VIA Chipsets Now Supporting UltraATA/100
Hard drives with the latest IDE interface (UltraATA/100) have been available for some months now. Technically, this new standard is not really a big step forward, as it was 'just realized' by tightening the bus timings. Former IDE standards used to double the transfer speed with every new generation: After 16,6 MB/s at PIO4, UltraDMA/33 enabled 33 MB/s. Today's standard UltraDMA/66 (or UltraATA/66) is now being replaced by UltraATA/100.
Thanks to those little changes, manufacturers could easily adopt the new standard for their current IDE hard drives. Still, there is no drive available which data transfer rate from the physical disk media would be fast enough to benefit from the vast interface bandwidth. Due to this, critics characterize the catchword "UltraATA/100" as being rather loaded with advertising appeal than with any substance. So far, people who wanted to have ATA/100 support were either forced to obtain a motherboard with Intel's i815e, i820e or i840e chipset, or a separate PCI controller card, with chips from Promise, AMI or HighPoint. With the inclusion of this standard to VIA's chipsets, UltraATA/100 should now be widely available, even in low-cost systems.
The first chipset with native ATA/100 support was Intel's Solano (815E), which was the initial product making use of Intel's ICH2 south bridge chip. Three months later, VIA has finally completed their south bridge chip VT82C686B. Thanks to VIA's modular chipset architecture, the new South Bridge can be used with VIA-chipsets for Athlon/Duron as well as for Pentium III/Celeron. We will now take a look at VIA's renewed products.
The chipset takes control of all components, which are found on the motherboard. Most sets consist of two chips, which are called North and South Bridge. The North Bridge manages all data transfers between the CPU, main memory and the Accelerated Graphics Port (AGP). Things like I/O-ports (serial, parallel), USB, keyboard controller, AC97 sound, floppy controller and IDE controllers are part of the South Bridge.
This graphic shows the block diagram of VIA's KT133 chipset for AMD Athlon or Duron processors. Here the North Bridge is called VT8363. VIA's new 686B is pin-compatible with its predecessors, so the one and only difference is the ATA interface.
The North and South Bridge talk to each other via the PCI bus. An exception is Intel's recent chipsets 810, 815, 820 and 840, which come with a different interconnect capable of working at 266 MB/s (twice as fast as the PCI). Also, Intel does not want their chips to be called North and South Bridge anymore, but (G)MCH and ICH (Graphics and Memory Controller Hub, I/O Controller Hub).
VIA vs. Intel
VIA and Intel have two South Bridge solutions each:
Besides their excellent modularity, VIA chipsets are very popular for their attractive prices. Today motherboards with Intel's 815 chipset can be up to $ 50 more expensive than products with a VIA chipset!
In addition, the 815 is not capable of working with two CPUs - only the 'shop-soiled' BX is. In contrast, VIA's 694X supports one or two Pentium III CPUs.
Dual Athlon systems are still not quite available yet. AMD presented a first working system a few weeks ago, but we don't expect systems to be available until Q1/2001.
