Introduction
New technology introductions in the semiconductor industry are usually treated with great enthusiasm or great derision, and Double Data Rate (DDR) SDRAM was no exception. Some trumpeted the release of DDR as the great saviour of the people from the rending maw of Ramtel (the Rambus, Intel tie-up). For those not familiar, Intel and Rambus combined to try to define another memory type (Rambus’ RDRAM) as the next generation memory standard. Unfortunately, SDRAM was yielding well and prices were low. RDRAM on the other hand was expensive, and the people were unhappy. Many saw DDR as the great saviour.
But there was confusion in the masses. Ramtel had turned the SDRAM memory naming conventions on its ear to gain a marketing advantage. PC-100 SDRAM was SDRAM that ran at 100MHz. PC-133 was SDRAM that ran at 133MHz. Along comes PC-800 and its brothers, PC-700 and PC-600! What?
In reality, PC-800 RDRAM runs at 400MHz, but uses double data rate technology! Hence the marketing boys and girls did their maths and came up with PC-800. Being the honest John and Jane’s that they are, <grin> they neglected to tell you that PC-800 only has 16 signal lines rather than 64. So, while RDRAM runs 4 times faster than PC-100 SDRAM (and is double pumped), all things being equal (which they aren’t), it can only transfer 1/4 of the data that SDRAM can transfer per clock. The net result is that PC-800 isn’t 8 times faster than PC-100, at best it is only twice as fast, and even that has serious limitations.
So the DDR faction got in on the act. RDRAM was claiming a bandwidth of 1.6Gb/s for PC-800 vs. “only ” 800Mb/s for PC-100 SDRAM. Enter PC1600 and PC2100. More marketing bull, but anyway, the numbers are bigger, so they must be better, right?. Well, we’re about to find out!
The purpose of this little intro is to set the scene. With the VIA hype (I can’t blame AMD for this one because they haven’t hyped DDR that much. AMD supports DDR, but they always qualify the statement with the “best performance for today’s processors, price ” blah, blah. AMD also hold a license to use RDRAM!), many people expected to see huge performance gains from moving from SDRAM to DDR. Certainly by now, many would realise that the gain isn’t there, and many analysts have said that this is the reason that the PC industry is in a bust cycle now instead of a boom cycle. You the people aren’t buying DDR motherboards! And it’s because DDR was over hyped! Bad people! Naughty people!
The truth as they say, isn’t that simple.
VIA (the Taiwanese chipset manufacturer) has more than half of the market for Intel processors and most of the AMD market. VIA has been pushing DDR for all it’s worth, but it did itself no favours by releasing the KT133A chipset for the AMD market. This chipset has been a killer, offering extremely good performance (very close to the best DDR platforms), and allowing you to use your existing SDRAM – so the upgrade cost is cheap. Sell your old motherboard and you can get great performance for very little cost. So why would you upgrade?
Cost. This is the killer. When DDR was first announced, the hype machine said that DDR should carry a price premium over SDRAM of no more than 5%! Well the reality was different. At its launch, DDR cost almost as much as RDRAM (sometimes more!) and that was several times what SDRAM cost. Now I don’t know about you, but I am not going to pay four times as much for very little performance benefit. Call me strange. The real reason (in my opinion) why DDR hasn’t set the world on fire is that we (the consumer) have been waiting for the cost to come down. In the meantime, those that couldn’t wait bought KT133A motherboards and are now happy with the performance. As other reviews have shown, the KT133A is no slouch. VIA shot themselves in the foot!
But times change. Led by Micron in the US, DDR prices have come down to the point where it is the same price as regular SDRAM (if you buy Crucial memory) or very little more. As well, motherboards that support DDR memory are now only slightly more expensive than KT133A boards. So if price was holding you back (and it held me back), then right now you have no excuse.
Which brings me to this comparison. I pit a KT133A based motherboard (the price/performance champeen) against “the contender ” – in this case the ALi MAGiK1.
What, on the the crack pipe again Campbell? Everybody “knows ” that the ALi is a poor performer. I say to you, some ALi motherboards are poor performers. Just like some KT133A boards are poor performers. But this is a price performance comparison, and the facts are this:
- AMD760 based motherboards are more expensive than both KT133A based motherboards, and ALi MAGiK1 equipped motherboards – enough more to make you think twice,
- Most AMD760 based boards do not offer the same “overclocking features ” as ALi boards (no multiplier adjustment, no voltage adjust etc.),
- KT266 based boards are buggy, and poor performers without a lot of tweaking – VIA is working on a revised version of the KT266. This is why you can’t find many.
