View Single Post
Old 10-11-05, 09:56 PM   #30 (permalink)
0gopogo
Aximsite All Star
 
0gopogo's Avatar
Elite Member
 
Join Date: May 2005
Location: Houston
Posts: 817
Thanked 0 Times in 0 Posts
Thanks for the link. Like I wrote on the other thread, cluster size does affect card performance in a very noticable manner.

Originally Posted by 0gopogo (on the other thread)
Aargh, don't know how I could forget about cluster size. A 2GB card must use a cluster size of no less that 32k when formatted with FAT16. The default cluster size for FAT32 is 4k. So it is really not too surprising that FAT16 performs better with many/most >= 2GB cards.

To x-check I ran the Ridata 150x 2GB formatted with FAT32 and 16k clusters through HDBench and Spb Benchmark just now.

The FAT32 results are clearly better now in both card reader and Axim than the ones I got for 4k clusters. See the other thread for details.

I don't seem to be able to format FAT32 with 32k clusters for a fair FAT16/FAT32 comparison. format.exe gives an error message saying that the cluster size is too large for the file system . SKTools does not work at all.

IMO the 2GB Ridata 150x SD is a really nice card; can't comment on the 4GB CF Kingston because I bever had one. The price is right - there even was a time when you could get it for $190 after @ newegg.com. And it should give you better write speeds in a card reader than the MicroDrive.
So here are the details I promised on the other thread.
Code:
Empty card FAT32, SC 1.1 card reader.

4k clusters                               16k clusters
--------------------------------------    --------------------------------------
17825  11211   17323    2844  Y:\1MB	  17825  11883   16817    2909  Y:\1MB
17860  11232   17290    2844  Y:\1MB	  17825  12379   16817    2844  Y:\1MB
17685   3737   17685    2618  Y:\2MB	  18188  13086   17210    2618  Y:\2MB
17834   5454   17860    2496  Y:\4MB	  18258  12468   17454    2519  Y:\4MB
18044   7080   17770    2519  Y:\8MB	  18703  10265   17466    2483  Y:\8MB
17685   7523   18188    2559  Y:\10MB	  18686  10068   17210    2481  Y:\10MB
18611   8902   17802    2539  Y:\50MB	  18720   9957   17612    2462  Y:\50MB
18723   8903   17756    2525  Y:\100MB	  18830   9913   17612    2457  Y:\100MB
I bet that with 32k clusters - which is what you get when a 2GB card is formatted with FAT16 - the results would be even closer to the FAT16 results. I also re-ran SPB Benchmark.
Code:
Empty card FAT32, x50v A03 ROM, on battery power.

4k clusters
-----------

Writing 1 MB file              1029 ms   995 KB/sec  
Reading 1 MB file               352 ms  2.84 MB/sec  
Writing 100 of 10 KB files     9591 ms   107 KB/sec  
Reading 100 of 10 KB files      564 ms  1.77 MB/sec  
Directory listing: 2000 files   147 ms  13.6 thousands of files/sec 
Speed index 214

16k clusters
------------

Writing 1 MB file               582 ms  1760 KB/sec  
Reading 1 MB file               349 ms  2.87 MB/sec  
Writing 100 of 10 KB files     8983 ms   114 KB/sec  
Reading 100 of 10 KB files      590 ms  1.69 MB/sec  
Directory listing: 2000 files   147 ms  13.6 thousands of files/sec  
Speed index 218
Note how the write speeds picked up.

I guess I'll recheck that 4GB Transcend card with FAT32/16k clusters sometimes. But I'd prefer to use 32k clusters. I'd appreciate any pointers to a tool that can format FAT32 32k clusters on WinXP.
0gopogo is offline   Reply With Quote