Original Link: https://www.anandtech.com/show/878
VIA's KT333: The Athlon gets DDR333
by Anand Lal Shimpi on February 20, 2002 1:57 AM EST- Posted in
- CPUs
Oh, how times have changed. Remember the days when it was shameful to have an Athlon chipset on display? Or when motherboard manufacturers weren't allowed to publicly talk about their Socket-A solutions, even while they were shipping to system integrators. Back then the future of the Athlon was uncertain, but times have changed.
Today the Athlon is easily AMD's most successful product line ever both from a business standpoint and an end-user standpoint. And it success can be attributed to three things: price, architectural design and VIA.
Indeed if it weren't for VIA, the Athlon would have definitely not been as successful of a CPU as it is today. The vast majority of Athlon platforms are running VIA chipsets in spite of the fact that there are three perfectly competent manufacturers out there other than VIA that make Socket-A chipsets. VIA brought the Athlon to the mainstream and more recently they brought DDR to that same market. Although clearly not as important as bringing the Athlon or DDR to the masses, armed with their KT333 chipset VIA will be bringing DDR333 and ATA/133 to the Socket-A world.
The KT333 chipset isn't a revolutionary Athlon platform, rather it's an evolution of VIA's outstanding KT266A design. The added features are obvious as we just mentioned them, DDR333 and Ultra ATA/133 support and thus the performance benefits are not huge. For this reason we won't spend much time covering the chipset's performance but enough to characterize the KT333 as the logical replacement for the KT266A.
DDR333: Marketing or Necessity?
When any chipset manufacturer gets to work on designing their next-generation core logic, do they try and solve today's problems or those that may appear in the future? Think back to the introduction of the Ultra ATA 66 spec into South Bridges, did it occur before or after hard drives demanded the added bandwidth? Or what about the introduction of DDR SDRAM to the Athlon platform with the AMD 760 chipset? Did the Athlon demand that sort of memory bandwidth back then? While it's true that necessity is the mother of invention, when it comes to designing chipsets it's what's perceived to be necessary in the future that gets included in today's new chipsets.
That brings us to the question of whether DDR333 is necessary or not. Given today's requirements, most likely not but remember that you could've said the same for DDR266 two years ago. You shouldn't have high hopes for the introduction of DDR333 nor should you rush out and replace all of your memory with DDR333 rated DIMMs, but realize the purpose of introducing it now is to be prepared when platforms like the Claw Hammer demand it down the road.
The specifications for DDR333 are pretty simple from a high-level; instead of operating on a 133MHz clock frequency the standard calls for a 166MHz operating frequency. Since we're dealing with DDR SDRAM here data is transferred twice per clock (as I'm sure you're tired of hearing) thus causing many to say it operates at 333MHz. You may have also heard DDR333 referred to as PC2700, both refer to the same basic specification of SDRAM.
With a 166MHz clock frequency, DDR333 SDRAM can offer another 25% of memory bandwidth over DDR266 SDRAM. We've already been through these calculations before when SiS introduced DDR333 for the Pentium 4 with their 645 chipset but as a refresher here's a quick table comparing all of the currently used Athlon memory subsystems:
|
Chipset
|
Memory Bus Width | Max Memory Frequency |
Peak Theoretical Bandwidth
|
| ALi MAGiK1 |
64-bit DDR
|
133MHz
|
2.1GB/s
|
| AMD 760 |
64-bit DDR
|
133MHz
|
2.1GB/s
|
| NVIDIA nForce 420 |
128-bit DDR
|
133MHz
|
4.2GB/s
|
| NVIDIA nForce 415 |
128-bit DDR
|
133MHz
|
4.2GB/s
|
| NVIDIA nForce 220 |
64-bit DDR
|
133MHz
|
2.1GB/s
|
| SiS 745 |
64-bit DDR
|
166MHz
|
2.7GB/s
|
| SiS 735 |
64-bit DDR
|
133MHz
|
2.1GB/s
|
| VIA KT333 |
64-bit DDR
|
166MHz
|
2.7GB/s
|
| VIA KT266A |
64-bit DDR
|
133MHz
|
2.1GB/s
|
The most limiting factor here will be the Athlon's 133MHz DDR FSB which will only allow a maximum of 2.1GB/s of data to travel to/from the CPU. So in spite of the fact that the chipset will be able to get, in theory, up to 2.7GB/s of data from main memory it will only be able to hand 2.1GB/s off to the CPU.
