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

10,000 write cycles isn't a lot, not at all, especially if the controller algorithm is bad.

JEDEC's future prospects looks at taking out the software aspect of the SSD controller and integrating it into the configuration software, possibly the chipset. It would further simply computers. Things like wear levelling would be done all on the CPU and looking at how under-utilized CPUs are nowadays and how much more powerful it'll be in couple of years, it looks good all around.

Things like Turbo Memory is one step closer to that vision. In Nehalem, the controller will be on the PCH.

 

Well that seems like there is misprint then.

 

Well yeah that's wise, I'm just reporting what I heard and what the Dell site says.
Like I said "maybe others can confirm this as well", as I dunno what to believe.
My biggest concern was that I wasn't getting SATA150 interface.

Looks like 'ofelas' got one and it turns out it was probably just a misprint, but the rep just couldn't believe that was the case.

Yeah reps know alot less than the people hanging in this forum about the laptops and SSD's.

 

Do you mean integrating it into the hardware?

 

Chipset software, like the chipset drivers.

 

I also bought the exact same one from Dell and it definitely is SATA 3Gbps

 

I always assume company like Samsung would have similar write algorithm as Intel SSD, distribution and spread the write cycles evenly across all flash chip.

Am I wrong?

 

No, I don't think so. The drive will move stuff around to evenly wear out the entire drive.

 

Someone posted they applied the Dell firmware upgrade to their geeks.com Sammy drives with a "hack" to change the update script so I just had to try it again.

1. It wouldn't work in my main computer so I had to remove the drives to try the update in a different computer.

2. I was lazy and broke one of the power connectors pulling out the drive with too much force! Luckily I was able to piece it back together but the power connection on one of the drives is weak.

3. The update script said it failed at the end, but when I tried it again it said it was updated, so I don't know what was going on.

4. I lost the DATA on my drives!! When I put them back in the computer they go in my motherboard Intel RAID did not recognize the RAIDED drives! And yes, I plugged the connectors into the right drives.

5. Fortunately I was "smart" enough to back up my system before trying this... so I was able to restore my system using Acronis.

6. The whole thing was a pain but it got done. I am not updated, but with permanent damage to a power connector on one of my two SSD drives... but at least it works.

7. I think my next build will be a full tower so I will have more room to work in there... plus I will try not to be so impatient when removing hardware and installing it.

 

I have the Samsung 256GB drive purchased early on from ebay from a M1330 pull. I have gotten very decent ATTO results in my XPS M1530, with 200MB/sec read and 175 MB/sec write on my OS partition.

You probably can't get more than 130MB/sec if the drive was SATA 150. It has to be SATA 300.

Most SATA 300 drives in a PC that only supports SATA 150 will downspeed to SATA 150 without a problem.

 

Is the new SuperTalent 128GB SLC a rebrand of the coming OCZ Vertex SLC 128GB? I am getting quite interested in a 128GB SLC as long as price is somewhat reasonable.

 

I like what I did - got two Samsung 64GB SLC drives and raided them for less than $400. You can probably do the same with the 64GB SLC drives on Newegg, but will need an adapter.

 

And now, for an entry in "Pimp my Vertex", I present the latest firmware (1.10) that dropped on Tuesday. This thing is pulling in holy **** numbers now.

 

WOW, nice Vertex pimping, some kind of numbers! WOW

 

That is awesome. I still find it odd that all these firmware changes cause major changes in speed. Are they sacrificing something else to get these bumps in sequential speed?

 

My ATTO and Crystal DiskMark results didn't show any big losses anywhere in their numbers, so whatever Indilinx is doing, they're doing a good job of it.

 

Just make sure you keep your system backed up every day if you're using Vertex... you never know with all these firmware changes.

 

IOMeter and IOPS?

 

Outstanding I/O's set to 3, 4k, 100% Random, 100% Write, Maximum Sector set to 200,000 (100 MB). I played around with this setting and I got the same results at 200,000 as unlimited, so I leave it on 200,000 to save hard drive wear. The more I use and bench this drive, the more clear it is to me that this is indisputably the best second choice behind Intel's drives.

