Hmm...would be an interesting topic, which soundcards are best for the bugged VIA southbridge 😁

Tetrium wrote:

Hmm...would be an interesting topic, which soundcards are best for the bugged VIA southbridge 😁

Hehe. It could be bugged. Also, some of these problems could be fixed with different BIOS timings. I must admit that I didn't try too hard with this.

I had some problems with an Terratec EWS64 XL sound card in a slot 1 board. The General MIDI music sounded wrong - chopped up sounding. I changed resources so that the SB emulation used IRQ 7 instead of 5. That fixed the MIDI music problems. Strange.

udam_u wrote:

There is one quote written by Ralph Waldo Emerson which I really like: 'Do not go where the path may lead; go instead where there is no path and leave a trail'.

Trailblazer. 😀

retro games 100 wrote:

For the 128MB Pro card, I get 11625 in 3DMark 2001 s.e. In my post above, I got 11728 using the 256MB 9800XT card. Percentage wise, the XT is less than 1% faster, which is a small amount.

This tells you nothing, except how much the poor old SDRAM is limited. I got almost 11000 marks 2001 on a Radeon 8500 back in those times... and we have to agree that the 9800XT is a MUCH faster card. To find the difference between the two cards 9800XT and 9800 pro you must use a newer CPU with much faster RAM, so that these limitations are not in the way. Or, test at top quality, with anti aliasing on.

BTW, what mutliplier is used on that T-bred? Sandra lists 16x but the board is using a lower one - but what exactly? I guess that T-bred is locked, right?

Re: limitations of SDRAM. OK, I understand. The next time I test this mobo, with the 2 different Radeons, I will use AA.

Re: T-Bred. The board is using the default 16x multi, and I believe that the T-Bred is locked, because I have not applied any mods to it.

retro games 100 wrote:

Re: limitations of SDRAM. OK, I understand. The next time I test this mobo, with the 2 different Radeons, I will use AA.

Re: T-Bred. The board is using the default 16x multi, and I believe that the T-Bred is locked, because I have not applied any mods to it.

Nah, MoBo is using 8x multi, but it just doesnt understand it 😜 The fact that you have a CPU with locked multiplier saved you the trouble we went through with mobile bartons 😀

elfuego wrote:

retro games 100 wrote:

Re: limitations of SDRAM. OK, I understand. The next time I test this mobo, with the 2 different Radeons, I will use AA.

Re: T-Bred. The board is using the default 16x multi, and I believe that the T-Bred is locked, because I have not applied any mods to it.

Nah, MoBo is using 8x multi, but it just doesnt understand it 😜 The fact that you have a CPU with locked multiplier saved you the trouble we went through with mobile bartons 😀

Unfortunately, I don't understand what you are saying. I think the CPU is operating at a real clock frequency of ~2.54 GHz, when I set the mobo's FSB to 159. To infer the CPU's multiplier, is the equation: CPU clock speed / mobo's FSB? That would be 2540 / 159 = 16.

retro games 100 wrote:

elfuego wrote:

retro games 100 wrote:

Re: limitations of SDRAM. OK, I understand. The next time I test this mobo, with the 2 different Radeons, I will use AA.

Re: T-Bred. The board is using the default 16x multi, and I believe that the T-Bred is locked, because I have not applied any mods to it.

Nah, MoBo is using 8x multi, but it just doesnt understand it 😜 The fact that you have a CPU with locked multiplier saved you the trouble we went through with mobile bartons 😀

Unfortunately, I don't understand what you are saying. I think the CPU is operating at a real clock frequency of ~2.54 GHz, when I set the mobo's FSB to 159. To infer the CPU's multiplier, is the equation: CPU clock speed / mobo's FSB? That would be 2540 / 159 = 16.

KT133A boards do not "see" the last L3 pin; they assume that the CPU has multipliers only up to 12.5x (or, four L3 bridges). So, depending on the fifth L3 bridge (introduced later, with Tbreds, bartons etc) you get multipliers 3-12.5x or 13-24x. If the CPU wasnt locked, you would have had it running only at 8x133 = 1064 Mhz, for the MoBo would ignore the 'unknown' fifth L3 bridge.

You can see the L3 (multiplier) layout here:
http://www.ocinside.de/go_e.html?http://www.o … d_painting.html
note the difference between 8x and 16x (bottommost L3 bridge) - the fifth bridge enables higher multipliers.

