KT266A vs KT333

[Silver]

BTW, I have found that lowering the multiplier and raising the fsb does increase bandwidth however I loose some processing power resulting in slightly (very small) loss in 3dmark2001se. Almost identical but just a shade less. Ram settings for the above were as follows. 2-4way-2-5-2-8-4-1-0-0-e-e. Much higher fsb can be attained (i.e. 192) at over 2601 (as tested by sisoft) on the samsung while (strangely) the crucial at 191 will score 2650 (3.2v). Though I presently test 3dmark2001 at 174x10 and have broken 10,100, I do feel that the best on this processor (xp1600) would be achieved at 182x9.5 as this would give me very strong ram timings while still making the processor work out hard. Make sense? I am not sure it does to me. My 'gut' feeling is this would be the "sweet spot" and would produce the absolute maximum results on this chip at 1.73 to 1.74. Unfortunately the chip needs to have the unlock redone and I am moving so it is on delay for now. Top score should be in the neighborhood of 10,200. Now a processor running at 1.95 to 2.0 should allow it to give most STOCK ti4600s a run for it. Really need to get one of those xp1900's.

 

[Sonny]

Well the only info that I can find regarding this test when done properly is that the 1/5 divisor does what it is suppose to which is to keep PCI & AGP bus speeds in check BUT it will kill your overall performance. The only bechmarks that I could find regarding the drop off of performance was with the lower multipliers.

1/3 Divisor;

1/4 Divisor;

If you guys have a 1/5 Divisor how about running the test or do you guys just don't want to admit the performance loss?

 

[Arkaine23]

benches

I'll post some AK35GT-R benchmarks around 1.4Ghz when I get home tonight... Its very hot in my house, so I may have a little trouble doing more than 150FSB even with a lower multiplyer.

 

[Malakai]

the new 3/14 bios for 8KHA gave me the same fpu mem score in sandra, but 120mb/sec more in alu. im now beating a kt333 at the same fsb. and she is DEAD stable. for some odd reasson this vrucial is fine 160fsb at 2.7 volts but wont do 166 even at 3.2 volts. i have good case cooling and active ram cooling
-Malakai

 

[Maximus Nickus]

I had the same problem, change the 1T command to 2T (leave everything else) and you should be able to get over 170FSB.

It doesn't result in a loss in performance either!

 

[Sonny]

Malakai - Unless you do the tests I have outlined we will not be able to use any of your info. It's important to have a standard so we can compare the peformance gains/loss of each of the chipsets

Maximus Nickus - Heard about your probs with your hardware. i hope you get back on track soon

 

[Malakai]

i cant do 166 at respectable timings 4 some reason, so ill post my sandra's at 133 and 150, all at 1.4.
and my 3dmark scores ill maybe post at those busses, but im so sick of looking at them. ill post em at like 4am when i get home from work
-Malakai

 

[NeoMoses]

wow, this is turning out harder than I expected. Is there ANYONE out there with an unlocked chip and kt266a motherboard that can reach 166 FSB? So far the only data we have is clock speed dependent, which throws off the numbers.

Unfortunately my friend sold his 1.4 TBird and got an XP. Both of our chips are locked, and neither of us plan on unlocking them.

 

[Malakai]

I just got home from work, im tired and goin to bed, ill definetely do it tomorrow. and maybe i will pull a stick out, if i do that i can do 166 max timings. I have 2 256 dimms installed(see sig)
but its a real biotch to do, my gf3 is in the way of the ram, so ill see
-Malakai

 

[Maximus Nickus]

I apologize for the delay, as soon as my rig is up I'll post some benchies.

 

[Sonny]

.

..

same thing

...

It would also help if you told us what the FSB is & exactly what the clockspeed your running.

 

[Sonny]

OK everybody let's get this right. Please read the first post on this thread. We will not be able to find out the advantages or disadvantages if the guidlines are not followed.

