swaaye wrote:

I assume you were on XP/2K because Win9x and Stability should not be in the same sentence. 😁

Oh come on. Its not a bug, its a feature 😉 I used to enjoy reinstalling Win98 every once in a while back then. Usually after cleaning up my room I would "clean up" the Win98 installation - quickly and efficiently 😁

swaaye wrote:

I assume you were on XP/2K because Win9x and Stability should not be in the same sentence. 😁

Yeah no way win98 would have taken that.

elfuego wrote:

...would you do another test for me? Go into the BIOS and loosen up the memory timings - set them as low as you can, but enable 4-way interleave. Then set the FSB to at least 150Mhz (or more, if your RAM allows it); then increase the multiplier to 15 (2250Mhz) or even more if CPU allows it, and then run the 3Dmark01 again.

Yes of course. Before I do this test, I think it would be a good idea to tell you about the hardware and BIOS options being used:

Mobo - Epox EP-8KTA3+Pro, PCB revision 1.1. Important note: some of the caps are domed. (Doomed?) Unfortunately, I'm confident that these (notoriously failing) caps would cause stability problems at high FSB speeds.

BIOS - Award V6.00PG 04/09/2003. Unlike the QDI mobo, this Epox mobo does not "fully understand" what an XP-M CPU is, because the BIOS POST screen reports it as an "AMD K7 Processor". (However, I prefer this kind of identication, as opposed to "Unknown CPU" which I have seen on other mobos.)

RAM - Infineon PC133 SDRAM, 1 stick, 256mb. Markings on the RAM stick say:

32Mx64 SDRAM
PC133-333-520
256mb sync 133mhz CL3 (unfortunately, it's not CL2)

BIOS memory options available are:

DRAM Timing by SPD = Enabled
DRAM Clock = 133 (No other options are available)
SDRAM Cycle length = 3 (2 is the only other option available)
Bank interleave = Disabled (2 bank, and also 4 bank are the other two options available)

DRAM PreChrg to Act CMD = 3T (2T is the only other option)
DRAM Act to PreChrg CMD = 6T (5T is the only other option)
DRAM Active to CMD = 3T (2T is the only other option)
DRAM Page-Mode = Enabled

Incidentally, I increased the FSB to 140 inside the BIOS yesterday, but I did not report the result on Vogons, because it failed to produce any result: 3DMark2001 failed to run correctly. (I did not alter any BIOS memory timings for this test.) I strongly suspect that this mobo will not operate correctly with an increase in FSB beyond about 140, because of failing capacitors, and also because the board is now about 7-8 years old. It's "on its last legs", so to speak.

BTW, Vogons user "5u3" stated in a Vogons thread that with his new(er) Epox board, he could achieve incredible FSB overclocking of around 170FSB, but as time went by, this impressive overclocking began to fail, and the FSB OC'ing could only reach about 150, before being reduced even further until the board died.

ADDiCT wrote:

My point of view (even when gaming): stability > performance.

I understand what you are saying. My nutty retro plan is to build a different machine for each major CPU release - just for a bit of fun! That means a 486, P1, P2, P3, T-Bird, XP, and XP64 machine. For each game I have, I'll try to select the most appropriate (or interesting) platform I have built. Two examples are:

Ancient "speed sensitive" DOS games: Either abandon them (the pre-VGA, pre CD-ROM ones) and use their potentially superior Amiga "counterparts" on a PC emulator, or use the DOSBox emulator on a fast modern PC, or use a "real" 486 PC (with BIOS L1/L2 caching options adjusted to reduce speed). Or why not all 3? Just out of interest/curiousity!

etc etc (P1, P2, P3, T-Bird) etc etc

For the XP-M machine with 1 ISA slot (KT133A), I would choose to play any (fast 3D) games that still use old (ISA) Soundblaster support. Eg: (3dfx) Redguard, some 3DRealms games, there must be some others...

If this KT133A machine crashes playing a "frivolous" game of Redguard etc, then I'll just shrug my shoulders, reboot the machine and start playing the game once more. No major problem!

IMO the KT133A was the best chipset of its time. I've built many systems with it, most of them running for years without problems.

I'm planning to build a KT133A system again, but currently the prices for working boards seem rather high, especially if you take into account that unmodified boards will fail because of rotting capacitors.

