prophase_j wrote:

What I'm not sure about ATM, is if you get a higher than 12.5x non-mobile, if it will boot as a it'd default, and also if you get a mobile and wire-mod it to get the higher setting.

Based on tests done on mobos with the same BIOS datestamp (09/04/03), all of which are on their "last legs" (which could affect test results) -

PCB revision number 1.0 will not boot with either a Palomino or T-Bred higher than 1800+. A Palomino 1900+ locks up the mobo within 30 seconds, and a T-Bred 2200+ causes the "debug LED panel" to instantly display "FF", and the BIOS POST screen does not appear.

However, PCB revision number 1.1 will boot OK with both of these CPUs - both are correctly identified on the BIOS POST screen, and they both work at 133fsb. (A T-Bred 2400+ boots OK too, but because the BIOS lacks a 15x multiplier for this particular CPU, the BIOS POST screen only reports it as an 1800+ speed CPU.)

I am waiting for 2 XP-M mobile CPUs to arrive (2200 and 2400). When they do, I will test them both on the PCB 1.0 and 1.1 boards.

The squeaking comes from the voltage regulation loop becoming unstable because of degrading capacitors.

Well sometimes it's just inductors that are vibrating at a high frequency and you can hear it. You can give the inductors a coat of epoxy to shut 'em up.

Looking at the Epox EP-8KTA3+Pro board with all of its onboard components (caps, coils, funny metal things, etc!), are there any components that I must not touch, when the mobo is operational? I'd like to prod at a few things, to see if I can identify the squeaking component(s), and also see if prodding them temporarily quietens them.

Get a long plastic stick if you want to prod live electronics :p

Capacitors can hold a significant charge even after turning the power off. Other than that, I don't think anything on a motherboard is typically higher than 12V (although the current draw can be high, especially for CPUs and video cards).

I hate inductors, they're not standardized enough to easily identify. Even when you know you're looking at one, it can be a pain to find out what its inductance value is.

I've just tested a mobile barton CPU, 2400 speed, code: AXMH2400FQQ4C, brown colour. Inside the Epox mobo's (PCB rev 1.1) BIOS set up area, I adjusted the multiplier setting to 13. (According to CPU world.com, this mobile barton has a 13.5 x multiplier, but unfortunately that value does not appear in the BIOS choices available.) The BIOS POST screen then reported this information -

AMD K7 processor 2000+ (133x12.5),FSB 266

In terms of speed, 2000+ is the best I could achieve. If I selected other multipliers inside the BIOS, such as 12 or 14, I got slower CPU speed results when the BIOS POST screen appeared. (I'm afraid I've been a bit dense. I didn't check beforehand whether the BIOS supported a 13.5 x multiplier.)

Edit: I really have been dense. I also ordered a 2200 mobile barton, code AXMD2200FJQ4C. I've just checked that CPU on cpu-world.com, and it has a multiplier of 12.5. Unfortunately, that setting also does not exist inside the mobo's BIOS set up area. However, when the aforementioned 2400 CPU information is displayed on the BIOS POST screen, it does mention 12.5, so maybe the mobo is aware of that specific 12.5 multiplier value?

I'm a bit disappointed that a) the mobo does not accept the 2400 barton's 13.5 multiplier, and b) I was stupid not to appreciate that before buying the CPU. Oh well. I decided to try out the famous Epox overclocking. I just picked a nice round number of 150. So, I input that number as the mobo's FSB value inside the BIOS set up area - usually this value is set to 133. Everything seemed to work OK. I got to the windows 98 desktop, and ran some benchies. Also, Prime95 didn't fail, although I only ran it briefly.

Sandra scores =

dry 5167
wet 2547

This is roughly on par with the 2200+ CPU comparison score, so no big deal really.

CPU temp = 36
BTW, the BIOS core voltage setting was set to its lowest possible value of -0.1v!
SpeedFan core reading = 1.52v (according to cpu world, it should be 1.45 for this barton CPU)

Edit: I increased the OC'ing FSB value from 150 to 166. The mobo POSTed OK, but Windows 98 quickly went "mad" because it told me that it had fixed a registry problem, and said it needed to reboot - at which point the FSB value went back to 133.

Edit 2: Perhaps I should get one of these: XP-M mobile barton 2500+ (AXMH2500FQQ4C). cpu-world says that its multiplier is 14x. That's the maximum multiplier setting allowed inside the Epox's BIOS.

I seem to have confused myself even more. I didn't think that was possible. I've just put in a t-bred 2400+ cpu (AXDA2400DKV3C), and the mobo has reported that it's running at its correct 15x multiplier value - but why is there no 15x multiplier option inside the BIOS settings? SiSoftware Sandra tells me that the CPU is running at 2Ghz (correct), and the FSB is 266, and the memory speed is 133, and the multiplier is 15. All systems go! Sandra gives me respectable scores of -

Dry 5521
Wet 2738

These values are better than the 2200+ comparison values (and so they should be). I don't know what's going on here!

