I wanted to retest the board, just once more. I ran the f.e. twice, and it didn't crash. Then, I ran the s.e. and it went mad. The textures on the tests were very corrupted. It's possible that some complex programs (eg 3DMark) don't clean up the system correctly after use, and the s.e. didn't like this. I had to reboot, and I then ran the s.e. 3 times (once with its settings maxed out), and it was OK.

Does anyone know which VIA 4-in-1 driver package is the most stable, and does it matter if

a) the board is either raid or non-raid
b) the revision is either 1.0, 1.1, 1.2 or the final 1.3

I find this very interesting: I have tried many times to install an ATI Radeon 7500 in an Abit KT7A 1.3 non-raid board. It uses the B4 BIOS, which I think is the very latest 1.3 non-raid BIOS. I gave up, and tested another KT7A non-raid board, this time a 1.2 version board. (I used the same HDD, stuffed full of failed installs, and unholy VIA 4-in-1 drivers.) It's got some old BIOS on there, not the latest "A9". It's the "64" version; dated 11/07/2001. Quite old. Well, the ATI Radeon 7500 installation was a breeze. No problems whatsoever. Isn't that odd.

retro games 100 wrote:

I find this very interesting: I have tried many times to install an ATI Radeon 7500 in an Abit KT7A 1.3 non-raid board. It uses the B4 BIOS, which I think is the very latest 1.3 non-raid BIOS. I gave up, and tested another KT7A non-raid board, this time a 1.2 version board. (I used the same HDD, stuffed full of failed installs, and unholy VIA 4-in-1 drivers.) It's got some old BIOS on there, not the latest "A9". It's the "64" version; dated 11/07/2001. Quite old. Well, the ATI Radeon 7500 installation was a breeze. No problems whatsoever. Isn't that odd.

Did you use the t-bred 2400 in the rev 1.2 board? 100mhz FSB?

Amigaz wrote:

Did you use the t-bred 2400 in the rev 1.2 board? 100mhz FSB?

Not yet. I've only just unpacked it, and booted it up. I decided to use the seller's pre-installed CPU, which was a t-bird. This was just to get the basics working. Now that it works, I will try the 2400 t-bred. I'll post back with the results...

It looks like I've solved the mystery of the Radeon 7500 card problem. I flashed the 1.2 board's BIOS, with the latest A9 version. I wanted to do this, before installing the t-bred 2400+. The 2400+ CPU went in, and I switched on power. The BIOS POST screen reports that it is an "unknown CPU type" running at 2000mhz. Several trips to the BIOS to perform mundane changes (such as disabling serial ports, etc) did not alter this "unknown / 2000mhz" BIOS POST reading. (As you recall from one of my earlier posts, doing these trips to the BIOS set up area for the 1.3 board, made this board increase the FSB from 100mhz to 133mhz, and so increase the CPU's real clock speed.)

So, the desktop appears and it's all corrupted again. The ATI Radeon 7500 card doesn't like the t-bred 2400+, even running at 100fsb. Using the t-bird 1200 (at 100fsb - 900mhz), the Radeon card was OK.

I'll have to swap out the Radeon card again, for an nVidia card, before I can continue testing this Abit KT7A 1.2 non-raid with a t-bred 2400+ CPU...

(I removed the Radeon 7500, and installed an nVidia MX 440.) Regarding the BIOS POST screen reporting that the t-bred 2400+ CPU is an "unknown CPU type", I can take a guess why this is -

The t-bred 2400+ CPU I'm using was released on the 21st of August 2002. You can check this on cpu-world.com The latest BIOS file applicable to the KT7A 1.2 revision board has a datetime stamp of 07/11/2002. That's probably why the BIOS program code doesn't understand what the 2400+ CPU is. On the other hand, that other Abit board I just tested, the 1.3, this had the "B4" BIOS, which was dated 08/22/2002. Look at that - just one day after the 2400+ CPU was released! When the BIOS POST screen appears on this 1.3 board, for the 2400+ CPU, it gets correctly identified as an "Athlon XP 2400+ CPU".

A quick go on Sandra 2002 Pro:

dry 5575
wet 2768

I'll run four 3DMark benchies next (99, 2000, 2001f.e., 2002s.e.) .....

retro games 100 wrote:

(I removed the Radeon 7500, and installed an nVidia MX 440.) Regarding the BIOS POST screen reporting that the t-bred 2400+ CPU […]
Show full quote

(I removed the Radeon 7500, and installed an nVidia MX 440.) Regarding the BIOS POST screen reporting that the t-bred 2400+ CPU is an "unknown CPU type", I can take a guess why this is -

The t-bred 2400+ CPU I'm using was released on the 21st of August 2002. You can check this on cpu-world.com The latest BIOS file applicable to the KT7A 1.2 revision board has a datetime stamp of 07/11/2002. That's probably why the BIOS program code doesn't understand what the 2400+ CPU is. On the other hand, that other Abit board I just tested, the 1.3, this had the "B4" BIOS, which was dated 08/22/2002. Look at that - just one day after the 2400+ CPU was released! When the BIOS POST screen appears on this 1.3 board, for the 2400+ CPU, it gets correctly identified as an "Athlon XP 2400+ CPU".

