Kingston V-Series - First look benchmarks and pics

Hi all,

Wanted to post this as I notice that many are waiting for some info on Kingston's new "value" series SSD's. I picked up the SNV125-S2BD/64GB yesterday to install in an HP 2140(HD) netbook. I actually wanted the notebook kit (S2BN), but they didn't have any on hand, so I just got the Desktop version (S2BD). Before I installed it, I decided to bench it first with some common tools (on my desktop system). While at it, I also disassembled it and took some pics (no warranty "seals" anywhere on the drive). Here's what I found - first, the pics:

Note: sorry for the slight color differences - the case is a single color. I should have set the camera's white balance manually.

The top and bottom:





Top cover taken off:




PC Board Top (note the Toshiba 4GBx8 MLC flash chips )




PC Board Bottom (note the *Toshiba* Flash Controller):




A closer look at the Flash controller:





So the Flash controller is manufactured by Toshiba, but according to Kingston, this SSD uses a JMicron controller with firmware modified by Toshiba and JMicron. The Toshiba part number has a 602 in it, so it very plausible that Toshiba is the actual FAB for JMicron or is manufacturing under license from JMicron. Either way, Kingston states that the firmware is custom engineered to eliminate the studdering associated with the JM602 series controllers.

I'll post some benchmarks in the next message.

- nn6o
​

.

 

Here are some benchmark numbers from ATTO, CrystalDiskMark, and HD Tune. All tests were run in Safe Mode (though the stats in normal mode looked the same).


First, the ATTO numbers:




CrystalDiskMark:




And HD Tune:




Added: IOMeter using Jackboot's 4K Random Write config:
(.CSV results attached in ZIP)




It's currently installed in an HP Mini 2140(HD) Netbook w/2GB RAM and WinXP Pro (SP3). Haven't noticed any studdering, though this is my first SSD and the Netbook isn't necessarily a powerhouse.

Hope this info helps out anyone considering this drive!

- nn6o
​

.

 

Nice stats and review + rep to yah. Also welcome to NBR! That is a pretty sweet budget MLC drive!

 

The 4 Kb writes are pretty good for a Jmicron drive.

 

Thanks for the report!

The V-series does come in as quite a bargain. For $125~ you get a 64gb SSD, a mounting bracket(or a enclosure) and a copy of a cloning program, and those itself sans the SSD is at least $20-$30 retail.

 

thanks for that info! I'm waiting for mine (notebook bundle), and was really worried about the crystalmark 4k. I'll run some IOMeter tests on it to see how it is but really, i just want a good user experience and not sit there and run benches on it all day.

 

Thanks for the post!

I'm surprised to see the Toshiba branded controller

Also, I *believe* that the original JMF602B could only use Samsung nand (and the same goes for the Indilinx). The fact that this drive uses Toshiba branded nand is very interesting.

Since the most important aspect of benchmarking this drive is 4K random writes (the cause of the JMF602B and earlier stuttering), a proper test with IOMeter is really important.

Do you think you could use IOMeter and test 100% random 4KB writes?

I think the most useful settings are similar to what Anand uses in his tests because they approximate a realistic but demanding usage scenario. Your test results will then be directly comparable to his benchmarks for some of the most popular drives on the market (X25-m and Indilinx drives (Vertex), among others). SanDisk also independently recommended a very similar test configuration in their vRPM white paper.

Here's the settings I've been using (a mashup of Anand and SanDisk):

4KB
100% write
100% random
~10 GB section of disk
4 outstanding I/Os (queue depth)
3 minutes test length

Note that benchmarks like Crystal Disk Mark and ATTO use a tiny portion of the disk (100MB and 256MB respectively) which greatly inflates small random write performance numbers. The length of the test for both benchmarks is also far too short which can inflate the results because of caching (we want to know what the sustained performance is, not peak). Crystal Disk Mark only indicates the *best* test result (out of 5 runs) and has a queue depth of only ONE! And if I understand correctly, ATTO tests sequential only...compare ATTO 4KB write with CDM 4KB write and you should be thoroughly confused - 28MB/s write vs. 2.0 MB/s write - how can such disparate results have any real meaning?

The original JMF602B got about SIX IOPS for 4KB random write (compare to I think 34,000 for the Intel and like 17,000 for the Vertex, IIRC). We need to know 4KB IOPS for Kingston's V-series.

To illustrate my points, look at the CDM of my JMF602B based SSD. According to this benchmark, the "new and improved" Kingston drive is basically the exact same as the "old and POS" JMF602B based drives. In fact, this benchmark suggests that the new Kingston drive is actually very slightly SLOWER. I'd be surprised if this is the case.



Next, look at this ATTO of this same drive with caching software installed (zFlashpoint). My old POS drive shows that 4KB writes are faster than the new Kingston version - which I cannot believe is true - because of a short-lived cache boost.



In short, ATTO and CDM are pretty useless for accurate and meaningful SSD benchmarks.