||Intel ICH2 (82801BA)
||810 (PC100 SDRAM), 815 (PC133 SDRAM), 820 or 840 (each RDRAM) for Intel Pentium III, Celeron or VIA Cyrix III
||Apollo Pro 133A (VT82C694X) for Intel Celeron, Pentium III, VIA Cyrix III or Apollo KT133 (VT8363) for AMD Athlon, Duron socketed or Apollo KX133 (VT8371) AMD Athlon slot model
||UltraATA/100 (ICH1 only ATA/66)
||UATA/100 (VT82C686A only ATA/66)
|Multiprocessor Support of North Bridge
||Apollo Pro 133A (VT82C694X)
||No/Yes (820+840 w/ RDRAM only)
||66, 100, 133 MHz
||66, 100, 133 MHz
||PCI 2.2, max. 6 Slots
||PCI 2.2, max. 6 Slots
||2x serial, 1x parallel
||2x serial, 1x parallel, floppy, keyboard, mouse, LM78
||Up to 4 Ports
||Up to 4 Ports
||approx. $ 45
||approx. $ 28
||Intel Celeron 500 or AMD Duron 700
AOpen AX3S Pro (i815 w/ ICH2)
AOpen AX6C (i820 w/ ICH1)
AOpen AX34 Pro (VIA 694X w/ 686A)
Azza KT133BX (KT133 w/ 686B)
||1x 128 MByte SDRAM (Corsair)
PC133, 7ns, CL2
||IBM DTLA DeskStar 75GXP
30 GBytes, UltraDMA/100, 7200 rpm
||NVIDIA GeForce DDR Reference Board, 32 MB
|Drivers & Software
|HDD/Motherboard/ AGP Drivers
||Intel: INF-Update for 815 and Busmaster IDE Treiber 6.0
VIA: 4in1 Update 4.24
||NVIDIA Detonator 3 (V. 6.18)
||Windows 98 SE 4.10.2222 A
|Benchmarks & Setup
||HD Tach 2.61
||SiSoft Sandra 6.0
||1024x768x16, 85 Hz
This time we only concentrated on the IDE interface performance. We used HD Tach 2.61 to assess the interface performance. What was interesting was the burst performance, which emerged from several read-outs of the hard drive's buffer memory. Of course, this result is far from realistic, but the meaning of these tests was not to determine the drive's performance.
UltraATA/100 - still not faster?
Owners of UltraATA/66 cannot anticipate higher performance upon exchanging the controller with an ATA/100 model. Only the ATA/100 drives might benefit, if they have been operated with a previous VIA chipset system and their sequential data transfer speed is above 30 MB/s (see benchmark chart on the next page).
Today's real meaning of UltraATA/100 starts with parallel use of several hard drives (e.g. in a RAID setup). Though no chipset supports RAID configurations by default, you may interconnect two or more hard drives in Windows NT 4.0 or Windows 2000 to something like a software-RAID. Stripe sets (RAID 0) will really increase the data transfer speed tremendously.
We expect database applications to run slightly faster with the faster IDE interface. In this case, lots of small packets are repeatedly read and written, which can be accelerated by the hard drive's cache memory. Again, this does not represent the hard drive's performance, but it makes sense to increase performance at certain applications. Cache memory is much faster than the hard drive and is only limited by the IDE interface.
All UltraATA/100 controllers score pretty much the same. The VIA chip is only two points behind Intel's product, which is absolutely negligible. As you can see, the data overhead of all three solutions is approximately 15%, proving that the maximum of 100 MB/s still is a theoretical value. We included the test result with VIA's 686A (old model), which is hardly capable of transferring 30 MB/s - even though this chip is supposed to support UltraATA/66. So far, this is not a disadvantage, since most hard drives do not exceed 30 MB/s. Even if they do, that only happens while sequential reading, which also occurs quite infrequently.
The performance of VIA's new VT82C686B is extremely close to Intel's ICH2 (82801BA) and the Promise Ultra100. It's also great to see that the new chip won't be more expensive than the known pin-compatible models 686A and 596B.
We expect that VIA chipset motherboards with UltraATA/100 support will not be more expensive than ATA/66 models (both 694X and KT133).
VIA's support of both Intel and AMD processors is to the advantage of all of us. The Intel ICH2 can only be used for Intel platforms, while VIA's South Bridge is suitable for Intel's Celeron/Pentium III and AMD's Duron/Athlon platforms.
Both the Intel and the VIA South Bridge can only be pushed to the limit in case you use two or more hard drives simultaneously. In turn that is only possible with Windows NT 4.0 and Windows 2000. Windows 98 users definitely need to purchase a PCI IDE RAID adapter card or a motherboard with integrated IDE RAID chip like, for example, the ABit KT7.
However, we can only recommend purchasing VIA motherboards with the new 686B South Bridge, as the predecessors, 686A and 596B, do not exceed 30 MB/s. That will definitely slow down next generation hard drives. The new 686B eliminates this bottle neck - at least for the time being.