Vertex 2 RAID 0 Vs Intel G1 Raid0?

Copyright

Supreme [H]ardness
Joined
Nov 17, 2000
Messages
6,004
Thinking of picking up a couple of these to replace my G1 Intel drives. Would I pickup performance? I know the writes would be better but curious if overall system performance would be worse or better over my G1 setup. Or if there are currently a pair of ~80gb drives out there I can RAID 0 with that would be an improvement or RAID 0 3 drives? Looking for the fastest option under 400$.

http://www.newegg.com/Product/Produ...0227601&cm_re=Vertex_2-_-20-227-601-_-Product
 
Last edited:
Vertex 2 RAID 0 Vs Intel G1 Raid0
Two 50GB Vertex2's weren't a match for my Intel 80GB G2s in RAID0.

I've also run two G1s in RAID0.

The Vertex2s would probably be a small improvement.

If you could swing a coupla C300s that would be the ticket. ;)

Check out 2x128mb C300s here.
 
Last edited:
The person above is clearly misinformed a vertex 2 ssd will beat an intel g2 ssd in every way with one possible exception and that's dealing with incompressible data.

That being said depending on what you do you still may not notice any real world difference.
 
The person above is clearly misinformed a vertex 2 ssd will beat an intel g2 ssd in every way with one possible exception and that's dealing with incompressible data.
The person above has had many SSDs and knows exactly what he's talking about.

This statement you made is from a "Stat Expert" that has no personal experience with SSDs and doesn't realize the difference between paper stats and real world experience. ;)

BUT, you can still have some turkey! :D
 
Would the diffrence beetween them be even noticeable without benchmark programs ?
 
The Intels will not be able to clean themselves while in RAID (though there was a rumor the latest Intel Matrix Driver might allow TRIM in RAID finally).

The Sandforce based drives should be able to clean themselves in RAID as they have OS-dependent garbage collection.
 
Would the diffrence beetween them be even noticeable without benchmark programs ?

It was to me.

The SF drive's aren't bad drives but don't live up to the paper specs in the real world.

Many SSDs excell in different aspects but without a side by side comparison they would all be about the same IF they're in the same class....ie...low/med/upper tier classes.

I can only give you my personal observations but I've had 8 or 9 SSD set-ups and a lot of personal experience.

The Intels will not be able to clean themselves while in RAID (though there was a rumor the latest Intel Matrix Driver might allow TRIM in RAID finally).
ALL SSDs have an alternative to TRIM in case is isn't avaliable and that includes Intel.

Some manfgs claim their garbage collection to be better/faster but ALL have an alternative.

Anyone that claims differently has no idea what's going on with SSDs.
 
So your saying that TRIM or short stroking is unnecessary? I mean I could just buy 3 to 5 80GB Intel 25-M G1 drives at approximately $105 each used and be better off than having any other SSD?
 
So your saying that TRIM or short stroking is unnecessary? I mean I could just buy 3 to 5 80GB Intel 25-M G1 drives at approximately $105 each used and be better off than having any other SSD?
I have no idea about that.

TRIM is an advanced GC tool and all SSDs have GC.
 
I know what TRIM is. The OS specifically tells the drive what blocks are not in use so the drive can forget about whatever data is on the block when it needs to rearrange data for wear-leveling or when it needs to write.

As for GC I am not sure how without the help of the OS it could figure out what blocks are not in use.

After reading the wikipedia entry I think I see the answer. GC does not do this. It uses a program to tell it specifically what blocks it can ignore.

Where TRIM is not automatically supported by the operating system, there are utilities which can send TRIM commands manually. Usually they list all free blocks as specified by the operating system and then pass this list as a series of TRIM commands to the drive. These utilities are available from various manufacturers (Intel,[16] G.Skill[28]) or as general utilities (hdparm since v9.17[29][30]).

I see the mention of hdparm. I am going to have to see how that will work since I am using my SSDs on linux.
 