- You can’t buy an SiS735 based board yet, and probably won’t be able to for months, and
- You can’t buy an NVIDIA nForce based board yet, and probably won’t be able to for months.
Smack down!
I also don’t get hardware given to me to review, I bought these products with my own ca$h. So this review is my experience with these two products.
But exactly who are they?
The Champ
Abit KT7A + 256MB Kingmax PC150 memory – YH BIOS
Abit, has earned a reputation as the overclockers motherboard. One could argue that with their venerable BH6, Abit created the overclocking market – at the very least, Abit showed that catering to overclockers could be beneficial to the corporate bottom line.
The KT7A is the latest in a succession of Abit motherboards for AMD’s Athlon processor. One look at the wealth of features and tweaks offered up for adjustment, and you will see that Abit have stuck to their roots.
The Contender
IWill KA 266 + 256MB Crucial DDR memory – KA0508B BIOS
IWill on the other hand, are from a different end of the marketplace. Better known for their workstation offerings – where stability and support are king – how will IWill’s KA266 offering, featuring the unloved ALi MAGiK1 chipset fare?
What am I using to benchmark performance?
Productivity
- Ziff-Davis Business Winstone 2001
- Ziff-Davis Content Creation Winstone 2001
- Ziff-Davis Winbench 99 (Disk test only)
Synthetic
- Cachemem 2.6
Game
- Quake 3 Arena v 1.16n
- Expendable – downloadable demo version
- Mercedes-Benz Truck Racing
- Giants: Citizen Kabuto v 1.398(b)
- MDK2 – gold version
- Evolva -Downloadable demo version
Benchmark set up
OS
- Windows 2000 (productivity) – Gold, FAT32 – Administrator rights.
- Windows 98SE (games) – Gold + IE 5.5 SP1 + DirectX 8.0a
Hardware
- Socket A Athlon “Thunderbird “, 1.33GHz (1333MHz – 266MHz FSB) with 256KB on-die, full speed cache
- Abit KT7A motherboard, YH BIOS or IWill KA266, KA0508B BIOS
- 256MB Kingmax PC150 (KT7A) or 256MB Crucial DDR (KA266)
- Hercules 3D Prophet II Pro 64MB (200MHz core, 400MHz memory)
- IBM 15GB 75GXP 7200 rpm with 2MB cache
- Creative Labs SoundBlaster Live! value.
- Diamond Data 52X CD ROM
- NEC 1.44MB floppy disk drive
Drivers (W2K)
- Microsoft pre SP2 IDE hot-fix
- NVIDIA Detonator driver v 6.31 – 1024 x 768 x 16bit @ 75Hz.
- VIA AGP driver version 4.05c (KT7A)
- ALi AGP driver version 1.70 (KA266)
- ALi IDE driver version 1.02 (KA266)
A note about the driver set up. VIA and ALi both recommend using the default Windows 2000 IDE/ATAPI driver. Unfortunately for this article, there are two problems. One is a known bug where the Microsoft driver reports the drive as running in PIO mode as opposed to UDMA mode (even though the drive is actually in UDMA mode), and the other is that some ATA-100 drives actually run (even with the pre SP2 IDE hot-fix) in PIO mode. Both bugs have been fixed in Service Pack 2. After extensive testing, I confirmed that the IBM hard drive was running in PIO mode. As it was too late to re-run the tests, I was stuck with using the ALi drivers under Windows 2000.
Drivers (Windows 98SE)
- NVIDIA Detonator driver v 6.31 – 1024 x 768 x 16bit @ 75Hz.
- VIA AGP driver version 4.05c (KT7A)
- ALi AGP driver version 1.70 (KA266)
Tweaks
- Sounds – Off (W2K, 98SE)
- Menu animations disabled (W2K, 98SE)
- 256MB permanent swap file (98SE)
- Abit BIOS Tweaks
- IWill BIOS Tweaks
Lets get to it…
Pages: 1 2 3 4 5 6 7 8 9 Next »
Be the first to discuss this article!