LSI 9211-8i versus LSI 9260-8i versus Areca 1231ML-2G

[trans am]

it doesnt work... it doesnt see the keydrive at all...

and you set the 1st boot device in mobo bios to the usb key and disabled everything else? You also browsed to the dos file in the HP usb boot utility when you formated the usb key?

you can't burn the iso on a cd and boot from that or its usb key only?

try this one

http://h20000.www2.hp.com/bizsupport...reg_R1002_USEN

[Tiltevros]

guid me plz tell me the steps one by one

[trans am]

guid me plz tell me the steps one by one

T we can't crap derail this thread any more out of respect for C.T. who started the thread. We are getting way off course here. Capt. Picard would crap if Riker was making Worf help Geordi do a level one diagnostic during a Romulen attack on the Enterprise. Just go back to this thread and follow it step by step until you have the usb boot disk formatted as a boot dos disk. If you are really out of ideas screw HD erase and read up on using santiary erase wiper by booting from the win7 disk and go to repair. I am sorry if I cant take much time on this but I am in the process of draining my watercooling loop.

http://www.ocztechnologyforum.com/fo...00&postcount=1

[One_Hertz]

Steve - still no results for game loading times?

[SteveRo]

Steve - still no results for game loading times?

Yes, by the end of the week.
Planning to use an mtron as boot drive and load the game/test level load times on the controller/8xacards - this ok?

[One_Hertz]

Yes, by the end of the week.
Planning to use an mtron as boot drive and load the game/test level load times on the controller/8xacards - this ok?

That's perfect.

[Computurd]

i think that should be fine. that will be the most anticipated results by far you have been dong an awesome job with these controllers. i see that your now listed very high in the pcmark vantage TOP 5!! nice result! did you try the MSM for the 9260-8i yet?
i appreciate your time very much as i know the other guys do.
here are some results from something i compiled on the 9211 boot array.
we must take these results with a grain of salt, as i am having issues with 'something' that is limiting me from max throughput.

[SteveRo]

Anyone know the cheat code to unlock levels in crysisWHD off the top of their heads? - I will look it up.

[Computurd]

i just load them by hosting up a multiplayer game. all the levels are there.

[ehorn]

Tyan S7025 is coming today but my CPUS will come tomorrow so... i will try to setup as soon its possible to make some results... do u think that we will see any difference?

Supermicro has MB's with LSISAS2008 onboard.

http://www.supermicro.com/SAS2/SAS2.cfm

I imagine this would help latency? I wish they had a board with (4)2.0 X16 slots and onboard controller... The X8DTG-QF has (4) X16 lanes but no SAS 2.0...

With the s7025, you get another onboard controller to add to the testing (LSI SAS1068E - albeit 3Gb/s-8 port), might be neat to see how this stacks up to the ICH10R, etc....

Curious to see the results from your new setup.

[Computurd]

@steve-ro--running stripe size of 16k with the 9260 i have been able to match the load speeds of the 9211 with crysis. i am installing l4d now to test as well.
i am going to do some targeted testing with 64k and 128 k as well. the 1mb stripe is probably best for the benchmarks, but not game loading.

[Anvil]

Well, I got the quote on the 9211 today, is the 9211-8i KIT worth ~$400 ex.vat, (half the price of the 9260-8i KIT in Norway)

Computurd,
Smaller stripe sizes almost always wins in real life performance.

[Tiltevros]

ok im back to the game dual xeons are on

[SteveRo]

i just load them by hosting up a multiplayer game. all the levels are there.

Figured it out - just restored a game profile from when I had completed the game sometime back.

[SteveRo]

@steve-ro--running stripe size of 16k with the 9260 i have been able to match the load speeds of the 9211 with crysis. i am installing l4d now to test as well.
i am going to do some targeted testing with 64k and 128 k as well. the 1mb stripe is probably best for the benchmarks, but not game loading.

I have already completed the CrysisWH bench using the 1MB strip!

