Kingston HyperX 3K Series 240GB SSD Review @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,601
Kingston HyperX 3K Series 240GB SSD Review - Kingston has realized tremendous success with their original HyperX line of premium SSDs for enthusiasts. With the SSD market becoming more value-driven, Kingston has responded with a new HyperX 3K model. Today we will look at the performance of the newest member of the Kingston family.
 
Last edited:
Good article, i liked the read. and i bitched about the terrible graphs without reading your post so sorry
 
Great to see more SSD reviews here! I sincerely hope that this is something you will be focusing even more on going forward. Very thorough review and a good read. Quite an interesting product. I hope the TRIM concerns will be sorted as promised.
 
Picked one of these up a few weeks ago for $180 shipped from Amazon and slapped it in my laptop. Sub-10s boot times make me feel warm and fuzzy inside.
 
I have the HyperX model that came before this one and it has been an outstanding and fast drive. This one looks to be just about as good.
 
How is $138 a super low price. There's a dozen 120GB SSDs on Newegg under $100.

Allen Malventano at PC Perspectives, a true SSD guru, considers the Samsung 830 series one of the two best SSDs in the consumer market (the other is the Intel 520).

A Samsung 830 128 GB goes for $129.00, more GB for less $$. How is this not a better deal?
 
The Samsung 830 is a good drive, no doubt about it. A difference of 9 dollars is not that large, and we have yet to compare the performance of the 830 in steady state.
We are really speaking of the drives that are in the testing in the article, which are among the top enthusiast solutions. We will be covering more as we go of course. I stand by the HyperX being a great deal at this price point.

Many of the sub $100 SSDs are older generation controllers and less desirable NAND combinations. These SSDs aren't comparable to this generation of SSDs, or the type of NAND used. Just because an i7-920 is cheaper than a i7-3930K......

I would like to extend an apology for the sub-par graphs. These definitely are not my strong suit. We did do some changing from the previous review. They are being amended.
 
The 830 is pretty fantastic, and I own several, but if you're looking at a SF2281... the HyperX is a good deal. Period. Kingston always has been slightly (to a lot) more expensive than other drives, but it wasn't until recently that Kingston started appealing more to enthusiasts.

A few years back, they were simply rebranding Intel drives, with some rebranded J.Micron controllers thrown in for good measure. When the 5K launched, they used their close relationship with Intel to buy NAND from Intel's private stock (so the story goes). I'm not 100% convinced that they're not using the same flash in the 3K that the Intel 330 currently uses -- supposedly rated for 3K, but Intel's Cherryville Jr. still seems to be sticking to 5K.

If you need better steady state random write performance, the HyperX is going to win every time over the 830, but the 830 is probably the better drive outside of specific applications. The 830 has strong TRIM and GC as well, something no 2281 can match at the moment. SF does the preponderance of it's GC on writes, meaning it has to get pretty bad before it gets better, with sustained write speeds all over the map. You'd have a hard time telling the difference between them in most usage scenarios, but if sequential writes are important, the 830 will wipe the floor with just about anything. The HyperX is faster on reads under some conditions, but the 830 has the edge on 4K random read performance.

The 830 has the edge on idle wattage (under .4w) but uses more on load (by a lot, like the Corsair). And for mobile applications that need 7mm chassis, the 830 is one of the few ways to go. Those are pretty rare though, and the installation kit is boss on the HyperX and 3K.
 
Last edited:
SF has several GC 'triggers' that will turn on the GC function. One is a certain capacity of fill, I'm not sure if the 830 shares that characteristic.
The 830 does have much more aggressive GC, from the tests that I have seen. I would like to test the 830 in RAID (or trim-free environment) in steady state just out of curiosity.
The 830 is a sequential writing banshee, that is for sure, and does NOT share the throttling characteristics that SF will use under certain scenarios (which most users will never see).
I also would like to test the 830 to see if it suffers read degradation in steady state... I'm hyping myself up for a 830 now LOL.
 
The biggest problem for me is trying to determine that price/performance/longevity formula... This drive looks great, but will it last as long as an Intel 520 or Samsung 830? I swore off of SSD's after a debockle with an OCZ Vertex Turbo as a primary drive going to crap on me in 2010, but am tempted to give it another go. Price still not being < $1.00/GB for most of the better known drives keeps pushing me back though. What's the word on Kingston SSDs?
 
