My "common sense" tells me that, given a fixed CPU frequency, a faster bus is generally better.

Generally useless on SDRAM systems, because RAM can work at 133 Mhz on both and can't saturate K7 DDR bus.

Not totally useless, you win about 5% between 100 and 133 on a Via SDR-SDRAM platform:
http://www.thg.ru/mainboard/20010104/images/a … 79_image001.gif

All things being equal the 1000C is the better choice, so long as your board supports 133MHz FSB of course. Even though the 133MB/s memory bus can't saturate either CPU bus, running asynch adds a bit of latency. Wouldn't surprise me if that's a significant part of the performace difference.

There are lot of KT133 boards out there and for those Athlon 1200B (200MT/s bus) is the best option due to max multiplier being 12.5. For KT133A use Athlon 1400C (266MT/s bus).

Am i wrong ?

Lies, damned lies and performance ratings... I recall someone did an overview of all SoA chipset platforms around the time of the AXP3200+ and recorded a 50% difference between slowest and fastest - at the same (100MHz FSB) speed. Until the memory controller moved to the CPU, chipset mattered massively. Those performance ratings weren't objective measures of speed, but a marketing tool, based on what somebody - optimistically - might expect to see vs a similar P4 from the same period.

dionb wrote on 2022-08-23, 16:32:

Am i wrong ?

Lies, damned lies and performance ratings... I recall someone did an overview of all SoA chipset platforms around the time of the AXP3200+ and recorded a 50% difference between slowest and fastest - at the same (100MHz FSB) speed. Until the memory controller moved to the CPU, chipset mattered massively. Those performance ratings weren't objective measures of speed, but a marketing tool, based on what somebody - optimistically - might expect to see vs a similar P4 from the same period.

Well yes, but not the worst marketing tool given that the way computers were marketed to your grandma etc. was always by comparing the amount of MHz. In the minds of most consumers more MHz would equal a faster system. You'd already loose a large chunk of your audience if you tried to tell them that you can only really compare two of the same exact type of CPU if you wanted to use MHz as a metric...

The Pentium 4 was a horribly inefficient architecture and performance per-clock was often pretty bad compared to a similarly clocked Athlon - true, not due to the CPU alone, but on the whole given the motherboard, memory etc. I'm not saying AMD was right with their PR, but it was a somewhat understandable attempt to wean most people off the MHz == performance perception.

leonardo wrote on 2022-08-23, 16:46:

[...]

Well yes, but not the worst marketing tool given that the way computers were marketed to your grandma etc. was always by comparing the amount of MHz. In the minds of most consumers more MHz would equal a faster system. You'd already loose a large chunk of your audience if you tried to tell them that you can only really compare two of the same exact type of CPUs if you wanted to use MHz as a metric...

The Pentium 4 was a horribly inefficient architecture and performance per-clock was often pretty bad compared to a similarly clocked Athlon - true, not due to the CPU alone, but on the whole given the motherboard, memory etc. I'm not saying AMD was right with their PR, but it was a somewhat understandable attempt to wean most people off the MHz == performance perception.

Disagree. It was completely counterproductive because it sounded like MHz speeds and so reinforced the idea that that was how to measure performance.

Intel was far more effective at weaning off by explicitly choosing names for the later P4 CPUs (500 series, 600 series, 800 & 900 D series) that had nothing to do with speed in MHz. No one batted an eyelid when the C2D was launched with completely different numbers. Meanwhile AMD kept PRing for years...

I am pretty sure I have Athlon Thunderbird 1000C model with Abit KT7-Raid 100MHz bus. It is set 1GHz 10 x 100MHz and works fine. I cannot comment on performance differences though.

The Thunderbird 1000B might be nice for KT133 boards, if you happen have one (KT133 runs max 200MHz FSB (2x100), the name is somewhat misleading). These CPUs are fairly easy to unlock btw.

The Serpent Rider wrote on 2022-08-23, 14:56:

My "common sense" tells me that, given a fixed CPU frequency, a faster bus is generally better.

Generally useless on SDRAM systems, because RAM can work at 133 Mhz on both and can't saturate K7 DDR bus.

Paired my 1400C with a ASUS A7N266-VM, which supports DDR 266mhz (PC-2100), controlled by nForce (1) chipset.
CPU and RAM buses synched.
Windows Millenium Edition, 1GB DDR 266 (over-512MB-RAM patch installed). MX440 (128bit).
Everything runs so smoothly. A very nice build to play anything 2000-2002.
Almost period correct (CPU is from 2001, motherboard and GPU are from 2002) perfect.

Tetrium wrote on 2022-08-23, 20:04:

The Thunderbird 1000B might be nice for KT133 boards, if you happen have one (KT133 runs max 200MHz FSB (2x100), the name is somewhat misleading). These CPUs are fairly easy to unlock btw.

KT133 - maybe for SDRAM support up to 133MHz?

Indeed, the p-Rating is a lie and always been. Even back in the Cyrix days for socket 7.
A pr-166 Cyrix was probably performing like a Pentium 166. BUT ONLY on non floating point operations.
So it was already half of a lie right in the beginning.

Additionally, the pRating was used for something it was not meant for : comparison between different ranges.
You can't compare an AthlonXP 3000+ with a Sempron 3000+ and an Athlon64 3000+.
The p-rating is assumed to be "per range". If you ignore this, you run into crazy things :
"Is is the Sempron 3300+ better than the AthlonXP 3200+ ?"
Probably not ! Same freq, Same cache, Same FSB, i's probably the same chip rebranded.
And the Sempron 3300+ is definitely not better than an Athlon64 3200+.

At most, it's a quick way to evaluate relative performances for chips in the same range having different cache capacity or fsb speeds.

For the Athlon 1000, unfortunately, i will not be able to test the difference since i only have the "C" model, but 5% is not that bad actually.
I will be using this CPU with an SIS 735 Chipset (K7S5A Pro Motherboard) with SDRAM.
It's not a very performant chipset. But i need it because it supports the Rage Fury Maxx.
For this system,i'm trying to reflect the performance level of a year 2000 Athlon system, even if the motherboard is more recent.
I would love to have a slot A system, be there are quite uncommon in France at least.

Ydee wrote on 2022-08-24, 07:34:

Tetrium wrote on 2022-08-23, 20:04:

The Thunderbird 1000B might be nice for KT133 boards, if you happen have one (KT133 runs max 200MHz FSB (2x100), the name is somewhat misleading). These CPUs are fairly easy to unlock btw.

KT133 - maybe for SDRAM support up to 133MHz?

Iirc that was the idea, yes.