ATA/100 interface stepped into the world together with Intel's i815E and i820E chipsets. This data transfer interface very quickly won popularity and acquired numerous manufacturers' support. ATA/100 interface developed by Quantum was a successor to ATA/66 and allowed transferring the data at 100MB/sec along the host-to-drive bus and hence unloading the HDD buffer memory. We have already discussed the history and peculiarities of this interface in our ATA/100 Investigation article that's why we won't dwell on the features of ATA/100 here. We would like only to remind you that besides data transfer rate, the main difference from ATA/66 is an 80-conductor cable. Besides, ATA/100 is backward compatible with ATA/33 and ATA/66 devices, which makes it really universal nowadays.

And now that we have said a few words about the interface, we would like to pass over to our today's topic: ATA/100 and ATA/66 controllers. Usually you don't really care about the controller you've got in your system. Mostly you are guided by the principle: the faster - the better. And which one is the fastest? Today we are going to take a look at the performance of nearly all controllers available in market currently and to find out who the winner is. Please, meet our participants.

CMD648

This is the first ATA/66 offspring of CMD Technology Company, which is pretty well-known thanks to its SCSI developments. Its specification looks as follows:

• 2 IDE channels (up to 4 devices)
• EIDE support (DMA and multi-word DMA, including DMA mode 3 and DMA mode 4)
• Supports all PIO modes
• Automatic recognition of 80-conductor cable
• Bus-Master DMA support (133MB/sec data transfer rate)
• External BIOS support

In our roundup we decided to test the integrated on-board CMD648 controller as well as the controller card built on this controller chip, since the result shown by these two different chip representations appeared different. So, we will look at ASUS CUBX mainboard with a CMD648 on-board solution and at ASUS PCI-66C controller card.

HPT366

This controller is offered by HighPoint Technologies Company. It features:

• PCI 2.1 support
• Up to 133MB/sec data transfer rate along the PCI bus
• 2 IDE channels (up to 4 devices)
• EIDE support (DMA and multi-word DMA, including DMA mode 3 and DMA mode 4)
• Supports intl3 extensions
• Booting with external BIOS
• Bus-Master DMA support (133MB/sec data transfer rate)
• Windows 96/98, Windows NT 4.0, Windows2000, DOS, Linux compliant
• Supports all ATAPI-devices (CD-ROM, DVD-ROM, CD-R, CD-RW, LS-120, Tape, Zip)
• Allows using hard disk drives with over 8.4GB capacity

To test HPT366 controller we chose a mainboard from ABIT - ABIT BE6-II with the integrated HPT366 controller.

HPT370

This is one more controller chip from HighPoint Technologies. Here is the specification:

• 2 IDE channels support (4 EIDE-devices)
• UDMA 1-4, DMA 0-2, PIO 0-4
• Up to 100MB/sec data transfer rate in Ultra ATA/100 mode
• Automatic recognition of the HDD type
• Allows using HDDs with up to 128GB capacity
• Built-in FIFO buffer: 256Byte per channel

This controller is represented in our roundup by ABIT HotRod100 Pro controller card.

Promise Ultra66

This controller from Promise is represented by a PCI controller card. Its features include:

• 2 IDE channels support (4 EIDE-devices)
• UDMA 1-4, DMA 0-2, PIO 0-4
• Up to 66MB/sec data transfer rate in Ultra ATA/66 mode
• Automatic recognition of the HDD type
• Allows using HDDs with up to 128GB capacity
• Supports all ATAPI-devices including DVD-drives

Promise Ultra100

One more PCI controller card from Promise:

• Controller/cable combination to support 100MB/sec burst data transfers of Ultra ATA/100 drives
• Backward compatibility supports Ultra ATA/66, Ultra ATA/33, EIDE, Fast ATA-2, and ATAPI devices
• Separate device timing per port allows running drives or devices at their highest rated speeds

Promise FastTrak66

• Supports Ultra ATA/66 and Ultra ATA/33 drives
• Allows using HDDs with up to 128GB capacity
• Mirrors data to second disk or disk pair on-the-fly on every save
• Supports "hot spare" replacement drive (using 3 drives)
• Rebuilds data in background with transparent data recovery

Promise FastTrak100

This controller is similar to the previous one with that only difference that it also supports ATA/100 data transfer interface.

