When we play with Athlons and Via KT chipsets, we are really putting the stress on the 5V and 3.3V rails. The 12V rail is Intels domain in that time. What are the AGP cards of the time feeding on?
I guess I need to aquire a really decent and quiet PSU for my Athlon/KT Retro. I am currently using an old 350W PSU by Q-tec. That name doesn't sound too good. My alternatives aren't any better. I had access to a few Fortron/Source PSUs which seem rather heavy and sturdy. Unfortunately, these PSUs all hum, and their fan instantly responds to changing demand: Just by opening an explorer window or a context menu, the fan will change its pitch audibly. Unfortunately, that is a no-go for me, no matter how nice the rest looks like.

Any suggestions for the KT league?

ux-3 wrote:

When we play with Athlons and Via KT chipsets, we are really putting the stress on the 5V and 3.3V rails. The 12V rail is Intels domain in that time. What are the AGP cards of the time feeding on?
I guess I need to aquire a really decent and quiet PSU for my Athlon/KT Retro. I am currently using an old 350W PSU by Q-tec. That name doesn't sound too good. My alternatives aren't any better. I had access to a few Fortron/Source PSUs which seem rather heavy and sturdy. Unfortunately, these PSUs all hum, and their fan instantly responds to changing demand: Just by opening an explorer window or a context menu, the fan will change its pitch audibly. Unfortunately, that is a no-go for me, no matter how nice the rest looks like.

Any suggestions for the KT league?

Using a qtec 500W PSU with my 8kta3+pro rig...very pleased with that PSu...super silent fan...35A 5v line

I am testing two Epox EP-8KTA3 boards. They are not "+" or Pro boards. I've noticed a couple of things about them -

Both have low PCB revision numbers of 0.3
Both will not post with a t-bred 2400+ rated CPU.
Both have difficulty getting to the desktop with a t-bred 2000+ rated CPU.
Both seem OK with a t-bred set to 1900+ rated speed in the BIOS.
Both seem OK with an XP-M set to 1900+. If you set it to 2000+, you don't get to the desktop.

Also, cpumsr won't work. This is interesting. I remember prophase_j reporting that he had problems with this utility, in conjunction with using a "plain vanilla" Epox board - not a "+" or Pro board. cpusmr freezes, no matter what multiplier you pick.

Also, I tried unlocking my 2400+ XP-M, by pushing a blade in to L6 (3). I had 2 attempts at it, and probably failed because cpumsr still reports that the max multi is 13.5, and not 18.

retro games 100 wrote:

I am testing two Epox EP-8KTA3 boards. They are not "+" or Pro boards. I've noticed a couple of things about them -

Both have low PCB revision numbers of 0.3

That's the same revision as my board. I was never able to change the multipler, and when I tried enabling stop grants (cpu cooling) it would always crash after a while.

My old PSU has turned up after some searching. It's an Enermax EG465AX-VE (+5V: 44 Amps), equipped with a popular hardware mod: Resistor R 82 is replaced with a 10KΩ potentiometer, allowing to adjust the output of the +5V rail.

The Radeon 9800 Pro still won't run. Not even when I power it from a second PSU. No problems with other cards (GeForce 4 Ti 4600, Radeon 9600XT, Voodoo5). I suspect gremlins. 🙄

Shodan486 wrote:

Just cool every little tiny chip which may overheat at some extreme conditions, like the ICS chip (think it's printed on it) near the DIMM slots to the right.

I just put a finger on this chip to check how hot it gets (on my board it's a PhaseLink PLL205-13XC), and wish I hadn't done that - Ouch! 😳
Now searching for a suitable heatsink...

Shodan486 wrote:

Just cool every little tiny chip which may overheat at some extreme conditions, like the ICS chip (think it's printed on it) near the DIMM slots to the right.

I have heard of that mod too. The ICS chip is actually the clock generator, and high temperatures will cause to drift and otherwise malfunction.

5u3 wrote:

The Radeon 9800 Pro still won't run. Not even when I power it from a second PSU. No problems with other cards (GeForce 4 Ti 4600, Radeon 9600XT, Voodoo5). I suspect gremlins. 🙄

🙁

BTW, what is the mobo's PCB revision number? (Top left corner). Also, what is the BIOS POST datestamp? (Is it April 2003? EDIT: 04/09/2003?)