Anyone willing to sell me a capacitor-plague-ridden, but otherwise undamaged KT133A board?

I have run some more benchies. Initially I was having a few problems, so as a precautionary measure, I set the BIOS settings to "failsafe mode", including the "default" core voltage setting, (but Sandra says the core voltage = 1.66!)

Multiplier = 17x (real clock speed = 2.27Ghz)
3DMark2001 s.e. (default settings) = 11547
3DMark2001 s.e. (max settings including 6x AA) = 7162 (CPU Temp, with the assistance of coolon.exe is now: 36.6)

I think this "1.45 volt" CPU "likes" a high core voltage setting (of 1.66). I tried lowering the core voltage setting inside the BIOS set up area, and when I tried benchmarking using 3DMark2001, it failed every time. I reset the BIOS core voltage back to "default", and my 3DMark2001 benchmarking worked, even at 17x multiplier with maximum 3DMark2001 s.e. settings.

I return to the BIOS set up area, and adjust various "failsafe mode" values to "optimised" values - mainly BIOS fields associated with AGP. I then rerun the first test listed above.

Multiplier = 17x (real clock speed = 2.27Ghz)
3DMark2001 s.e. (default settings) = 11552. What's interesting about this score is that it is virtually identical to the score above (11547), which was done using BIOS "failsafe" values.

Also, I attempt once more to set the multiplier to 18x, but the PC goes mad almost instantly. (I seem to have a stable 17x system, but 18x is totally unstable.)

I return once more to the BIOS setup area. Because I am currently only using 1 stick of SDRAM, I don't think I can set the bank interleave value to "4 bank". Instead, I set this BIOS field to "2 bank" and reboot, then rerun the second 3DMark2001 "max settings" test again.

Multiplier = 17x (real clock speed = 2.27Ghz)
3DMark2001 s.e. (max settings including 6x AA) = 7164

OK, it's time to return to the BIOS again, and attempt to increase the FSB. Unfortunately, I can't increase it by much, because this Epox board really does have rotten caps. I will try: 137.

BTW, note to self: when you've set the multiplier using cpumsr, rebooting hangs the system. (Maybe the BIOS gets confused because the multiplier has been changed without it being fully "in sync" with cpumsr?)

Multiplier = 17x @ 137FSB = real clock speed = 2.33Ghz.
3DMark2001 s.e. (default settings) = Unfortunately, this test fails.

This is interesting - inside the BIOS set up area, I increase the core voltage by a small amount (+0.025). Sandra tells me the core voltage is now 1.68. I set the multiplier to 18x, and I can access the desktop OK! Sandra tells me that the CPU is running at 2.4Ghz. But 3DMark2001 fails and I have to reboot. So, perhaps the 18x multiplier requires lots of extra juice?

Related question: is it possible to cut the 24 pin power plug on the Enermax PSU with a sharp knife, in order to "push away" the 20 to 24 power pins section, which is not required on this mobo - and only prevents me from using this PSU, because a "coil thing" on the mobo is inconveniently located immediately next to the power socket? If I can get these unwanted 20 to 24 pins out of the way, I would be able to plug the Enermax (40Amp 5V rail) PSU in to the mobo using the first 20 pins only, and get "better" power in to the mobo.

Edit: Regarding the PSU "24 to 20 pins conversion", I found a converter on ebay -

http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewIt … em=290338678466

The only problem is that the photo doesn't make sense. The plug for the mobo (on the left) has 24 pins. It should only have 20.

I have made 3 changes to the hardware -

1) I have removed the Epox "caps of death" board, and replaced it with an Abit KT7A non-raid mobo, revision 1.3

2) Luckily, the Enermax "Athlon friendly" PSU will fit this KT7A mobo, so that is being used to power the mobo, and the lower powered Nexus 350W PSU now powers all of the components.

3) I have removed the 1 stick of Infineon 256mb SDRAM, and replaced it with 1 stick of 512mb "no name brand" SDRAM. Maybe this is a bad decision? I might change my mind about this..

I set the multiplier to 17x, and run 3DMark2001 s.e. at default settings. The test unexpectedly quits back to the desktop. I go to the BIOS set up area, and increase the core voltage by "two notches" to 1.625. I then rerun this 3DMark test. This test now works. I get a score of 11176.

