SSD Thread (Benchmarks, Brands, News, and Advice)

I was just replying about Intel's not dying randomly - then remembered - you got unlucky

Well... I suppose the same as my Vaio's battery dying after 2 weeks.

But nothing like the stanby/hibernate failures from other companies on Intels

Only issue they had was the firmware upgrade once - but that's not standard use - there is always a risk upgrading it.

 

locked in a 160GB Gen 2 X25-m for $345 today.

am well pleased.

let the games begin.

for my own reference sake, is this not a good Sammy?

Samsung 256GB SSD (Solid State) Hard drive - BRAND NEW - eBay (item 220601526731 end time May-09-10 18:21:43 PDT)

 

Yeah crap one.
MMDOE56G5MXP-0V A <--- OLD version
MMDPE56G5DXP-0V BD7

Wish you all the best

 

I just bought a Intel x25-m G2 160GB from a guy with lots of heatware feedback for $325 shipped. Installed it this weekend in my M11x and was so happy with it that I was on the hunt for another.

When I told the guy that I left him heatware feedback, he asked if I happened to be interested in another.

Now I have a second one on the way for my desktop.

 

Nice choice....

 

and nice price too

 

I am looking for a solid performing SSD 256gb and I was offered the following:

Samsung OEM (Dell) MMDPE56G5DXP-0VB - $695
Samsung OEM (Dell) MMDPE56G5DXP-0VBD7 -$745 (Full Encryption?)

I need no less than 256Gb.

I was wondering if these SSDs performance is good, if they are new models from Samsung, If they work well with Win7 TRIM and if the price is right, the seller is very respectable so I don't mind paying a bit more and have a peace of mind.

 

Samsung is worse than Intel...

 

How much slower in real world work would the Samsung newest drive be slower than the Intel ?

There is one major advantage with the Samsung is the 256gb capacity which is so much more than 160gb. 160gb won't cut it for me. 256gb is the sweet spot for me. So if the Samsung will be only just a bit slower I don't mind.

 

I only said it slower because of the slower 4K read vs Intel's (CDM bench test) but in real world usage, it's hard to notice unless you put them side by side.

On a side note, I can get a lil under 13 seconds Windows boot time (basic setup). This thing flies

 

4K read and writes are key - Dave claims a very noticeable difference in snappiness, and the most common access is 4K.
Oddly enough Software seems to gobble a lot of that up too... (reduces the Intel from 70Mb/s write down to about 30MB/s - not in Safe Mode though where it get's 60 odd MB/s) I don't want to know how the Samsung fares.

About size - I wonder why you need 256GB... - I use 35GB, that's with Office and the Adobe Web Suite + Vista (not slimmed, I like my drivers handy).

SSDs aren't for storage - carry a WD passport for storage

 

As others have said, you need to compare the performance for 4-8kb random writes on the Samsung compared to the Intel and there is lots difference.

Having said that, if its your first ssd you will NEVER know this as the difference in moving from a hard drive to a ssd is comparable to getting rid of your VW for a Porsche.

Its simply the nuts like us who pick and pull performance results apart and do what we can do squeeze every last drop of power from the SSD. Either way you are choosing a ssd with a reputation for stability and no surprises can be expected.

If you want it for the space, go Samsung. If you want to know you have the best performer, go Intel. Either way, you as a new user could never tell the difference...unless of course you become a SSD nut like the rest of us here.

 

3x100gb Vertex LE's RAID 0
All 0xFF




Default random

 

The 4K read speed actually shows how rubbish sandforce is at 4K speeds...
And didn't you once say you use easily compressible tests?

Try one that's incompressible.

 

how to explain this:

 

Modern 4k aware OS's like Vista/Windows 7 and the latest Mac OSX's read and write aligned on 4k boundries. If you run an IOMeter 4k random read test aligned on 4k boundries instead of 512byte ones like XP does you will get different results.

Here is an IOMeter 4k random read aligned on 4k boundries.

 

I assume that's 3 SSDs? I wonder how the Intels would look

 

I'm sure they would look great, you can expect 50% performance scaling for 2 drives over 1 on ICH10R and 16% for 3 drives over 2. There might be a few % overhead in there too, not sure.