Regarding KT133A + AGP + "gremlins" - I cannot get a 3DPower nVidia Geforce3 Ti500 to work in a QDI Kinetiz 7E-A mobo. I reach the Win98 desktop OK, but the mobo is unstable. I tried various "archived" nVidia driver versions, but none are stable. I suspect some odd compatibility problem with the mobo and the Ti500 card. The mobo functions well with other AGP graphics cards (and the caps look OK too).

Quick benchie -

Mobo: EP-8KTA3+PRO, PCB 1.1, with at least two "bad looking" caps

CPU: XP-M 2400+, running at 13.5 multi, 133FSB = 1800Mhz - I'm using CPUMSR.exe to set these settings. Unfortunately, I have failed to unlock the L6 (3) bridge. My two attempts at pushing a knife in to it has not disconnected this bridge. (But the CPU is not damaged, so that's OK.)

Gfx: nVidia 3DPower GeForce3 Ti500, not OC'd.
Driver: nVidia 30.xx (I think 45.xx would be faster.)

Edit - added: 3DMark 99 Max: 12251, 28730
3DMark 2000: 10863 3DMarks

I finished my testing, and decided to uninstall a few things. Reboot time. That's funny, the screen has gone blank. Switch off power. Power on again. The onboard LED has gone a bit mad. It reads 88. 88 looks like all the LED's "light bars" are illuminated. I remove the Ti500, and replace it with an *identical* Ti500. That works. I try the first "dead" Ti500 again. No joy: the LED says 88 again. I give the card's contacts a clean. I also pull out the noisy onboard fan cable. I retry the card. It works! It's amazing how many cards you can "bring back from the grave" by cleaning them.

5u3, have you tried cleaning the contacts on your Radeon 9800? And also given the AGP slot a blast of "canned air"?

My 8KTA3+ Pro is PCB rev. 1.0, maybe that prevents the Radeon 9800 Pro from working. I also tried cleaning the contacts, raising AGP voltage, selecting different AGP speeds, and disabling AGP fast writes.

The BIOS date was 02-05-2002. I upgraded to the 04-09-2003 version, but alas this does not help with the Radeon card.
The newer BIOS identifies my TBred CPU, but the drawback is that I can not select FSB frequencies below 133 MHz any more.

Ah well, these are just minor issues. I prefer this rather than Intel BX any time 😉

5u3 wrote:

My 8KTA3+ Pro is PCB rev. 1.0, maybe that prevents the Radeon 9800 Pro from working. I also tried cleaning the contacts, raising […]
Show full quote

My 8KTA3+ Pro is PCB rev. 1.0, maybe that prevents the Radeon 9800 Pro from working. I also tried cleaning the contacts, raising AGP voltage, selecting different AGP speeds, and disabling AGP fast writes.

The BIOS date was 02-05-2002. I upgraded to the 04-09-2003 version, but alas this does not help with the Radeon card.
The newer BIOS identifies my TBred CPU, but the drawback is that I can not select FSB frequencies below 133 MHz any more.

Ah well, these are just minor issues. I prefer this rather than Intel BX any time 😉

I can confirm that my Radeon 9800XT works in my 8kta3+pro rev 1.0

5u3 wrote:

My 8KTA3+ Pro is PCB rev. 1.0, maybe that prevents the Radeon 9800 Pro from working. I also tried cleaning the contacts, raising AGP voltage, selecting different AGP speeds, and disabling AGP fast writes.

The BIOS date was 02-05-2002. I upgraded to the 04-09-2003 version, but alas this does not help with the Radeon card.
The newer BIOS identifies my TBred CPU, but the drawback is that I can not select FSB frequencies below 133 MHz any more.

Regarding the Radeon not working - have you tried a different CPU? Anything will do, eg Duron or a T-Bird. I remember I couldn't get a Radeon 9800 Pro working with a T-Bred CPU. It was 2400+ rated.

Regarding the FSB issue: does this only happen with your current T-Bred CPU? I think that if you cleared the CMOS, then tried a different type of AMD CPU, you will regain access to 100 FSB. I'm using the same BIOS dated version as you, and more often than not if I install a 133FSB T-Bird, the initial BIOS POST screen states that the system is running at 100FSB.