The SF drive's aren't bad drives but don't live up to the paper specs in the real world.
You sound like one of those people who ran benchmarks using incompressible data and then wonder why the results came out different than the specs.

If you use compressible data the vertex 2 drives do indeed get close to their specs.
 
As for GC I am not sure how without the help of the OS it could figure out what blocks are not in use.

There are several different techniques that go by the name of "garbage collection" or "idle time garbage collection" or "real time garbage collection". None of them are well documented, so there is a lot of confusion.

One concept that is needed to understand garbage collection is that SSDs generally have a smallest addressable I/O size of 4KB ("page"), but the smallest erasable block size is generally 512KB ("block"). Before writing to flash memory, the SSD needs at least one block that has been erased. And erasing a block is a relatively slow process. The point of GC is to try to always have some pre-erased blocks available. The trick is how to manage that.

The most complex to implement GC is ITGC-NTFS that is aware of the NTFS filesystem structure. When the SSD has been idle for a while (for some definition of idle) It examines the NTFS metadata to see what sectors/clusters are unused, and then it collects the unused sectors into erased block sized chunks and pre-erases some blocks, so that the next time the SSD writes, there will be blocks available that are already erased. I've seen claims from OCZ and Anand about some SSDs that use this technique, but I am skeptical whether all SSDs that have been claimed to use this technique actually use it. Obviously, it will only work if you are using an NTFS filesystem. And it will NOT work in RAID, since the NTFS metadata can be on a different SSD than the one doing the ITGC-NTFS.

Another form of ITGC works with all filesystems. It simply looks for blocks that have lots of unused ("invalid") pages. Then it collects the used pages together into erase-block sized chunks, thus freeing up other erase blocks to be pre-erased. Note that this technique requires reserved flash memory capacity, since without a reserve, once the full specified capacity of the SSD has been written, there would be no unused pages (even if files have been deleted, the SSD does not know it without TRIM). But since all SSDs have reserve capacity, there will always be some unused pages. Sometimes this process can be made more efficient by overprovisioning -- secure erase the SSD then partition it so that some sectors are never written to.

Another type of GC is real time garbage collection. This is similar to ITGC, except it only runs when there is a write to the SSD. As far as I know, this is the type of GC that Intel uses. If you check Anand's site, you can find some HD Tune scans where you can see a high-frequency periodic waveform for write speed for Intel SSDs, which I interpret as the RTGC doing its job -- some writes are slower as it pre-erases some blocks. RTGC may also be an explanation for why the sequential write speed of Intel SSDs have lagged some other SSDs. The upside of RTGC is that the drive never seems to get into a seriously degraded state (performance-wise). Another way of saying it is that Intel has worked to optimize the worst-case performance.
 
Last edited:
You sound like one of those people who ran benchmarks using incompressible data and then wonder why the results came out different than the specs.

If you use compressible data the vertex 2 drives do indeed get close to their specs.

The problem is that most data looks incompressible to the Sandforce SSDs. Sure, they can do a great job of compressing a string of zeros, but who writes nothing but zeros to their drive?

Keep in mind that the Sandforce SSDs consume only a few Watts of power max. Compare that to typical desktop CPUs, which consume 70-130W. Even laptop CPUs are typically around 15-30W. Also, the Sandforce SSDs have a specified write speed of 270 MB/s. How sophisticated a compression algorithm could it possibly be running that only consumes 1-2 Watts and yet has a throughput of 270 MB/s ? The answer is that is cannot be very sophisticated. My guess is that it is doing compression of running strings of identical bytes, and little more.
 
The problem is that most data looks incompressible to the Sandforce SSDs.
What are you basing that off of?

Vertex 2 drives do a wonderful job with real compressible data such as anything relating to the os and game loading which are some of the main points of having an ssd for some people.

The only real incompressible data I can think of are things that would be better suited for large hard drives anyway.
 
What are you basing that off of?

I explained that in my post. Perhaps you should read more carefully.

