Even the fastest IDE hard disk only just reaches a sequential data transfer rate up to a maximum of 45 MBytes/s. Considering this actually an Ultra-ATA/66 interface should be sufficient to provide full performance for the current drives.
Why then a new interface at all? The initiator of Ultra-ATA/133, Maxtor presupposes an increase in performance of 40 percent yearly for hard disks while according to our Hard Disks Tests 20 percent are much more realistic. Due to the command overhead and the status informations the full bandwidth of the interface can not be used, anyway. Refering to our measurements the Ultra-ATA/100 interface permits maximum burst rates of 85 MBytes/s at best. Maxtor even maintains only 62 percent of the theoretical interface speed, that is 62 MBytes/s, can be fully used. Future Ultra-ATA hard disks with sequential data transfer rates of 62 MBytes/s or more are meant to be braked at this point already.
With Ultra-ATA/133 Maxtor wants to counteract this right now. Whether the new interface will find a similar spreading as its predecessors still remains uncertain. Especially as the expected follow-up model of the parallel IDE hard disk interface is ready for the off by now: Serial-ATA.
We held a little poll to learn about the manufacturers' opinions on Ultra-ATA/133. Will Ultra-ATA/133 remain a solo effort by Maxtor? And in addition we tested whether the users will benefit right away from Ultra-ATA/133 and we asked where the problems may be found.
Specifications
Maxtor's step from Ultra-ATA/100 towards Ultra-ATA/133 is not very spectacular from a technological point of view. Concerning the specifications basically the timings changed. Both the data width of 16 Bit and the 40-pin plug are identical with the present ATA standards. At the moment the T13-Committee is working on the inclusion of Ultra-ATA/133 into the official ATA/ATAPI specifications.
Ultra-ATA/133's higher data transfer rate of 133 MBytes/s (calculating on base 1024 that is 127.2 MBytes/s) is achieved by shortening the cycle time from 40 ns (ATA/100) to 30 ns. Within this window of time the interface transfers its data in burst mode with the falling and the rising timing edge. This DDR procedure has already been used since Ultra-ATA/66.
Mode | Transfer Rate | Cycle Time | Cable |
|---|---|---|---|
The transfer rates are calculated on base 1000. | |||
ATA/133 | 133 MBytes/s | 30 ns | 80-core |
ATA/100 | 100 MBytes/s | 40 ns | 80-core |
ATA/66 | 66 MBytes/s | 60 ns | 80-core |
ATA/33 | 33 MBytes/s | 120 ns | 40-core |
For Ultra-ATA/133 operation you need - as well as for Ultra-ATA/66 and /100 - an 80-core ribbon cable. Each signal line and data line is assigned to an own mass line here. That is necessary to reduce "crosstalk" between the signal lines during high frequencies. In case of connecting a 40-core IDE cable the interface will automatically switch over to the much slower Ultra-ATA/33 operation mode.
Compatibility
While developing specifications for Ultra-ATA/133 Maxtor has taken into consideration full downward compatibility to older ATA standards. The operation of IDE drives only mastering PIO-Mode 0 therefore also is possible with the Ultra-ATA/133 interface. For example, in our test a seven year old Quantum ProDrive LPS with 52 MBytes capacity worked with an Ultra-ATA/133 controller without any difficulty. And there is no problem using Ultra-ATA/133 PCI controllers in old mainboards. Our tests did not show any incompatibilities.
Because of their downward compatibility Ultra-ATA/133 drives run without any problems with Ultra-ATA/100, ATA/66, and ATA/33 interfaces as well. The drives will adapt their interface speed automatically to the slower interfaces.
In the Master/Slave operation mode of an Ultra-ATA/133 hard disk with an Ultra-ATA/33 or /66 drive at an IDE channel each of both can work in its highest transfer mode. In our test we did not find any difficulties.
Ultra-ATA/133 hard disks work with the current 80-core IDE ribbon cables. According to Maxtor no new cables with better sreening features are necessary for a stable Ultra-ATA/133 operation. Using a 40-core ribbon cable for the interface means forcing the drives automatically into the Ultra-ATA/33 operation mode. On the other hand, both types of cable have the 40-core plug in common.