Intel 820 (ICH1)

This chipset from Intel features integrated ATA/66 support. As you know, the chipset consists of three hubs: 82820 Memory Controller Hub (MCH), 82801AA I/O Controller Hub (ICH1) and 82802 Firmware Hub (FWH). And ICH1 hub is responsible for the operation of HDD subsystem. Here is what it can boast:

• PCI 2.2 support
• 2 IDE channels (up to 4 devices)
• EIDE support (DMA and multi-word DMA, including DMA mode 3 and DMA mode 4)
• Bus-Master DMA support

The work of this integrated controller will be tested on ASUS P3C-2000 mainboard.

i815 (ICH1)

This Intel chipset is provided with ATA/66 support integrated into ICH1 hub. We will test its performance on ABIT SL6 mainboard

i815E (ICH2)

One more chipset from Intel. This time - with ATA/100 support. With ASUS CUSL2 mainboard we will see what this controller is worth.

VIA Apollo Pro133A (686A South Bridge)

South Bridge of this VIA chipset supports ATA/66 interface. It features:

• PCI 2.1 support
• Up to 133MB/sec data transfer rate along the PCI bus
• 2 IDE channels (up to 4 devices)
• SupportsDMA and multi-word DMA 2, including UDMA33
• Bus-Master DMA support
• Compliant with all operation systems (on the drivers level)
• Supports all ATAPI-devices including DVD-drives

We will check the performance of VIA 686A South Bridge with the help of Chaintech 6ATA4 mainboard.

Performance

Well, now let's pass over to the most interesting part of our roundup: the tests. The testbed was configured as follows:

• Intel Pentium III 600E CPU;
• ASUS P3B-F mainboard (as default) with 1005 BIOS;
• 256MB PC133 SDRAM by Hyundai;
• Creative 3D Blaster Annihilator 2 (GeForce2 GTS) graphics card (32MB DDR);
• WD273BA and IBM DTLA 307015 HDDs;
• Windows 98.

All the tests were run three times each for each testing participant and then the average value was taken for the charts and diagrams. As you have already noticed, there are ATA/100 controllers as well as ATA/66 ones among our testing participants that's why we decided to run all the benchmarks for HDDs with ATA/100 and ATA/66 interface as well. For a better comparison we also added the results obtained for i440BX based mainboard with ATA/33 support, so that you could get a better idea of the today's performance gain compared to the good old veteran.

The benchmarks were run with the following drivers:

As you see, we used the latest drivers version 3.13 for our ASUS PCI-66 and version 2.2.3 for ASUS CUBX mainboard. This will be a perfect example reflecting the influence of the drivers on the HDD subsystem functioning.

We used two hard disk drives in our tests: WD273BA and IBM DTLA 307015. We tested the work of all the participants with the first HDD and as for the second one, we focused mostly on ATA/100 controllers with it.

Both HighPoint controllers left all their rivals quite far behind. Besides, the controller integrated into i815 chipset also performed very well. However, the difference between the ATA/100 controller (i815E with ICH2) and ATA/66 controller (i815 with ICH1) is very small. ASUS PCI-66C on CMD648 chip managed to jump in-between, which it undoubtedly owes to version 3.3 drivers. If you look at the results shown by the same controller chip integrated onto ASUS CUBX mainboard, you will see that the gap between them is quite tangible. As for Promise controllers, they appeared somewhere in the middle.

Almost all the controllers integrated into the chipsets (i440BX, i820, i815) have suddenly yielded to their competitors here. But this background appeared very favorable for i815E, which managed to retain the third position. It almost caught up with HPT366. Promise FastTrak66 and FastTrak100 performed similarly in this test.

