SATA disk advanced performance?

YAYitsAndrew · Feb 11, 2008, 01:57 PM

Tons of sites tell you you can speed up vista if you're using a ups backup by turning on the disk read/write caching. I was wondering if anyone had any benchmarks on the feature? How long does it delay the reads and writes, how much hard drive improvement do you see using it?

**ChaosMinionX** · Feb 11, 2008, 02:44 PM

I believe this feature is only available when running drives as IDE in the BIOS, if you select SATA you need to install SATA drivers, most these days on new boards (if intel) use Intel Matrix, which by default turns on all the advanced features.

Also there is a chance for data loss if power is lost while the information is still cached.

Zelig · Feb 11, 2008, 03:01 PM

Quote:

Originally Posted by ChaosMinionX

I believe this feature is only available when running drives as IDE in the BIOS, if you select SATA you need to install SATA drivers, most these days on new boards (if intel) use Intel Matrix, which by default turns on all the advanced features.

I see the option selectable for both my SATA drives and my SATA RAID arary.

Quote:

Originally Posted by YAYitsAndrew

Tons of sites tell you you can speed up vista if you're using a ups backup by turning on the disk read/write caching. I was wondering if anyone had any benchmarks on the feature? How long does it delay the reads and writes, how much hard drive improvement do you see using it?

I don't notice any performance difference with it enabled/disabled. Disk performance benchmarks like hdtach don't show any difference, I haven't tried any I/O benchmarks.

I've got a UPS, so I leave it turned on, FWIW.

YAYitsAndrew · Feb 11, 2008, 03:36 PM

So perhaps you've got a hardware implementation of the caching feature either on your hard drive or your motherboard (my components in both these cases are over 2 years old) and when you enable it with vista, it's doing nothing for you.

I'm looking into buying a ups soon and I was hoping I could activate this for a little boost in hd speed but I was afraid it did absolutely nothing. I guess I'll have to benchmark and find out how it works for my particular setup. I wish someone would cover this with different hardware though and find out exactly what you should expect to gain from this feature.

mkk · Feb 11, 2008, 06:27 PM

The Enable advanced performance setting is actually a compatibility setting for software that took advantage of a bug in early versions of Windows. The bug let programs believe that a piece of data had been written to the physical layer of the disk when it could still in fact be processing in the drive buffer. When Microsoft later fixed that bug there was so many reports about database programs beginning to perform slowly that they had to bring the buggy behavior back in. This settings is a leftover from that period and no properly programmed software of today should retain this behavior, but very old database engines might.

Their choice of name for this setting is quite unfortunate.

The regular write caching is enabled by default and could be turned off for troubleshooting purposes.

Judas · Feb 12, 2008, 09:53 AM

i've tried to enable this and run test... but after checking it and hitting apply.... even a restart doesn't enable it....

this is even in the case of using IDE HDs.... i think the setting is broken.... or it only works on older chipsets.

Zelig · Feb 12, 2008, 04:08 PM

Quote:

Originally Posted by mkk

The Enable advanced performance setting is actually a compatibility setting for software that took advantage of a bug in early versions of Windows. The bug let programs believe that a piece of data had been written to the physical layer of the disk when it could still in fact be processing in the drive buffer. When Microsoft later fixed that bug there was so many reports about database programs beginning to perform slowly that they had to bring the buggy behavior back in. This settings is a leftover from that period and no properly programmed software of today should retain this behavior, but very old database engines might.

Their choice of name for this setting is quite unfortunate.

The regular write caching is enabled by default and could be turned off for troubleshooting purposes.

Do you have a source for this? I looked in the MS kb articles, but could only find stuff regarding Windows 2000 pre-SP3.

mkk · Feb 13, 2008, 12:39 AM

Certainly.

http://www.microsoft.com/technet/tec...l/default.aspx

Quote:

Windows 3.11 introduced "32-bit file access," which was a 32-bit implementation of the low-level file I/O interface. But the implementation of the Commit function contained a bug that effectively ignored requests to flush file buffers. If you took a program that flushed its file buffers and ran it on Windows 3.11, the flush call had no effect. As a result, if you lost power at just the wrong time, you ended up with a corrupted database.

The folks working on the Windows 95 file system fixed this bug, but new bug reports started to trickle in. Somebody’s accounts payable program started running really slowly. Then somebody else’s database program did the same. What was going on?

It turned out that these programs constantly issued flush calls. The programmers noticed that flush calls were really fast on Windows 3.11, so they sprinkled them liberally throughout their program. Write a byte, flush. Write a string, flush. Since the flushes were so fast, the app could commit the data to disk after every operation with no noticeable performance degradation. But once Windows 95 fixed the bug, these programs started to run very slowly since the Commit calls were suddenly doing actual work.

Of course, if the file system team had done nothing, these programs would have continued to run slowly and users would have jumped to the conclusion that Windows 95 was the problem. "Windows 95 runs like a dog," they would tell others. On the other hand, if the file system returned to the old Windows 3.11 behavior, they would have been reintroducing a bug that could lead to those pesky "file integrity problems."

So they concluded that the solution was to leave the bug fixed but add a check-box—albeit buried in the Troubleshooting page—to return to the Windows 3.11 behavior for those people who were running programs that encountered problems due to the bug being fixed.

It turned out that history repeated itself. In Windows Server® 2003, the I/O folks found a bug where requests tagged as Forced Unit Access (FUA) would lose the FUA tag and be performed as normal I/O. It was the modern-day version of ignoring flush requests! They fixed the bug but left an option to return to the old buggy behavior. The Windows Server 2003 version of this checkbox is called "Enable Advanced Performance," but now you know it really just means "Restore old buggy behavior."

Zelig · Feb 13, 2008, 05:12 AM

Great, thanks.

YAYitsAndrew · Feb 13, 2008, 03:08 PM

Man, they really spun that feature. Almost every vista tips guide on the net mentions how you can check that box for improved performance. Clearly no one's ever benchmarked this, and the only source of truth is an obscure tech article from Raymond Chen.

Maybe DH could get some publicity "benchmarking" this feature and highlighting the article and exposing this fake vista tip.

edit: Blogged about it

Vista “Enable Advanced Performance” Benchmark at YAY! it’s Andrew!