Data rate roughly halved on written areas / writes capped?

I bought a CSSD-F40GB2-A, using it in Windows XP SP3. (not for the OS, just running programs from it)

Abit P35 chipset mainboard. AHCI enabled.

 

Created a fresh primary partition and aligned it properly and it is confirmed being correctly aligned. (My harddisk partitions are not, so there should be no false positive.) Then did a benchmark using AS SSD Benchmark:

 

Seq: R=165.07 W=61.21

4K: R=18.72 W=60.52

4K-64Thrd: R=85.02 W=59.29

Acc.time: R=0.066 W=0.274

 

I suppose that the writes are not as they should be. All writes look like capped at 60 MB/s. (And are the 4K reads supposed to be slower than the writes?)

 

For comparison, sequential with the same benchmark on my HDs:

HD1 system partition: R=87.04 W=96.31

HD1 temp partition: R=102.10 W=101.99

HD1 data partition: R=79.82 W=74.64

HD2 data partition: R=100.80 W=90.91

 

HD Tach gives weird read results over the span of the space:

Up to 2 GB 130 MB/s, then up to 8 GB 240 MB/s, then up to 27 GB 130 again and from there to the end at 40 GB 240. The area with the low read speed is roughly the amount of space currently used on the drive, so AS SSD Benchmark's sequential read figure of 165 MB/s is probably the average of both speeds, since HD Tach's average figure is nearly the same.

 

What happened here?

 

 

P.S.: I wanted to directly write to technical support, but after I got a list of general FAQ that didn't help, I immediately got to a page that said I have to send in my product. :-/

please use ATTO to benchmark, and post a screenshot. (link on left) this is the standard Corsair uses to publish their specs IIRC.

 

also please fill in your complete specs in your profile, i cant tell what board/controller you have. some Marvel controllers cap the bandwidth, Intel does not so at this time I/we cannot advise you accurately.

I added info to the system specs profile.

 

The Abit IP35 Pro XE has Intel P35 Express and Intel ICH9R.

 

ATTO doesn't show any cap, so maybe it's really due to the resting method of the other benchmark, but the HD Tach result still confuses me. I think I've read somewhere that using HD Tach is officially accepted by Corsair, so I attached a screenshot from that result, too.

ATTO looks fine.

yup, HDTach IMHO is not for SSDs, its a Hard Drive Tachometer, an SSD is not a HardDrive. (in the spinning platter sense i mean)

It could still mean that ATTO does a more artificial test while HD Tune a more practical. There must be a reason why ever test yields the same transfer rate profile over the length of the drive. And the specialized SSD benchmark probably does a sequential read over the whole length and thus gets the same result as HD Tune.

 

But I will do another test soon: I will copy more data onto the SSD and see whether the curve changes.

running too many benchmarks can temporarily cripple some drives, dont go overboard.

Look, as I suspected...

 

This is after I copied 5 more GB onto the SSD. (see attachment)

 

Tell me if this can't be, but could this somehow be related to the data compression logic that's used? I've not read details about how that works, but it's supposed to internally compress data without the external seeing it. Maybe HD Tach shows the compressed read stream before the data is inflated to its natural size, although that would mean that HD Tach circumvents it, that it knows what's going on, which would be unlikely.

I just have no other explanation why areas of the SSD that contain data show a decreased read speed.

 

This might be helpful in solving the mystery, too: After I deleted the additional data, those regions still had half transfer rate. Maybe because a garbage collection hasn't been done there yet, I'm not sure.