Please explain Samsung SSD choice?

[Anthony2816]

It seems everywhere I look I see reviews and opinions saying that the Samsung 840 Evo is the best SSD. But the Kingston HyperX seems to have better stats, as far as I can see:

Samsung:
Sustained Sequential Read
540MB/s
Sustained Sequential Write
410MBps
4KB Random Read
Up to 94,000 IOPS
4KB Random Write
Up to 35,000 IOPS
MTBF
1,500,000 hours

Kingston HyperX:
Max Sequential Read
Up to 555 MB/s
Max Sequential Write
Up to 510 MB/s
4KB Random Read
Up to 85,000 IOPS
4KB Random Write
Up to 73,000 IOPS
MTBF
1,000,000 hours

So what am I missing? The Kingston HyperX seems mostly faster, especially for writes. Its MTBF is obviously lower, but a million hours is 114 years, so that suggests they'll outlive me.

Please educate me on why the Samsung is a better choice.

 

[RU482]

Evo isn't the best, Pro is

 

[creed3020]

I also would not trust any Kingston SSD. They tinker around too much changing the NAND.

If you want top tier performance consider the Samsung 840 Pro or Evo, or SanDisk Extreme II or Seagate 600. You couldn't really go wrong with any of those. The differences at that level is measurable but hard to actually notice.

http://www.anandtech.com/bench/SSD/730

 

[Mand]

There was the time that Kingston changed its NAND source, on the same model, without telling anyone, and the performance went down. Brand can sometimes matter more than the specs.

 

[rsutoratosu]

First line, there's your problem. If you look at the link Creed3020 offered above, you see only 1 ex manufacture used sandforce on that list, and they were near bankrupt and got sold to toshiba.

http://www.kingston.com/en/ssd/hyperx

- SandForce controller technology


My biggest gripe is compressible data - almost all data format are near incompressible now, adobe docs, jpgs, etc. Using SF, it's wear on incompressible data

SandForce's SF-2281 controller handles compressible data very effectively, and the preceding page illustrates its best-case behavior nicely. However, the controller vendor openly admits that performance drops when it comes to addressing incompressible information.



Off wiki
-

The first generation SandForce controllers do not fully comply with the SATA specification[23] and are incompatible with the Intel Haswell platform.[24][25]

After the introduction of the SF-2000 series controller, some customers using drives with that controller reported issues such as BSOD and freezing. In early June 2011, Corsair Memory issued a recall on the 120 GB Force 3 with specific serial numbers, but not on any other Force 3 drive with a SandForce SF-2000 controller, therefore that recall does not appear to be related to the controller.[26] In October, 2011, SandForce sent out firmware updates through their manufacturing partners such as OCZ that fixed the reported issue.[27] In August 2012, a website known as Tweaktown identified an issue with SandForce-based SSDs using firmware 5.0.1 and 5.0.2 wherein TRIM support did not perform optimally when fully erasing the SSD, but also confirmed that the 5.0.3 and 5.0.4 firmware resolved the issue.[28]

In 2012, SandForce SF-2000-based drives were discovered to only include AES-128 encryption instead of the advertised AES-256 encryption. It was speculated the lower grade encryption was used to qualify for US ITAR licences which are precluded for products featuring certain levels of encryption heading for a selected list of US-ambivalent or actively unfriendly countries.[29][30] Products such as Kingston SSDNow V+200 and KC100 were re-documented to state the use of 128-bit AES encryption.[31] Intel offered refunds for affected users of Intel 520 Series SSDs until 2012-10-01, while Kingston offered exchange program to cover the cost of shipping for customers who request a swap.[32]

 

[h9826790]

If money doesn't matter, 840 Pro is better than Evo.

 

[G73S]

I had a Kingston HyperX 3K SSD last year. On the box it said Read/write speed 550/500 MB or so...it was nothing near that in my benchmarks. Only in ATTO Disk Benchmark it was because that uses compressed data to test.