As I said - you are lucky your CPU is locked 😉 Locked CPUs override L3 bridges 'hint' and work at default multiplier. So, dont unlock the CPU - less you want to underclock it 😜

elfuego, thanks very much for the explanation! 😀

I decided to race two different Radeon 9800 cards. One is a 128MB Pro, the other is a 256MB XT. I used 3DMark 2001 s.e., and used all of its default settings, except for AA which I set to 6x. The FSB was increased from 133 to 150, which gave the 2600+ rated T-Bred a real clock speed of 2.4 GHz.

Pro = 9104
XT = 9460

The XT is 3.76% faster than the Pro card, when 6x AA is used on this 2.4 GHz system.

I removed the T-Bred 2600+ rated CPU, and tried a T-Bred 2400+ rated CPU. I set the FSB to 160, to give this CPU a real clock speed of 2.4 GHz. I then ran Quake in pure DOS, and its timedemo demo1 test finished in 3.3 seconds scoring 293.0 FPS. The BIOS core voltage was set to 1.850, which is its highest setting.

BTW, on the mobo there is a voltage jumper with 5 settings: 0.0, 0.1, 0.2, 0.3, 0.4. This jumper can only be found on the Pro boards. Does this jumper work in conjunction with the core voltage setting found inside the BIOS set up area, or does the core voltage BIOS setting override and effectively ignore this jumper?

With the FSB set to 160, please can someone tell me what speed the AGP bus is running at? I understand that some AGP cards don't like to be run at high speeds. The card installed at the moment is a Radeon 9800. Do some AGP bus slots adjust their speed, so that they implement a kind of "ratio" speed, in conjunction with the speed of the host bus speed? I have seen some mobos with a jumper to set this "AGP to host bus speed ratio". I wonder if the AGP bus speed on some mobos is automatically adjusted by the BIOS, using some kind of "ratio" value, so that the AGP slot does not run too quickly?

It would be great if there was a utility that told you the speed of the slots on a mobo. For example, the AGP, PCI, and ISA slots.

Edit: I need to point out that the 2400+ rated T-Bred @ 160 FSB = 2.4 GHz configuration goes nuts if I attempt to run Windows 98. This overclock is too much for the CPU to cope with. DOS is OK however, but you do need to send a lot of extra voltage through the CPU in order to run Quake successfully.

retro games 100 wrote:

TheMAN wrote:

the Thoroughbreds get very unstable once you overclocked the shit out of it... I have an XP 2100, but I made it run at 2.1ghz re […]
Show full quote

the Thoroughbreds get very unstable once you overclocked the shit out of it... I have an XP 2100, but I made it run at 2.1ghz real speed.. so that was basically "XP 2600"... I up'd the FSB but lowered the multiplier... it was a bit faster but overtime the system got more and more unstable
I used it this way for 6 of the 8 years that it was in service

now it's in pieces, and I'm using an Athlon 64 X2 now... as of 2 months ago 😁

if you want a stable overclock, get a Barton

Maybe the performance degredation you were experiencing was either partly or possibly entirely related to capacitor failure? Caps can still go "rotten", even if they look OK.

However, it is understandable that continuous "over volting" of a CPU could permanently affect its stability. Visibly, do these particular CPUs show any sign of damage? For instance, do they "crumble" a bit at the edges, or is that entirely due to heatsink mounting/dismounting problems?

I didn't remove the CPU so I don't know
the CPU was run 0.1V over before... don't think it'll do too much damage

and yeah, the caps could have gone bad.... I don't know and I don't really care... the board was weird and screwed up enough that for the past few years, the network connectivity just flakes out after 3 days... like clock work... didn't matter what I did in XP, new drivers, tweaks, etc... all did NOTHING

I even threw in a old 3C905B-TX card in there hoping that an actual NIC would cure it... same thing... so something was up with the board or XP

I don't really care to find out now... it deserves to be sent out to the pasture

Asus A7V8X board btw

What a coincidence. This weekend, I took out my Epox 8KTA-3pro (thanks to RG100) to build just the same thing.

I am using an Athlon XP-M 2600+ (mobile). As expected, I can't boot to full speed. In bios, I can only get to 12.5 multi (by setting 13). Once in Win98, I can switch the multi anywhere between 3 and 15 it seems. As I obviously try to build another final all in one Dos/Win98 Box, this is perfect for me. 300-1666 MHz in DOS, and 2000 and more (via fsb) in Win98 should let all games fly.

As I plan to disable L1 and/or L2 cache to shift speed, I am forced to use ISA sound. When I was dumping stuff, I kept the AWE64 gold, since RG100 reported her as good match before. I also have a cap-modded Maestro 16/32 with build in wavetable, which I could try. Of course, I will also give my SB16s and SB32s a chance.

I have given up my Geforce 6800 though. So most likely, I will power a Voodoo5 &2 or a Geforce 4200 & V2 SLI.