I considered getting some Intel drives for my desktop build but took the risk with Sandforce. You know you are getting excelent drives and great support with Intel but I went with the hype on the Sandforce. The jury is still out as to whether I made the right decision or not.

 

@ ComputerMinder: I didn't notice any difference between Samsung (slowest 64 GB one) and Intel (slowest X25-V 40GB)

@ DetlevCM: here is results for "super fast" 3x X25-M 160 GB in RAID0

 

I need help possibly picking an SSD for my Dad's new computer. My dad is finally replacing a desktop from Medieval times with a new moderately powerful desktop he got for a steal (i3 CPU, 1TB hard drive, 8 Gigs of Ram, other goodies). Yeah it's only an i3 but I don't think he needed much more. Anyway he primarily does his home tax clients on this machine. I've investigated the professional tax software he uses...
1. It's single threaded and the developer has a "multi threaded recompile TBD" in an email. It's not
2. It uses a lot of random reads and writes to turn on and create tax returns. The tax returns can become huge tax documents that then benefit from large sequential reads on subsequent openings.
3. The amount of disc access I've seen on his desktop seems to confirm it likes to constantly write to disc. The tech guy I emailed made it seem like it has behavior of a server as far as writes go
4. My Dad says he doesn't have an issue with the CPU being powerful enough or needing more RAM: he says to him it feels like the hard drive takes forever to load things on his old computer.

Ok so my question is: If I wanted to buy him an SSD primarily for using his tax software, what the heck can I do for one? Would he be a candidate for an SLC drive? Is there a more cost efficient SLC drive than an Agility EX?

Edit I know for most SSD's I'd need a 3 inch adapter to use it in a desktop (well securely mount it anyway)

 

I wouldn't get one of the new controller manufacturers Micron or Sandforce for your dad's work machine as neither is proven yet. That leaves only Intel because it has the best random access IOPS performance(by a wide margin). He should be able to get by with the much cheaper MLC model if he sets the toolbox to run on a frequent schedule to offset the random access degradation of his server like usage pattern.

Right now Sandforce has good random access out of the box but degrades once every block has been written to, then they suck wind...hard. There is no toolbox yet and the only way to restore performance is HDDErase. I didn't have that problem with the single drive configuration with TRIM but didn't test it extensively either.

Knowing I was getting a third drive today and was going to HDDErase the two drive configuration anyway, went ahead and hammered it with benchmarks. Once reaching the point of writing to every cell, 512k random access fell to 5900 IOPS, thats worse than any other SSD that I have tested. Did not need to run a benchmark to know performance was slower, as there were noticeable, HDD like delays when right clicking and opening programs.

 

I would like to explore this especially since I am once again receiving ssds for review. I presently have the OWC Mercury Extreme RE 100Gb installed which has 28% overprovisioning. Does this not accommodate for that slowing at all?

Do you feel this is the same effect with all Sandforce drives and, if so, why hasnt it been recognized just yet? Anandtech?

 

Strange indeed... why don't 4K writes reads improve? - that's just slightly above a single SSD....

(And would mean SSD raid is pointless if you don't queue)

 

I am sure the 28% overprovisioning helps alot with staving off degradation. Anand uses and tests his drives normally and probably secure erases in between tests. Anand does not use CDM3 or AS SSD benchmark for benchmarking, he uses IOMeter, Anandtech storage bench and PCMV. Anand also uses a single drive configuration that is TRIM enabled.

I suspect that CDM3 default random and AS SSD Benchmark could be causing the Micron and Sandforce drives problems. If you run those benchmarks excessively on any SSD it will bork it's performance.

You can get improved 4k read speeds with a pci-e RAID controller and use a small stripe size.

Here look through the SSD benchmarks on this site--all the top performing drives, including the Intels show about the same 4k read speeds with CDM3 default random and AS SSD Benchmark.

Storage - XtremeSystems Forums

 

Looks like a messy forum to me... is there no comparison?

 

On the OCZ site in a thread pertaining to the new vertex 2 I had read that, specifically with SandForce controllers, some bench programs will not give accurate readings (ie Crystal Score with SandForce) because of the way the info is compressed but I dont understand how that would make things different between my OWC and my Intel? Does the ssd work in a different compression format?

 

Intel doesn't compress at all.