Fact is, that drive sucks so much due to its sandforce chipset and only performs well using compressed data that's why it shines in ATTO Disk Benchmark but takes a huge hit getting around 170MB read/write in Crystal DiskMark

In real world usage, data is UNCOMPRESSEd so that was in fact the slowest SSD I have owned and Kingston did nothing about it when I complained they just gave me a refund

 

[dave_the_nerd]

In real world usage, data is UNCOMPRESSEd so that was in fact the slowest SSD I have owned and Kingston did nothing about it when I complained they just gave me a refund

Actually, most data is pretty compressible. The drive compresses it and that's where the speedup comes in. (It received 500MBs of data, compresses it to 200MBs, writes it in one second, and tells you it wrote 500MB/sec instead of 200MB/sec.)

It's only when you're dealing with completely random or uncompressible data (zip files, movies, or benchmarks) that you see the massive speed hit with Sandforce drives. But those are also data types which are typically not edited a lot - write once, reads over and over and over.

That said, Sandforce was the only company that tried the compression gimmick, and it worked fine. But their last controller design is a couple generations old at this point and there are plenty of controllers and SSDs on the market that get high speeds the "right" way.

 

[G73S]

Actually, most data is pretty compressible. The drive compresses it and that's where the speedup comes in. (It received 500MBs of data, compresses it to 200MBs, writes it in one second, and tells you it wrote 500MB/sec instead of 200MB/sec.)

It's only when you're dealing with completely random or uncompressible data (zip files, movies, or benchmarks) that you see the massive speed hit with Sandforce drives. But those are also data types which are typically not edited a lot - write once, reads over and over and over.

That said, Sandforce was the only company that tried the compression gimmick, and it worked fine. But their last controller design is a couple generations old at this point and there are plenty of controllers and SSDs on the market that get high speeds the "right" way.

so do you mean to say that it only does bad in benchmarks but is equally or slightly less faster than the new generation controllers?

but why does it get bad reviews then?

 

[dave_the_nerd]

so do you mean to say that it only does bad in benchmarks but is equally or slightly less faster than the new generation controllers?

but why does it get bad reviews then?

They got good and bad reviews because, for a while, like 150% of the drives on the market were Sandforce, so they were both the performance king and the bottom-tier budget option at the same time. Manufacturers liked them because they benchmarked well (if using compressible data), were cheaper to build (no DRAM cache needed) and the controllers were cheap to license. (Manufacturers didn't need to have their own controller tech.)

There were also, as has been mentioned, some teething issues with Sandforce firmware. Those problems were addressed by the end of the product cycle, but it was a gritty period. OCZ, which came first to market with the new drives and priced them very aggressively, cut a LOT of corners, and went out of business as a result. Despite promising initial reviews (fastest on the market!), their drives had very high failure and return rates.

Intel's SF-2281 based drives, however, got excellent reviews, for instance. They came out later, had newer firmware, and basked in the reflected glory of the earlier (very popular) Intel X-25M models.

They do well in benchmarks, if the benchmarks use compressible data. They were the fastest available when released (2-3 years ago now.)

They do less well in benchmarks, if the benchmarks use incompressible data, but in real-world use, they're quite capable.

I wouldn't go out and buy a SandForce drive now, because it's probably yestertech. Newer gen controllers are slightly faster for the same money, and use higher density and/or better performing NAND.

But if you already have a Sandforce drive with average-or-better NAND, it's probably fast enough that there's no point in upgrading unless you need more capacity.

 

[tweakboy]

Grab the 2 model and your set.

 

[Dman8777]

I recently put a SSD in my laptop and considered a couple different drives. I ended up with a 120 GB EVO and avoided Kingston due to the recent fiasco where they switched out the NAND and drastically lowered the performance of their drives without any warning (this was done after the reviewers had all gotten faster drives and given kingston good ratings of course). Their response to disappointed customers was basically suck it and a company like that won't get my money.

 

[blastingcap]

I hope whoever made the bonehead decision at Kingston got fired, because that one instance destroyed untold amounts of goodwill because it made a LOT of people decide against buying Kingston products of any kind after that.

 

[Dman8777]

I hope whoever made the bonehead decision at Kingston got fired, because that one instance destroyed untold amounts of goodwill because it made a LOT of people decide against buying Kingston products of any kind after that.

Yup. I can't imagine the money they saved from switching the NAND comes close to the damage done to their reputation.

 

[BonzaiDuck]

Yup. I can't imagine the money they saved from switching the NAND comes close to the damage done to their reputation.

I've so far used Patriot (Pyro), Mushkin (Chronos), and now Samsung.

