retro games 100 wrote:

Damn, it didn't work. I tried the Nexus PSU. The f.e. test got slightly further this time, before freezing.

Get some more Juice 😜

Amigaz, it's great that you've got the t-bred 2400+ installed OK. I have found that 3DMark 2001 first edition is the version which is the "toughest" to run. In order to feel comfortable that your system is stable, please try running this f.e. at least 3 times in a row, preferably 4 or 5. It doesn't take all that long - the f.e. runs quicker than the s.e.

Thanks.

retro games 100 wrote:

Amigaz, it's great that you've got the t-bred 2400+ installed OK. I have found that 3DMark 2001 first edition is the version which is the "toughest" to run. In order to feel comfortable that your system is stable, please try running this f.e. at least 3 times in a row, preferably 4 or 5. It doesn't take all that long - the f.e. runs quicker than the s.e.

Thanks.

Will do

Just ran 3dmark2001se with everything maxed (32bit, 4xaa etc) at 1280x1024, no crash...got 6747 3D marks 😀

*off to d/l 3dmark "oldie version" *

I'm gonna shut this horrible contraption down for a bit. The sound of all these fans is giving me tinnitus! Well, I guess my testing ends on a positive note: I swapped out the t-bred 2400+ CPU, and put in a t-bred 2000+ rated CPU instead. (For some odd reason, I've never got a 2200+ to POST in it, and I don't have a 2100+). I ran the f.e., and it didn't crash. I got a score of 8784, with this utility left on all its default settings. I also cleared the CMOS, and selected the BIOS "fail safe" setting.

I'll find some earplugs and retest the f.e. 4 times in a row tomorrow. I'll also set the 5th test to max settings (anti aliasing, etc...)

So, why does it work? Who knows. Maybe using a t-bred 2000+ doesn't use so much juice. Maybe the board is unstable at 2400+ @133fsb? After all, it was hardly designed for it!

Ok

I've run 3dmark2001 f.i now 4 times....no crash but I noticed the score degraded a bit after each run

I've used the latest Nvidia drivers (GF FX 5950 Ultra) and Windows XP when testing both 3dmark2001 and 3dmark2001se

Amigaz wrote:

Ok

I've run 3dmark2001 f.i now 4 times....no crash but I noticed the score degraded a bit after each run

I've used the latest Nvidia drivers (GF FX 5950 Ultra) and Windows XP when testing both 3dmark2001 and 3dmark2001se

Windows XP? That's cheating! 🤣 I was using Windows 98SE, with the 71.84 driver package for a GF FX 5950 gfx card. Also used: win98 unofficial service pack, and VIA 4.35 4-in-1 driver package.

Is it possible for you to slap together a temporary win98se HDD? 😀

retro games 100 wrote:

Amigaz wrote:

Ok

I've run 3dmark2001 f.i now 4 times....no crash but I noticed the score degraded a bit after each run

I've used the latest Nvidia drivers (GF FX 5950 Ultra) and Windows XP when testing both 3dmark2001 and 3dmark2001se

Windows XP? That's cheating! 🤣 I was using Windows 98SE, with the 71.84 driver package for a GF FX 5950 gfx card. Also used: win98 unofficial service pack, and VIA 4.35 4-in-1 driver package.

Is it possible for you to slap together a temporary win98se HDD? 😀

sure, I'll test it in win98se too...this machine dual boots win98se/winxp
but I'm not sure I'll do that test today....might be tomorrow

retro games 100 wrote:

So, why does it work? Who knows. Maybe using a t-bred 2000+ doesn't use so much juice. Maybe the board is unstable at 2400+ @133fsb? After all, it was hardly designed for it!

I'm pretty confidant it all comes down to power. Not matter how fast the processor is, the FSB is going to stay the same. I too used to think it was correlated to the stability, but those issue are long gone.

Here's where I'm currently at with this project:

PSU: 350W Nexus; 3.3v=28A, 5v=30A (220w combined), 12v=18A
(Removed Corsair 450W; 3.3v=20A, 5v=20A [just 130w combined!], etc)
CPU: t-bred 2000+ rated
(Removed t-bred 2400+ rated)
FSB: 133
Gfx: FX5950 (71.84 drivers)
(I will remove this soon, because it's too power hungry for my Nexus PSU)
OS: Win98SE, VIA 4-in-1 drivers: version 4.35.
RAM: 1 stick of 256mb PC133 (Infineon)
BIOS: "optimised" setting selected, but I didn't manually "max out" any settings myself.

4 lots of 3DMark2001 first edition testing = 8802, 8969, 8960, 8970
Then 1 lot of 3DMark2001 f.e. testing (every setting maxed out, 1280x1024) = 7413!