 

I don't think the compression algorithms are as much of an issue with the CDM3 and AS SSD Benchmarks as the way the data is aligned on 512 byte sectors instead of 4k boundries. If you are installing and using apps written for modern 4k aware OS's, they will be highly compressible anyway.

This is the kind of technical stuff that is way over my head, been asking, and still waiting for one of the more technical oriented posters on a couple of different forums to respond but haven't got an answer yet.

I am thinking that the slow transition away from older OS's like XP is making it difficult for engineers to develop the next gen controllers because the way modern OS's read and write data is different from the way older OS's like XP.

XP aligns I/O's on 512 byte sector boundries and modern OS's align I/O's on 4 kilobyte boundries. Not sure if this is as problematic as I am thinking but the benchmarks are much higher for the latest controllers when I/O's are aligned on 4k boundries. Could be that the way some of the older HDD benchmarks are written (aligned on 512 byte boundries) could be causing problems with controllers designed to work best with OS's that read and write data on 4k boundries.

 

Thanks...this place remains to be a wealth of knowledge untouchable by any other.

 

SSD knowledge rules! BTW, I am liking my Corsair NOVA so much I think it may be illegal

 

How can I achieve this kind of score on my X25-M 160gb G2 Lenovo T500. My current score is 388

 

You can't. As far as I know the T500 does not support 3 drives. You could do a Raid with 2 SSDs and that would get you 80% of the way there or so.

 

Wow, I thought I had seen some long threads in my time!!! And this is a 2nd version, dang!
It seems like there should be a general consensus on what are the top 3 drives (in terms of speed, heat, and battery use) for each popular size (40gb, 80gb, 260GB)...Does such a doc exist? Would really like a decent SSD for my new i7 proc laptop with 8gb of ram!

 

Doesn't exist - but generally the best choice is Intel
Reliable too

They take a beating and don't loose performance.

 

So...X25-m?
Newegg.com - Intel,intel ssd

 

Generally a good choice

X25-E are server drives.

On the X25-M you can use the Toolbox to manually trim it under XP and Vista.

 

What about for win7, it automatically does the trim functions?
There is also a kingston...50gb more for $60?

http://www.newegg.com/Product/Product.aspx?Item=N82E16820139113

 

Yep. Windows 7 support TRIM feature

*You can still use the toolbox on Windows 7*

 

QUESTION....

If you download the newest Crystal Disk Mark you now have test options which are random, 0Fill and 1Fill. Can anyone explain this and what bearing it has on scoring?

 

Thanks for the advice. I'll see how my Dad likes the computer before I consider dropping in an SSD.

 

I think this is to allow those Sandforce controllers to have compressible data and incompressible data.

 

@DetlevCM
I have to agree with that because on mine it makes no difference

 

Intels don't compress so that fits the patter too

 

Well......

When I test the new OWC with the normal random filter, I get scores that are low around 200MB/s seq read and 135MB/s write but when I change it to 0 Fill or 1 Fill I get their claimed result almost dead on....funny thing eh.

I wrote the guy who created it but he cant really speak English well.

 

I think he's Japanese (Sourceforge.jp )

Well, its not surprising that you get the claimed result only with easily compressible data - its the same with CF cards... on most cases the speed they advertise is the read speed... continuous read speed... never the write speed...

You'll pick the best values you can obtain in artificial conditions - if someone complains you just say "according to our test procedure..." tell them how to recreate it and send them packing.
Not matter if there is any real life value or not in it.

 

Ya hate to say it though but the OWC blows the Intel away in PCVantage...

Check this out; OWC first and Intel second. I had to rerun the Intel a few times to believe it and even manually optimized it just to make sure...

 

Why do you hate to say it?

 

Figure of speech......

 

This method of secure erase might be of some interest to laptop owners with SSD's as many laptops BIOS can't be configured for IDE compatible mode for HDDErase.
http://www.ocztechnologyforum.com/forum/showthread.php?73002-Secure-Erase-for-Windows&
Haven't tested it myself, I use HDDErase whenever wanting to reset my SSD for a re-image or clean install as my desktops mobo has IDE and compatible SATA configuration options.

 

Is the Corsair Nova V64 a good alternative for the Intel G2?

The review of the V128 looks quite good:
A look at new SSDs from Corsair, Kingston, Plextor, and WD - The Tech Report - Page 6