Originally Posted by Jayner
We just have to accept that those are the scores-- nothing will improve them, except an upgraded cpu I guess. Sucks how the 830 I had previously had higher IOPS in random read, oh well. Like many have said before, the scores are 'fine'-we have to lower our expectations.
It's not the CPU that is the problem, it's the AMD SATA chipset, SATA driver, or both.
AMD had SATA III (SATA 6Gb/s) support on their SATA chipsets before Intel did. But for some reason, they don't perform at the level that the Intel SATA III chipsets do. I'm not an AMD expert so I'm not sure why, such as what buss they are using (PCI-E, etc) for the SATA interface, or how much of an interface is allocated to the SATA controller. That can make a big difference.
It's possible that the relatively older AMD SATA portion of their chipsets, which have not been updated along with the changes to support the newer AMD CPUs is the issue. AMD has some new chipsets now, like the A8x series, but I haven't seen any news about it being equal or superior to the Intel SATA III chipsets.
PC hardware reviewers know the Intel SATA interface is faster than the AMDs, and they use Intel boards for all their testing. I get it, AMD owners see all the benchmarks that were run on Intel boards, and don't have similar results in some areas. In some cases, the Samsung Magician software won't run properly on AMD boards when using AMD SATA drivers. Why that is the case, ask Samsung and AMD.
Regarding the benchmarks, the reality is the level of stress used in benchmarks is far more than ever experienced on a PC. So you won't be experiencing any performance loss with the results you get. The IOPs on my Seagate Barracuda HDD are below 1000, actually below 500 read and below 200 write.
The random read and write tests create multiple I/O requests that are sent to a drive in groups of up to 32 at a time. That is what QD or Queue Depth means (a queue being a line, like a line at the checkout lane at a store), with QD = 32. That means there are 32 outstanding (not yet processed) I/O request. That cannot happen without AHCI "mode", or the enhancements it provides.
like AMD's implementation of AHCI is less than optimal, and can be confusing to enable in a BIOS. I've seen forum posts that claim AHCI on AMD boards is not functioning well unless RAID mode and RAIDExpert is used. If that is true or not, I don't know. Note that your sequential speeds are right on spec, and over a bit. The IOPs of a sequential test are ~32 at a speed of ~510MB/s. Your random tests (using small files, likely 4K) are doing 50,000+ IOPs.
A possibility with the difference between an 830 and 840 Pro is, the Magician software has changed over time, several version have been available. Do we know if the benchmark program is identical in all versions? I don't know either way, but how do we know it's the same?
Checking my saved AS SSD benchmarks of my 256GB 830, vs the same size 840 Pro and 840 EVO, the 830's IOPs scores are lower across the board (uhm... on an Intel board) and never reach 70,000. Benchmarks programs are all different, and are affected by many different things. If the environment of the tests are not identical (OS drive vs empty drive, CPU power saving options enabled, background processes using I/O resources, etc) results will be different and not comparable.