PDA

View Full Version : KT333CE Performance


TW
06-15-04, 02:25 PM
The issue of performence on the KT333CE depends on a choice between stability and a few meagre horses under the hood. At the speed the system is running, there is no perceptible difference. What the user gets with the kt333 is stability. The 266a was not built for 166/333 operation, whereas the 333ce is. Via techs have as much as admitted the 266a overclocks more efficiently, especially memory bandwidth. But the 333CE will overclock to 166/333 with a much better level of stability due to design and divider issues. If you wish for a 200 fsb and do not care about pci/agp compatibility, the 266a is the best chipset. The 333 can be tweaked by VIA for better than 266a performance, but the MB would need to be reworked with higher spec components, lifting the price considerably.<BR><BR>What the KT333 provides is the foundaton for a 166 fsb T'bred. When the 333 was designed a 166 T'bred was not on the horizon, but VIA bet on it. I don't like to gamble much but I do know the 166 T'bred is almost a sure thing. One of the reasons the T'bred is late is for core mods to allow 166 operation. As AMD starts to struggle at the low end against Intel, a quick speed boost must come along. Higher speed CPUs are fine, but as any overclocker knows, upping the fsb/memory to 166/333 gives a good performance gain, particularly in regard to bandwidth. It also has the advantage of allowing faster CPUs without upping the multiplier.<BR><BR>AMDs roadmap will almost certainly for the next year have the 133 T/bred badged as a Duron and the 166 as their Athlon, until Clawhammer pricing ends one of them.<BR><BR>So we are back to Via, who do not have a stock 166 CPU to put in the KT133 board yet, but expect it to arrive soon. Thus the CE is a test platform (get VIA to admit it though!). It allows VIA to work on 166 stability and performance, and MB manufacturers to start designing for 166 stability. The Shuttle board tested in the KT133 article is a 266a unmodified by any noticable standard, hence the 3 legged horse performance. A ground up redesign will show better stats, but without a CPU to slot in, there is no justification to many companies to go into production. But the CE allows the design teams to start tackling the signal/heat issues needed for 166 operation.<BR><BR>When the 166 T'bred arrives (any bets on SSE2 inside!), they will go into what is at the moment called the KT400. You will have 8x AGP, the DDR controller from VIAs new P4 chipset (which appears to be a nasty piece of work), and O/C for 200/400 with memory performance probably 10-15% better than the 266A.<BR><BR>So, for you nutzoid overclockers who love their AGP and PCI buses interfering with TV reception, stick with the 266A. The average buyer will be happy with their KT333CE at present, and as DDR2 looms on the horizon, the KT400 will be left to hold open the final curtain on DDR. Southbridge development will I feel be the final chapter for the KT series, allowing Serial ATA and a few other integrated options. <BR><BR>The CE performance issue is there, but it is overblown by the hardware community. The real issue is the fact that despite the Athlon introducing DDR to the world, the best chipsets for performance, from VIA and SIS, are on the Intel platfrom (the P4 has more to gain from DDR refinement thus development went in that direction). VIA will at least be updating their DDR controller for AMD . <BR>

Colin
06-15-04, 02:25 PM
Hey there. Yes, the KT333CE was built for 166/333 MHz operationk, however VIA has not publically stated that that the KT266A is faster then the KT266A. At least I don't see it anywhere on their <a target=new href="http://www.via.com.tw">website</a> or on any other hardware site.<BR><BR>Yes I agree that overclockers should not have problems hitting 166 Mhz+ FSB's now with the KT333CE however honestly what's the point if performance is so lousy? I'm not sure if a 166 MHz T-bred will happen... AMD might be better off focusing on the Hammer line of CPU's instead.<BR><BR>While what you're saying is correct, having a faster FSB can produce faster systems even though CPU speed might be slower, AMD is still in a MHz battle with Intel. The common user doesn't know that AMD processors have a higher IPC and that makes up for the difference. All they know is MHz means it's faster.<BR><BR>I'll bet ya the T-Bred doesn't have SSE2! <img src="i/expressions/face-icon-small-smile.gif"border=0><BR><BR>Also, we're the first to bring up this KT333CE VS KT266A issue. Where else have you seen it? I certainly haven't so I'm not sure how it can be overblown by the hardware community if only one site has reported this.<BR><BR>Anyway thanks for your input, I'd be interested in what other people around the web think as well.

TW
06-15-04, 02:25 PM
Yor site was the first to officially post on the issue, and in a damn fine manner, unlike some places who mention it then leave everyone adrift. The comment from a VIA person on OC performance (a shy anon type)was related to the issue, posted by many people from different sites, that the 333CE sux compared to the 266a when oc between 133 and 166. No real info as to why, it is technical. This doesn't worry VIA, as the 333 at 166 fsb, and a ground up board design, performance goes up 5-15% depending on the apps. This is what the 333 is meant to be used for when oc. <BR><BR>As to the 166 T'bred, it is already here. Most T'breds can run at 166. All is required is slight change at FAB, multiplier down, bus up, maybe a nudge on voltage. T'breds have a lower voltage, thus room for a bit more. Multipliers are important, at 166, the 15* limit gives a lot more room above 2200.<BR><BR>A 333 mb with t'bred support should only require a bios update, providing it was designed with 166 in mind. Reviewers around the sites have found that not all 333 boards were created equal, many have issues at 166 fsb.<BR><BR>In the end, the Marketing Dept will likely have final say. <BR><BR>As to SSE2, it is a remote chance with Barton, along with the 512 cache. Barton is a mobile solution, time will show this. It serves no purpose with hammer close, except as P4M compeitor. Onboard cache uses little power for the performance boost, and SSE2 may be needed by the marketing suits. Hi bus=heat & power,thus no chance at 166.<BR><BR>Considering Claw looming, we all still win. I like everyone else though, would like to see 166. The catchcry of tomorrow is Bandwidth Counts!

MUSTANGPC
06-15-04, 02:25 PM
I had read somewhere on another Forum that KT333 is not compatible with DDR333. Is this true?