Maxtor DiamondMax D540X
With DiamondMax D540X and D740X Maxtor launches the very first hard disks with Ultra-ATA/133 interface available.
DiamondMax D540X turns with 5400 RPM and is available with 120 and 160 GBytes capacity. In this drive Maxtor uses platters of 40 GBytes size. A data puffer of 2048 KBytes size comes along with the Ultra-ATA/133 interface.
Maxtor offers DiamondMax D540X as well in a Ultra-ATA/100 version. The capacities here are limited to 20, 40, 60, and 80 GBytes. Fortunately, no additional charge has to be payed for these Ultra-ATA/133 versions with the new interface.
To address the 160 GBytes model with full capacity controllers with a 48 Bit addressing are necessary. According to the ATA specifications 128 GBytes maximum are to be addressed over the - until now usually - 28 Bit wide address room. Maxtor enlarges the address room to 48 Bit. At this very moment the T13-Committee is including the 48 Bit addressing into the ATA/ATAPI-6 specifications. So now up to 128 Petabytes can be addressed. In the BIOS of the Ultra-ATA/133 controllers the 48 Bit addressing is already integrated. This article separately deals with the problems the hard disks nevertheless still show.
In the benchmarks DiamondMax D540X reaches a maximum sequential data transfer rate of 35.3 MBytes/s, what's a really good result for a hard disk with 5400R/min. And even in the practice test the drive showed a convincing performance.
Interface | Ultra-ATA/133 |
|---|---|
| |
Revolutions | 5400 RPM |
Cache | 2048 KBytes |
Platter Capacity | 40 GBytes |
Available Capacities | 120 / 160 GBytes |
Price | 300 / 380 Euro |
Maxtor DiamondMax D740X
Maxtor's second Ultra-ATA/133 series is DiamondMax D740X. These hard disks belong to Maxtor's high performance class and correspondingly work with 7200 RPM. Like DiamondMax D540X these magnetic disks take 40 GBytes. The manufacturer offers the drives as 1 and 2 platters versions with capacities of 20, 40, 60, und 80 GBytes. The Ultra-ATA/133 interface of DiamondMax D740X can fall back upon a data puffer of 2048 KBytes size.
In the low-level benchmarks DiamondMax D740X reaches a maximum sequential data transfer rate of 40.3 MBytes/s - a top result. And in the practice tests the hard disk also walks off with a good mark. Along with the cache the quick access times of only 6.99 ms within the first 500 MBytes attract attention.
Interface | Ultra-ATA/133 |
|---|---|
| |
Revolutions | 7200 RPM |
Cache | 2048 KBytes |
Platter Capacity | 40 GBytes |
Available Capacities | 20 / 40 / 60 / 80 GBytes |
Price | 110 / 130 / 170 / 200 Euro |
ATA/133 Controllers
The first mainboard chipsets with Ultra-ATA /133 support are offered by VIA: For KT266A and P4X266A VIA uses the southbridge VT8233A. The southbridge VT8233 (without A) only provides the user with Ultra-ATA/100 support.
Among the other chipset manufacturers ALi and SiS have announced support for Ultra-ATA/133 as well. Their first chipsets will be launched very soon. Intel refrains from Ultra-ATA/133 for the time being and will do an about-turn until the end of 2002 Serial-ATA.
If you do not want to buy a new mainboard, you can use an Ultra-ATA/133 controller card. The PCI cards add two independent supplementary channels to the two IDE ports on the mainboard.
For our test we had ACARD, HighPoint and Promise Ultra-ATA/133 PCI controllers at our disposal. Maxtor offers Promise Ultra133 TX2 also as bundle version with the 160 GBytes hard disk DiamondMax D540X.
While Promise Ultra133 TX2 and ACARD AEC-6280 only have two normal IDE channels, HighPoint RocketRAID 133 and ACARD AEC-6880 show up with RAID functions. In the controllers' BIOS the connected hard disks can be combined either to RAID-0 (stripe set) or to RAID-1 (reflection). Particularly with RAID-0 a transfer rate of 133 MBytes/s makes sense. Two Maxtor DiamondMax D740X including overhead are already reaching the limits of Ultra-ATA/100. Altogether it is possible to connect up to four hard disks to one RAID with the controllers.