A quick go on Sandra 2002 Pro:

dry 5575
wet 2768

I'll run four 3DMark benchies next (99, 2000, 2001f.e., 2002s.e.) .....

Is it running at 20x100mhz = 2000mhz now on the 1.2 board? or are you only using it on the 1.3 board?

Amigaz wrote:

Is it running at 20x100mhz = 2000mhz now on the 1.2 board? or are you only using it on the 1.3 board?

I am currently testing the 1.2 board now. And yes, the t-bred 2400+ CPU is running at 20x multiplier, at 100fsb = 2000mhz real clock speed. I did not "select" these CPU speed settings from the BIOS set up area. I simply allowed the BIOS to pick this running speed for me.

BTW, this will definitely work for a 1.1 board. I know, I've tried.

retro games 100 wrote:

Amigaz wrote:

Is it running at 20x100mhz = 2000mhz now on the 1.2 board? or are you only using it on the 1.3 board?

I am currently testing the 1.2 board now. And yes, the t-bred 2400+ CPU is running at 20x multiplier, at 100fsb = 2000mhz real clock speed. I did not "select" these CPU speed settings from the BIOS set up area. I simply allowed the BIOS to pick this running speed for me.

BTW, this will definitely work for a 1.1 board. I know, I've tried.

Awesome! 😀

Will be interesting to see how your 3dmark etc benches will work....

Benchies, as promised:

3DMark 99 Max = 9808 (3DMarks), 25220 (CPU 3DMarks)
3DMark 2000 (version 1.1) = 6773 3DMarks

I then ran 3DMark 2001 (f.e.), but during the first jeep chase test, it displayed an error message warning me that my graphics card was trying to do something it couldn't. (Sorry, I forgot what the exact message was.) So, I went to the mobo's BIOS set up area, and disabled AGP fast write support (as a guess), then resumed testing. Unfortunately, 3DMark 2001 f.e. then dropped back to the desktop during testing, and so I made another trip to the BIOS setup area, and disabled as many chipset features I could see - mostly to do with AGP, PCI, caching, etc. Then, testing resumed once more. (I notice that the animation is quite often a bit poor - jerky and non fluid.)

3DMark 2001 (f.e.) = 3605. I reran this test, to get a 2nd score, but the graphics got corrupted half way through, and so I pressed the ESC key to abort this test.

3DMark 2001 (s.e.) = 3309. Again, the animation is quite often jerky.

I will now compare those last 2 scores above, with the scores obtained using the 133fsb 1.3 board. On this "1.3 133fsb board", using the f.e. test, I was getting scores ranging from about 4100 through to 4800+. Using the s.e. test, I was getting more "stable" scores of about 4500. As you can see, the 3600 and 3300 scores on this "1.2 100fsb board" are much lower. Curiously, the board seems more unstable than the 1.3 133fsb board! I started to wonder if it was the fast speed of the 2400+ (2000mhz) t-bred. But then Amigaz reports that his 1000mhz t-bird isn't stable either, in his 1.1 board. I suppose the problem could be related to either the 4.35 or 4.56 VIA 4-in-1 driver, but I have been using the 4.56 driver OK in other KT133A boards.

I'm beginning to wonder if the problem is to do with Abit.

One final thing - whenever I run these benchmarks, there's always a curious faint "click" sound that comes from either the graphics card, or something on the mobo very close to it. I don't think I've ever heard this noise on other boards.

I think what I'll do is swap out the nVidia MX 440 card, and put in a Radeon 9800 Pro card, then rerun all of the 3DMark benchmark tests again...

I can definitely confirm this unusual problem with the ATI Radeon cards. I tried an ATI Radeon 9800 Pro card, with the same 6.2 catalyst driver I used with the 7500 card. When I get to the desktop (using the t-bred 2400+), it looks corrupted. Then, all I did was swap out the t-bred CPU, and replace it with a slower t-bird 1200 (133fsb), and the problem vanished.

This means I can't rerun all of the 3DMark tests with the t-bred 2400+ using any of the Radeon cards. So I'll get rid of it (the Radeon), and use an nVidia 6800GT instead. That should boost the 3DMark scores a bit! Who knows, maybe that will make the board more stable? Hehe, it'll probably makes things worse...

