Adaptec 1200A Raid array/Matrox2500

[Mike Newman]

I am using an Adaptec 1200A raid card with two identical Seagate 75GB drives to provide extra storage on my Matrox RT2500 system. The system is running on a GA60XM7E motherboard with a 1GHz Pentium III processor and 612 MB Ram. The basic system all works well, with smooth video and no dropped frames. The Adaptec array flies past the Matrox benchtests and captures video fine but when I use it to playback it stutters. Has anyone used this raid card successfully or have any good ideas on what to try?

[Dr Mordrid]

One problem with the Adaptec 1200A is its CPU utilization.

Storage Review tested it vs. several other boards, including the Promise Fasttrak100. They got these numbers using IOMeter & sequential writes - heavy (similar to the mode used for video);

Adaptec 1200A: 41.29% CPU utilization
Promise FT/100: 12.43% CPU utilization

And this is using Win2K!

IF this CPU utilization for the Adaptec holds for sequential reads as well it spells trouble, especially if you're running a bus-hoggish audio card like the SBLive or an AudioPCI based card.

By the time the Adaptec and such an audio card would be done with the CPU there would be precious little left for the rest of the system.

Dr. Mordrid

[Mike Newman]

Many thanks for your informative reply. I have raised the question with Adaptec - although I guess they are unlikely to admit anything! It may be that my best option is simply to try another raid card and see if it is any better. I'm slightly put off, not just by the cost, but by the fact that I think I may have to reformat the drives on another machine before I can configure them with a new raid adapter. Isn't life with computers fun!

[Dr Mordrid]

All you need to do is to un-partition the array from DOS with FDISK (or from Windows with Partition Magic) before removing the Adaptec card. This should return the drives to out-of-the-box condition and ready for a new controller.

I've used nothing but Promise Fasttraks since day one of the original RT-2000 beta and have been running a Promise Fasttrak TX4 during the 3.1 driver tests. No sweat, no strain and it scores over 90 MB/s writes & reads in the Matrox Benchmark with a 4 drive RAID0 array. I think this array would score slighly higher in a 64 bit PCI slot.

Dropped frame? Whazzat?

The TX4 is better for a 3 or 4 drive array, or if you might end up there later. The TX4 can run on either the current 32 bit PCI slots or the 64 bit PCI slots when they become more common. It's blistering fast in either.

If you're going to use a 2 drive array and not go higher then the Fasttrak100 or TX2 is just fine.

Another good choice (better than the Adaptecs, at least) are the 3Ware Escalade cards, but their throughput is slightly slower than the Fasttraks using the same drives. The difference is about 10-12%.

Avoid the Promise SuperTraks. They are geared to server duty, having an abundance of security and parity features but not the raw speed of the Fasttrak100/TX2/TX4's. This makes them unsuitable for video work.

Dr. Mordrid

[Mike Newman]

Sounds like time to scour the retailer's catalogues.

[JerryH]

Doc, are the Escalade and Promise cards you talk about a "true" hardware RAID solution, or software/hardware like, say, a HPT370?

[Mike Newman]

I checked CPU useage whilst doing a replay off the raid array and it was typically less than 25% and never went above 50% - so it seems unlikely that the problem is with over intensive CPU demands. What issues are there if any with onboard cache on the ATA drives? What else is there which might be causing the problem?

[Dr Mordrid]

25-50% is high when you consider the other cards only use 12%, give or take, plus it's a pretty wide variance. That makes one wonder.

IF you're also using an older PCI audio card it too can use 30-40%, leaving the system not much. What audio are you using?

In terms of other things that can cause throughput problems, the first that comes to mind is if you have S.M.A.R.T. turned on in the cards utility program, if it has one.

S.M.A.R.T. isn't necessary and can cause wide swings in the cards throughput as the drives status will be polled evey few seconds, most often interrupting the I/O. Not great for streaming video.

The array drives caches don't really enter into this since when you're streaming large chunks of data (ex: video) they pretty much saturate. Some controllers, the Fasttrak's included, use their own cache in the system memory since it can be much larger.

The Fasttrak's are HW/SW mixes like the HighPoints, but the HighPoints are far less stable and tend to have twitchy BIOS's. I won't use one.

Pro-sumer HW RAID often uses chips like the Intel i960 and i2O (Intelligent I/O) to offload the processing from the CPU to the card. These chips cores typically max out at 80 mhz +/-, and can be as low as 10-25 mhz. As such they are often slower than the system CPU at doing the tasks assigned to them.

Their drivers are typically complex having high level administration features for security, array integrity and remote administration. These can also comprimise performance by using up all those cycles, and then some, that the RAID chip took off the CPU's hands.

These cards mainly use RAID3 (similar to RAID0, but with at least one parity drive) or RAID5 (similar to RAID0. but with the parity information striped across the array). Writing the parity does slow the array somewhat in both configurations, further reducing it's utility for video.

Since these cards don't have the fast streaming performance of the HW/SW cards in RAID0 these cards are perhaps best reserved for secure servers.

They do have very nice programmable intrusion alarms and are cheap, given their advanced features, at only a few hundred USD each

I can tell you chapter and verse on this since I have a SuperTrak100 with a 66mhz i960, i2O, 6 master channels and a 64 meg onboard cache running under Win2K/XP, as well as a couple of others that I can't mention right now.

The benchmarks change somewhat from board to board...but are never even close to what the FastTrak TX4 in my main system delivers.

One othert thing: those HW RAID's are not for small cases. The largest of the bunch, the SuperTrak100, is over a foot long.... 12.5 " (31.75 cm).....and that's with the rack-mount handle removed !!

Dr. Mordrid

[Mike Newman]

When I reported 25% CPU utilisation I meant total CPU use. I can't find a figure just for the adaptec as there is nothing showing up on the task manager/processes list which uses any significant amount other than Premiere and System idle.

[Dr Mordrid]

Still, it can run afoul of SMART. This is a big problem and if you can disable it for the array, you should. When SMART checks a drive the throughput of the array can drop by up to 75% or more for the duration of the check. Can you say "frame drops"?

Dr. Mordrid

[Mike Newman]

Unfortunately there does not seem to be any option to disable SMART on this card.

[Dr Mordrid]

Have you tried doing it from the system BIOS yet? Some have an option there, but if it affects the card or not is another matter....

Dr. Mordrid

[Mike Newman]

I'll give it a try. Meanwhile here is a helpful reply from Adaptec:

Hello from Adaptec. Thank you for the response. I would agree that the CPU utilization is very high and I have forwarded this information onto the product focal. I have been told that they are aware that this is an issue in some (but not all) cases and are looking into it, possibly to address the problem with a new version of the driver in the future. I have no information as to an ETA for it at this time. Any future updates for the card will be posted on the following page.

http://www.adaptec.com/worldwide/support/drivers_by_product.html?cat=%2fProduct%2fAAR-1200A&prodkey=AAR-1200A

[Dr Mordrid]

Wow.....

You can't get more helpful than to acknowledge their product has a major problem and that they have no idea when it'll be fixed

That's the same kind of user friendly attitude that made Easy CD Creator such a pleasure to use.....

Dr. Mordrid

[Mike Newman]

No luck either on SMART - it's already dissabled in BIOS. However.... I have now upgraded the mobo BIOS and , even more significant I've realised that both disc drives are on the same connector - I imagine I did this as I planned to add more drives but it would certainly be quicker if they had a cable each. Off to the store to get another cable!