During testing, I could hear the computer make "hissing" sounds. No problems occurred: no dropping back to the desktop, no lockups, no freezes, and no reboots required.

Until I get a better PSU, I'm going to sideline the FX5950. I'm going to swap it out for an nVidia MX 440 AGP card. I'll also swap out the t-bred 2000+ CPU for a t-bred 2400+. Now that I'm using the Nexus PSU (instead of the Corsair with its lower 3.3v and 5v rails), I'm hoping that the 3DMark2001 f.e. testing will be OK with the 2400+ rated CPU - that is, now the FX5950 has been removed as well. Sure, the 3DMarks score(s) will be much lower, but I think the key test here is to see if the mobo is stable running @ 2000mhz/133fsb...

Following on from the post above - the only thing I've now done is swapped out the FX5950, and replaced it with an nVidia MX 440 (using the 45.something drivers). Everything else stays exactly the same.

4 lots of 3DMark2001 first edition testing = 3665, 4372,

On the 3rd test, the benchmark utility froze on the very opening splash screen. I had to reboot. Also, check out those 2 scores. It's weird the way they differ so much. I think it's time to abandon the MX 440. I've had too many problems with it. I need another basic AGP card. Something that's going to use even less juice. How about a Voodoo3?! So I put one in, and run the f.e. again. It warns me that my new graphics card is crap (well not in those words), and that the settings have been "downgraded" accordingly. The default settings still look like they're running at 1024x768x16 though. (Maybe it was referring to its "maxed out" option settings, which now cannot be set to 32-bit colour?)

4 lots of 3DMark2001 first edition testing = 1991, 2002, 2009, 2008. (Consistent values)

Then 1 lot of 3DMark2001 f.e. testing (every setting maxed out, but that means no AA, 16-bit for everything, and also 1280x1024) = 1188

So, no surprises. The thing that really needs to be done now is to swap out the t-bred 2000+ CPU, and replace it with a t-bred 2400+. I've just done that. Now for the benchies again -

4 lots of 3DMark2001 first edition testing = Well it froze on the first test. Something's clearly wrong. What could the problem be? Could the t-bred 2400+ @ 133 be drawing too much juice? With just a Voodoo3 AGP card, and absolutely nothing else in the mobo apart from 1 stick of RAM? I wonder if it's because the board itself can't cope with the 2400+ CPU running at 2000mhz @ 133fsb?

Edit: BTW, I will get a better PSU, and redo all of the tests.

Edit 2: Sorry to bore everyone even further. Hehe. I decided to increase the mobo's FSB setting by 10mhz. Everything else was left alone. The f.e. score with "max settings" improved from 1188 to 1368. The animation always appeared OK. (I'm using a Voodoo3.) I wonder if all the other tests had jerky animation because there's something not right about the AGP 4x mode?

I've swapped out the 1.3 board for a 1.2 board. The latest BIOS on this board is I think "A9" - (July 2002). It's older than the 1.3 board's BIOS, which is "B4" - (August 2002) Here's some "quick test" results -

* Like the 1.3 board, this 1.2 also will not POST with a t-bred 2200+ (I must check if it's broken!)

* This 1.2 board works OK with a t-bred 2000+ CPU, and at 133fsb. Real clock speed = 1666mhz.

* The "strange Radeon problem" (see previous posts) does not occur using the "slow" t-bred 2000+ CPU.

* A little disappointingly perhaps, the BIOS POST screen reports this "old" t-bred 2000+ as "Unknown". I guess the 1.2 board's "A9" BIOS didn't include this CPU in it's "CPU check list", unlike the 1.3 board's "B4" BIOS, which reports a t-bred 2400+ correctly on the BIOS POST screen.

A quick go on Sandra 2002 Pro:

dry 4649
wet 2309

I maxed out all of the mobo's BIOS settings, then ran 3DMark2001 first edition 4 times using a Radeon 7500 AGP card:

4112, 4116, 4123, 4126 (This is promising: consistent values)

Then, I swapped out the ATI Radeon 7500, and put in an FX5950. I reran 3DMark2001 f.e.:

No test results. The FX5950 is completely lifeless. No power, no POST. I'm not sure what the problem is. I'll have to swap out the 1.2 board for the 1.3 board again, and try again. Phew, that's a relief. The FX5950 grinds in to life on this 1.3 board. I max out the mobo's BIOS, then run 3DMark 2001 f.e. -

8975, 9141, 9143, 9154 (This is promising: reasonably consistent values, that do not "degrade")

Then, I max out every setting inside 3DMark2001 f.e., and run at 1280x1024: 7543

The machine makes a hissing sound, as if something sounds under strain. I think that noise is coming from the FX5950.