[Computurd]

@steve-o...gah i feel for ya! i just found out what the problem may have been with the 9211.. will have to check it out later. i am getting testing fatigue
so i am going to play around some to rest before i put it back on to check. i am sorry to spring the 16k thingy on ya after you already pounded away at it. if you dont get around to it that's cool.

for reference someone asked me what the idea behind my game loads testing is. basically i reboot, let the system set for about 30 seconds, then load game.
then reboot, and repeat. that is 1 nonconsecutive and 2nd nonconsecutive runs.
the consecutive 1 2 and 3 are just repeated runs one after the other after one of the non-consecutive runs. that is to check reload times. they are much faster of course
the reason i feel they are relevant is that you never go in and play one level, then reboot and play another. even if you load one map, say beach, then immediately load the next map, even though the map is different, you will get the speed of the 'consecutive' runs. so after the first map is loaded the loads are all screaming fast, i think it is cause the respective controller, or operating system or whatever, has already 'mapped' that portion of the drive. i know it isnt cache, cause the 9211 doesnt have cache and it really excels at this portion of it. however i have beaten the crysis load level times for the 9211 with the smaller stripe on 9260. go figure.
however for reasons such as wear leveling, or evenly using the drive, i am not sure if 16k is going to be a good 24/7 stripe size, because of the LBA and such. however i am no genius on that, maybe someone with a clearer idea of how that would effect the wearing of the drive could reply?

[One_Hertz]

^^

It puts more stress on your drives. Say a 128kb file gets written to your array, 16kb gets placed on each of your drives and each drive has to erase a block to make this write. Total writes done = 8 times the block size. If stripe is 128kb or more in the same scenario then your total writes are 1 block size or 8 times less.

It wouldn't make wear leveling less effective, but it would increase writes quite a lot and you know the consequences of that yourself.

[Computurd]

yes of course. so ideally then the best stripe for longevity purposes is 128?

[One_Hertz]

yes of course. so ideally then the best stripe for longevity purposes is 128?

Not 128 exactly, just larger is better. In my example having a larger stripe size than 128 would not make things better or worse, but if I were to use a larger block size in my example then it would matter.

[Computurd]

cool, after i am done with this playing around im gonna split the array between the two controllers and retire the drives. TBH the 1/2 to 1 sec difference i will never notice....just need them to last after that, at least until the 18xx arrives, then hopefully by then i will have some 6gb/s devices to play with.

[Ao1]

Dr. Busch talks about raid performance with ssd's in the Q&A on this podcast @ 47min 20 seconds. It's a shame he does not elaborate more but clearly he is not impressed with hard or soft raid.

http://intelstudios.edgesuite.net/id..._MEMS002/f.htm

[SteveRo]

^^

It puts more stress on your drives. Say a 128kb file gets written to your array, 16kb gets placed on each of your drives and each drive has to erase a block to make this write. Total writes done = 8 times the block size. If stripe is 128kb or more in the same scenario then your total writes are 1 block size or 8 times less.

It wouldn't make wear leveling less effective, but it would increase writes quite a lot and you know the consequences of that yourself.

This is probably n/a for acards? dig, dig.

[Anvil]

Guys,

Don't get caught up in just writes, most operations are read and in order to utilize all (or at least most) SSD's in a raid, the stripe size should be according to the size of your files.

With a stripe size of 1MB and a file size of <1MB you'll only be reading from one of the SSD's, this does'nt take advantage of raid-0.
The main point would be reading from >1 drive.

Of course this puts a bit of a strain on the SSD's but again most operations are read, at least on a desktop computer.
Also remember that most SSD's (and raid controllers) are performing write combining to reduce the number/amount of writes.

[One_Hertz]

With a stripe size of 1MB and a file size of <1MB you'll only be reading from one of the SSD's, this does'nt take advantage of raid-0.

That is not true because of read ahead/queue depth. It is true only if you are doing a single request, for example you want a single small file. But if you want like 10MB of small files then it will read from all SSDs even if stripe is like 1MB.

Thi is probably n/a for acards? dig, dig.

Of course.

[Anvil]

One Hertz,

Of cource, QD must be taken into consideration, but as a single (atomic) operation it is correct.

Have you tried to determine the actual QD during normal/heavy workload on your system?