Great review. It's hard to find thorough reviews like this on RAID cards. I have seen a lot of people on forums say that 3ware cards and generally only good for streaming-like applications, and this seems to be the case. If you are right about the PowerPC processor being only 266mhz, surely that makes sense when the latest Areca and Adaptec cards feature dual-core 1.2ghz Intel IOP chips.

I would love to see some data on the Multi-threaded reads and writes, namely 1) with more drives and 2) with other controllers. A way to compare them (chart/table) would be absolutely priceless.

Well done and thank you for the interesting read :)

I would say it would have been nice to see a comparison with Areca cards, who are seen as the fastest cards around

Maximum PC did a review on a SATA raid card that has a 1.2GHz dual core processor!

I can't remember the name and model though

I’m a little disappointed with this review. There were no comparisons to other cards on the market such as those from Adaptec, nor were there any CPU utilisation scores. I was surprised at some of the points as well, for example, when the reviewer suggested it was “nonsense” that write speed on RAID-0 was faster than read speed on RAID-0, when this is to be expected with write cache enabled, because you don’t have to factor in HDD seek time. I also found the comparison between the performance of different RAID levels completely pointless. Seriously, what was the point? A comparison of RAID levels between RAID controllers would have been far more helpful. It would have also been nice to have a comparison of performance between operating systems such as Windows and Linux. I’m not sure why Windows 2000 PRO was chosen for this review over Server 2003 or XP.

I have another missing info.
Win 2000 ?
why not to test it with XP/Vista/2003/Linux ?
Btw, Linux test is must for this kind of hardware

It's a shame you haven't tried more disks with this controller. In my tests, adding more disks doesn't improve performance. I tried up to 7 and was unable to go beyond sustained 250 MB/s read on RAID0.

Well it looks like pretty much everyone agrees that this review could really have used more drives (to see how it scales and where the limits of the card are) and a comparison of other drives. However, the review really was quite long as it was, and I don't think it's necessarily feasible to have a 30-page or whatever review. Plus if you look at some of the charts that reviews on tomshardware sometimes have, there's like 20 lines on the same graph and it really can be difficult to tell what's what - so an interactive chart (like say the disk head-to-head comparison feature at StorageReview) would be fantastic.

On a side note, the 3ware cards really don't perform that well lately it seems. I think this would have been even more evident in say a 10+ drive comparison of RAID5 on this 3ware and equivalent generation LSI/Adaptec/Areca. One really good thing about 3ware is the support and KB though, a lot of tuning tips here:
http://www.3ware.com/KB/article.aspx?id=15244

Can get very much improved *streaming* speeds from that - IOPs improvement is more marginal. RAID5 12-disk on a 9550SX for example, without tweaks gets 142mb/s sequential read. After the blockdev command alone - 406mb/s (keep in mind my PCIX is limited to 533mb/s theoretically and even lower in practice). I'm sure the 9650 can do much better, at least for the sequentials.

If I typed any more, I'd have to make a whole article on the subject :p

Was curious - so I tested my cheap Areca 1120 + 4 Raptor 150GB in Raid0.
I got more than twice better results in WinBench99...

There is no mention in the article whether or not the card supports the Port Multiplier specification. If you were to use port multipliers with all 16 ports this gives a theoretical maximum of 80 disks on one controll (or 120 disks on the 24-port card).