HDTach CPU usage

Schlotkins · Post by **Schlotkins** » Sun Mar 29, 2009 9:12 am

I have a quick question on HDTach's CPU Usage. Of course, the smaller the number the better. However, it seems to me this is pretty much a number that is correlated with the performance of the drive.

For example, see:

http://techreport.com/articles.x/16291/10

Why am I interested? I'm running a program which is pretty memory stupid and uses a lot of HDD for temp files. I'd like to speed up the temp file drive. I basically have three options:

1) Some old i-RAMs RAIDed together
2) some fast SSDs
3) RamDisk

2 seems to be the least appealing because I know I am going to do a lot of writing to this disk. Given SSDs have a finite number of writes and I don't need to actually store any of this data, it seems a bit silly to go that route.

#1 is OK, but I either need to get the box option or have some PCI slots. I probably don't need more than 8 GIGs of RAM so 2 would work. Of course it's a bit silly beacuse you only get 250 MB/s RAIDed together when the memory could do 3 times that.

#3: I downloaded RAMdisk 9 plus (the demo). It gets about 1600 MB/sec on my laptop which is fine. Since I have 12 GB in my box I'd run this on, I could easily allocate 6GB for a RamDisk. The problem is HD Tach told me CPU usage was 50% when this was going on. That's huge and of course, the program needs some CPU as well. I'm trying to get an idea if that 50% number really means anything outside of you have a really fast drive.

Thanks for any assistance!
CHris

Kepakko · Post by **Kepakko** » Sun Mar 29, 2009 12:28 pm

HDTach is terrible old benchmark that tells really nothing about real life perfomance.

tackle · Post by **tackle** » Mon Mar 30, 2009 10:24 am

Kepakko wrote:HDTach is terrible old benchmark that tells really nothing about real life perfomance.

Can you recommend any better harddrive benchmark tool?

Blue_Sky · Post by **Blue_Sky** » Mon Mar 30, 2009 2:49 pm

Iometer is as close as you can get to real world performance. Hdtach and Hdtune test sequential read AFAIK, which isn't something that occurs often, naturally.

Munters · Post by **Munters** » Tue Mar 31, 2009 7:01 am

Not a problem at all: if hdtach pushes the cpu load from nearly nothing to about 50% you really found (and probably removed) the bottleneck.
In fact it would be perfect if hdtach uses 100% cpu, as this would mean the processor never stalls waiting for I/O, just running the hdtach code as fast as possible.

Schlotkins · Post by **Schlotkins** » Thu Apr 02, 2009 8:19 am

Thanks for the replies!