KT266A tweaking

[LiamC]

Hi all. Is there any secret to tweaking the KT266A chipset?

I have been playing around with an EPoX 8KHA+ and can't seem to get it to go much faster than my 8K7A+. In fact, in Q3A, the AMD760 is faster!

I have done things like upgrade to the latest BIOS, set the most aggressive memory timings, 1T command rate etc. Am I missing something?

The web reviews I have seen praised the KT266A chipset, but I can't see much difference at all...

 

[The JoJo]

What kind of memory are you using?

The AMD760 is still a good performer. I have the impression that generally speaking the performance difference with the chipsets is quite small.

 

[LiamC]

Crucial PC 2100 @ 2-2-2 timings for both

 

[Clocker]

Are you expecting to see a difference or be able to measure a difference? You won't be able to feel a difference, IMO.

If you are measuring a difference with a benchmark of some sort, what are you seeing?

C

 

[CougTek]

4-bank interleave. Memory setting in the BIOS.

 

[Prof.Wizard]

Strange you don't find your system faster than the AMD760... Every benchmark I ran with my ASUS A7V266-E (which is based on the KT266A) overcomes anything AMD/SiS/ALi/nVIDIA/etc on the same processor.

Have you tried good old FSB O/Cing?

 

[LiamC]

CougTek, measure it, I don't expect to see it. AMD760 is faster in Q3A, KT266A is faster in ScienceMark. I'm just running the Winstone bench's now - appears to be a dead heat so far.

4-bank interleave is set.

ScienceMark reports the memory latency and I can see that the KT266A is better - but as cache should satisfy 93~95% of memory requests, the few percentage points of better latency shouldn't make a huge difference - and this is exactly what I'm seeing. But if you read AnandTech or Tom's (I know, I know) you get the impression that the difference is huge - 10% ~ 20% faster. My findings are leading me to think that they are full of crap. I think I know why (they get the numbers that they do), but I will do the test numbers first before commenting...

Prof. Wiz, it's not about o/cing, I want to see which one is faster in stock speeds. Mind you, I have tweaked both BIOSes and setups for speed - to give a fair comparison.

 

[CougTek]

CougTek, measure it, I don't expect to see it. AMD760 is faster in Q3A, KT266A is faster in ScienceMark. I'm just running the Winstone bench's now - appears to be a dead heat so far.

Unfortunately, I don't have two motherboards with those chipsets in my hands to compare. I only have a KT266A here and it's been a while since I have stopped to consider the AMD760-based motherboards when I assemble new systems (less features and everyone else say/show them as being slower).

I'm looking forward to your findings. Keep us informed and if you indeed find something, please try to confirm with someone else to be sure you are not just victim of an under-performing KT266A motherboard.

 

[Prof.Wizard]

...But if you read AnandTech or Tom's (I know, I know) you get the impression that the difference is huge - 10% ~ 20% faster. My findings are leading me to think that they are full of crap...

Liam, the difference is real. Sandra isn't the best benchmark but as a synthetic one, it gives me a comfortable edge over AMD760s and other chipsets.

I would still love to see your own conclusions though... if there is a catch etc.

 

[Adcadet]

I'm sure that most hardware sites exagerate the speed differences between chipsets. Suggesting one chipet over another since it's slighly faster is fine. Suggesting that people run out to buy a KT266A when they already own an AMD760 is misleading and irresponsible. Hopefully you didn't make your last motherboard move just cause a few people made some wild claims.

 

[LiamC]

Nope! (Hopefully) not that stupid. I was actually concerned by the statements before I bought the 8KHA+, and specifically bought it to test my theory. But I was expecting a little more performance out of the box. The article I write should go into things in a little more depth.

 

[LiamC]

OK, after the Deathstar died, I had to re-install everything on a Maxtor D740X. Performance is now up to scratch. Maybe the install went bad because of a bad sector or two? Very strange. I would have thought that the whole thing would have went down the gurgler, not just slow down a little...

 

[time]

Are you saying that the dodgy drive affected Q3 scores?

What about low level tests?

 

[LiamC]

Yep. After the install, I did my usual BIOS tricks and patches and then started running some benches. Q3A scores were slower than the EPoX 8K7A. ScienceMark was a little faster (could be within the margin of error) and the Winstones were line ball - slight victory to the 8KHA+.

MS Flight Simulator was marginally quicker, Giants was way slower. SANDRA wasn't affected though Winbench 99 scores were 20% low. ScienceMark mem latency tests showed that the KT266A was much better than the AMD760.

I checked my setting, patches and results several times (except for the Winstone scores)

Reinstall on the Maxtor and everything is as expected, and it does appear that the KT266A has the edge over the AMD760.