AMD hasn't announced plans to move to a higher speed FSB anytime soon and while the added memory bandwidth would be useful on a point-to-point multiprocessor platform (each CPU gets its own 2.1GB/s bus to the chipset) VIA has already stated that they won't enter this market for the Athlon.
What about Ultra ATA 133?
A
new feature being introduced alongside the KT333 North Bridge is support for
the Ultra ATA 133 standard through a new South Bridge. As the name implies,
the Ultra ATA 133 spec gives ATA/IDE devices a maximum transfer limit of 133MB/s
which is up from the 100MB/s of the previous Ultra ATA 100 spec.
While it's very clear that even the Ultra ATA 100 specification bought hard
drives a huge amount of headroom to work with, there are some situations in
which even the Ultra ATA 100 specification can be exceeded. When transferring
data purely out of a drive's buffer it is possible for the transfer to occur
at speeds greater than 100MB/s, but with the largest of buffers on IDE drives
being 8MB we're not talking about an incredible amount of data here. Even then,
the performance difference will hardly be noticeable since the vast majority
of data will still be dependent on the time it takes to find and read/write
data on the platter(s).
We've been playing around with benchmarks internally to see if we can ever create
a situation in which we sustain transfer rates greater than 100MB/s thus requiring
an Ultra ATA 133 compliant controller. Unfortunately the only situation we've
been able to dream up is with a 4 drive RAID-0 array of Maxtor D740X drives
where transfer rates just barely peaked above 100MB/s. Until single hard drives
get that fast, Ultra ATA 133 won't buy you any additional performance.
The Chipset
The KT333 chipset is comprised of two parts as usual, the KT333 North Bridge and the VT8233A South Bridge.
The KT333 North Bridge is identical to the KT266A North Bridge with the exception of the memory controller which now officially supports DDR333 SDRAM. The FSB interface remains unchanged and only officially supports 100/133MHz operation.
The VT8233A South Bridge is identical to the VT8233 South Bridge with the addition of Ultra ATA 133 support.
The two chips are connected using VIA's V-Link interconnect bus and as usual the chips are pin compatible with the previous KT266A chipset.
The Motherboards
We had two KT333 boards to test with, VIA's reference KT333 design and Gigabyte's GA-7VRXP. The two performed very close to one another so we decided to use the Gigabyte board in order to test something that would actually be available for purchase in the near future.
VIA's microATX KT333 Reference Board
Gigabyte's GA-7VRXP KT333 Board
In order to compare to the competition we chose the ASUS A7V266-E, ASUS A7N266 and ECS K7S5A to represent the KT266A, nForce 420/220-D and SiS 735 chipsets.
The Memory
Officially, there is no such thing as DDR333 SDRAM. Many manufacturers have already been shipping DDR333 or PC2700 SDRAM which is essentially memory that they have tested internally to work at DDR333 speeds. In the next month JEDEC will finalize a DDR333 specification and afterwards you'll begin to see officially sanctioned DDR333 DIMMs on the market.
There are many DDR266 modules that will work at DDR333 speeds and thus we were able to test the KT333 without any problems.