 

The drive is cheaper. There are reasons for being cheap. Yea there is a brand name value or whatever but the difference is still much too big in this case. There's always a sacrifice to it. For the Intel drive to achieve near 1x write amplification(reliability) they have a greater degradation of write performance.

Another thing is write performance is related to reliability features.

 

Someone kill the troll posts above. Another single post poster pimpin' (spamming) his site. Sheesh.

Back on topic. Does anyone have an Intel X25-E here?

 

Thanks ZERO for spamming like 20 times but X25-M is MLC. And your link only shows sequential throughput which showed that it has very little relation with random write performance. X25-E, the SLC version is out of reach of any of the drives out there(including price).

 

Keep in mind a lot of SATA II devices sold to retail come with a jumper installed to make them SATA I. I always thought this was the stupidest thing ever until I ran into a system where I had to reinstall the jumper to make the drive work. Most SATA Is will work fine with a SATA II drive with just the drive doing the "backwards compatibility" but occasionally you actually need to install the jumper. I'd imagine it'd be more likely these are SATA II drives that just come with the SATA I compatibility jumpers installed. It might well be that Dell's individual drives are SATA I out of the box until you remove the jumper.

 

Ok so should I use the OCZ VERTEX 1.10 drivers for my SUPER TALENT UltraDrive ME FTM32GX25H 2.5" 32GB SATA II MLC Internal Solid state disk (SSD)??

 

They aren't drivers, it is firmware, it's slightly more...how should I say it...intrusive? Check TW1's experience above.

I don't know if the drives accept each other's firmware updates. They may or they may not. It depends on how the OEMs have configured and shipped them.

 

yea i meant to say firmware =D Thanks for the info! I wonder if the manufacturer has drivers.

 

firmware rather =D

 

Yeah, regular may do it with a jumper or even a utility, I've used a utility to set the drives operation which is probably a firmware entry being changed.

SSD's have jumpers too?
I assumed they didn't but since I don't own one I guess I wasn't sure how the whole master slave relationship was going to be handled without a jumper.

 

sata doesn't have master slave.

haven't seen any jumper on sata drives yet actually.. and none on any ssd for sure.

 

NO!!!!
1.10 is made specially for OCZ product, u may have a chance to brick ur SSD

 

It just keeps getting better eh?

Compare your results with Anand's test with the 1275 firmware:

IOPS: 1656
Transfer rate: 6.47 MB/s
Average latency: 1.81 ms

The new firmware shows improvements in every respect. Still far behind the Intel, but I highly doubt that the difference in a desktop environment is even noticeable at such a high level of performance. With the vertex we are talking about IOPS performance that is like 16X faster than a raptor HDD...and fits in a notebook.

BTW, I *think* the max sector setting will only make a difference on HDDs where there is varying degrees of mechanical movement of the heads depending on how large or small the area of the disk being used is.

 

Ahh yes, makes sense.
Come to think of it my latest SATA have no jumpers either that I can recall.
SATA did away with that.

 

Yep! Backed up is a REAL good idea! Dave

 

Just received this from good old Dell. I am quite certain you can apply these universal 'HDD' savings to your brand spanking new 256GB SSD by Samsung. Go crazy

US Only--sorry my international compatriots.

Edit: Reference link

 

indeed, espencially for firmware updates which always invalidate all data on disk great to simply boot with the usb stick in, connecting to homeserver and restoring the last image. really great

 

Yes, once you get into this level of performance, while the Intel drives are still significantly faster, you reach a point again where the hard drive is not the bottleneck. 100% random 4K writes with large queue depth are pretty much absolute worst case SSD scenario, so getting 7.2 MB/s and 1850 IOPS is enough to cover 99.9% of all consumer usage scenarios.

