I totally forgot to say this. In REAL life testing I was transferring 5GB file (a mkv) from RAID0 Drive (3xWD Caviar Black 750GB Each) to this USB 3.0 Card using that addon card and it started with over 400mb and within 10 sec it dropped around 25 or so ,,,,not 130 or anything close to that.
Raptor on SATA addon card. (Why I am getting this spikes ?)
Seagate USB 3.0 on Addon card (HD Tach quick test)
Seagate USB 3.0 on Addon card (HD Tach long test)
My triple RAID0 on motherboard RAID controller
Let me know if there are some other tools I can use to test things It's just odd from one side I see numbers are good from other real transfer speeds are not that good. I'll do screenshot of actual file transfer in few.
Just did a few tests and I am a bit confused. I single 7z archive file around 3GB transfered at 130mb /sec without losing a single point. Yet ISO and MKV within I'd say a minute goes from burst of 400 to around 30mb/sec ......
Windows caches data in RAM - that's why copies start out fast (as the RAM buffer is filling) and then slow down to the speed of the output drive (when the RAM buffer is full and data can only be put into it at the same rate that it's emptying to the output drive).
If you copy one file, then immediately follow it by copying another file to the same place, then it may be that the disk is having to thrash the access arm between the two files, which will slow down the transfer. I'm thinking that this might happen if the first copy completes and you start the second copy while Windows is still dumping the the first copy's data from the RAM cache to the drive. Try waiting for all drive activity to cease before testing a new copy.
I feel your pain. My Raid controller promised much at a low price but was a let down.
I have an identical Raid controller to you (SiI 3132) in an identical card. The most I have ever managed to get out of it is SATA I speed. For some reason those cards are limited to 150MB/s. That is a total for both drives running a RAID 0. I have tested it with 2x Barracuda 7200.12 (I can get just under 300MB/s peak on other controllers) and other configs.
The card just does not perform, but it was cheap compared to the alternatives.
You bought a masscool SATA2 harddrive controller for $20....did you really expect it to work miracles?
As for the USB3 controller & external harddrive....don't expect any USB3 devices to run at full speed for at least another 6-9 months. Most "USB3" devices currently on the market are really nothing but USB2 devices. The same thing happened when USB2 went mainstream...
1) First, I am impressed by the usb3.0 speeds. What a great way to do external baclups.
2) Hard drives will have differing data transfer rates, depending on where the data is coming from.
Data on the outer rings or cylinders will transfer faster. The outer rings have more blocks, so more data can be transferred during one revolution.
The pattern of declining throughput can, I believe be largely attributed to this.
3) Benchmarks are designed to test the maximum throughput that can be achieved. They do negligible processing, and issue I/O commands as rapidly as possible, even before the previous operation is complete.
They bear little resemblance to what you or I will do. Normal operations will read some data process it, and then do the next I/o(in or out) Usually, the application is synchronous in nature, meaning that the first operation must be posted as complete before the next can be initiated. If the processing is not negligible, a hard drive can miss a rotation and really slow down. That is why big blocks are better for sequential operations.
4) Raid-0 is often a disappointment. The benchmarks make you think you are doing super, but the reality is that it does not help much unless your app is written like a benchmark. It is often better to arrange your data on separate drives, the source on one, and the target on a different drive. That way, there is no access arm stealing, and the drives can operate at their mechanical best.
5) It is possible for a discrete raid card to detect sequential operations and read ahead data. Or, write behind if it has sufficient cache, and a batter backup protection. Needless to say, they are very expensive.
6) If you have the time and inclination, do some testing of your most used apps, and see what arrangement works best for you. Create a personal benchmark that is valid for you.