EDIT:

For comparison with nn60's IOMeter results, here is my IOMeter results using the same config file on a JMF602B-based drive:

 

BTW, here is an IOM test config that I created based on Anand and SanDisk's recommendations. Note that the 4KB test config file going around from the OCZ forums is made to inflate results.

 

i would love to see Jackboot's test and also a HD Tune Random Access test (needs Pro version)

 

I second that.

 

Hey all,

I originally wanted to include IOMeter tests, but had no clue how to set it up. Thanks to Jackboot, I used his 4K config file I added the results to the original post. At 17, 3x better than "stock" JM602B, but still far behind Intel and Indilinx. Of course, this is targeted as a "value" drive - something to keep in consideration!

Sorry, don't have HD Tune Pro , so we'll have to wait for someone else to post it's Random Access Test results.

For reference, I also played with a 32K and 64K offset - didn't seem to make any real difference compared to the test numbers already posted - all were within 5% tolerance.

 

Thanks a tonne for the information! I edited my last post to include a "stock" JMF602B IOMeter result.

But man your results are really disappointing...

A regular 5400rpm notebook HDD will get 50 - 60 IOPS at 4KB random write. So Kingston's claim that this SSD which is "50% faster than a HDD" is just plain marketing BS.

With such low IOPS performance I think that this new Kingston SSD will still be at risk for stuttering during certain tasks.

It could still be a descent budget drive, but the price is still way too high in my opinion to not spend a bit more to get a drive that will absolutely blow it away. Indilinx based drives can be had for around $190 USD shipped and have something like 750x better IOPS performance - which is what really matters for perceived system responsiveness.

So the decision in terms of ~60GB SSDS:

Kingston V-series for about $135 shipped
vs.
Indilinx drives (Falcon, UltraDrive ME, Vertex) that are 750x faster for $55 more...

Not a hard decision in my opinion.

With that said, I'm tapping out at about $150 for an SSD. Once I can get Indilinx performance for that price, I'm in. We should be there in a month or two. JMicron's upcoming controller (JMF612) will likely drive SSD prices WAAAAY down as all sorts of 3rd tier manufacturers get into the market since they will finally be able to use ANY nand flash modules (not just Samsung, Intel, or Toshiba). Look at the USB flash drive market to get an idea of what kind of pricing can be expected (i.e., 60GB for $100 to start things off).

PS: you should post the IOPS result as a NewEgg review to let others know to stear clear.

 

Eh... The controller chips cost on order of $5 per SSD, the new JMicron isn't going to dramatically change anything. The only thing that will drive the price of SSDs lower is cheaper flash, and that's coming with the finer process technologies.

 

disappointing results, looks like i'll keep waiting for prices to drop and speeds to increase

thanks for the benches!

 

That's exactly what I said...

JMicron's upcoming controller will allow the use of ANY MFR's flash - cheap flash - not just flash from Samsung, Intel, or Toshiba. It's not the controller cost that is keeping SSDs priced high (but for the record, the JMF602B is ~$10 and the Indilinx is about ~$15, no controller sells for $5 AFAIK) but rather the limitation of all current controllers only accepting one of the aforementioned MFRs flash.

 

Do these results look better?

I used your random 4k config file.





Tested on my X61s (L7500, 4GB RAM) with Windows 7 x64 RC Build 7100, fresh install.

 

Those are GOOD results! That IO performance is about 3x better than a regular HDD.

Those are with the Kingston V-series SSD?

I don't understand why there is such a huge discrepancy between you and nn60's results. Can you offer any insight?

 

yep, 64GB notebook bundle from Newegg. Came out to $146 after taxes and all. Pretty happy with the results, to be honest.

I did a few of the OCZ tweaks like the lastdatestamp, disabling indexing/prefetching. Pagefile might have been at default (4GB?), i did eventually change that to 500MB since I don't game on this and don't need it big.

I just did various updates (Windows Update, Avira AV, downloading Windows Live components, installing firefox) and no stuttering. Very happy with this.

I also installed Photoshop CS4 while still using the system for other things, and it didn't stutter.

 

With IOPS at ~200 you will never experience stuttering

And you are not using any sort of disk caching program, like zFlashpoint?

Those tweaks you mentioned won't inflate a benchmark score - the reason OCZ ever recommended them is to try to limit stuttering on their JMicron-based drives. Most of those tweaks won't change your user experience, but I personally feel that prefetch is best left enabled as it should increase performance (RAM is faster than SSD and you have 4GB). You aren't at risk of stuttering with the level of performance this drive seems to be providing you so there is no point in tweaks that will decrease your performance elsewhere.

You should run the IOMeter benchmark once more - but this time after doing a safe boot so that your startup programs etc don't run and potentially affect the benchmark.

Also, if you are so inclined, it would be interesting to see if you have a different controller on board compared to nn60. I read one reviewer on NewEgg that claims that Kingston support told him that they are using 2 separate controllers and that which one you get is luck of the draw. I find this hard to believe, but it would certainly explain why you and nn60 are getting such different results.