The Test
We tested the KT333 with the memory frequency set to 166MHz (DDR333).
|
Windows XP Professional Test System |
||||||
|
Hardware |
||||||
|
CPU(s) |
AMD XP 2000+ | |||||
| Motherboard(s) | ASUS
A7V266-E (KT266A) ASUS A7N266 (nForce 420/220-D) ECS K7S5A (SiS735) Gigabyte GA-7VRXP (KT333) |
|||||
| Memory |
256MB DDR333 Kingmax DDR SDRAM (CAS2) |
|||||
| Hard Drive |
80GB Maxtor D740X Ultra ATA 133 |
|||||
| CDROM |
Phillips 48X |
|||||
| Video Card(s) |
NVIDIA GeForce4 Ti 4600 |
|||||
| Ethernet |
Linksys LNE100TX 100Mbit PCI Ethernet Adapter |
|||||
|
Software |
||||||
|
Operating System |
Windows XP Professional |
|||||
| Drivers |
|
|||||
Memory Bandwidth
As we move into the new year we start encountering the 2002 versions of many popular benchmarks. One such benchmark that has actually been out for a while now is SiSoft Sandra 2002. The one thing we have used Sandra for in the past is as a useful measure of memory subsystem performance, in particular its memory bandwidth test is usually fairly accurate.
In the 2002 version of Sandra, the benchmark now includes the ability to perform the conventional memory bandwidth test but also enables using SSE instructions to move the test data around. The end result is that you can get a much better idea of what the true peak theoretical bandwidth of a particular platform happens to be. We performed both the conventional memory bandwidth tests with the enhanced instructions disabled and with them enabled.
|
1986Using the enhanced data movement instructions it's clear that we run into the 133MHz FSB limitation of all of the platforms. The 2% difference in memory bandwidth between our KT333 and KT266A test platforms is small enough to be attributed to variations in board designs. Interestingly enough is that although the nForce 420-D platform has twice the theoretical bandwidth of the KT266A, its attainable memory bandwidth is actually around 8% lower. This could be because of NVIDIA's implemention of the EV6 FSB interface or because of their memory controller in general. The reason this doesn't translate into an 8% reduction in real world performance is because there are never any situations where you'll hit this sort of memory bus utilization.
This perfectly illustrates the first point we made about DDR333's usefulness in today's platforms; you shouldn't expect to see a tangible performance improvement because in the end you'll be limited by the Athlon's FSB.
|
Disabling the enhanced instructions simulates a much more down-to-earth usage pattern that is still a bit on the high-side for the memory bandwidth utilization characteristics of most applications. Here we can see that although the peak theoretical memory bandwidth figures don't increase by much, the actual attainable bandwidth figures do go up a bit.
The reasoning behind this is that, all things kept the same, as clock speed increases latency naturally decreases. Lower latency memory accesses results in greater memory bandwidth utilization and thus the increased figures you're seeing out of DDR333 SDRAM used with the KT333 vs. DDR266 SDRAM used on the KT266A.
What's even more interesting is that the nForce's dual channel memory controller setup on the 420-D is able to provide an additional 7% of bandwidth over the single channel setup.
As we've seen numerous times in the past however, you cannot base your real-world performance expectations on synthetic tests like these because in the real world, the ideal scenario is never the case.
From this point on we've omitted the nForce 220-D results since they are identical to the nForce 420-D results in all real-world performance tests where the integrated graphics are not used.
Overall Performance
The next updated benchmark we're using is SYSMark 2002. The SYSMark 2002 test suite remains relatively unchanged from 2001 in the manner in which the results are produced. What has changed however is the applications used; they have all been updated to the latest versions and include the following:
The Internet Content Creation Suite tests these applications: Adobe Photoshop® 6.01, Adobe Premiere® 6.0, Microsoft Windows Media Encoder 7.1, Macromedia Dreamweaver 4, and Macromedia Flash 5.
The Office Productivity Suite tests these applications: Microsoft Word 2002, Microsoft Excel 2002, Microsoft PowerPoint 2002, Microsoft Outlook 2002, Microsoft Access 2002, Netscape Communicator® 6.0, Dragon NaturallySpeaking Preferred v.5, WinZip 8.0, and McAfee VirusScan 5.13.
