RAID controllers are of little interest to enthusiast users, but the entire IT industry relies on them. The new Adaptec RAID cards target all sorts of business storage solutions in the areas of DAS, NAS and SAN.

http://www.tomshardware.com/2008/0 [...] index.html

Do you do any testing of the heat generated from this card? I currently have an Adaptec 31605 that I ripped out because it was just generating too much heat. I basically had to put a fan on the botton of my case to blow air directly on the processor of the card. The card fried once and I had to get it replaced. I am using water colling and all silent 800RPM fans so there is not a ton of airflow inside the case but there is some airflow. I cant find much about how hot the cards get on the manufacturer's website. I was about to try the ARECA ones and see if they at all run cooler.

I am wondering why testing was done only with the 8 port internal card. It is the only port configuration that does not offer external expansion. The 4 port, 12 port, 16 port and 24 port internal cards all offer 4 more ports external. The strongest feature point of this card is the fact that it is compatible with so many SAS expander boxes. The internal and external ports can be used with either discrete connections or SAS backplanes and then the external port can be used in the future with expansion boxes. I have tested 3 of the 5 series controllers so far with several expander boxes and it works well. I will agree that it is very fast in discrete but it is also fast when using expansion.

Impressive. I think it might have been useful to compare all the RAID cards to the integrated Intel ICH9R where possible since that seems to be a topic being debated a reasonable bit.

-mcg

Great Article,
I was shocked to see the Areca controller score much lower in all raid configs at iometer in all patterns (file server/web server/database/workstation) except streaming i/o then the new Adaptec and LSI logic and to a lesser degree the AMCC.

Note that Areca has a new one (the ARC-1680ix) that has a 1.2GHz IOP 348 instead of the 800MHz on the previous model (the ARC-1680) so that one will probably get the Areca on the same level as Adaptec/LSI/AMCC.
Areca, I really don't like a small fan on such expensive cards, we used to have that crap on cheap mainboards that would die in a couple of years because of fan failure. So Arcea get us passive cooling, for these prices we expect the best!

Very interesting is that (at least in the Netherlands) most LSI Raidcontrollers are also available as Intel branded for a lower price.
Instead of the LSI MegaRAID SAS 8888ELP you could buy the Intel SRCSASJV for less and get 512MB cache instead of the 256MB of the LSI one.
If you need only internal connectors getting the LSI MegaRAID SAS 8708 instead of the 8888 would be intersting since its much cheaper and is the same hardware except it has 128MB cache.
The Intel version of the 8708 is the SRCSASRB which is also cheaper and gets you 256MB cache.

In the review's comparison table the above 4 controllers miss the "XOR-engine" column and Netware 6.5 should be added as supported OS on the LSI SAS controller (the driver was released later so its not listed in LSI PDF's)

I compiled some pricing for these controllers in the Netherlands in Euro's including 19% tax because the prices here are sometimes quite different from the US.

AMCC 3Ware 9690SA-8I € 489.-
Adaptec RAID 3805 € 415.-
Adaptec RAID 5805 € 485.-
Areca ARC-1680ix-8 € 699.-
Atto ExpressSAS R348 € 819.-
ICP Vortex ICP5085BL € 795.-
Intel SRCSASRB € 475.-
Intel SRCSASJV € 685.-
LSI MegaRAID SAS 8708ELP € 515.-
LSI MegaRAID SAS 8888ELP € 839.-

It seems Adaptec has the best/price performance at the moment, while Atto ICP-Vortex and LSI have the worst price/performance ratio.

Makes me want to get rid of my 31605 that I just got. I agree with ^^^^, I'd like to see some temp results. The 31605 runs quite hot !!! I was anxiously awaiting a review of these new dual core cards from Adaptec, and looks like I should have waited to get the new series, but I didn't know and I bought it only a week or so before the new ones came out.

Can you provide the config files (*.icf) for your tests? I'd like to make some comparisons.
Thanks.

So now to make the decision between this new Adaptec one and the new ARECA one. I miss the performace of the savvio drives and want to go back once I figure out which card runs cooler which would translate to less cooling needed which means less noise too.