Probably the CPU is able to report its desired multiplier to the motherboard, and the motherboard is automatically using that value since it exceeds what could be manually set. It's possible that no CPUs available used 15x at the time your motherboard's BIOS version was released but that the clock hardware still supports it anyways.

The mainboard doesn't set the multiplier, this is done on the CPU.

Normal Athlons ignore multipliers set by the mainboard and are hardwired on a certain multiplier (encoded in the L3 and L4 bridges, if I recall correctly). It is possible to bypass the stock multiplier by closing the L1 bridges, which is easily done on a Thunderbird/Spitfire, but hard work on a Palomino/Tbred/Barton/Morgan. Some variants. like the multiprocessor and mobile versions already come factory-unlocked.

Early Athlons use a 4-bit ID to set the multiplier for the CPU. This gives us 16 states, enough to set a range from 5.0x - 12.5x in 0.5x steps.
Newer Athlon cores use a 5th bit to differentiate between a low (5.0x - 12.5x) and a high range (13.0x - 16.5x in 0.5x steps, 17x - 24x in 1x steps).
Older mainboards designed for Thunderbird/Spitfire cores can only use the 4-bit ID, thus limiting multiplier settings to the low range. However, this this doesn't stop them from working if the CPU is locked to a multiplier in the high range.
The only thing that can happen is that the BIOS reports nonsense or even crashes, but that is usually fixed in later BIOS revisions.

I hope this clears the confusion a bit...

I've just discovered that T-Bred 2400+ CPUs are rather versatile. They work in every KT133A board I've tried! The 2 Abit boards (KT7A rev 1.1 & rev 1.3) can run them @ 100 and 98 FSB: the CPU's real speeds are reported as 2000 and 2200 respectively. The rev 1.0 epox board (8KTA3+Pro) can run them, even at 133 FSB. The QDI Kinetiz 7E-A can also run them at this top FSB speed. All of them perform well, without exception. I'm quietly confident that if you picked any KT133A chipset based board, these top performing CPUs would work. Here are some examples -

Abit 1.1 @ 100 FSB

Sandra: dry = 5471, wet = 2711. These values comfortably beat the 2200+ comparison scores.

Abit 1.1, OC'd 107/35

Sandra: dry = 5945, wet = 2949. Not bad!

Abit 1.3 @ 98 FSB

Sandra's CPU scores are 6055, 2985. These are great values, IMHO.
Quake2 @ 1024x768 = 176.9 fps (using "crappy" nVidia 440 MX)
3DMark 99 = 7274 3Dmarks, 25959 CPU 3Dmarks (440 MX)

Epox 1.0 @ 133 FSB

Sandra: dry = 5503, wet = 2744.
3DMark 99 = 5926 3Dmarks (edit: v-sync may have been on), 29469 CPU 3Dmarks (using "crappy" Radeon 7500)

QDI Kinetiz 7E-A @ 133 FSB

Sandra: dry = 5469, wet = 2727.
3DMark 99 = 9081 3Dmarks (v-sync was off), 29490 CPU 3Dmarks (using "crappy" Radeon 7500)

Also, it's possible that for some of the tests above, I had v-sync on. Maybe ignore the 3Dmarks scores, and just look at the CPU 3Dmarks scores instead?

I should have posted the message above, along with this one below in my other thread called "More fun and games with VIA's KT133/A chipset". Nevermind.

From the opening paragrah in the message above, I said: "I'm quietly confident that if you picked any KT133A chipset based board, these top performing CPUs [T-Bred 2400+] would work." Well I was wrong. I have just discovered that not all KT133A chipset based boards are the same! I've just tested an IWill KK266 rev 1.2 board. I flashed its BIOS to the latest: KK1022.bin, which is the 22nd of October 2001. That's quite an old BIOS. I tried a whole pile of CPUs. All T-Breds (1700, 1900, 2000, 2400) failed to POST. For each T-Bred tested, I set the mobo's FSB jumper to 100 and 133. Somewhat unsurprisingly, a mobile barton 2400 also failed to POST. A Palomino 2000 failed to POST too, but a Palomino 1900 POSTed OK, at 133 FSB. The BIOS POST screen correctly identified this CPU as an XP 1900+ CPU. Inside Windows 98, Sandra reports its real speed as 1.60Ghz. Here are some Sandra CPU scores -

dry 4414
wet 2190

Using Sandra's comparison CPU checks, these 2 scores sit between the XP 1800+ and the XP 2000+, which is to be expected.