Thanks for the input guys! 😀

I already tried the 9800 Pro with the Thunderbird CPU. Now I've also tried it in an Abit KT7A, and neither does it work there. So I reckon the card is defective in some weird way.

The FSB "issue" looks rather straightforward to me. The BIOS seems to determine the possible FSB speeds from the CPU model it detects.
With the Thunderbird installed, I always get the full range of FSB speeds (100-180 MHz). With the TBred, it depends on the BIOS version: The older BIOS lists the TBred as "Unknow" [sic] and gives me all the FSB speeds as well. The latest BIOS identifies the CPU as Athlon XP, and only then the lowest selectable FSB is 133 MHz.

5u3 wrote:

[...] I reckon the card [Radeon 9800 Pro] is defective in some weird way.

A-ha! You could look at this situation really positively. Radeon 9800 cards are popping up all the time on ebay. They aren't cheap, but they never go for "Voodoo5 amounts", for instance. OTOH, your specific board (with the good caps + OC'd FSB tests) seems really good, and it's unlikely there's anything wrong with its AGP port. 😀

Recently, I bought some "electrically conductive silver paint" from ebay. Today, I just tried to use it. I shook the bottle, and the silver "particiles" mixed in with the other clear liquid in the bottle to produce a silver looking liquid. Then, I dipped a needle in to the bottle, then put this needle on to the laser hole on the L5 bridge on the XP-M CPU. I tried this 4 times, to attempt to fill the laser pit with silver paint. Unfortunately, hardly any paint was adhering to the needle. I reinstalled the XP-M CPU in to the Epox mobo, but the CPUMSR program only "saw" the multi as a maximum value of 13.5x, and not 24x.

I guess what I need to do is get more paint out of the bottle, on to the L5 bridge laser pit. This laser pit is very small. A needle is a perfect size for it, but it is not picking up any of the liquid in the bottle, when I dip the needle in to the bottle.

Idea: I think I need to temporarily cover the area surrounding the laser pit, then perhaps just brush on the paint? Any ideas as to what I can cover the area surrounding the laser pit with please? And also, any ideas as to what kind of "applicator tool"/brush I could use, to brush the paint on with?

Thanks a lot for any ideas.

Edit: I've got a thin (as in not wide) roll of decorator's masking tape. I've also got a very fine art brush. Both items are probably a bit daft to use, but it's what I've got at the moment. I might try this tomorrow..

I thought that you're supposed to fill the pit with a non-conductive material and then draw a conductive line over it.

prophase_j wrote:

I thought that you're supposed to fill the pit with a non-conductive material and then draw a conductive line over it.

Regarding the use of glue - it says on this website, that glue is required:

http://technobabble.com.au/technobabble/html/ … XP_Palomina.htm

However, if you look at that first photo, there are big laser marks between the "dots". On the other hand, if you look at an XP-M CPU, there are no big laser marks between the "dots". Instead, there is a much smaller, neater looking tiny pit mark. It's clearly in the middle of the "dots", with space either side. Just big enough for a needle, for instance. Also, the "dots" on the XP-M are smaller than the size of a needle head!

On this website:

http://fab51.com/cpu/barton/athlon-e24.html

It says: How to set the bridge of [Open] to [Close]
We know the Laser pit should be filled with an insulator before anything else. Then, it is made to [Close] by using electric conductive material.
In the new package(#27648), it might be difficult to connect two points in the bridge. In this case, it can be solved by filling the laser pit directly with the electric conductive material.

The number written on the corner of the XP-M I have is 27493 (not 27648), but I'm wondering if the 2nd line above is applicable to this XP-M CPU. I wonder if I just need to fill the pit with conductive material.

I'm going to try cutting the L6 bridge once again on this CPU - my 3rd attempt! If that fails, I will try brushing some paint in the L5 laser pit. If that fails, I think I'll leave it!

Solved. The other day, I bought a new magnifying glass from ebay. It has 2.5x and 5x magnification, and using the 5x magnification "option", I was finally able to see what I was doing. (Previously, I was using a magnifying glass that came from a christmas cracker!) My 3rd attempt at cutting the L6 bridge has succeeded. I managed to make a decent cut using a "box cutter"/Stanley knife. Now, CPUMSR says that the multi is 18x, and not 13.5x. I'm happy with that. 😀

I wasn't aware that in one of the packages it would actually work. That's awesome that you were finally able to bump up that multiplier. Kudos buddy!!