Also, you can look up benchmarks. The only ones that have the Sandforce drives performing close to spec are ones like ATTO that write strings of zeros. If you look at benchmarks like copying real files, the Sandforce SSDs perform at 50% of spec or less.
 
I explained that in my post. Perhaps you should read more carefully.
You explained nothing in your post other than your "guess".

Also, you can look up benchmarks. The only ones that have the Sandforce drives performing close to spec are ones like ATTO that write strings of zeros. If you look at benchmarks like copying real files, the Sandforce SSDs perform at 50% of spec or less.
Completely false.
 
Fine, you can go on drinking the Sandforce Kool-Aid and believing in magical compression chips that can compress most data 50% at 270 MB/s while consuming 1 Watt. The rest of us will be in reality. Join us when you are ready.
 
LOL what happend to my post. I guess the real questions I shoudl ask is are there really any talks of TRIM working in RAID. I would hold out for that in a heart beat. The only single drive I think I might consider woudl be the 256gb C300 but it's not cheap. Right now my Intel G1's are almost a year old but still bench the same as they did on day one. OS still feels fast but I have a buddy in need of an upgrade do to failure and he always ends up with my old stuff. I honestly don't even see anythign out that would just be worth jumping on. Maybe ill just hold out to the G3... my buddy is screwed lol
 
If you want to get high performance with a RAID 0 of SSDs, then get an LSI 9211 controller, and a couple Crucial C300 128GB SSDs. You should be able to get 600+ MB/s sequential reads, 260 MB/s sequential writes, and incredible random read and write performance. But I doubt you will notice much of a difference in everyday usage compared to Intel G1 SSDs.
 
Last edited:
The most complex to implement GC is ITGC-NTFS that is aware of the NTFS filesystem structure. When the SSD has been idle for a while (for some definition of idle) It examines the NTFS metadata to see what sectors/clusters are unused, and then it collects the unused sectors into erased block sized chunks and pre-erases some blocks, so that the next time the SSD writes, there will be blocks available that are already erased. I've seen claims from OCZ and Anand about some SSDs that use this technique, but I am skeptical whether all SSDs that have been claimed to use this technique actually use it. Obviously, it will only work if you are using an NTFS filesystem. And it will NOT work in RAID, since the NTFS metadata can be on a different SSD than the one doing the ITGC-NTFS.

This is done totally inside the SSD with no help from the OS? If so I do not think this could work on raid being that each SSD in raid would only have a fraction of the NTFS filesystem metatdata.
 
Yes, that is why I wrote "And it will NOT work in RAID..." in the part you quoted!
 
why bother just get 1 SSD with windows 7 and use AHCI, RAID+ SSD is completely not needed More so with the SF-1200 based SSDs as there random access speeds smoke most stuff thats currently around even makes my M225 look slow , one of my customers got an corsair F60 or F40 and the speed that boots the desktop is insane you lucky if you can even see the welcome screen (bit just before the desktop, if you have or have not passworded your pc),

vertex 2 180gb
http://www.newegg.com/Product/Product.aspx?Item=N82E16820227602 $299
http://www.newegg.com/Product/Product.aspx?Item=N82E16820227612 $399

some wacky prices on newegg ,some one tell me why the $100 more one it looks the same (one mite be retail boxed ? or other one old stock)
or
http://www.newegg.com/Product/Product.aspx?Item=N82E16820227552 $419
http://www.newegg.com/Product/Product.aspx?Item=N82E16820227591 $449
the 240gb as thats just above by $19

buy the SSD you need (update the drive firmware if needed) and do not bother to raid them, no need unless your doing server loads or something that Requires stupid amount of IOP

if your using it for games or norm programs you be subsystem (CPU or somthing els) limited when games are loaded (bit-tech do not bother listing game start up times any more or OS as they norm around the same time apart form the C300 that one was 3-4 seconds faster)

not sure but think GC does work under RAID very well for the sf-1200 that is (sure anandtech tested it i recheck on that) still no point thought
 
That all makes sense to me. Im actually sick of RAID so it would be nice to go with a simple single drive. Now I have to decide if I want a SandForce 180gb or Crucial C300 256gb. Curcial will cost me more of course but seems to run better. Then there is the Intel G3 coming soon.
 