Hey Turas,
I don't know how hot the Areca and Adaptec are, since we use mostly Intel raidcontrollers and they get quite warm but not extremely hot. The controllers are however always used in Server chassis with enough airflow, still our servers don't generate a lot of noise but cannot be called exactly silent. Problem is these cards are designed for servers in which there is sufficient airflow. Still I would rather put a silent 80mm fan close to a passive raidcontroller then get one with a 40mm fan. Maybe Intels SRCSASRB would be an interesting option since its cheap and fast (same spec as LSI 8888 without external ports) I could probably get you temparatures of this card but not very soon and without comparison to other controllers which I don't have.

Also I 'found' this comparison:
http://www.adaptec.com/NR/rdonlyre [...] aperv5.pdf
Its obviously made by Adaptec so it will be optimised, but u can compare at least all other cards with each other. The test seems solid, its done with IOmeter and al used cards/drivers/settings are specified.

I can tell you that the 5 series cards I have tested do run a little hot. If you were to touch the heatsink on the bridge chip, (all the cards with the external port have it) you would not be able to keep your finger on it for more than 1-2 seconds. I also had the 5805 card but I didnt get to do as many tests with it so I can not speak for its temp.

edit:

I just tested the temp of one other card and both chips/heatsinks are quite hot. It does not appear to be a problem but quite hot at least to the touch.

Thanks for all the info.

Rozar, I love your sig "Generic Case with no blue lights"

Hey, thanks for getting to us on the temps. That's good to know that the 3 series seems to run just as hot as the 5 series, so I didn't get jacked there. But then it's sad to see that the thermals haven't improved much. I might just get a small 30 or 40mm @~1000rpm fan and put it on there so it'll be quite but help some.

muk, do you have the config files for the tests to share?

Thanks for the benchmarks. Any chance of getting one of the new Areca 1680ix cards in for testing? Also, what bios was used on the Areca card, as it appears they have a faster performing one (?) out:

http://www.areca.us/products/pciet [...] rmance.htm

THG tends to put Areca's card below all others...which seems quite unfair.
I do not own such controller, but there are plenty other reviews out there that can show more realistic numbers. Even 'normal' users have posted at least some write/read RAIDx transfer rates on Areca's and they are above all THG has ever posted as results on Areca's 12xx+ series. And lets not forget the upgrade option for the cache up to 2GB.

Of course,there is no perfect controller - Areca has problems with some drives and still somehow 'cheap' firmware. Some of 3ware's models have problems with some motherboards ( slots incompatible ), BAD official driver support ( no win 2k8 drivers ). Adaptec is reported to have very poor transfer rates.
Should CPU assisted controllers to be considered 'enterprise' class at all ?

I have the 51245 card attached to Supermicro SC933 EL2. I'll put up some benchmarks when I get a moment, but so far it looks impressive.

Let me know how it goes with the 51245 in that chassis. I just did 2 836E1s with a 5 series card in each and I had quite a bit of trouble getting it to work.
Backplane rev matters
Driver version matters
Firmware virsion needed to be 15738
and even motherboard slot mattered

Also my test 51245 card runs REALLY hot, check yours for temps.

Curious why that card for that chassis too??

I bought this card based on TG review and have it installed in a new server with the following config:

2x XEON Quad Core 2.33 GHz processors
8GB of DDR2 667
1x Seagate 80GB (system drive)
8x Seagate 15K.5 Cheetah SAS drives (ST373455SS)
Windows Server 2003 x64

In my testing I've varied the stripe size from as low at 32k all the way up to 512k, ran iometer 2006.07.27 and have not come close to TG test results. I think they are getting PAID off by Adaptec to post higher numbers than are actually being produced. The only difference between our systems is the processors and since we are testing a RAID card the processor shouldn't make a difference. On paper my drives are rated faster than the Savvio 10K so my test results should be faster. If anyone has any suggestions it would be greatly appreciated.

I had a 15k hitachi drive that was about 80mb a second tranfer, and the 300mb is much faster at 115mb. I was using hdtach for a simple read test. Try 1 drive as a volume and run hdtach and see what the transfer is with those 73gb drives. Just run the short test to get an idea of the speed.

Ran test 3 time average score 114.93

That sounds about right so it does not sound like its the drives. Of course this only speaks for the speed of 1 drive and not the array. But I can tell you that the 5000 series cards are very fast.

What driver are you using and do you have 15728 or 15738 firmware?