 

zFlashpoint doesn't support x64 as of yet, from what I have understood. Nope, nothing like that, cannot even set the write caching on the drive.

I'll try with prefetch enabled, should I enable Superfetch also? I remember there being 2 entries for fetching in the registry.

You want me to go to safe mode and run the bench?

At some point, i might open it up but i have limited time this weekend to do that.

 

There are lots of opinions on prefetch and superfetch so take mine with a grain of salt...I leave them both on. They are designed to utilise all of your available memory - and you've paid for 4GB so you should use it. RAM is faster than SSD, and you have no need to worry about stuttering with your system from the looks of it so I'd say leave them both on.

If you run the bench in safe mode then you can be sure that none of your programs are interfering with the SSD's capabilities and you can get a more accurate picture of what it can do.

I can't think of a good reason why you are getting 20x the performance of nn60 so it would definitely answer a major question if you verified what controller your SSD is using (in light of the fact that there is some rumour going around that the v-series uses at least 2 different controllers). So you've got my vote to open er up and let us know what's inside

 

I'll pop it open one day, not sure when
Not feeling like running anymore benches on this thing though, it's half full now, and working like a clock. I just installed XP in a VM (Vmware Workstation) while surfing using Firefox (cache on C:\, no ramdisk) with about 5 tabs active, no stuttering. AV running in the background, and all, same thing I used to do with my HDD installed.

 

just installed my v-series 64GB ordered from newegg (notebook kit)
ran jackboot's random4k config file, fresh 32-bit windows7RC install, IOPS 15.99
running a lenovo s10e with 2GB RAM, no tweaks

 

What is the IOPS with the old hard drive? Just wondering if the Lenovo is the bottleneck.

 

Don't worry about benchmarks, worry if you have problems/dissatisfaction with using the drive.
Did you buy it to benchmark it or to use it?

 

Hey, who wants to check their V-series drive to find out what the firmware revision number is? I got my hands on an updated firmware from JMicron and I'd like to find out if it is newer, older, or the same as the one being used in the V-series.

 

I don't mean to revive a long dead thread, but I posted some results (and a lot more questions) somewhat relating to this post on the desktop review forums. HERE

In addition to that... it's probably of little use now, but since noone ever replied to Jackboots last post, I was just going to mention that Crystal Disk reports my firmware as C091126A.

 

Hi there,

I recently purchased a 64GB Kingston V-series SSD as detailed above for my netbook, but the benchmark figures I'm seeing are well down on the results that nn60 has posted here, in particular for the Seq read and write my figures were coming back at around 78MB and 55MB respectively... (screenshot attached)

AHCI isn't supported on my machine (although from what I've read it won't boost performance significantly, certianly not for sequential access)

I'm running Win7, I've done a fresh install so the SSD's offset is 1024KB. The machine is an Acer aspire One AO751h.

If anyone has any ideas as to why I'd be very grateful, I have queried Kingston technical support who have tried to be helpful but sadly without success!

Thanks for taking time to read my post,

Spadger

 

Can't anybody help this guy? I know we have a lot of very smart members in the Hardware forum.

 

Spadger,

Are these results from "safe mode", regular boot, or are they roughly the same in both?

I just Googled (hehe, Topeka'd, it's April 1st) the AO751H and its a lower powered netbook. The Z520 CPU is single-core and running 1.3GHz - it's possible that you've hit a hardware performance limit.

Another hit...check this thread:

http://forum.notebookreview.com/showthread.php?t=392238&page=2


Scroll down to runtohell121's reply - he posted his benchmarks of the AO751H with a Kingston 64GB SSDNow drive and they basically match your results (actually a little slower). So it seems that the performance is limited by the hardware.

Sorry, I know that's not what you wanted to hear...


BTW, as an update, I returned the Kingston drive I originally benchmarked, but picked-up another one a few months ago for $89. Firmware is newer (sorry, don't have the actual version handy), but the benchmark numbers were more or less the same. It's currently in my HP Mini-2140 netbook running Win 7 Pro. Works like a charm, no stuttering, and infinitely quiet. I'd like to use it in one of my larger notebooks, but 64GB just won't cut it! Starting to look at some of the Indilinx based SSD's, but still too much quid.

C'ya!

- nn6o

 

Hey Grenade01,

The model you have is a SVN425 - it's in the same SSDNow "V-Series" series of drives, but the specs would indicate that it uses a different controller. The SVN425 notebook upgrade was on sale this week (ends today) at Fry's for $279 - I was considering picking one up just to bench, but decided not to.

I just Topeka'd the SVN425 and found a review of the 128GB version here:

http://www.pureoverclock.com/review.php?id=945


It uses a Toshiba branded JMF618 (so it's still JMicron based), but this controller is faster (specs say 200MB reads/160MB writes) and supports TRIM. Also includes 64MB cache. Head over and take a look at the review. Overall, decent read performance, but, as expected for JMicron based drives, underwhelming 4K write performance.


- nn6o