MiniKnight wrote>"... I don't have the link handy but thought it was a forum post here? Also iirc it was resolved using iometer? "
I think this was the link:
http://www.servethehome.com/lsi-sas-2308-hba-ssd-windows-server-2012-turn-write-cache/
Patrick's only reference to IOMeter was>"... I did want to point out that IOMeter managed to provide almost 17K on QD1 and over 40K on QD4 sequential 4K write tests so the above seems correct. ..."
---------------------------------------------------------------------------------------------------
Interesting quotes are:
"... flashed to IT mode, for the results displayed below. IR mode showed very similar results. ...". That contradicts mrkrad "... the HBA should be used in IT mode only ..." ??
" 840 Pro Anvil 4K Writes (I assume these are QD=1) went from approx. 2MByes/sec to approx. 66MBytes/sec by checking Anvil's>Setting>Write Through box.
My 2MBbyte/sec score does not change whether Anvil's Write Through box is checked or unchecked.
Question: Why should Write Through be so much faster.
I thought the Write Back is the conceptually faster option, as it would allow Samsung's internal 512MB Ram and controller to do its own thing and hopefully overlap writes using its multichannel write capability?
--------------------------------------------------------------------------------------------------
mrkad wrote>"... Return the 840 pro and use the proper SSD like server GRADE S3500 ..."
That is not a possibility. Note than anandtech measured its 4KB Random Write as 47.49MBytes/sec.
AnandTech | Intel SSD DC S3500 Review (480GB): Part 1 - Print View
---------------------------------
mrkrad also wrote>"... I bet you would get better performance in IT mode with software raid-0 almost as good as software raid-0 with intel ports ..."
That would mean using Microsoft Disk Manager's dynamic/striped disk capability for my two Samsung 840Pros. BUT THAT WOULD MEAN NOT BOOTABLE. Unless someone can provide how to make Window Disk Manager dynamic/striped raid0 bootable, that is a deal breaker?
Also, I suspect the LSI sata3 provided raido sequential read rate would drop from approx. 1GByte/sec to 512MByte/sec, if I abandoned the LSI 9212 4i43 HBA and used Dell T110 sata2 3gbps for Windows Disk Manager raid0 using dynamic/striped volume.
--------------------------------------------------------------------------------------------------
mrkrad also wrote>"... the HBA should be used in IT mode only.
According to Patrick article linked above "... IR mode showed very similar results ..."
--------------------------------------------------------------------------------------------------
I am waiting for LSI to deliver on their promise to provide an ECN that fixes what LSI admitted via email is a defect effecting LSI HBAs in regard to 4K, QD=1, Writes.
If anyone as any ideas why Anvil write through setting made a substantial difference for Patricks testing, but no difference for my P16 IR Mode LSI 9212 4i4e HBA, I am all ears.
Thanks to STH for sometimes being the cutting edge on these things
.