Kingston has a solid rep imo.
The majority of SSD failures aren't due to the NAND, it is other components.
Also, in real world tests guys have written over a Petabyte to 3k cycle NAND!
Remember, the NAND is spec'd at a 3 percent failure rate.
They expect that 3% of the nand will fail at around 3,000 P/E cycles, but the device can withstand multiples of that in failures and still function fine.
Intel and other NAND manufacturers are just over-zealous with the endurance specifications.
I would have no reservations using this drive whatsoever.
 
comador said:
The biggest problem for me is trying to determine that price/performance/longevity formula... This drive looks great, but will it last as long as an Intel 520 or Samsung 830? I swore off of SSD's after a debockle with an OCZ Vertex Turbo as a primary drive going to crap on me in 2010, but am tempted to give it another go. Price still not being < $1.00/GB for most of the better known drives keeps pushing me back though. What's the word on Kingston SSDs?
Click to expand...

There is nothing wrong with a 120GB SF2281 drive running Windows 7 x64. Have a 1 or 2TB disk drive too. Move games back and fore between them (like a cache) when you want to take advantage of the SSD. Do a dual boot too. Put your downloads onto the 2TB disk drive.

As long as the OS does not do too much page-file action, then the drive should last a long time. Myself, I have been using a Corsair Force 3 120GB that way for a SMART of 8000+ Power On Hours with only 886GB Lifetime Writes. That is less than 8 complete drive erases in more than 6 months. Considering that maximum complete drive erases is around 3000, it should last longer than 10 years.
 
Understandably there can only be so many drives thrown in the review, but it's still fair and important to consider others that weren't in the review.

You can get the 256 GB ADATA Premier Pro for $175, comes with a drive sled and with Acronis http://www.newegg.com/Product/Product.aspx?Item=N82E16820211603
This drive boasts up to 550 MB/s read and 530 MB/s write.

What is interesting is when you consider the longevity of the drive. The HyperX 3K 120 GB at $140 vs the ADATA 256 GB at $175 - given the same usage would not the ADATA last *much* longer because it has more free area and can spread the writes/deletes out over that free area? And the $ / GB ratio is ridiculously better.

I've also seen the Crucial M4 256 GB on sale for $180 at MicroCenter - and that was not a "walk-in" sale.

SSD prices are definitely in a serious downward trend, so if you're going to buy make sure to look around. I personally don't think the Kingston drive is worth the higher $ / GB that it's asking. Kingston is marking up based on its name recognition.
 
Yo, what's up with the multiple shades of red? These graphs are useles! Sorry for the harsh criticism.
 
were all of the drives tested the same capacity (240 or 256)? i didnt see it noted anywhere what the other drives were.

also the numbers for the m4 seem off. i only have a 128gb model but i am getting faster speeds then you seem to be getting on as-ssd, especially in the copy benchmark.

as-copy-benchM4-CT128M4SSD2711201210-42-13PM.png


compared to:

iso: 294
programs: 153
games: 221

i am getting twice the speed of your drive in programs and games.
 
ChrisRam said:
The Samsung 830 128GB has really bad latency when writing data.
Click to expand...

ChrisRam said:
The Samsung 830 128GB has really bad latency when writing data.
Click to expand...

Well, if you're talking random writes, it's not going to scale with QD, so latency will just rise without an increase in throughput.

But it's not like that's the only reason to get one...

I have almost 2.8 million GiB written to my 256GB 830. That's almost 2.7 binary petabytes, AKA a metric shitload of writes:

attachment.php


Thats almost 12,000 PE cycles. Not too bad.

That's 2,784,523 binary GiB by the way, or almost 3 quadrillion bytes.
 
Last edited:
Glad to see a mention of the drive height, even if it was brief and with no discussion. 9.5mm drive height means this drive is worthless to anyone trying to upgrade one of the newer thin/light laptops that need 7mm height drives.

I wish the drive manufacturers would come up with a consistent description for these drive heights. Some SSD manufacturers put 7mm right in the name or description, some even list the height as some weird fraction of an inch! No matter how it is marked, this really should become part of the normal marketing product description because so many laptops are now limited to 7mm height drives.

