LSI 9265-8i

[alfaunits]

I'll tell you. That experiencie was an a 3ware 9650. Although I shutdown the system when the controller was install an update of Windows, to see what happened.
On 3ware 9690, everything run fine, just did the checkdisk (with UPS, no BBU). Than, when I put the BBU and disable the UPS,did the same things and just run fine on a variaty of situations, the BBU was there to save everything.

If that is how you figure "cause->outcome", no need to talk any more.

[Jose Teixeira]

If that is how you figure "cause->outcome", no need to talk any more.

Relax, this was an example. I know that are situations more deeply than mine, of course...I just see what happened, more than one time,and on alarms tag of controller must talked to herself...

[vxr]

recently I got some messages about the processor core of LSISAS2208....
its prototype of the processor is different with the LSISAS2108......
LSISAS2208 uses the new 32-bit PowerPC476 core....
internal bus utilizes IBM PLB 6 technology....
PLB6 is for CMP archtecture....
at 800MHz, crossbar architecture, full-duplex, 128-bit....
it will be gotten a massive bandwidth number for internal bus.....
25.6GB/s......

LSISAS2108 is the same with LSISAS1078...
it is based on the 32-bit PowerPC440 core.....
if it is at 533MHz, 1066DMIPS...
but I was not aware its internal bus architecture...
I guess it uses IBM PLB4 technology....
at 166MHz, crossbar, full-duplex, 128-bit...
data bandwidth number: ~10.4GB/s...


another case....
about 3ware 9690SA...
it adopts dedicated StorSwicth architecture and use PPC405CR at 266MHz with 16MB SDRAM(not cache, decompression code area....)...
PPC405CR is 32-bit processor....
it is the same with LSI RAID controller to use PLB technology..
but it is older version....
no crossbar, 64-bit, at 133MHz....
the data bandwidth is the same with the PCI-X...

[Jose Teixeira]

If that is how you figure "cause->outcome", no need to talk any more.

Hello alfaunits:
Look, this is not so simple because WB + UPS or WB + BBU is not the same.
Write cache on/off is a characteristic of the disk driver – the driver indicates to the OS how it feels about write caching. In a RAID setup without battery backup, you are risking array integrity corruption if the cache isn't properly flushed (e.g. in the case of a power fail, lockup, BSOD or other unclean system restart). OS system like a Windows itself has nothing to do with it; the Windows kernel simply operates according to what the driver tells it the hardware can do.
Arrays functions is not a separate case, it's all connected because windows has to know there is an array but in write back context, it doesn't know anything.
I'm not saying that your arrays lost of Areca wich means this case, I'm just saying that array *can be* not should losed due to not have a BBU.

Regards to all

[Jose Teixeira]

recently I got some messages about the processor core of LSISAS2208....
its prototype of the processor is different with the LSISAS2108......
LSISAS2208 uses the new 32-bit PowerPC476 core....
internal bus utilizes IBM PLB 6 technology....
PLB6 is for CMP archtecture....
at 800MHz, crossbar architecture, full-duplex, 128-bit....
it will be gotten a massive bandwidth number for internal bus.....
25.6GB/s......

LSISAS2108 is the same with LSISAS1078...
it is based on the 32-bit PowerPC440 core.....
if it is at 533MHz, 1066DMIPS...
but I was not aware its internal bus architecture...
I guess it uses IBM PLB4 technology....
at 166MHz, crossbar, full-duplex, 128-bit...
data bandwidth number: ~10.4GB/s...


another case....
about 3ware 9690SA...
it adopts dedicated StorSwicth architecture and use PPC405CR at 266MHz with 16MB SDRAM(not cache, decompression code area....)...
PPC405CR is 32-bit processor....
it is the same with LSI RAID controller to use PLB technology..
but it is older version....
no crossbar, 64-bit, at 133MHz....
the data bandwidth is the same with the PCI-X...

Hello vxr are you ok?

Yes, the 3ware 9690 has old PPC405, I think this processor was released in 2004

[vxr]

another LSI engineer told me and confirmed LSISAS2208......
its PPC prototype still is PPC440....
not PPC476...