 

Wow, while investigating SSD upgrades, in conjunction with the offer mentioned above, look at what I came across. It's hilarious because of the superior RPM speed. Jackboot TOLD you vRPM was a good idea. At least we wouldn't have 1800 RPM SSDs!

Oh, hey, don't let the grand and third price tag dissuade you from purchasing this refurbished marvel!

 

I thought you guys might find this interesting! A new driver-level write filter is being developed. It is currently in beta 3. Basically it uses system memory to turn all random writes into sequential writes. It is supposed to only be for XP, but I read that someone tried it on Windows 7 and it worked...so I did a system image and installed it.

OCZ Solid Series 60GB
Windows 7 Beta build 7057

IOMeter 4KB 100% Random Write, 10GB portion of the drive, 2 minutes, queue depth = 4

Without Flashpoint:


With Flashpoint:


The above indicates that flashpoint seems to be able to completely alleviate the JMicron stutter...essentially turning "useless" old JMicron drives and making them faster than raptors.

As far as I can tell the above benchmark is accurate and is not merely the result of Flashpoint "tricking" IOMeter into reporting incorrect results. I need to do a bunch of testing to evaluate my newfound performance. At this early stage I can at least report that when IOMeter prepares the disk prior to conducting the test (writes the 10GB test file) it takes only a *fraction* of the time to do so with flashpoint compared to without. I'm not clear on if this would be a sequential or random write task though, or what the size of the writes would be...

 

Wow--any differences in real-life usage? Where the heck do you find these things? (Great find!)

 

Oops, forgot the LINK to the developer's website. It is freeware.

If you have a JMicron based drive, you have to try this out.

EDIT: Do NOT try this out on Windows 7 - it is currently NOT supported and although the performance increase works as it should, any changes changes made to the disk are lost at reboot. This also means that it is impossible to uninstall the software since the changes you make to the registry to uninstall are lost at reboot. At least I couldn't figure out how to uninstall and so just restored from a system image. So for now this software is not suitable for a live Windows 7 environment. However, I hear that it is working just great on XP.
/EDIT

It is too early to say whether or not I can see any real-life performance increase other than the creation of the IOMeter test file as noted above. I intend to do some testing to see. Specifically, my solid series has been terrible at dealing with uncompressed audio editing. I'd like to see what happens now.

Oh, and for anyone interested, I calculated the vRPM spec for the solid series with this data

OCZ Solid Series (JMicron): 600vRPM
OCZ Solid Series (JMicron) with flashpoint: 32,000vRPM
SanDisk G3 (advertised, not yet released): 40,000vRPM
OCZ Vertex (firmware 1275): 73,000vRPM
Intel X25-M: 414,000vRPM

Here is the 4K random read screen that I used for the read portion of calculating vRPM:

 

Hmm, sounds like something very similar to Windows SteadyState which did completely eliminate the JMicron stutter but at the cost of long reboot times and a large chunk of disk space for extra cache. Does Flashpoint have either of those two issues? If not, this would be a great budget solution.

My Core V1 and V2s were both great using Windows SteadyState.

 

Anyone have any more info about that Solidata drive that NBR "reviewed"? I remember someone mentioning that Solidata made their own controllers, but the temps seem to be in line with those of a dual-JMicron drive. Nonetheless, the price is extremely high for a dual-JMicron drive, so I don't know what Solidata is smoking...

 

Jackboot, might want to update your posts with the little problem you ran into.

Basically, under Windows 7 with Flashpoint, none of the cached writes are recommitted to the drive, so you're working with effectively a LiveCD for a system. I BSOD under XP SP3, but on another installation it worked for me just fine. It's definitely not ready for the prime time, but if it gets sorted out, it should eliminate any drive stutter. My Vertex is currently truckin' along on firmware 1.10 so I'm in no hurry to apply any experimental tweaks until I'm sure things are worked out.

