How important are 4K random performance? Are sequential numbers pointless?

None! That's why I'm interested in reviews.

 

I've tested a lot of SSDs, including the current generation SATA III SSDs. The idea that higher random performance will feel more snappy is a myth in my opinion. Simply because all the SATA III drives, including Intel 510, have high enough random performance for single user scenario's. If you don't believe that you need to experience it yourself.

A good illustration that random performance isn't all that important is the Kingston V+ 100. It's random performance is bad. It's real world performance is fine. I have that drive here too. It actually beats many drives with much higher random performance.

By the way, PCMark Vantage is far from real world. It doesn't use real applications, it only uses traces. Further more the traces are from Windows Vista. If you're running real applications under Windows 7 you may as well ignore it.

 

So then tell us, once and for all....what is the one factor that most thoroughly accounts for how "fast" an SSD feels or how "snappy" it is.

Mr. Mysterious

 

That's the thing, there is no such one factor. Real world performance is a complicated mix of random and sequential read and write speeds at different files sizes and queue depths. Latency and controller specific behavior will also influence the end result.

So since there is no such one factor, I look for real world benchmarks.

 

Give it up.
We have reached the point where the 4Ks of the SSDs are more than what every day operations use. Vertex 3, M4, Intel 510, it doesn`t matter.

And who cares about benchmarks? They are pretty much useless because HardwareHeaven and Anandtech have said and shown that the differences are so small although the scores in benchmarks varies between the drives a lot. I rather buy a reliable and proven brand that you can install and count on every single day, than get a second faster operation here and there...

 

I think the selling point now becomes longevity. How long will a particular, specific single SSD last before it bites the dust.

Because after price, speed and reliability, what else is there? lol

Mr. Mysterious

 

maybe it would be best that new benchmarks for ssd's be developed. if the present benches arent exactly reflecting performance irl or the differences are so small then these benches could be of little value as most here say. maybe its also because a lot of these benches were developed for mechanical drives.

 

Other than a few heavy users, most consumers will never find longevity an issue with SSDs. Most companies use the same few NAND chips and the same few controllers, so I wouldn't expect anything drastically different between them. In my opinion, for most notebook users should care about reliability first, battery life second, and price third. Most of the new generation SSDs have sufficient speed and I'd take a more reliable and/or lower power consuming SSD for a slight price increase when compared to a similar model.

 

Some of the Sata III ssd's are now using 25nm nand instead of the 34nm nand. From what I have read, that changes the write cycles from 5000 to 3000 (40%). I have no idea when (if) we will see this come into play, but that and the minimal speed difference I see in the different ssd's is why I chose a 34nm ssd. Once I moved from a hhd to my first ssd, I noticed a huge difference. Since then I have used a Corsair Nova with Indillinx, Intel x25-m, Intel 310 in my tablet and 510 in this laptop. All are very much faster for my work, but I can't really tell a lot of difference between them over the course of a week or month. All are much faster than the Scorpio Black I used. All.

 

As someone who owns three of these darn V+ 100 drives, I tend to agree with Phil here. I also started out thinking that random read 4K would have "snap" and sequential non-compressibles would do the best video editing. I still think these basic ideas are generally correct, and these numbers are still the most important to consider when buying an SSD. However, just focusing on the numbers and insisting that you must have good randoms for a snappy system isn't true for the reasons that Phil mentioned. I have firsthand evidence that the V +100 sucks at benchmarks, but it's plenty fast.

Ironically, it's for the same reasons you stated. You don't care that Office opens up in 4 vs. 5 seconds or boot happens a second faster. That's precisely why the numbers in drives 2010 and newer don't matter so much. Any of the reputable drives are fast for most people.

Between Intel X25-M, Vertex 2 and V+100, honestly they all feel about the same for most stuff. Only the RAID-0 felt like an actual upgrade and that speed is limited to Adobe Premiere. iTunes is still slow because it's crap software.

My most relevant statistic is Anandtech's Heavy Storage Bench. High performers of that test will not tell you which drive is the fastest for all scenarios. But it does tell you with reasonable certainty which drive will not have some huge bottleneck in one critical form of computing.

That's the primary impetus for upgrades, at least, to me: to remove bottlenecks. Driving 90mph on the freeway doesn't make me happier than driving 80mph. But driving 15mph on the freeway makes me want to punch people in the face. As long as I can get rid of those, I'd be happy to go the speed limit.

Ultimately, I think we all agree on this. Heavy Storage Bench is the best tester I've seen so far. Name a drive that scores high, but suffers huge bottlenecks? If you've got a better test, name it.

 

If that's true for most ppl, then I'm exactly reversed. Even if a Vertex is at 3% failures and Intel is at 0.5% failures, Vertex is 6x more likely to fail but still highly unlikely overall. Just by not having moving parts makes them more reliable than HDDs in general, IMO. I care about price per GB first, then speed for my particular applications and lastly, reliability.

 

I agree with you there on that last part.

But I want to say that just because reliability is last does NOT make it not important or less important. A company has to be rock-solid pretty much before I buy from them.

Mr. Mysterious

 

The top brands are fairly close in price, especially when you do some comparison shopping. The real speed you will see and feel is very close. Most people can not notice a real difference. Unless you live and die by the benchmark. So just stay away from the brands that have major issues, and I think use your requirements to pick a ssd. This is not like picking between a C2D and Sandy Bridge in a cpu.

 

I don't think prices are similar at all. I got three V+100 96GB at $100 each after rebate. Nothing comes close to that.