For that matter, why are these SSDs even coming in 9.5mm heights? They can pretty much all fit into a 7mm case, and it is really easy to add a spacer if the drive would rattle around in a larger drive bay. A handful of 7mm SSDs come with small snap-on plastic spacers to make them fit snugly in 9.5mm drive bays, but you simply can't cram the 9.5mm drives into the 7mm bay. So why the hell are they stuck on using the older/larger 9.5mm thickness? Especially when using an SSD in a desktop, there is no reason for the larger size.
 
ckryan said:
Well, if you're talking random writes, it's not going to scale with QD, so latency will just rise without an increase in throughput.

But it's not like that's the only reason to get one...

I have almost 2.8 million GiB written to my 256GB 830. That's almost 2.7 binary petabytes, AKA a metric shitload of writes:

attachment.php


Thats almost 12,000 PE cycles. Not too bad.

That's 2,784,523 binary GiB by the way, or almost 3 quadrillion bytes.
Click to expand...

How many MB/s is that?:p
 
DW-UK said:
How many MB/s is that?:p
Click to expand...

2.8 billion-ish! Typing that many digits in a calculator gets old.

Now that I think about it, the first solid state storage device I owned was a PS1 memory card. I remember seeing that a PS2 men card was good for 1M erases, with only 8MB.

It's writing 25,000 GIB a day.
 
Last edited:
Nice write up! I thought at first it was just me not being able to tell the color differences in the graph :D Looking forward to updates graphs!
 
Nice results Ckryan, this is actually one of the tests I am referring to with guys pounding away on SSDs with write data! that particular SSD is a legend :)

For that matter, why are these SSDs even coming in 9.5mm heights? They can pretty much all fit into a 7mm case, and it is really easy to add a spacer if the drive would rattle around in a larger drive bay. A handful of 7mm SSDs come with small snap-on plastic spacers to make them fit snugly in 9.5mm drive bays, but you simply can't cram the 9.5mm drives into the 7mm bay. So why the hell are they stuck on using the older/larger 9.5mm thickness? Especially when using an SSD in a desktop, there is no reason for the larger size.
Click to expand...

You make a good and relevant point. There really isnt a clear answer on that!

were all of the drives tested the same capacity (240 or 256)
Click to expand...

yes they are all the same capacity. I believe that the 128 M4 is a bit faster than the 256 :) The 128 is great in some regards. The different capacities will handle different depending upon nand config, etc.
The only thing that gives me pause with the M4 is the mixed read/write workloads. I would be interested to run that capacity through the steady state testing.

256 GB ADATA Premier Pro for $175
Click to expand...

That is a zero-provision drive, which means there is no spare area. This is included on SSDs to smooth out performance over the long run, especially in steady state. Most drives have a minimum of 7%. Some have more. This drive also has lower quality flash with Async NAND. These two issues combined, imo, would relate to poor performance over an extended period of time.
I will admit I did not see this needle in the SSD haystack, but I can see why the price is so low.

From newegg price histories:
http://camelegg.com/product/N82E16820211603

Current $174.99 Jun 28, 2012
Highest * $309.99 Apr 11, 2012
Lowest * $174.99 Jun 28, 2012
Average + $241.41

I believe the new low price may be due to some scathing reviews on this SSD.
 
Last edited:
comador said:
The biggest problem for me is trying to determine that price/performance/longevity formula... This drive looks great, but will it last as long as an Intel 520 or Samsung 830? I swore off of SSD's after a debockle with an OCZ Vertex Turbo as a primary drive going to crap on me in 2010, but am tempted to give it another go. Price still not being < $1.00/GB for most of the better known drives keeps pushing me back though. What's the word on Kingston SSDs?
Click to expand...

The SF-2281 controller depends on compression to gain much of its performance (as you probably know). Myself, I believe that the SSD market will be moving away from that, internal compression algorithm, due to its disadvantages and the fact that a faster controller will be more limited by the SATA bus transfer anyway. The problem with internal compression is that data which has already been compressed, within your OS filing system, can’t be compressed any further, so it is slower on SandForce . Additionally, as the drive gets bigger, you are likely to have bigger media files and game data files taking up most of the space, which are compressed already. Then there are files, which you can have the OS compress, such as documents. Reading these OS compressed files on the non-compress SSD controllers will give you the full benefit of that. Basically, SandForce cheated with the performance numbers by hoodwinking the OS filing system compression opportunity.
 