I made me choices for the first two makes-models-sizes on the basis of price, availiabity, and particularly -- performance. Especially, I'd noticed that some SSDs have specs showing close to SATA-III maximum throughput for reads, but with half that speed for writes. So I avoided drives that fall down in their specs for sustained write throughput.

The Samsungs have been touted since they produced the 830 drives. I'm especially excited about migrating my accelerated-HDD to a standalone SSD, and particularly the Samsung. I'm hoping that the promotions and benchmark tests for performance of the "RAPID-Mode" feature of Magician in conjunct with the Samsung SSD is true.

"RAPID-Mode" shows some of the throughput specs as near-double what the Sammy 840's show in their specs without the ingredient of Magician.

 

[Wall Street]

Watch out for the max speed claims. Drives often perform differently when "dirty" with a lot of used sectors compared to a new empty drive. That is why it is important to read reviews instead of just looking at the headline "random 4k" read and write speeds (where each manufacturer only shows the most ideal scenario).

 

[BonzaiDuck]

Watch out for the max speed claims. Drives often perform differently when "dirty" with a lot of used sectors compared to a new empty drive. That is why it is important to read reviews instead of just looking at the headline "random 4k" read and write speeds (where each manufacturer only shows the most ideal scenario).

I'm aware of that, but the available media and sources are filled with reviews on the Sammys, more than one probably published here at anandtech.

SO far, my choices based on "specs" bear out pretty much with CrystalDiskMark. I find the "Magician" "RAPID-Mode" claims to be encouraging, and at least want to try it.

 

[Anthony2816]

Thanks for all the info, people.

 

[MongGrel]

Evos still work very well though.

 

[Cerb]

Samsung has managed to do a great job at marketing. They are good drives, but, "best," is quite a stretch.

What makes them better than Sandisk's? Crucial's? Intel's? Plextor's? If one's faster at something, by how much and how much and how much will that affect your usage? They are good, but don't go paying extra over some phantom of superiority over other good makers.

 

[BonzaiDuck]

Samsung has managed to do a great job at marketing. They are good drives, but, "best," is quite a stretch.

What makes them better than Sandisk's? Crucial's? Intel's? Plextor's? If one's faster at something, by how much and how much and how much will that affect your usage? They are good, but don't go paying extra over some phantom of superiority over other good makers.

Because I want to economize on power-consumption in my server (running 24/7), I'm doubling the size of the HDDs and replacing the separate "C: drive" boot-system disk with a Sammy 840 EVO 120GB.

I'm going to post another thread on "PCI-E Marvell SATA controllers and the native MSAHCI windows drivers. These weren't installed properly after I moved "SATA-mode" drives off my nForce controller so I could disable it (the nFarce). So I had the system go to the CD of the Marvell drivers. They installed the controllers as "SCSI storage devices." OF COURSE!! Samsung Magician threw up an error message that it couldn't communicate with the SSD. It said that "AHCI wasn't configured." I became panicky about my drivers.

I got that all sorted out -- nicely, I have to say. But while Magician seems fully functional, it still won't successfully implement "RAPID-Mode." There must be . . . . some trick to it. No matter.

SOMEBODY must have RAPID-Mode working. All I can say is this: for an old 680i motherboard and chipset with an (old) Q6600 C2Q processor, my server really rocks! Like . . . Lightning!!

 

[Cerb]

...and since you couldn't even get RAPID mode working, there's nothing Samsung is doing that any other SSD wouldn't (notwithstanding that RAPID offers fairly little in the first place, and uses perfectly good RAM to do little more than write cache). Not that they're bad, just that you'd have been equally well off with any of several others, too, as would the OP.

 

[Topweasel]

deleted

 

[johny12]

Cerb pretty much summed up what i had in mind. I mean there have been instances where Ive read the Samsung failing too. There have been mixed reviews about the sandforce all over the place but why would a company like intel choose to go with the sf controllers. In the end, the best ssd is the one that works for your usage, Samsung or SF.

 

[DigDog]

i have a kingston (a small one too) and it was dirt cheap. it performs well. so well in fact, that i cannot tell the difference between it and a better samsung.

tbh if i hadn't been broke when i bought it, i would have gone with samsung. because after all, the difference in $$ is little.