Depends on model and brand

Id say cf is too slow for windows

ps: the controller has a big effect on speed

ps: my 512mb klick photprint sd card has a transfer rate of 3.8mb/sec according to hd-tach my hard drive does about 100mb/sec

Ahem we're connecting them through ATA, not through USB.

true but I can imagine the bottleneck to still be the card
my pretech 128mb usb thumb drive has a transfer rate of 11.9 mb/sec

and usb 2.0 theoretically is rated at 60mb/sec but in the real world because of overheads ect, you wont get over 40mb/sec

How can I benchmark the speed of my CF device in Windows 98? I DL'd HD Tach, but an error message told me that it was for Windows NT. So, I DL'd an older version of HD Tach (2.61), but you cannot select Write testing in the trial version! Thanks a lot for any info.

ps: the controller has a big effect on speed

You say that, and yet at the same time you seem to be entirely overlooking that point. That Lexar 133X card, running properly through an IDE adapter, should be getting somewhere near 25MB/sec read and 20MB/sec writes. Those USB multi-readers are notorious for being slow, and the results you're showing are certainly limited severely by the reader rather than the cards.

A decent 133X or greater CF card will significantly outperform any Win98-era hard drive, and should at least hold its own against more recent hard drives due to the low access times. Particularly old or cheapo CF cards wouldn't be good for Windows though, since they often sport much lower transfer speeds, and tend to lack DMA support.

I should have provided more information about the hardware I'm using. The Compact Flash device is a Transcend 133x 4 GB card. It's connected to a CF reader, which has a sticker on it saying DMA. That CF reader device is attached to an 80-pin IDE cable, which in turn is attached to a UIDE Raid port on the mobo. How can I check if DMA is "activated" for the CF device?

Also, I tried downloading some more software for Windows 98, in order to benchmark the writing ability of the CF device. I tried Emsa DiskCheck, and DiskSpeed 32. Unfortunately, both don't appear to have write testing functionality.

retro games 100 wrote:

How can I check if DMA is "activated" for the CF device?

Windows hardware manager

Compact flash cards are native IDE devices, and DMA is activated the same way as any other hard drive... check the little "Use DMA mode" box in the disk properties.

As far as benchmarking the write speeds, that's a tough one... Maybe try a Linux LiveCD and something like Bonnie++?

Please note that the CF apparatus is connected to the mobo's UIDE Raid port. If I go to the Win98 hardware manager area, I cannot see the DMA "checkbox" in the properties section. This is because the HDD is being managed by the HighPoint Raid controller, and not the mobo's integrated IDE port / IDE controller. If I right-click on the Raid controller, it doesn't tell me if DMA is active or not. I installed a Raid administrator and then examined its options. There appears to be an option which says "Set transfer mode". This option is disabled/greyed out. It looks like I need to create a "raid array" before I can do anything, but I've only got one HDD, and so it won't let me create any kind of array. (Please note - I've never worked with Raid before, and so I don't understand the various Raid options available.)

I'm currently running a full HDD test using "Speedsys". It's taking a long time. I can't remember if that reports any write testing information, but if it does, I'll post back with any info...

It could very well be that the controller's drivers natively have DMA enabled... but if you want to be sure, there's a registry hack to bring back the setting. Here's a link, scroll down to the "ENABLE UDMA/RAID HARD DISK DMA" section about halfway down the page.

I found a HDD benchmarking utility. It's called Michael's Disk Benchmark v.1.11, and can be found at the bottom of this webpage -

http://www.acnc.com/04_02_02.html

I ran it on Windows 98 SE, and used the following options...Please note that the options I picked were purely "guess work". I don't know if the options I selected were well chosen, or not.

File size = 256 mb, Test data = Random.

The results were:

_______________Cached_____Non-cached
Sequential Write: 7781 kb/sec 7408 kb/sec
Sequential Read: 29594 kb/sec 34583 kb/sec
Random Write: 37871 kb/sec 206 kb/sec
Random Read: 2215 kb/sec 34379 kb/sec

Do these results look OK? Wow, look at that non-cached random write value. Has this particular write test pin-pointed where the CF device cannot perform well?

Those results look a bit screwy... cached random reads shouldn't be 15x slower than non-cached ones, for example. Non-cached random writes are going to be slower, but 206KB/sec seems way off.

I tried that program on my system with the same settings and got equally odd results... like, 889MB/sec random reads on a single hard drive. I don't really know the ins-and-outs of hard drive benchmarking, but something ain't right.

That's a pity about the program's unreliable looking results. I DL'd another utility, and again it was found on this webpage -

http://www.acnc.com/04_02_02.html

It's called Adaptec ThreadMark, and I'm running it now. It warns the user that it takes about an hour to complete. I'll post back later with the results...

The ThreadMark benchmark utility has finished, and it says:

Data Transfer Rate (MB/sec) = 20.18
Average CPU Utilization = 52.45

The test also produced a report file, and I have attached a copy to this message. It's in standard .txt text format. Its original file name extension was .rpt (report), but I renamed it to .txt, in order to successfully upload it to Vogons. The write tests appear to be the tests where the "Destination" is a Write .txt file name. What's interesting is that when these tests are done, the CPS and MB/sec values are reported to be 0. I don't know what CPS means, but something doesn't seem right to me.

I can tell you that my EeePC 900 with its 4GB and 16GB SSDs suffers from similar stuttering problems. It is due to very slow write speed with small files (esp many small files at once) and the lack of a cache on the SSDs. Old hard disks usually had a data cache even if it was only like 64KB. The 16GB SSD's speed is similar to a floppy drive when it is writing small files!

I'm pretty sure the CF stutters are the same problem.

How are DOS games with and without smartdrive?

There is a little partition tweak you can do for flash drives that gets a bit of a speed boost. You align the partition starting point to the erase block size of the flash memory so that writes don't overlap more flash cells than necessary.
http://www.ocztechnologyforum.com/forum/showt … ve-working-life.
I recommend offsetting the partition by 1MB as suggested here because it works with basically all flash chips.
http://www.techpowerup.com/articles/other/157

The lack of cache on SSDs seems a curious design decision. I can understand its absence on a Compact Flash device, because I guess these are primarily designed as a cheap digital camera medium. Unfortunately at this moment in time, I haven't got the resources to test a DOS based CF set up. Perhaps someone else reading these messages can try this?! Also, it seems that partition alignment importance for a SSD is associated with Windows XP, rather than Win9x.

win 9x has a disk cache called vcache
some details
http://www.putergeek.com/vcache/

Thanks. I tried setting the vcache, but unfortunately the Mysteries of the Sith demo still freezes about 10 seconds in to the game. I wonder if that particular game needs to write to a disk file during gameplay, for whatever reason. It's no big deal, but it does highlight the potential downside to using a compact flash device. I reckon if I swapped out the flash drive for a real HDD, that "shudder" would be almost unnoticeable. (I'd like to test this, but I don't have the time right now.)