That all makes sense to me. Im actually sick of RAID so it would be nice to go with a simple single drive. Now I have to decide if I want a SandForce 180gb or Crucial C300 256gb. Curcial will cost me more of course but seems to run better. Then there is the Intel G3 coming soon.

Judging from Oldie's and john4200 posts, I would not even consider a Sandforce at the moment due to the issues they mentioned. The data and performance results that I've seen here on the forums backs up their statements. I haven't seen anything that would back up xeonMP.

So either go Intel or Crucial C300.
 
ALL SSDs have an alternative to TRIM in case is isn't avaliable and that includes Intel.

Some mfg's claim their garbage collection to be better/faster but ALL have an alternative.

Anyone that claims differently has no idea what's going on with SSDs.

Care to enlighten me as to why my 3 year old Patriot Warp JMicron drive which does not support TRIM nor allow Secure Erase runs like absolute fucking dog shit now when makeing writes? I mean you'd think 3 years of use would be enough for it clean it self up with the undocumented garbage collection it has........
 
I ran the Benchmark.exe benchmark on the 2 SSD's in my sig. But before I go on, I must explain a catch to the results. The Intel 160GB has Windows based disk compression enabled to save writes and space. The OCZ Agility 2 does not have disk compression enabled because it already does that.

So I used diskbench and read one of the data files from World of Warcraft. This was about a 3GB file. Different computers, but the file is the same. They both read very close to each other...I think it was right about 200MB/s, they were like maybe within 10MB of each other. So that is a real file that has whatever kind of data I am likely to see on a regular basis. I did some writes, but I can't remember the actual numbers. They also weren't too far from each other, but I don't wanna just make up some number.

So they were close to each other, but I was kinda impressed that Sandforce with less then half the NAND was able to keep up with denser Intel.
 
Care to enlighten me as to why my 3 year old Patriot Warp JMicron drive which does not support TRIM nor allow Secure Erase runs like absolute fucking dog shit now when makeing writes? I mean you'd think 3 years of use would be enough for it clean it self up with the undocumented garbage collection it has........
I'm sure if I had kept my old GSkill Titan it'd be the same way.

They weren't screamers in the first place but without some kinda GC it'd run "absolute fucking dog shit" slow after the first 72 hrs. :D

There's ALWAYS been some kinda GC/Cleaner/TRIM in EVERY SSD.

If not, they would all run like yours after the first time you've written to every cell.
 
Fine, you can go on drinking the Sandforce Kool-Aid and believing in magical compression chips that can compress most data 50% at 270 MB/s while consuming 1 Watt. The rest of us will be in reality. Join us when you are ready.

LOL!
 
Looks like 2x128gb and a single 256gb are about the same price. Makes me wonder if I would be better off with a single drive so that TRIM can be used.
 
Looks like 2x128gb and a single 256gb are about the same price. Makes me wonder if I would be better off with a single drive so that TRIM can be used.

Decsions, decisions........:D

I sold my 2 Intel 80gb in RAID0 for a Crucial C300 256 because I felt it was a better overall situation....no TRIM in RAID, prices were falling on the Intel drives and the writes were slower.

Had I known that 2x128 Crucials were going to perform so well, I would have gone that way if the prices were close to a single 256GB unit.
 
Looks like 2x128gb and a single 256gb are about the same price. Makes me wonder if I would be better off with a single drive so that TRIM can be used.

Decsions, decisions........:D

I sold my 2 Intel 80gb in RAID0 for a Crucial C300 256 because I felt it was a better overall situation....no TRIM in RAID, prices were falling on the Intel drives and the writes were slower.