Sequential read\writes for 1 drive are just about below par for my Raid 0 SSD setup. I think Kyle's review is seriously making me want to upgrade, especially since I do quite a bit of audio production and Ableton Live 8.4 (Finally implementing 64bit support) is right around the bend.
 
Compression has its pros and cons. One pro is the reduced amount of actual host writes that hit the NAND. This results in lower write amplification and higher endurance.
Another pro is the fact that hardware level compression does not incur host system clock cycle overhead. NTFS compression will come at the cost of some CPU overhead.
The cons are many though, and the variability in performance is certainly chief among them. Another is the increased latency at low QD that we observe commonly with SF SSDs.
 
Hugh_Briggs said:
Compression has its pros and cons. One pro is the reduced amount of actual host writes that hit the NAND. This results in lower write amplification and higher endurance.
Another pro is the fact that hardware level compression does not incur host system clock cycle overhead. NTFS compression will come at the cost of some CPU overhead.
The cons are many though, and the variability in performance is certainly chief among them. Another is the increased latency at low QD that we observe commonly with SF SSDs.
Click to expand...

Really, SF's compression is more like a bonus than something that should be relied upon for keeping WA low. SF uses cheap hashes on blocks because of both latency and controller horsepower. It's likely future iterations with beefier controllers could really crank up the compression, but it's hard to predict whether latency would take a hit.
 
I believe the SF controllers have the lowest WA. relying upon it is another subject of course, esp since the read degradation makes me wonder if there is some sort of 'unpacking' necessary to read its own compressed data. There is definitely something going on there, and I think compression is the culprit.
 
I see the m4 is at the bottom of the perf charts ... is that why it's been so cheap recently? (As low as $379 for 512GB.)
 
Hugh_Briggs said:
I believe the SF controllers have the lowest WA. relying upon it is another subject of course, esp since the read degradation makes me wonder if there is some sort of 'unpacking' necessary to read its own compressed data. There is definitely something going on there, and I think compression is the culprit.
Click to expand...

See, I think clt he compression adds overhead with hard to compress data, pushing actual WA over one even if raw/host writes are less than 1. Who knows?
 
Hugh_Briggs said:
That is a zero-provision drive, which means there is no spare area. This is included on SSDs to smooth out performance over the long run, especially in steady state. Most drives have a minimum of 7%. Some have more.
Click to expand...
I see. If you format the drive and leave some capacity unformatted does the unformatted space work as spare area? I thought even free partitioned space was used as spare area of sorts.

In light of scathing reviews I'd switch over the the Crucial M4 anyway, but I'm still curious about whether a drive's unused space will be used by the firmware as "spare area".

Edit: by the way - DIDN'T KNOW ABOUT CAMELEGG - THANKS!
 
yes, if you leave the area unpartitioned the free space will work as OP on SSDs. it depends upon the controller and firmware as to how well it uses it, but they all do it to my knowledge.
After roughly 47% OP you will lose steady state performance and endurance scaling (yes it makes them live longer too), but imo the point of diminishing returns is around 25%. this includes the OP that is already set aside by the manufacturer themselves, so add that to your calculations. There are tools to manually resize the SSD yourself if you want to go down the rabbit hole.
BUT you must do a secure erase, then leave the unpartitoned area, or it doesn't seem to work as well imo. So simply shrinking the available partition will not work as well in some cases. So SE, then leave unpartitioned space.
I personally leave a 20% OP on my boot drives.

Yeah camelegg is cool stuff, i did get zinged on pricing by the guys the last go round, so figured I would do some more research :)
 
I just read the review. I found the red graphs very difficult to read.

I read the disclaimer to be patient. Ok, fine. Wait, it's 2 days later. ??
I imagine that a large percentage of your readers won't go back to read the article 4-5 days later to see the "fixed" graphs.
 
J Macker said:
I just read the review. I found the red graphs very difficult to read.

I read the disclaimer to be patient. Ok, fine. Wait, it's 2 days later. ??
I imagine that a large percentage of your readers won't go back to read the article 4-5 days later to see the "fixed" graphs.
Click to expand...

The data on the graphs is not incorrect, simple formatting changes that will be made. I do not have any expectations of you going back to see prettier graphs.

SHOW ME what you want the graphs to look like please.
 
You must log in or register to reply here.
Back
Top