 

Dug up info on the Solidata X1:
http://forum.ssdworld.ch/viewtopic.php?f=4&t=85&sid=95f11416502e055b9849c551e82c6422

 

According to SSD review Solidata's newest SSDs are SLCs with the JMicron controller. It's not as slow as the MLC drives because SLCs are at transistor-level 2x faster with writes and the write amplification factor is typically 1/2 to 1/4 of the MLC devices due to the nature of SLC architecture.

 

Yes, the idea is similar to steadystate or MFT in that all random writes are turned into sequential. The big difference is that Flashpoint attempts to solve the problem of committing changes to disc (problem with steadystate) and operates on the OS volume (problem with MFT). And although steadystate is free, MFT is not. Also, Flashpoint uses 32MB of memory and no disk space whereas both MFT and steadystate use the disk as cache.

So yeah...if this developer can keep trucking away on Flashpoint it may turn out to be the best solution yet for SSDs with poor random write performance. The developer actually initiated the software for netbooks with the small proprietary SSDs built in that cannot be upgraded.

Err yeah, thanks for posting this! The forum wasn't loading for me for some reason and then I had to restore my system from an image due to my discovery that Flashpoint is not working correctly with 7 and cannot be uninstalled due to that problem!

The Vertex is more or less cheap enough now ($190 shipped from Newegg for 60GB) that I wouldn't bother with a cheapy drive again. But lots of cheap JMicron drives were sold in the last while and software like Flashpoint can actually make them perform very well as shown by the benches I posted earlier! Perfect for a HTPC, CarPC, SATA-I notebook, older computers, etc.

Here is a question I've wondered about for a while: since creating software that changes random writes into sequential writes is obviously possible and works well (MFT and now flashpoint), why doesn't Microsoft build this into the operating system? They have already produced software that does this very thing (Steadystate) but for a different reason. Why in the world wouldn't MS build this functionality right into Windows 7?

 

because when win7's around, it won't matter anymore? half a year, another price drop to half. one year, another. good working ssd's will be around by then everywhere (just look at the current state. indilinx, intel, samsung, mtron all create perfect working ssd's (except for maybe bugs in firmware or hardware)).



i can't wait for the next intel update, as stated by anandtech. as the vertex now gets very close i'm ready to get another bomb dropped by intel go intel, go!! drop the price to half again, increase the storage by two, make it even faster.. *boom de yada boom de yada i love the whole world*

 

I can speculate all I want but I can tell you this, Intel is rocking at the moment. I think the Nehalem mainstream release at the end of this year will be much of a storm as Nehalem was to servers. Currently the lacking part for Intel products are Integrated Graphics and Turbo Memory.

Nehalem is about solving the I/O bottleneck. What do you think would happen if "Braidwood", which is currently only known as 3rd generation Turbo Memory was really the next generation SSD controller integrated into the PCH itself?? So the only difference between the Turbo Memory version and the SSD version are the chips itself? It would allow pricing to potentially plummet.

I'm totally speculating here yes but that is the future of the SSD controllers, integrated into PCH.

 

i hope that we soon get completely away from sata in notebooks. pcie works very well in netbooks for ssd's as well, so it's not like it's not doable. maybe we will all have 1gb/s disks then

 

Left: ASUS flash_con using modified pci-e type connector containing USB, PATA and SATA pins
Right: SATA I/O pins of interest marked in red,green and blue on an ICH8-M southbridge chip. Want to trace these to accessible points on the systemboard

It is possible. in theory, to run 4 SATA I/O pins to a 4mm runcore SATA SSD (designed for eee PCs..) plus the 3.3V and GND to run it. It could sit somewhere in the chassis where there is room (eg: spare pci-e slot, pc-card slot)

Many of the ICH8-M/ICH9-M chipsets have 3 or more sata controllers on the southbridge, with only 1 (HDD) or 2 (HDD+optical drive) of the ports having connectors. Meaning if someone can run wires directly off the southbridge OR find points on the systemboard to tap off, can run more SATA storage. See here for an idea of what too look for, if you are impartial to soldering onto your systemboard to do this.