I'm coming to one early conclusion - if you want to ensure system stability, stick with a t-bred 2000+ rated CPU. They're OK at 133fsb. You might be OK with the 2400+ if you've got an excellent PSU with *more than* 30amps on its 5v rail. I really want to test this. But I can't ATM, because I don't have a really good PSU. Hopefully I can do this in a couple of weeks...

Edit: To conlude, in terms of all the speed tests- A t-bred 2400+ rated CPU running @ 100fsb was approximately 10-11% slower than a t-bred 2000+ rated CPU running @ 133fsb. (Using an FX5950)

I wanted to see why the FX5950 failed to POST in the 1.2 board, so I plugged this board back in again, and continued to use a t-bred 2000+ CPU with it. Rather than try the FX5950 once more (I see no point), I tried a Radeon 9800 Pro instead. And I get the same thing: no POST. But here's another indicator that the PSU (or rather, simply the power) is the "smoking gun". All I did was remove the 2000+ t-bred CPU, and replace it with a 1700+ t-bred. Now the board POSTs OK.

Edit: Running the 1700+ rated CPU (at 133fsb), with the Radeon 9800 Pro, I get these results from 3DMark2001 f.e. -

8298, 8301 (Seems consistent + OK; I couldn't be bothered to run any more of these tests)

Then I maxed out all the f.e. settings (including 6x AA!), and ran it at 1280x1024: 6207

I just tried a mobile barton in the 1.2 board. It's 2200+ rated. Looking at cpu-world, it has a core voltage of just 1.35v. The CPU code is AXMD2200FJQ4C. I took absolutely no chances, and popped out the mobo's button battery and also cleared the CMOS jumper! The BIOS POSTed first time. But to my horror, the BIOS set up area told me that the core voltage was 1.525! Is that bad? (Please note that at that point in time, the CPU was only running at 1250mhz - something like that - and it was on 100fsb). I quickly adjusted the CPU core voltage to 1.35, and at the same time, I set the CPU speed to multiplier "above 12" and fsb 133. That seemed to work. I'm now running at 1666mhz (real clock speed) with a fsb of 133. The BIOS POST message tells me that the CPU is "unknown".

I maxed out all of the BIOS settings, then ran 3DMark2001 first edition with a Radeon 7500. I get 4282.

A quick go on Sandra 2002 Pro:

dry 4566
wet 2263

Edit: I tried a mobile barton 2400+, but it looks like the 1.2 board only has an understanding of a 12.5 multiplier, and not the required 13.5 multiplier for this specific CPU, because the BIOS POST screen reports that its speed is 1666mhz. That's equal to 133x12.5. So, I guess the good news is that it works at 133fsb. But the 2200+ CPU seems a better bet, because that runs at a lower core voltage of 1.35, and also works at 133fsb. BTW, this 2400+ CPU runs at core 1.45v.

Now I'm using a mobile barton (1.35v, 2200+ rated, 1666mhz), instead of a t-bred (1.65v, 2000+ rated, 1666mhz), the FX5950 POSTs OK in the 1.2 board. I guess as the mobile barton is consuming less juice, there's more left over to get the FX5950 up and running?! I run 3DMark2001 f.e., but after a few minutes the screen goes black!

I go to the BIOS, and increase the CPU core voltage from 1.35 to 1.4, and increase the IO voltage from 3.4 to 3.5. I've no idea if I have to do this to get 3DMark2001 to run correctly, or if these new values are the best ones to try.

I rerun 3DMark2001 f.e., but this time I max out all of its settings. (I can't be bothered to run 4 lots of "normal setting" tests again.) After about 3 minutes or so, the screen goes black again!

Oh well. It's really interesting that this 1.2 board can't seem to cope with the FX5950, even with a lower powered mobile barton. It worked OK in the 1.3 board, with a t-bred. I say "worked OK", the animation in 3DMark2001 f.e. looked jerky and "unconvincing", and the FX card was making peculiar hissing sounds. Perhaps the "power lines" (for want a better word) on the 1.2 are in a poorer condition than the ones on the 1.3 board. I expect both boards could do with being fed better power from a better PSU! 😀

Even though the mobile barton was only 1.35v, I couldn't get the FX5950 to work sensibly in the 1.2 board without problems. So I removed it, and replaced it with a t-bird 1200 (133fsb). It has a much greater core voltage of 1.75v, but I guess it uses less PSU power, as it's a much less powerful CPU? (Really, I have no idea about volts and watts and stuff! - as if you hadn't already guessed.)

I max out all 3DMark2001 f.e. settings, and finally get a score without it locking up: 6535. I rerun again and get 6648. I rerun for a third time and get 6657.

So, now the t-bred and mobile barton have gone, replaced with a humble t-bird, it looks like there's just enough juice to keep the FX5950 going, even on max settings. And what's interesting is that the 3DMark2001 f.e. scores are fairly respectable.