There was some fireballs that were complete dogs. Got one, won't barely do 15MB/sec on anything, 20GB one, think it's an LE. Supposed to be ATA133 interface too 🤣

Edit: Thought you were complaining about speed overall, looks to be in the ballpark for it's age/series. Remember it's ATA33, so it's running sync on a PI but async on a PII which could be a difference.

not sure what you mean, the 430TX i also tested this on comes with the exact same PIIX4 southbridge as 440LX on the PII, yet the performance difference is quite significant.

certainly wouldn't call this drive a dog for 1996, especially for this low RPM it performs quite well IMO. but yeah, the point is really about the difference between platforms here.

Specs for the Fireball TM series are here : https://web.archive.org/web/19971119181435/ht … fireball_tm.pdf and here https://web.archive.org/web/19970713092328/ht … /fbl_tm_ch4.pdf

It's a 4500rpm drive with an ATA-2 interface (16.6MB/second max) and a max disk to buffer (internal) transfer rate of 10.7 MB/second (I wouldn't be surprised if that was raw and includes ECC and possibly other overhead) .

It is a product of its time .

nobody an idea why a p1 would beat a p2?

auron wrote on 2021-05-19, 21:50:

nobody an idea why a p1 would beat a p2?

It may be a quirk/fluke in Speedsys . Have you tried comparing results using other disk benchmarks ?

Not kidding by what guys says.

These drives that are smaller than 10GB usually are slower due to density of the platters used. When you go to 80GB or larger, then ATA 133 is realized.

Cheers,

darry wrote on 2021-05-20, 00:07:

auron wrote on 2021-05-19, 21:50:

nobody an idea why a p1 would beat a p2?

It may be a quirk/fluke in Speedsys . Have you tried comparing results using other disk benchmarks ?

here's some hdtach results from win95 osr2 on that pii system with dma off and on for this drive - quite a difference to be sure and actually more than i've expected.

even though i've read the contrary, maybe the award BIOS actually is employing IDE DMA in DOS - i'm not sure where that higher transfer rate on the p1 systems should come from otherwise. i don't have a 440LX board with award BIOS to test this with unfortunately.

This why I switched my Win 98 system to SCSI. I was running a Seagate 80GB 7,200 RPM drive on the Triton II chipset and I never could get more than 7.8 mbps. With an Adaptec 2940UW, I get a consistent 29 mbps with a period correct 9GB 10K SCSI hard drive. Makes booting prtetty nippy even on an Pentium 166 MMX.

well, using that p2 233, even this "slow" drive boots a bloated old win95 osr2 install that's been passed around through numerous boards in 35s (measured from end of post screen to finishing loading on desktop), so it's simply not an issue at all in my book. i wouldn't be surprised if the SCSI BIOS scan alone would nullify a lot of the speed gains vs. a similar IDE drive.

if you were only getting 8mb/s with a much newer drive on a controller that does 16mb/s in theory, that doesn't seem right. backwards and even forwards compatibility issues are possible on IDE - for instance the quantum drive talked about in this thread simply isn't recognized when i plug it into a promise ultra100 controller. they probably figured that nobody is going to buy a new controller to use ancient drives with anyway.

SCSI's still quite nice for its uniformity, as the adaptec controllers just force busmastering everywhere, but the IDE busmastering controllers on intel chipsets actually stole quite a bit of its thunder - i thought the HDtach results posted before show it in a very good light and adaptec's threadmark also shows significant gains on CPU efficiency. intel actually advertised the PIIX IDE busmastering as offering "SCSI-like" performance on their page back then and there's some truth to that. SCSI really shows off its advantage when running huge numbers of drives on a single chain or when leveraging its commang queing through heavy multitasking, and let's face it, neither of these things are very likely when playing around with this stuff these days.