Inside the BIOS set up area, you can set the CPU clock speed to as high as 200! The clock ratio's highest setting is 14x. I decided to adjust the CPU clock setting from 133 to 150, but I got no POST. I increased the core volts inside the BIOS set up area, and also gave the IO a 5% boost using a jumper on the mobo, but still no POST.

To conclude, this IWill board seems the least powerful of all KT133A boards I have tested, in terms of its ability to accept fast CPUs. I'm surprised, as the reviews of this board were very positive. A review of this thing can be read below. In technical terms, it seems a bit outdated, probably because the reviewer was using either an earlier board revision, or an earlier BIOS, or both.

http://active-hardware.com/english/reviews/ma … board/kk266.htm

Is this is really bad idea: flash the IWill mobo's BIOS with a BIOS file intended for an Abit board? (Both are KT133A boards. Both have roughly the same look about them.) I have successfully flashed both boards with their correct & latest BIOS files, using the same awdflash utility. Both the IWill and Abit BIOS files are exactly the same size. The IWill BIOS is "old and crusty" - October 2001, whereas the Abit BIOS is newer: July 2002. I was hoping that if I could flash the IWill board with the newer Abit BIOS file, then I could get the IWill board to "understand" what a T-Bred CPU is. Currently, using the IWill mobo, no T-Bred CPU will POST in it.

Here is another example of IWill's "lack of understanding" about newer CPUs: A Palomino 1900 CPU works, but a Palomino 2000 does not. The P-1900 CPU was introduced in November 2001 - and the IWill's last BIOS is dated October 2001 - very approximately the same time. The P-2000 CPU was introduced in January 2002, a couple of months after the last BIOS update for the IWill board. Perhaps that is the only reason why the P-2000 CPU does not work?

Perhaps I could experiment by physically removing the IWill's BIOS chip, and replacing it with the BIOS chip from the Abit board. That way, if it doesn't work, I can simply "reverse" the situation - and I won't have corrupted the IWill's BIOS chip with a "bad flash".

It probably won't work out well unless they both have all of the exact same on-board hardware.

Swapping chips and clearing the CMOS is the safest way to test, assuming the Abit BIOS chip is compatible with the IWill mobo.

I noticed that an EP-8KTA3 board was sold on ebay just a couple of days ago. Anyone here on Vogons buy it? If so, please let us know how you get on with it!

retro games 100 wrote:

I noticed that an EP-8KTA3 board was sold on ebay just a couple of days ago. Anyone here on Vogons buy it? If so, please let us know you get on with it!

I'm "innocent" 😀

Hey Amigaz, I thought you already had a bunch of those motherboards lying around. Did you ever get a chance to test them? I hope your not planning some army invasion armed with exploding capacitors 😀

prophase_j wrote:

Hey Amigaz, I thought you already had a bunch of those motherboards lying around. Did you ever get a chance to test them? I hope your not planning some army invasion armed with exploding capacitors 😀

hehe

Yep, have four of them....three pro+ models and one "plain" model

Haven't got them running 🙁
Going to give them a go with a cpu I got recently which they SHOULD boot with if the work....a t-bird 1000
If that fails I positive these boards have gone to heaven

One thing I have found to be really important with these motherboards is the power supply. They draw power for the processor from the 5 volt rail, instead of the 12 volt that is seen in P4 and later Athlons. That's why none of these boards have the 4 pin ATX12V power header.

Here is a good example on why this can trip you up. I have a 400 watt power supply that has a 5 volt rail rated for 28 amps. It's not enough to run a KT333 based board with a 1.8ghz Athlon XP. It's currently using a 480 watt with a 38 amp rail. I can show you mountains of current power supplys that are rated for 500 watts and over, that only supply 20 amps on the 5v, simply cause that's the way the new spec is setup. A quick thought, and you'll realize how far your going to get with that.

prophase_j wrote:

One thing I have found to be really important with these motherboards is the power supply. They draw power for the processor from the 5 volt rail, instead of the 12 volt that is seen in P4 and later Athlons. That's why none of these boards have the 4 pin ATX12V power header.

Here is a good example on why this can trip you up. I have a 400 watt power supply that has a 5 volt rail rated for 28 amps. It's not enough to run a KT333 based board with a 1.8ghz Athlon XP. It's currently using a 480 watt with a 38 amp rail. I can show you mountains of current power supplys that are rated for 500 watts and over, that only supply 20 amps on the 5v, simply cause that's the way the new spec is setup. A quick thought, and you'll realize how far your going to get with that.

Have tested with two different psu's that came from working PC setups

I'm gonna check what amp's it pushes out from the 5v line 😀