[Jose Teixeira]

another LSI engineer told me and confirmed LSISAS2208......
its PPC prototype still is PPC440....
not PPC476...

I don't search the PDF yet...don't know.

May be 2x PPC440

Although this LSI2208 rocks on 4-64k test

[Anvil]

There is a new firmware for the 9265 (dated April 15th) Link

Bug Fixes and Enhancements:
===========================
Firmware:
LSIP200139711 (DFCT) fw hit montask at pd.c when rebuilding from raid 6
LSIP200154034 (DFCT) Kill Adapter after 62 cycles of cable pull/push with I/O
LSIP200143213 (DFCT) I/O miscompares when issue OCR.
LSIP200143179 (DFCT) Return status code 'Success' to certain obsolete SCSI commands
LSIP200121969 (DFCT) Liberator FW 10M09 cause system (sles11sp1/rhel6) reset while lspci -vvxxx is used
LSIP200139742 (DFCT) sscd show 0 byte in debugger and MSM
LSIP200139758 (DFCT) Drive comes up as unconfigured bad while running power cycle and error injection script
LSIP200139756 (DFCT) ekms bios is not enable
LSIP200155390 (DFCT) Port fixes for LSIP200112913 and LSIP200116269
LSIP200071988 (DFCT) Inconsistent drive state after drives are changed to different slots.
LSIP200139713 (DFCT) Clear configured with SSC degraded causing FW went to mon_tsk()
LSIP200139759 (DFCT) With multipath setting, found data miscompare while IOs across 2 enclosures
LSIP200153318 (DFCT) FW doesn't detect the iBBU08 if it is remotely mounting with OEM controller
LSIP200123091 (DFCT) DateOfFirstuse is Incorrectly Causing BBU Deep Discharge Requirement
LSIP200139722 (DFCT) Controller hit montask during BGI
LSIP200139762 (DFCT) Configured PDs become not supported PDs, LDs dropping offline.
LSIP200145892 (DFCT) Liberator FW 10M09 cause system (sles11sp1/rhel6) reset while lspci -vvxxx is used
LSIP200104775 (DFCT) Controller reports "phy is bad" for single port with any SATA device
LSIP200139764 (DFCT) HSP commissioned alarm beep goes away after reboot
LSIP200154055 (DFCT) Port fixes
LSIP200155340 (DFCT) port 2TB error handling changes
LSIP200139728 (DFCT) FW aborts the recovery resulting SES not responding
LSIP200142547 (DFCT) Provide mechanism to bypass settings delayPOST and envDelaySOD
LSIP200123036 (CR) Deliver HII in Q1 2011 Liberator maintenance

[Computurd]

hey guys the Raid 5 review with 8 c300 just went up, check her out!

http://thessdreview.com/our-reviews/...ron-c300-ssds/

[Anvil]

Thats great CT

Will read it later today.

[waylander_druss]

Hi there.

First time poster. Guess I'm more of a "Lurker" ;-) I recently purchased this card for my Array setup.

Details:

OCZ Vertex 3 120GB - X4
LSI-9265-8I (With Fastpath Enabled)
Raid 0 with Stripe size of 64K
Windows 7 Ultimate 64bit

As am pretty new to this card I am unsure of the exact set-up I should be using going by the numbers AS-SSD is kicking out. For example:

First Screenshot: No Read Ahead, Direct I/O, Write Back, Disk Cache Enabled.

Second Screenshot: Fastpath recommended settings which is same except Write Thru enabled.

Seq Read/Write speeds take a fair knock with Fastpath settings but the 4K-64Thrd increases massively. I suppose my main question is which setting should I be using that gives the best overall performance? I do use my system mainly for gaming.

Apologies if I have left anything out information wise. If I have just ask and I will add it.

Many thanks in advance for any advice.

[mbreslin]

I personally run mine geared for 4k random, if you don't have a need for the highest possible sequentials (and as primarily a gamer it sounds like you don't) I would just stick with the fastpath settings. You may consider trying other combinations as well as smaller strip size.

In general I'd say your array is beast.

[Nizzen]

Asssdbench is not the best test dor a controller with cache. With 1880 4gb the results of asssd bench is almost just cache results. I prefer iometer with BIG testfile. 16gb+.

