so what about kt400....some confusion

[zhena]

ok so with all the reviews on the net going on about how the kt400 has no performance improvement over kt333...
there are some things i don't understand...

people are saying how the memory standard for ddr400 is not there yet.. how it's not stable yet.. yada yada yada..
well this is all true, but once you finally get it working using tested samples of memory (like the reviews on tomshardware and anandtech)
at the same timings shouldn't the kt400 be clearly faster than the kt333?

i don't think i've yet seen a clear comparison of two boards running the memory at the exact same timings using the exact same cpu just one running at kt333-ddr333 and the other one running at kt400-ddr400.

the athlon cpu is definetly bandwidth hungry, not as much as p4.. but still. there is no way that ddr333 memory is more bandwith than it needs (which is basically what the tests are saying)

second of all kt400 running the momery at ddr333 should be basically exactly the same as kt333 running the memory at ddr333 shouldn't it?

this is of course true only if the chipset is designed correctly.

anyhow...
any thoughts on this?

 

[zhena]

well hmm

 

[cmdrdredd]

I dunno...

wondering what is faster...Kt400 running PC3200 at 400Mhz or KT333 overclocked to run PC3200 at 400Mhz.

 

[zhena]

up

 

[mechBgon]

Actually, the 266MHz-bus AthlonXPs are a precise match to the bandwidth of PC2100 memory. When KT333 arrived with official PC2700 support, people were squinting to see any improvement by using a 266MHz-bus AthlonXP with PC2700 memory, but there really wasn't much difference. Why not? Because the northbridge now had to try to play "middleman" between two busses running at two different speeds. The penalty of this juggling act just about counteracts whatever benefits PC2700 would theoretically have to offer. PC3200 just takes it a step further. The CPU can still make use of only 2100Mb/sec worth of the bandwidth.

Naturally, people began clamoring for AMD to bump the CPU bus to 333MHz so it would match up with PC2700 and run synchronously, giving the best of both worlds, and that's what AMD's doing, starting with the 2700+ and 2800+. Many people, myself included, have taken matters into our own hands and overclocked a 1600+ to the 333MHz bus on a KT333 or KT400 board, for an effective 2100+ performance level.

I like the approach we're seeing with nForce2, where the width of the data bus itself gets doubled when you add a second identical stick of RAM. Latency penalty? None. Bandwidth improvement? Noticable. Performance boost? We'll see when some motherboard reviews come out. Evan Leib says AnandTech will have some full reviews within a couple weeks.

To come back to the question of "why does KT400 exist, why isn't it called KT333A or something," it doesn't really have much of a claim to fame... AGP 8x (of questionable value), double the bandwidth between the northbridge and southbridge (also of questionable value), potential for using DDR400 (although VIA doesn't make this claim) and official support for all 200MHz-, 266MHz- and 333MHz-bus AthlonXP's (definitely a point in its favor). The fact is, VIA is going to disable AGP 8x on a variant of KT400 and start selling THAT as KT333CF (replacing the existing KT333CE), so the line is getting blurred. Confused yet?

 

[PUNKtotalled]

Originally posted by: zhena
i don't think i've yet seen a clear comparison of two boards running the memory at the exact same timings using the exact same cpu just one running at kt333-ddr333 and the other one running at kt400-ddr400.

I've seen that comparison on

that german site.... benchies speaks for themselves....


....a "perfect" equality.