR2 for less than two of those SSDs with the performance (besides seek) being most likely higher. (4x of those 320GB drives in RAID 0+1 gives about 300MB/s.. if all 4 were in RAID 0, I would expect around 400-475MB/s).

your performance would only come close when dealing with large files. the main speedups would be in smaller files.

for what it's worth, you can hit 700mb/s tops on ich10r with just two c300 drives, you just won't have as much space to work with.
 
I know that AMD AHCI driver supports TRIM under single SSD, but what about RAID config?
 
I was going to post that exact graph, Maximus ... might as well start the graph at 182.16 lol
 
This was a great read and I will definitely be reviewing it for my future SSD purchases.

As for the graphs, Maximus825, you totally beat me to it. But yes, the top graph makes it appear that Synchronous is over three times faster.

In this graph below, I adjusted the minimal horizontal axis to be a value of 0

1312424624fz0ZfhgN4o_4_7.png

Hardocp_chart_example.png


In defense of the author, Microsoft Word charts do not automatically start at zero. If you were to input the values yourself, you would get a chart resembling the one from the review. Microsoft should not have this be the default method because it is not the proper method to compare data via bar graphs. It can be very misleading.

I hope this helps to clear the air on this issue.
 
Last edited:
I know that AMD AHCI driver supports TRIM under single SSD, but what about RAID config?

I believe that you're boned for using TRIM under a RAID configuration... I'm surprised that this was not noted in the conclusion. RevoDrives being the only exception.
 
Now I see what you all are yammering on about! It does look misleading, although admittedly I always just look at the numbers. In life not everything is to scale, those bar graphs clearly arent an exception.

Seriously though, there are numbers on the bars. If all you're doing is looking at the pretty pictures and not the data and the facts included in it then whose the sucker and whose really to blame?

I still think it was a good article, as are most of them over my many years here, and just dont see the point in all the bitching and complaining over a bar graph? Was the data and the facts in the article accurate? Was there plenty of data to backup there claims? Was it impartial and unbiased? Was the conclusion spot on? Those are the questions you probably outta be asking yourself. Not, "Gee these bars are purrty but they dont match up..."
 
I believe that you're boned for using TRIM under a RAID configuration... I'm surprised that this was not noted in the conclusion. RevoDrives being the only exception.

the revodrive is hardly an exception for now. as quoted here: http://www.nordichardware.com/news/...-revodrive-3-with-internal-raid-and-trim.html

Sadly TRIM (on the new revodrive) doesn't work in Windows 7, due to a limitation in the operating system that can't send the TRIM command to a SCSI/SAS controller and this is a problem that only Microsoft can solve. There will be a fix for this wit the next generation Windows Server, but when the TRIM command will work on RevoDrive 3 with Windows 7 is uncertain. The TRIM command should work with Linux-based operating systems as long as they can send it to a SCSI/SAS controller.

more details confirming that the trim doesn't work as well on http://www.theinquirer.net/inquirer/news/2094151/revodrive-x2-ssd-drives-light-day

in order to "trim" a revodrive your OS has to be capable of a feature called "scsi unmap" because trim is an ata command, and you can't send ata commands to pci-e devices. so if they ever add support for scsi unmap to windows 7, then yes you will essentially be trimming your revodrive. just not right now.
 
Great article, very good info.

Actually I was under the impression that Sync vs Async would only see a difference in benchmarks and that real world use would be pretty much the same. I'm surprised that there is a tangible performance increase with Synchronous NAND.
 
Back
Top