Well, ABIT HotRod100 Pro managed to make a huge jump forward in this benchmark. All the others are running neck and neck with each other and even i440BX doesn't fall too much behind.

The higher gets the HDD subsystem workload, the smaller gets HPT370's advantage over its rivals. All the participants show very close results. i440BX appears on the last position but it doesn't look an absolute loser. With IBM DTLA HDD it performed much better than any other ATA/66 controller with WD273BA HDD. Of course, it owes this victory to the increased data density of IBM DTLA model.

Well, here Promise Ultra100 proved the best of the best and then comes i820, which is really surprising. As we have expected, HPT controllers turned out outsiders in this benchmark. Note that new HPT370 controller chip in ABIT HotRod100 Pro appeared even slower than its predecessor, HPT366. All the other controllers performed very close to each other. And the last thing: CMD648 managed to show slightly better results with the newer drivers. However, the performance difference here was not so noticeable as in Disk Winbench99.

We would like to draw you attention to the fact that the controllers without RAID support perform better than those with RAID. It was probably caused by the fact that RAID controllers emulate the work of two HDDs (1+0) even if they work with only one physical harddisk. That's why the driver (DSP-controller microprogram) needs additional time to switch between two "quasi-HDDs". i815E controller proved not bad at all in this benchmark. Even i40BX looks nice against the background of HPT370 controller, which performed dramatically low here. Yes, that is exactly where the lopsided driver optimization tells.

We do know what advantages we will get installing an UDMA controller. And now let's take a look at the performance gain resulting from the use of a HDD with higher data density. We will compare performance of our testing participants on WD273BA and IBM DTLA 307015:

Not all controllers managed to make real use of the advantages provided by IBM DTLA's higher data density. Strange as it might seem, but ABIT HotRod100 Pro gained nearly the same 25% of performance as i440BX did. In fact, we expected it to show better results, because all modern controllers should theoretically perform better here.

Well, again i440BX took the lead. The second position is occupies by Promise Ultra100. All other ATA/100 controllers managed to gain only about 7-9%.

In this test, the performance gain on all controllers appeared nearly the same. The highest, though, was achieved on Ultra100 and the lowest - on HotRod100 Pro.

Conclusions

Well, it's high time we announced our verdict.

We were surprised to find out that i440BX still keeps fit and shows very good results even when working with high-performance HDDs.

The controller integrated into i815E chipset also didn't arouse any reproach and proved a good average. However, unlike most of its successful competitors it can boast one very significant advantage: you can get it almost for free, because it is integrated into the chipset.

CMD648 in ASUS PCI-66C performed just excellent. And as we have expected, the newer drivers version 3.13 ensured a relatively high performance gain.

Promise FastTrak66 and FastTrak100 didn't work wonders but showed stable high performance. However, their comparatively high cost may turn not in their favor for single-harddisk systems.

Promise Ultra100 has every right to be called the best: a combination of high performance and low cost made us reward it with the Editor's Choice. Unfortunately, Promise Company has changed quite a lot in the Ultra100 controller compared to its FastTrak100 that's why if you are fond of soldering you'd better think of something else to have fun with.

And finally, the most unlucky buddy of our roundup: HPT370 in ABIT HotRod100 Pro. At first we had stability problems with IBM DTLA HDD. The controller didn't recognize the harddisk correctly and tried to assign it UDMA2 mode. Having reflashed the new BIOS (version 6.22) we managed to eliminate this problem. In fact, we have to admit that the driver of this controller is best of all optimized for work with file system. And our tests proved this point: the controller took the lead in all file benchmarks. However, the synthetic benchmarks appeared a stumbling stone for HPT370 and it dropped down to the last position there.

All in all, we have to say that the performance difference between ATA/66 and ATA/100 controllers isn't great at all. If you are a happy owner of an ATA/66 controller, replacing it with an ATA/100 one doesn't make any significant difference. We would recommend deciding on Promise Ultra100 or CMD648 (ASUS PCI-66C, for instance). And those of you who are planning to get a new low-cost system, i815E or the good old buddy i440BX will be a nice solution.