2Mb and 8Mb cache size on HDs tested in RAID

jmke · 16th June 2003, 23:58

Quote:

How does a 2mb cache harddisk compare to its bigger 8mb cache brother? We test them in different RAID setups, using real word benchmarks to show you the actual difference! Software RAID 0 / Hardware RAID 1 / 0 with Stripe sizes: 1 - 4 - 16 - 64 - 512.
HDTach/Sisoft Sandra/File Copying/UT2003

http://www.madshrimps.be/gotoartik.php?articID=69

17th June 2003, 11:21

Would have been nice to see some ATTO benchmarks and maybe copying of ISO size files and or large AVI files.

Mind copying from your source drive (you say 20Mb maxtor) probably isn't a good idea as it'll be slower that the destination drive depending which way around your doing it. so that really becomes the limitation. It should have been a faster drive that the tested (say some SCSI's).

But it is a good general review of stripesize effect.

Oh could you explain how you did the timeing of file transfers and Map loadup?

jmke · 17th June 2003, 11:34

for the file copying I used a small VB app that times the time it takes to copy the whole folder
UT2003 maploading however doesn't have such a function, so I used my trusty chronometer for that one, repeating each test 3 times

I didn't have any faster disk available, as I know it was a limiting factor during the tests, but still shows the performances differences. I was unable to run ATTO benchmark due to time constraints as I already had to hurry up alot to get the drives back to their owner (Comtech).

In the future, HD tests will be more complete, we are here to learn

thanks for your comments!

17th June 2003, 22:40

Different RAID Controllers would have been a very useful comparison. A few IDE RAID cards actually use some sort of software RAID instead of a true hardware RAID.
The controller you mentioned doesn't include info on the chipset, so that could say a lot.
Also, the major vendors like Promise and HighPoint should be considered. Their products may out perform other cards, or may be the same. That would be very useful.

17th June 2003, 23:01

indeed, but that's a whole different review you're talking about!

RichBa5tard · 17th June 2003, 23:13

Quote:

Originally posted by Unregistered
Different RAID Controllers would have been a very useful comparison. A few IDE RAID cards actually use some sort of software RAID instead of a true hardware RAID.
The controller you mentioned doesn't include info on the chipset, so that could say a lot.
Also, the major vendors like Promise and HighPoint should be considered. Their products may out perform other cards, or may be the same. That would be very useful.

The chipset used was a silicon image 680 chip. It's not slower than a comparable promise/highpoint chip.

18th June 2003, 12:24

Well, I'm the first unregistered user (post No.2), not the rest.
As you gave a nice clear reply with no skipping the issue I thought I'd register.

One reason I actually read the articel was becuase I have 2x80Gb DM9+ 8Mb in RAID0 config on my onboard Higpoint Controller (also using a simialr 1800+ CPU. I like the fact you didn't use the latest greatest and more a more common CPU, although I bet it's because that's what you had

)

Thing is my single IBM 180GXP outperforms it

Well i'll look into it someday and figure out why...

jmke · 18th June 2003, 12:48

Quote:

Originally posted by FluffyChicken
also using a simialr 1800+ CPU. I like the fact you didn't use the latest greatest and more a more common CPU, although I bet it's because that's what you had )

I chose to run the XP1700+ @ default speeds, to keep everything running as it should. The CPU can easily reach 2.2ghz on aircooling and 2.5 now in my Prometeia

thanks for registering,

What stripe size are you using on the RAID 0 array?

18th June 2003, 19:46

To be honest I've forgotten

but will check Block Size = 64K (default) haven't had a play with them yet.

What I would also be interested in as I use the NTFS file system would be how the block size/clister size does to the equation for different files and if it actually makes a difference.

19th June 2003, 13:10

That last post was me but for some reason even though logged in it posted as unreged

17th June 2003, 11:34	#3
jmke Madshrimp Join Date: May 2002 Location: 7090/Belgium Posts: 79,022	for the file copying I used a small VB app that times the time it takes to copy the whole folder UT2003 maploading however doesn't have such a function, so I used my trusty chronometer for that one, repeating each test 3 times I didn't have any faster disk available, as I know it was a limiting factor during the tests, but still shows the performances differences. I was unable to run ATTO benchmark due to time constraints as I already had to hurry up alot to get the drives back to their owner (Comtech). In the future, HD tests will be more complete, we are here to learn thanks for your comments! __________________

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
2Mb and 8Mb cache size on HDs tested in RAID	Sarcastro	WebNews	1	15th June 2003 02:43

17th June 2003, 11:21	#2
Unregistered Posts: n/a	Would have been nice to see some ATTO benchmarks and maybe copying of ISO size files and or large AVI files. Mind copying from your source drive (you say 20Mb maxtor) probably isn't a good idea as it'll be slower that the destination drive depending which way around your doing it. so that really becomes the limitation. It should have been a faster drive that the tested (say some SCSI's). But it is a good general review of stripesize effect. Oh could you explain how you did the timeing of file transfers and Map loadup?

17th June 2003, 22:40	#4
Unregistered Posts: n/a	Different RAID Controllers would have been a very useful comparison. A few IDE RAID cards actually use some sort of software RAID instead of a true hardware RAID. The controller you mentioned doesn't include info on the chipset, so that could say a lot. Also, the major vendors like Promise and HighPoint should be considered. Their products may out perform other cards, or may be the same. That would be very useful.

17th June 2003, 23:01	#5
TeuS Posts: n/a	indeed, but that's a whole different review you're talking about!

18th June 2003, 12:24	#7
FluffyChicken Posts: n/a	Well, I'm the first unregistered user (post No.2), not the rest. As you gave a nice clear reply with no skipping the issue I thought I'd register. One reason I actually read the articel was becuase I have 2x80Gb DM9+ 8Mb in RAID0 config on my onboard Higpoint Controller (also using a simialr 1800+ CPU. I like the fact you didn't use the latest greatest and more a more common CPU, although I bet it's because that's what you had ) Thing is my single IBM 180GXP outperforms it Well i'll look into it someday and figure out why...

18th June 2003, 19:46	#9
Unregistered Posts: n/a	To be honest I've forgotten but will check Block Size = 64K (default) haven't had a play with them yet. What I would also be interested in as I use the NTFS file system would be how the block size/clister size does to the equation for different files and if it actually makes a difference.

19th June 2003, 13:10	#10
FluffyChicken Posts: n/a	That last post was me but for some reason even though logged in it posted as unreged