😀
Time for some benchies, and some of the results are exciting!

Mobo: Epox "caps of death" board, EP8KTA3+Pro. 2 caps near the SDRAM slots look a bit bulged, but all others look OK.
CPU: mobile barton XP-M "2400 rated" (1800mhz real clock speed, requires 1.45v)
Ram: 1 stick of 256MB. Nothing special.
Video: Voodoo3 AGP. (Light weight card.)
Benchmarking: 3DMark 99 Max, with V-Sync set to Off. (The Voodoo3 is unable to do one of these tests - the "bump mapping" test.)
Power: 2 PSUs used, just to be sure!
Software: wpcredit used to alter the "55 bit 2" field, coolon.exe used for cooling, and cpumsr.exe used to alter the CPU's multiplier value.

6x, 133fsb = 800mhz -> 6541, 16772 (6x is the CPU's start up multi on this Epox mobo)
13.5x, 133fsb = 1800mhz -> 6904, 29284 (13.5x is the CPU's maximum official multi)

14x, 133fsb = 1867mhz -> 6906, 29867
15x, 133fsb = 2000mhz -> 6926, 30866
16x, 133fsb = 2133mhz -> failed. Bios change: -0.1v to -0.075v. 1.50v to 1.54v
16x, 133fsb = 2133mhz -> 6911, 31999 (retry)
16.5x, 133fsb = 2200mhz -> failed. Bios change: -0.075v to -0.050. 1.54v to 1.56v
16.5x, 133fsb = 2200mhz -> failed. (retry) Bios change: AGP = +0.1v.
16.5x, 133fsb = 2200mhz -> failed. (retry) Bios change: -0.050v to -0.025. 1.56v to can't remember.
16.5x, 133fsb = 2200mhz -> 6901, 32186 (retry)
17x, 133fsb = 2267mhz -> Immediate failure using CPUMSR. Bios change: -0.025 to “Default”. CPU voltage is now 1.61v
17x, 133fsb = 2267mhz -> failed. (retry)

Voltages unaltered from testing above; so now the vcore = "Default" (1.61v), and AGP = +0.1v.
Now the FSB will be increased...(from 140, to 169!)

14x, 140fsb = 1961mhz -> 6901, 31340
15x, 140fsb = 2101mhz -> 6901, 32338

14x, 144fsb = 2017mhz -> 6911, 32500
15x, 144fsb = 2161mhz -> 6901, 33258

14x, 148fsb = 2077mhz -> 6904, 33304
15x, 148fsb = 2225mhz -> failed.

14x, 152fsb = 2130mhz -> 6899, 33831

14x, 155fsb = 2171mhz -> 6907, 34690

14x, 158fsb = 2214mhz -> failed.

13.5x, 158fsb = 2135mhz -> 6916, 34730

13.5x, 161fsb = 2175mhz -> 6892, 35291

13.5x, 164fsb = 2217mhz -> 6900, 35876

13.5x, 167fsb = 2259mhz -> failed.

At this point, I run Sandra 2002 Pro. CPU = 6218, 3137. Multi media = 12413, 13771. Memory bandwidth = 1283, 1270

13x, 167fsb = 2175mhz -> 6895, 36163

6x (mobo's boot up multi), 170fsb = mhz -> failed - no BIOS POST!

13x, 169fsb = 2202mhz -> 6912, 36476 (Best score)
Sandra 2002 Pro: Memory bandwidth = 1298, 1283 😎

13.5x, 169fsb = 2286mhz -> failed.

Conclusion - it seems that if you can successfully increase the mobo's FSB, then you don't need to push the CPU's multi too high. The ironic thing is, the best scores above were obtained with the CPU's multi set to below its official multi max value!

A note to 5u3: I understand your "old new" Epox mobo could do 170fsb. I just couldn't reach that value! I could get to 169fsb!

Also, I've tested quite a few of these Epox boards. None of them could manage more than about 135-140 fsb. (One of them can't even do 133fsb!) This one however, seems different. I guess the caps on it have got a bit of life left in them. Lucky find, perhaps?