Had I known that 2x128 Crucials were going to perform so well, I would have gone that way if the prices were close to a single 256GB unit.
 
I want to sell my 2 x 80GB Intel G2s to get a Crucial c300.................
 
The problem is that most data looks incompressible to the Sandforce SSDs. Sure, they can do a great job of compressing a string of zeros, but who writes nothing but zeros to their drive?

Keep in mind that the Sandforce SSDs consume only a few Watts of power max. Compare that to typical desktop CPUs, which consume 70-130W. Even laptop CPUs are typically around 15-30W. Also, the Sandforce SSDs have a specified write speed of 270 MB/s. How sophisticated a compression algorithm could it possibly be running that only consumes 1-2 Watts and yet has a throughput of 270 MB/s ? The answer is that is cannot be very sophisticated. My guess is that it is doing compression of running strings of identical bytes, and little more.

Your logic is 100% faulty here. You can't compare power consumption of a compression algorithm on a general-purpose CPU to a compression algorithm implemented in hardware.

The Sandforce controller has circuitry dedicated specifically to compressing data without the arbitrary limits of a general-purpose CPU. Just like cell phones have dedicated circuitry to decode MPEG-4 video without consuming much power.

Or for another analogy, look at how much more efficient GPUs are at crunching graphics-related operations than your CPU is. Dedicated hardware will always be far, far more efficient power-wise.
 
Fine, you can go on drinking the Sandforce Kool-Aid and believing in magical compression chips that can compress most data 50% at 270 MB/s while consuming 1 Watt. The rest of us will be in reality. Join us when you are ready.

Do you have any benchmarks or evidence to support your claims? Or are you just arguing based on your incorrect assumptions from above?
 
Do you have any benchmarks or evidence to support your claims?

Of course. There are benchmarks all over the web. I've linked to some before in this forum.

As for your claims, please provide a link to the magic compression chip that can compress most data 50% at 270MB/s while consuming 1W or less of power.
 
Last edited:
Decsions, decisions........:D

I sold my 2 Intel 80gb in RAID0 for a Crucial C300 256 because I felt it was a better overall situation....no TRIM in RAID, prices were falling on the Intel drives and the writes were slower.

Had I known that 2x128 Crucials were going to perform so well, I would have gone that way if the prices were close to a single 256GB unit.

On a SATA III controller those 2x128's would scream even more but wonder how time will effect it only using GC for cleanup and no TRIM support in RAID. I am really surprised my intels with an almost year old OS still bench over 500 for the read and 140 for the write. To many decisions. :(
 
Of course. There are benchmarks all over the web. I've linked to some before in this forum.

As for your claims, please provide a link to the magic compression chip that can compress most data 50% at 270MB/s while consuming 1W or less of power.

Could you do us all a favor and link to them or provide references? I have no doubt that data that doesn't lend itself well to compression will be transferred at a lower rate than highly compressible data, but all we really care about is the real-world impact. If the real-world impact is minimal, then we don't care.

You probably won't find your magic compression chip on the market. :p Lossless streaming data compression is a trivial problem in VLSI these days, and would definitely just be integrated into the controller chip.

However, if we look to hardware H264 HDTV encoders (a MUCH more complicated problem than simple lossless data compression) there are plenty of chips which will compress 1080i/60fps YUV422 (approx 124MB / sec uncompressed) at 1 Watt:

http://www.ntt-electronics.com/en/products/video/products/lsi_and_modules_solutions/pinea/index.html
 
Who is this "all" you are talking about? Many people here have either looked up some benchmarks themselves, read past threads discussing this, or are capable of doing some simple searches to find it. I suggest you try that.

Also, I guarantee you that the chip you linked to will NOT compress most data people write to their hard drives by 50% at 270 MB/s. If it were as easy as you imagine, there would be plenty of HDDs available with such magic chips -- doubling the HDD capacity would certainly be worth a few bucks and 1W of power.
 
Back
Top