Hello! New To This Forum, but was looking for a guide to help a friend Flash his 9211 Card I just had him buy, because I didn't have time to walk him through the process, and I found this guide to be INCREDIBLY informative and exactly what he needed! So awesome guide.
However, as a pedantic point, I just had to point out, I am sorry I just couldn't stop myself to signing up to this forum just to say this, I just felt compelled to fix this one small point.
PCIE Gen 1 is 2.5GT/s (or GigaTransfers a Second) which after the 8/10 Encoding works out to 250MBps per Lane. PCIE Gen 2 is 5GT/s, which works out to 500Mbps Per Lane. Your Guide says that PCIE Gen 2 is capable of 250GBps per lane, which is actually about 500 times faster then PCIE Gen 2 is capable of.
In the grand scheme of this guide, this is really an irrelevant point, as this guide is still one of the best cross flashing resources I have seen on the net, but I still could not stop myself from pointing that out. Sorry, character flaw I know...
Anyways, Awesome fricking guide, thank you for making it, you have literally saved me a nights work that I rather take off to play WOW Classic. Have a great night!
But as best I understand it, the substantially-lower-than-theoretical-maximum bandwidth is correct. Specifically, the realistic figure *is* likely to be more like 250 MB/s than 500 MB/s per lane when a PCIE card (of any speed) tops out - the reduction I've included seems to be correct in principle even if the exact reduction isn't well-defined.
Apparently - and I'd love to be corrected if wrong - the problem is that PCIE 2 bandwidth 5 GT/s is based on PCIE 2 full duplex (2.5 GT/s inward and 2.5 GT/s outward??). So it assumes that at maximum bandwidth usage, whatever's using the card can make full use of that full duplex capability to achieve a total of 5 GT/s. But apparently disk storage use doesn't make good use of full duplex at all efficiently. SATA can't use it at all, and apparently SAS should be able to use it but in practice doesn't make very good/efficient use of it. I can't find a hard answer but that's the suggestion.
I'm not an expert on bandwidth/efficiency aspects of SATA/SAS, so I'd love to either see substantiation or correction. It's based on web pages that note this issue. Ideally it needs corroboration from online benchmarks of maximum bandwidth achieved on LSI 9211 based HBAs hosting 8-16 SSDs (testing with both SATA+SAS), and a suitable benchmark that tries to write to all of them at the highest rates, or something. I imagine it would be quite sensitive to setup as well since there are multiple potential ways it could bottleneck.
So in a *realistic* sense, if you attach a bunch of SSD/HDDs to a PCIE 2 HBA, I gathere from these, that the *effective* bandwidth ceiling might well be closer to the half-duplex figure based on 2.5 GT/s = 250 MB/s per lane, or at least significantly less than the full duplex 500 MB/s headline bandwidth for PCIE 2 x 8 with optimised use of the PCIE bus. But like I said, I can't find a rigorously tested answer, just hints that that's the case. I've clarified this point in the resource so it won't confuse anyone.
I'd love to have this checked and thank you for picking up on it.