For an Ultra-ATA/133 PCI controller, like Promise Ultra133 TX2, you have to pay round about 80 Euro at Alternate's. The ATA/100 version on the other hand is available for fifty percent of this price already. The RAID controllers remain an expensive bit of fun: HighPoint's RocketRAID 133 for example will cost you 129 Euro according to price lists. And for ATA/100 RAIDs you still have to pay round about 110 Euro.
Benchmarks: Transfer Rates
In our previous tests of Ultra-ATA/100 hard disks the maximum interface speed was approximately 87 MBytes/s. In the burst tests a hard disk transfers its data from the fast data puffer.
But you can not expect clearly more from Ultra-ATA/100, though. On the one hand the theoretical maximum is 95.4 MBytes/s because the PR effective 100-detail works on base 1000 and therefore names millions of Bytes instead of MBytes on base 1024. On the other hand beside the data also commands and status informations are transferred which occupy bandwidth.
Calculating on base 1024 Ultra-ATA/133 provides a theoretical interface speed of 127.2 MBytes/s. The Ultra-ATA/133 hard disk Maxtor DiamondMax D740X with a gauged burst rate of 117.4 MBytes/s therefore clearly benefits from the new interface. In comparison to the operation at an Ultra-ATA/100 controller the burst rate jumps about 30 MBytes/s. This approximately corresponds to the increase from ATA/100 to ATA/133.
The top result of 117.4 MBytes/s we only reach with Promise Ultra133 TX2, though. HighPoint's RocketRaid 133 only allows a transfer rate of 101 MBytes/s. In comparison to an Ultra-ATA/100 controller this is just 14 poor MBytes/s more.
Controllers | Burst Data Transfer Rate [MBytes/s] |
|---|---|
All results are determined with Maxtor DiamondMax D740X. | |
ACARD AEC-6820 | 104.5 |
HighPoint RocketRaid 133 | 100.5 |
Promise Ultra133 TX2 | 117.4 |
VIA VT8233A Southbridge | 106.3 |
Compared with that current IDE hard disks show sequential data transfer rates of slightly more than 40 MBytes/s. The actual Read-/Write-Procedure onto the magnetic disk therefore is carried out far more slowly as the interface allows it. Do you add on the command overhead to the maximum sequential data transfer rates, then only Ultra-ATA/66 controllers are nearly reaching their limits at the moment.
So just on the basis of the figures it is proved that using an Ultra-ATA/133 controller - even if it is used as RAID-0 - only is worthwhile if you run more than two disks.
Benchmarks: Practice Results
In our practice tests the high interface speed of Ultra-ATA/133 should show some effect. While reading, writing and copying files the cache of the hard disk is of importance because it has to puffer data. Our benchmark tecMark reads, writes and copys under Windows XP files of different size. Ideal requirements to use the cache of the hard disk.
The results are a disappointment, though: The reading of files in Ultra-ATA/133 operation takes place only 0.7 MBytes/s faster. That almost makes no more difference to the range of the benchmarks. And the benefit of Ultra-ATA/133 is even more low down while writing and copying files. The following table gives you a survey of all results. We tested Maxtor DiamondMax D740X under Windows XP with Promise Ultra66, Ultra100, and Ultra133 TX2. So we were able to test directly the effects of the different interface speeds on the practice performance.
Controller | Read [MBytes/s] | Write [MBytes/s] | Copy [MBytes/s] |
|---|---|---|---|
All results are determined with Maxtor DiamondMax D740X. | |||
Promise Ultra133 TX2 | 28.2 | 14.9 | 16.1 |
Promise Ultra100 | 27.5 | 14.7 | 15.7 |
Promise Ultra66 | 26.4 | 14.4 | 15.5 |
Manufacturers back Serial-ATA
Our little poll among the manufacturers of hard disks on the question of future interfaces in desktop hard disks showed a very clear tendency in favour of Serial-ATA:
Fujitsu
In August 2001 Fujitsu stopped its production of desktop hard disks with ATA interfaces. The manufacturer will concentrate on SCSI drives for server as well as for notebook hard disks. But nonetheless Fujitsu showed a 3.5 inch hard disk with a Serial-ATA interface at the Comdex 2001. Fujitsu is planning to establish the Serial-ATA technology for its 2.5 inch hard disks halfway through the year 2002. The interface shall have an integrated power connection.