 

[res0r9lm]

8.5X166.1(1412)

 

[res0r9lm]

9.5x150.3(1428)

 

[res0r9lm]

10.5x133(1401)

 

[Malakai]

ill post mine tonight. i work so damn much i never have time.
i need to try to o/c more anyways
-Malakai

 

[UnseenMenace]

Has anyone figured out if the DDR333 memory support is so badly needed on a Socket A platform. The Socket A processor systems has a bus bandwidth that makes 2.1GB/sec (in case the bus frequency is equal to 266MHz). However, the memory bus bandwidth provided by PC2700 (DDR333) DDR SDRAM makes 2.7GB/sec. which surely means that a CPU working with 266MHz bus will be simply unable to use all the memory bus bandwidth available. Increasing the memory bus bandwidth over 2.1GB/sec doesn't bring in any tangible advantages. That is why the implementation of DDR333 memory support in KT333 can be in the first place regarded as a marketing move.

Although the memory bus in the new Socket A chipsets can work at 333MHz, AMD is not going to increase the CPU bus frequency so that the processor could work synchronously with the memory and use the DDR333 memory bandwidth to the full extent. Therefore, all the new DDR333 chipsets for Athlon XP, Athlon and Duron processors support only 200MHz and 266MHz processor bus.

What else does the VIA KT333 offer the user besides the obvious DDR333 support, not much in fact. The only new feature compared with VIA KT266A, is the memory controller. However, VIA not only made it possible to clock the DDR SDRAM bus at 333MHz (166MHz DDR), but also enhanced the entire controller. The memory controller of KT333 fhas far superior timings than that of the KT266A, which in turn has lowered the memory subsystem latency and increased the effective bandwidth even when PC2100 DDR SDRAM is used. Moreover, they also increased the buffer size. That is why KT333 should be slightly faster than KT266A even when working with PC2100 DDR SDRAM.

All the other features of the newboy are just the same as by KT266A, including the support of AGP 4x and 266MB/sec V-Link bus between the North and South Bridges of the chipset. The only other thing worth noticing is that VIA will provide its KT333 North Bridge with VT8233A South Bridge supporting ATA/133 protocol

 

[Maximus Nickus]

I agree, you need a relative CPU bus to use the memory's bandwidth to its full extent.

Well Said.
M_N

 

[Malakai]

true UnseenMeneace, but i think we all increase the fsb and the ram synchronously.
i know i do.
and ive never even heard of an athlon not being able to take ANY given fsb, even 400+(DDR)
-Malakai

 

[Sonny]

Has anyone figured out if the DDR333 memory support is so badly needed on a Socket A platform. The Socket A processor systems has a bus bandwidth that makes 2.1GB/sec (in case the bus frequency is equal to 266MHz). However, the memory bus bandwidth provided by PC2700 (DDR333) DDR SDRAM makes 2.7GB/sec. which surely means that a CPU working with 266MHz bus will be simply unable to use all the memory bus bandwidth available. Increasing the memory bus bandwidth over 2.1GB/sec doesn't bring in any tangible advantages. That is why the implementation of DDR333 memory support in KT333 can be in the first place regarded as a marketing move.

Although the memory bus in the new Socket A chipsets can work at 333MHz, AMD is not going to increase the CPU bus frequency so that the processor could work synchronously with the memory and use the DDR333 memory bandwidth to the full extent. Therefore, all the new DDR333 chipsets for Athlon XP, Athlon and Duron processors support only 200MHz and 266MHz processor bus.

What else does the VIA KT333 offer the user besides the obvious DDR333 support, not much in fact. The only new feature compared with VIA KT266A, is the memory controller. However, VIA not only made it possible to clock the DDR SDRAM bus at 333MHz (166MHz DDR), but also enhanced the entire controller. The memory controller of KT333 fhas far superior timings than that of the KT266A, which in turn has lowered the memory subsystem latency and increased the effective bandwidth even when PC2100 DDR SDRAM is used. Moreover, they also increased the buffer size. That is why KT333 should be slightly faster than KT266A even when working with PC2100 DDR SDRAM.

That was the reason for this test. So that people realize what they are actually gaining & losing with the newer KT333 chipset. Still a lot of people do not realize the difference of Asynchronous & Synchronous running bus speeds. All we need now is for people who are interested in it to actually follow the giudelines of the test.

All the other features of the newboy are just the same as by KT266A, including the support of AGP 4x and 266MB/sec V-Link bus between the North and South Bridges of the chipset. The only other thing worth noticing is that VIA will provide its KT333 North Bridge with VT8233A South Bridge supporting ATA/133 protocol

The newer KT266A133 has the VT8233A like the KR7A-133RAID.