You'll find that the vast majority of your day-to-day tasks fall in the Office Productivity Suite where memory bandwidth differences are even less pronounced.
|
There's not much to say here other than it's impressive to see how far VIA's DDR memory controller has come since it was originally introduced. There was a time when the SiS 735 was outperforming it but now with today's more mature KT266A and now KT333 boards, the 735 is easily outperformed by 10% or more.
|
Nothing new here, the additional memory bandwidth doesn't do much for the office applications mentioned before although VIA's chipset performance lead continues to be 10% over the 735. Also remember that although we haven't listed them here, the AMD 760 and the older KT133/A chipsets are even slower than the SiS 735. In many cases a mere upgrade to a KT266A/KT333 platform will yield close to a 20% performance improvement for owners of older chipsets.
Gaming Performance
One of the best tests for chipset/memory controller performance has been today's 3D games. To kick things off we've selected the latest Serious Sam demo available from Croteam's website. The benchmark was run with the Extreme Settings add-on executed at 640 x 480 x 32 using the first built in demo.
|
There is absolutely no tangible performance difference between any of the top three performers (we're dealing with gaps of less than 3%), and again the 735 is approximately 10% slower than the KT266A/KT333.
Our last choice for a benchmark is the Unreal Performance Test 2002. This build of the Unreal Engine is a far evolution of the engine found in Unreal Tournament and is being developed as the base for games such as Unreal Tournament II and Unreal 2. For more information on the benchmark feel free to read about it in our first two tests involving it here and here.
|
Even using a next-generation gaming benchmark such as the UPT2002 the results do not change at all. With the results being virtually identical across all applications and tests we chose to severely limit the number of benchmarks run to save us time and to save you boredom from flipping through pages upon pages of < 2% differences in performance.
Onto the conclusion, shall we?
Final Words
When the KT266A was launched it completely blew us away; the performance of the chipset was spectacular and it was clear that it would quickly become a top pick for all Athlon owners. The KT333 doesn't have nearly as great of an impact but the reasons behind that are understandable; both new features supported by the chipset, DDR333 and Ultra ATA 133 aren't features that will result in tangible improvements in performance today. Instead the KT333 is more of a technology enabling platform for VIA. The chipset will not cost any more to manufacture than the KT266A and thus motherboards won't increase in price. While DDR333 SDRAM isn't officially available today (the specification isn't complete), when it is first made available it will carry a price premium over DDR266 SDRAM.
Remember that the KT333 will function just fine with DDR266 SDRAM, meaning that it will eventually replace the KT266A in all forthcoming motherboards out of Taiwan. Although the performance improvement is non-existant, there's nothing to complain about from the motherboard manufacturer's perspective. Outside of ensuring stability at 166MHz DDR memory bus frequencies, it costs no more to manufacture a KT333 board than it does a KT266A motherboard.
Obviously to current KT266A owners there's no need to upgrade, nor is there any need to feel outdated. What should be taken away from this review however is the presence of VIA at the top of the Athlon chipset market. Moving forward the biggest threat to VIA's Socket-A marketshare will be the nForce 415-D which is supposed to be very aggressively priced, however it seems like it will have to be some very impressive pricing from NVIDIA in order to convince both motherboard manufacturers and end users to make the switch. The high-quality audio system and integrated LAN would be the only reason to lean towards the 415-D as the performance between the 415-D and the KT333 should be relatively similar.
That pretty much sums up all that can be said about the KT333; it wasn't as hard hitting of a launch as the KT266A but that's understandable given the limited improvements. The KT333 is a natural replacement for the KT266A and if AMD does ever move to a 166MHz FSB you can expect VIA to be there with a KT333A chipset to match. The bottom line is that today, DDR333 doesn't result in any tangible performance improvements but if paired with a 166MHz DDR FSB the boost could be quite tangible.
Facebook
Twitter
RSS