IBM
IBM is not going to produce any Ultra-ATA/133 hard disks. In January 2002 Deskstar 120GXP will be available and this hard disk still will have an Ultra-ATA/100 interface. For future products IBM, being a member of the Serial-ATA Working Group, backs Serial-ATA. But the manufacturer does not expect broad availability before the end of 2002. Possibly the first Serial-ATA drives will get a parallel interface in addition. The company thinks that will make it much easier for users to migrate to serial IDE interfaces.
Maxtor
Maxtor probably will remain the only manufacturer of Ultra-ATA/133 hard disks. But of course Maxtor like its competitors is planning Serial-ATA too. However, the first of these products will not be available before the end of 2002. For a transitional period Maxtor will then offer its hard disks in two versions - with serial or parallel interface.
Seagate
Seagate is proceeding like IBM and will directly do an about-turn to Serial-ATA. The date for the first availability of the new Serial-ATA hard disks Seagate gives is much more optimistic though, the first quarter of 2002. The manufacturer sees no necessity for offering Ultra-ATA/133 drives inbetween. Seagate like IBM is a member of the Serial-ATA Working Group.
Western Digital
Western Digital will switch over to Serial-ATA for its future products as well. According to the manufacturer the first drives shall be launched halfway through the year 2002. For a problem-free connection of Serial-ATA hard disks to standard parallel ATA controllers adapters shall be available. An integration of serial and parallel interfaces directly on the drive is not planned because of the costs according to Western Digital. Western Digital would only produce Ultra-ATA/133 hard disks if there were a great demand of it by OEMs.
Details on Serial-ATA
The Ultra-ATA/133 interface will probably be the last parallel interface for IDE hard disks. At the Intel Developer Forum in February 2000 Intel has shown for the first time a serial connection for ATA hard disks with a bandwidth of 150 MBytes/s. In August 2001 the Serial-ATA Working Group has introduced the final specification 1.0.
Halfway through the year 2002 Serial-ATA shall replace the parallel IDE interface. The manufacturers of PCs never ever really liked the broad ribbon cables. They are expensive, unwieldy, and error-prone. Serial-ATA is based on only two pairs of lines: one for reading and one for writing. The new hard disk interface is a point-to-point connection, per drive one connection is necessary. Therefore the Master/Slave and the CS jumper on the terminals are dropped. There shall be up to four direct connections for devices at the Serial-ATA controllers.
Apart from the hardware implementation for Serial-ATA no other adaptations are necessary. The firm- and software work with the same registries and commands as the parallel ATA interface. With that simple adapters can be realised, which convert the signals of the previous 40-core ribbon cables to the four-line cables. In February 2000 Intel has already shown first protoptypes of these adapters. And so old boards or ATA devices would stay compatible to the new standard .
Serial-ATA version 1.0 allows a data transfer rate of 150 MBytes/s. The second generation of Serial ATA shall already work with 300 or 450 MBytes/s. For full information on the basic structures of Serial-ATA standards please do read this extra article.
Conclusion
Our benchmarks speak for themselves: No person at all really needs an Ultra-ATA/133 interface. This result was to be expected, for even an Ultra-ATA/66 interface almost is sufficient for the currently fastest IDE hard disks.
Using a Maxtor Ultra-ATA/133 hard disk you should not get dazzled by the high interface speed. The benefit of higher burst rates falls flat in practice. And the money for an additional controller should better be saved, after all an Ultra-ATA/133 PCI card costs round about 80 Euro at the moment. Is the Ultra-ATA/133 integrated in the mainboard chipset, like in VIA's VT8233A southbridge, there's nothing against the new interface. In this case though no additional charge is asked for.
But even the VIA chipsets are obstructing the success of Ultra-ATA/133 cards. With their low PCI performance the VIA chipsets reduce the hardly measureable benefit of Ultra-ATA/133 to absurdum. Given this situation even an Ultra-ATA/100 controller is faster. We have already reported in detail on these problems, please do read our (English) extra article VIA Chipsets Slow down PCI Cards.