Nice results anyway

[waylander_druss]

Thanks for the replies folks. I'll keep it on the Fastpath settings. I'll try the suggested smaller stripe size as well when I get the chance. If there are any major differences I'll post again for the benefit of anyone else in similar situation.

Cheers

[Levish]

Seq Read/Write speeds take a fair knock with Fastpath settings but the 4K-64Thrd increases massively. I suppose my main question is which setting should I be using that gives the best overall performance? I do use my system mainly for gaming.

just a comment since I was in the same situation

Most of the games that I've taken the time to Perfmon show little / no disk activity outside of changing areas / zones if you have sufficient RAM.

The three that I tested (ME2, FF14, DaO) had some usage while loading but most of the time spent loading was CPU intensive (80-100% use) and not storage limited/intensive.

A vertex3 can load a entire 10GB game in ram in approximately 30 Seconds (500MB/sec gives 20 Seconds for 10GB but at file sizes below 512KB you'll dip below the 500MB/sec peak rate of the drives).

The rest of the time loading a level is processing, the rest of the time gaming there is very little IO hitting the storage system.

[bivaughn]

Does anybody know if the Battery Back-Up (BBU) unit is required for any functionality on this card? IE, is Writethrough not possible unless you have the BBU?

Thanks!

[Computurd]

writethrough is always possible. there are 2 different writeback settings, you have "writeback" then "write back with BBU". writeback is more aggressive with the BBu, but works without the bbu as well

[thebad]

How and where do you get the FP licence for the 9265? Is it still a hardware key?

[Computurd]

its a software key.

[thebad]

So you buy it direct from the LSI website?

[waylander_druss]

Hi Thebad. Not sure if you got yourself sorted for the FP software license key as yet. Ig you live in the UK then I got mine from Lambdatek. I've linked it below. This was a real for me to find and all places but these guys hadn't a clue what I was talking about. Ordered up and had the license via email within a couple of days.

http://www.lambda-tek.com/components...prodID=B538100

[thebad]

Yeah I found out that I need to order the LSI00266 P/N. Some resellers have it in the US and it can be emailed...

[thebad]

posting my results with 4 C300 128gb 64kb SS:

No FP/FP with NRA DIO WT/FP with NRA DIO WB.

With Writeback the Sequential and low queue depth is actually better but the longer queue depth suffer. I am really wondering if WB is not better for an OS drive.

[WonderSausage]

writethrough is always possible. there are 2 different writeback settings, you have "writeback" then "write back with BBU". writeback is more aggressive with the BBu, but works without the bbu as well

No, "write back" means force/always write back, "write back with BBU" means enable write back only when the BBU is present and functional (e.g. no write back enabled without BBU, and/or auto disable write back if BBU is present and fails).

There aren't two kinds of write back, the performance and "aggressiveness" of write back is identical in both cases, the setting merely determines under what circumstances write back is turned on.

"Write back with BBU" is the default but if you want to live on the edge, you can force write back on without a BBU. Personally, I use a BBU

WS

[WonderSausage]

Computurd,

A sincere thank you for all of the effort you put into these reviews, and your insightful comments. I really hate to quibble under these circumstances, but ultimately I don't find performance results "in a vacuum" to be useful. There needs to be a direct comparison with some known competitor (especially the 9260, which ironically I see pictured in your photos, so you clearly had one on hand). The 9260 review on thessdreview site was done by a different person on a different hardware configuration.

I currently have a 9260-8i with 4x Intel X25-M 160GB RAID0 64k. Is the 9265 worth the money to upgrade? I don't think this has been answered. What is the real-world application impact of FastPath, especially in gaming and content creation? [we have data for server apps, assuming we trust LSI's benchmarks]. While your review card came maxed-out, FastPath is $150-$165 retail, which is not insignificant (it's nearly as much as the LSIiBBU09, which incidentally adds cost because one can't reuse the BBU from the 9260). It sounds like all of your tests were run with FastPath enabled, so you weren't testing a $700 controller, you were testing an $850 controller. The Adaptec 6805 is in the neighborhood of $500.

So a 9260 comparison would be eminently useful, thanks.

WS