Change of plan: forget the 6800GT, I'll use an FX5950 instead. I maxed out many of the mobo's BIOS settings, without understanding what I was doing. (Nothing new there.) Then I reran the benchies:

3DMark 2001 (f.e.) = 8156. After I got that score, I shut down the machine, added another fan to cool down the system, and reran the test. I got 8115

3DMark 2001 (s.e) = 8348, then I ran it again and got 8346

So far, so good. That's 4 tests run without any "incident", 3 of which were run one after the other. Next test:

3DMark 2000 (version 1.1) = Well I spoke too soon. It instantly dropped back to the desktop. Who knows why? Perhaps running both 2001 f.e. and s.e. has messed up windows 98? I rebooted the machine, then restarted this 3DMark 2000 test. Same problem - it drops back to the desktop. I think 3DMark 2000 is incompatible with the FX5950 running nVidia's 71.84 drivers.

This is interesting: 3DMark 99 Max won't run either. I guess I can only run the 2001 f.e. and s.e. versions using the FX5950. Well, I have done this and the results are above without any problems or crashes.

I guess the next set of tests is to do all of the above on the 1.3 board, running at 133fsb and compare all of these 2001 scores...

I maxed out the 1.3 board's BIOS, just like I did with the 1.2 board. (Well, I tried to remember as many of these settings as possible.) Then I ran the f.e. test. Within about 5 seconds, it completely locked up. I shut the machine down, then swapped over the RAM sticks: I used the stick of RAM I was using in the successful 1.2 board testing. I then reran the f.e. test, and it locked up again, about 3 minutes in to the test.

Oh well. Maybe I set the mobo's BIOS settings too "aggresively"? Perhaps the board is unstable at 2000mhz @ 133fsb?

🙁

Are you still using the Cosair power supply in this setup?

prophase_j wrote:

Are you still using the Cosair power supply in this setup?

Yes I am. 😊 (Right now, I wish I had a better replacement.) Do you think it's likely to be a PSU problem? Do you think it's worth me getting something with more power on the 5 amp rail? I don't want my testing to indicate failure of the board and/or its components, when in fact it was due to lack of power on the 5 amp rail.

Another way of looking at this is - I could remove the power hungry FX5950, and put something primitive in its place. If that fails as well, then could we discount the requirement to change PSUs?

I could reuse the nVidia MX 440 card. I can't use the ATI 7500 - it doesn't like the high speed 2400+ CPU for some reason. Or I could use another "simple" AGP card perhaps?

Hang on a minute! I've just dug out my Nexus NX-3500 PSU. It's "only" 350W, which is 100W less than the Corsair, but the label on it reads:

+3.3V = 28.0A
+5V = 30.0A
+12V = 18.0A

It also says:

+3.3v&+5v=220w max

That should be ok, shouldn't it?!

That should do much better. Even if your pulling all 30 amps of the 5v, you'll still have a little over 21 amps for the 3.3v, and that's okay. The 3.3 is used for a lot of things, like memory, chipset, and PCI devices. When you get a big video card, mainly one that has external power, they use the 12v. That rating on this power supply might struggle there. Another thing to note, since the card pulls from the 12v, its safe to say putting in a smaller card wouldn't have the same effect.

Ho-ho

Just threw in the t-bred 2400, cleared the CMOS and booted it up and setup the BIOS to my likings 😀

CPU-Z sense the CPU correctly

prophase_j wrote:

That should do much better. Even if your pulling all 30 amps of the 5v, you'll still have a little over 21 amps for the 3.3v, and that's okay. The 3.3 is used for a lot of things, like memory, chipset, and PCI devices. When you get a big video card, mainly one that has external power, they use the 12v. That rating on this power supply might struggle there. Another thing to note, since the card pulls from the 12v, its safe to say putting in a smaller card wouldn't have the same effect.

I guess i'm "safe" with my PSU I'm using in the KT7A box...it has 35amp's on the 5v line

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

BTW, the Corsair PSU worked OK with the FX5950 on the 1.2 board. The only fundamental difference between that 1.2 board and this 1.3 board is that it was running at 100FSB and not 133FSB. Really, that's the only "power related" difference.

I'm going to mess about with the BIOS settings again. I'll set them to "fail safe" or something similar, then retry the f.e. test. I think if that fails, one of my very early conclusions is that a CPU as fast as a 2400+ (2000mhz) combined with a 133fsb leads to instability. I don't actually know if I can "force" the board to use 100fsb. There's no jumper for the FSB. Maybe I can set it in the BIOS. Perhaps I could try that: lower the FSB. If I can't, or it doesn't work, I could try a CPU with a bit less speed, something like a 2100+ rated CPU...