The manufacturers of hard disks are sceptical towards Ultra-ATA/133. Except Maxtor all of them want to switch over directly to Serial-ATA. Even for Maxtor Ultra-ATA/133 is nothing more but an intermediate stage and the manufacturer already works on the serial interface version as well.
And finally, for a résumé: There is no obstacle to a hard disk with Ultra-ATA/133 interface because in this case Maxtor does not ask for an additional charge. An additional Ultra-ATA/133 PCI controller is not recommendable and the higher price can in no way be justified. Merely RAID solutions can benefit from Ultra-ATA/133, if you want to work with more than two hard disks. (cvi / bmu)
Test Setup
We carried out all our tests of Ultra-ATA/133 components in an exactly fixed configuration. As operating system we used Windows XP. The Windows XP installation was streamlined on performance and ran with the classic surface. In addition, all energy-saving modes were active.
Component | Detail |
|---|---|
| |
Mainboard 1 | MSI 845 Pro2 |
Class | Socket 478 |
Chipset | Intel i845 |
BIOS | from 24.10.2001 |
Mainboard 2 | Shuttle AV45GTR |
Class | Socket 478 |
Chipset | VIA P4X266A & VT8233A |
BIOS | v6.00PG from 06.11.2001 |
Processor | Intel Pentium 4 |
Clock Speed | 1.5 GHz |
Socket | Socket 478 |
Hard Disk 1 | Maxtor DiamondMax D540X Model 4G120J6 |
Capacity | 120 GBytes |
Interface | Ultra-ATA/133 |
Hard Disk 2 | Maxtor DiamondMax D540X Model 4G160J8 |
Capacity | 160 GBytes |
Interface | Ultra-ATA/133 |
Hard Disk 3 | Maxtor DiamondMax D740X Model 6L40J2 |
Capacity | 40 GBytes |
Interface | Ultra-ATA/133 |
Hard Disk 4 | Maxtor DiamondMax D740X Model 6L80J4 |
Capacity | 80 GBytes |
Interface | Ultra-ATA/133 |
PCI Controller 1 | HighPoint RocketRAID 133 |
Class | Ultra-ATA/133 |
Firmware | v2.1 |
PCI Controller 2 | Promise Ultra133 TX2 |
Class | Ultra-ATA/133 |
Firmware | v2.20.0050.10 |
PCI Controller 3 | ACARD AEC-6280 |
Type | Ultra-ATA/133 |
Firmware | --- |
PCI Controller 4 | ACARD AEC-6880 |
Type | Ultra-ATA/133 RAID |
Firmware | --- |
PCI Controller 5 | Promise Ultra100 |
Type | Ultra-ATA/100 |
Firmware | v2.00 (Build 12) |
PCI Controller 6 | Promise Ultra66 |
Type | Ultra-ATA/66 |
Firmware | v1.12 |
Graphics Card | Leadtek WinFast GeForce3 TD |
Graphics Chip | NVIDIA GeForce3 |
Graphics Memory | 64 MBytes DDR-SDRAM |
BIOS | 04.32.2001 |
Driver | Detonator 23.11 |
Circuit Board | Rev: B |
RAM 1 | Corsair XMS2400 |
Capacity | 256 MBytes |
Type | PC266 CAS=2,0 |
Chips | MT 46V16M8-75A |
RAM 2 | Wichmann WorkX AG |
Capacity | 256 MBytes |
Type | PC133 CAS=2,0 |
Chips | MT 48LC16M8A2-75b |
CD ROM Drive | LITE-ON LTN-382 |
Speed | 40x |
Firmware | -- |
Interface | EIDE-UltraATA/33 |
Disk Drive | Teac FD-235HF |
Capacity | 1.44 MBytes |
Mains Adaptor | Delta Electronics |
Model | DPS-300KB-1A |
Power Output | 300 Watts |
Format | ATX |
Keyboard | Cherry RS 6000 M |
Interface | PS/2 |
Mouse | Logitech M-S35 |
Interface | PS/2 |