Showing results for 
Search instead for 
Did you mean: 

A call to reviewers!

Level 14
The SATA 3.0/600 specification is rated to support up to 6Gbps, which (after error-correction) equates to 600MB/s maximal throughput.

I notice that the peak speeds offered by the fastest available SATA drives do approach but don't actually reach this (theoretical) 600MB/s.

Some of the fastest SATA SSD performance specs I've found (for peak sustained-access Read and Write, ignoring random-access IOPS and other complications):
Mushkin Striker = up to 565MB/s and 550MB/s
Crucial Neutron XT = up to 560MB/s and 540MB/s
Samsung 850 PRO = up to 550MB/s and 520MB/s
SanDisk Extreme Pro = up to 550MB/s and 520MB/s
Intel 730 = up to 550MB/s and 470MB/s

Crucial's engineers have commented that they didn't design faster memory/controllers for their MX200 SSDs because they were already hitting SATA3 limits, they chose to not decrease component yields or increase component prices past this threshold. (Although Crucial's MX300 and subsequent SSD models were faster, lol.)
Transcend's engineers have commented that their SSD370 consistently encountered performance bottlenecks on many SATA3 controllers.
Mushkin's advertising (somewhat bombastically) claimed their Striker could exceed the maximum SATA performance capacities of almost any motherboard.

And I admit that 550MB/s~565MB/s is awfully fast. But it's still not 600MB/s. Apparently some professional SSD device engineers have reached a consensus that SATA3-capable mobos can't actually attain 600MB/s speeds. Do they deliberately build their SSDs with slower controllers and slower NVRAM? Do they then also use these deliberately slower (cheaper) components in their PCIe, NVMe, and M.2 offerings?

It turns out that some of the other hardware guys at work are developing a variant SATA-based storage interface (with a nonstandard/extended pinout to provide proprietary hardware key lockout and encryption protocol) for our proprietary medical computing platforms. (This isn't quite as evil as it sounds: the objective is to ensure that the platform and the drives are not interchangeable with standard SATA counterparts as a layer of hardware security on confidential medical data, although of course we do make our money, lol.) I'm not part of this rather unexciting project and I'm not directly involved with their work. Although I do get to peek in from time to time, see what it's about, draw upon their SATA-related expertise if needed.

And I've been thinking about these performance limits. It seems that some PC motherboards might have faster SATA3 than others. Maybe some max out at 550MB/s or less ... and maybe others can even exceed the required 600MB/s.
There's no way to measure and compare maximal SATA3 performances without having SATA drives capable of exceeding them. You can't see where the motherboard/chipset/firmware bottlenecks things until it happens.

I can put together an "overclocked SATA SSD" capable of attaining >600MB/s (maybe >750MB/s!) peak throughput rates. Not built for capacity, density, or longevity. No wear-levelling or overprovisioning or compression or other fancy algorithms. Probably a puny 4GB~16GB total capacity, able to mount an OS and benchmark apps and not much else. Heck, it'd probably just be a GHz Cortex board packing DDR3L instead of NVRAM, lol. And not packaged very pretty, lol, SATA soldered onto PCB bolted onto plastic would suffice. Built for burning-hot benchmark speeds and nothing else. Ugly and useless but functional enough for this one purpose.

What I can't do is measure and compare maximum SATA3 performances on many motherboards. I just don't have access to more than a few (actually eight), plus they're restricted to just a few (only three!) different models. Apparently our C612 machinery is off limits for experimental hardware prototypes, lol.

I asking if any reviewers would like to put such a device to the test across a wider variety of motherboards. Hopefully lots of ROG mobos, PCH and ASMedia ports alike if possible, since these specs interest me most, lol, although good to compare vs Gigabyte/MSI/EVGA/ASRock mobos too. Test if this maximum capacity indeed varies between manufacturers and models (and even individual units). I would test it on my own mobos first, of course, and post my performance scores. But not a very worthwhile project if I can't go any further, such a limited list wouldn't be very useful. Any interested takers?
"All opinions are not equal. Some are a very great deal more robust, sophisticated and well supported in logic and argument than others." - Douglas Adams