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

[Phil]

The link on Anandtech:
SandForce Announces Next-Gen SSDs, SF-2000 Capable of 500MB/s and 60K IOPS - AnandTech :: Your Source for Hardware Analysis and News

[min2209]

Hey guys,

how well do PCMark Vantage scores correlate with real life performance?

Suppose one drive gets ... 100MB/s in the application loading suite whereas another one gets 200MB/s. Am I actually going to see a complex program open 2x as fast?

Thanks

[Phil]

Hey guys,

how well do PCMark Vantage scores correlate with real life performance?

Suppose one drive gets ... 100MB/s in the application loading suite whereas another one gets 200MB/s. Am I actually going to see a complex program open 2x as fast?

Thanks

No where near that. I've actually seen examples where one drive gets a lower score than the other but opens an application faster.

The Vantage Total score is a nice indication for performance, but nothing more than that.

[min2209]

No where near that. I've actually seen examples where one drive gets a lower score than the other but opens an application faster.

The Vantage Total score is a nice indication for performance, but nothing more than that.

hmmm... really. that's interesting. Well, so how does that benchmark work? Doesn't it load a pre-written set of applications etc?

Another question: for normal home use (starting up, launching programs, playing games, etc) is random write as significant as random read? Would I actually see much of a difference between 10MB/s and 50MB/s 4k random write speeds?

Thanks

[Tomy B.]

I didn't see any difference between my SSD (around 4 MB/s random 4k writes) and Intel X25-V (over 40 MB/s random 4k writes).

[tilleroftheearth]

As Anand states:

The four cornerstones of an SSD depend on Sequencial R/W's, and Random R/W's.

He narrows the Random R/W's to 4K blocks, but I suggest that all block sizes are important in the 'feel' of the system.

Some SSD's are untouchable at the 'popular' 4K level, but fall flat when other sizes (512B, 1K, 2K, 8K, 16K, etc.) are analyzed.

No benchmark 'works' to predict real world usage because as soon as a benchmark is popular enough, a hardware manufacturer will have hired someone to learn how to 'cheat' it.

Another thing to keep in mind is that at queue level depths between 1-3, most SSD's do not 'shine' as brightly as their makers would like us to believe.

The bad part is that almost everyone from a netbook to a workstation worker 'lives' in the 1-3 queue level depth (and this is why mechanical HD's are still a viable alternative currently - performance/productivity-wise).

[Jstarnino]

The link on Anandtech:
SandForce Announces Next-Gen SSDs, SF-2000 Capable of 500MB/s and 60K IOPS - AnandTech :: Your Source for Hardware Analysis and News

So basically Intel's g3 will be lagging behind the competition again...

I still don't understand why Intel didn't go SATAIII with the g3...

@TIlleroftheearth... do explain, as i am not current with the ssd-terms you speak of. What exactly are 1-3 queue depths? And how are mechanical HDDs still a viable alternative currently? I always thought even the crapiest ssd will outperfrom a hdd, in any test,benchmark and real-world application... or so i was told..

[Phil]

I had a Kingston V+, a C300 and a Sandforce drive in my Acer 1830T. Going by the 4K speeds the Kingston V+ should be much slower than a Sandforce drive. In reality it was (at least) equally fast, if not faster.

In Techreport's benchmarks the Kingston V+ manages to outperform the Sandforce drives in the majority of the real world benchmarks.

So much for synthetic benchmarks.

There are a couple of review sites that use real world benchmarks that reflect real world usage: Techreport, Techspot, Hexus, Behardware and a couple of more. These are the best in my opinion.

Then there are sites like Anandtech and Storage review. What they call real world benchmarks doesn't actually reflect real world usage. Instead they run a scenario and measure IOPS. These benchmarks favor controllers that do well in heavier multi tasking like Sandforce.

Lastly there are sites like Tomshardware that don't do any real world benchmarks. Least helpful reviews imo.

[Hayte]

I don't rate Techreport. Woefully out of date on every front except cpu/gpu development, gamer centric and the forums are terrible. So many threads with lots of views and zero replies. I always liked xbit labs for their storage reviews and monitor roundups. Anandtech does good features and I like Anand's enthusiasm but everytime theres a big press release and the likes of Intel and Sandforce are preparing product launches, their marketing divisions start pumping out the war rhetoric and invariably, theres Anand off to the side banging on the war drums. Christ...

I don't know what SF1200 drives he was reviewing with what firmware but he managed to instantaneously 'unthrottle' his drives with a full TRIM?! I have no idea how he managed that.

[lakondas]

Hi there

I recently bought a Corsair Force F60 and now int's in my Acer 3820TG-s tummy. It' s relly fast, but i've tested it with ATTO (Corsairs recommended test tool for getting " speeds they promise" ) but i'm not glad with my results.

Features (from Corsair website)
Maximum sequential read speed 285 MB/second
Maximum sequential write speed 275 MB/second
Random 4K write performance of 50,000 IOPS (4K aligned)

ATTO Results:
R W
0.5K 11.1 14.4
1K 22 27.7
2K 43.8 41.9
4K 88.1 165.5
8K 143.7 195.6
16k 179.7 221.8
.....
4096K 241 225


F40 results: Corsair Force 40GB Solid State Drive Test System Setup and ATTO Baseline Performance :: TweakTown USA Edition
Reaches 285/270

I have "intel rapid storage " installed, and trim is running. Anybody, any advice?