Also, I try and set the multiplier to 18x, but the PC goes mad. I increase the core voltage again inside the BIOS set up area, this time to 1.65. I try and set the multiplier to 18x once again, and this time I get Sandra to tell me the CPU benchmark scores without the machine freezing or rebooting! Sandra says -

dry = 6594
wet = 3312

However, poor old Sandra goes mad again, after I ask her to tell me the multimedia scores. So, it's another trip to the BIOS set up area, and I increase the core voltage *again*, this time to 1.675. (This is getting a bit silly!) But wait, Sandra remains sane and provides the scores for both the CPU and multimedia tests. The multimedia scores are -

13141
14518

(Also, Sandra tells me that this "1.45 CPU" is running at 1.70 volts. Sandra also says that the CPU temperature is an amazing 20.something at the desktop. BTW, I am running coolon.exe.)

I then run 3DMark2001 s.e. with the default settings, and it instantly fails. Oh well. I can't really keep increasing the core voltage, otherwise something is going to break. This was fun though. I am impressed by the fact that the desktop is still stable. I try running a "simple" game such as JK:MotS (Jedi Knight - Mysteries of the Sith), and it fails. But I simply lower the multiplier by "1 notch" to 17x, and rerun it and it's fine! 😀

Edit:

I tried these BIOS settings -

FSB = 133+7=140 (Any higher, and I doubt whether the system would remain stable - this Abit board looks "old and tired".)
3 memory banks = "Fast"
Interleave = "2 bank"

Then I set the multiplier to 16.5x (Real speed was 2.31Ghz) 3DMark2001 s.e. (on default settings) got to the "8 golden dragon" test then failed.

retro games 100 wrote:

I have run some more benchies. Initially I was having a few problems, so as a precautionary measure, I set the BIOS settings to "failsafe mode", including the "default" core voltage setting, (but Sandra says the core voltage = 1.66!)

Not too big a deal. Non-mobile bartons usually spec out to 1.65v. The only difference with a mobile is that AMD tested them to be stable at lower voltages, it just means they came out of the "oven" a little better.

I've made one hardware change - I have removed the ATI Radeon 9800 Pro 128MB card, and replaced it with an nVidia 6800 GT 256MB card. (I use the 77.something driver.)

I set the multiplier to 16.5x (2.20 Ghz real clock speed), then set 3DMark2001 s.e. settings to maximum (including 4x AA), and get a score of 9891. I then set the multiplier to 17x (2.27 Ghz), then set 3DMark2001 s.e. settings to "default", and get a score of 12053.

Also, with the multiplier set at 17x, I run Jedi Knight: Jedi Academy, and set everything to maximum (but not detailed shaders), and the game plays well, and the system remains stable.

Conclusion: opening the L6 bridge on an XP-M CPU gives this curious retro platform a significant power boost.

Damn I found an interesting disadvantage to cutting/painting the bridges.

Depending on the BIOS maker it will have different reactions with XP-M and modified L6 bridge. Award BIOS basically ignores the L6, and AMI will boot to the setting of the L6. This is such a big deal for me since my KT333 board uses AMI. In what I believe to be a consequence of this, my board always boots at the chip's factory setting 14x, no matter what I change the BIOS to. When I set my board to 166FSB, it gives me a frequency of 2.3ghz which isn't stable, at least with the max 1.675 voltage selectable by me in the BIOS.

Edit: Damn, I should have bought an Asus or Abit instead, as they use award.

Hi guys!

I've got some hardtimes understand a few things. BadCaps guys here always recommend to replace the failing caps with better quality ones and of course, with the same attributes or above, such as more uF and voltage...I am to show the opposite in this thread and I'd like to ask for your clarification/approval.

Ya'll definitely know what the Abit VP6 looks like - 2 FC-PGAs, SDR, AGP...

So I have these 6 of them (I just love this one), including a one that has never been used and still has those Jackon caps (what a rare)...the other 5 have been already recapped - but differently ...

See this:

On this picture, I show the original one - the circled caps are the ones that have been replaced. I picked one of those 5 which behaves the best (and stable). And now we begin:

Red (for CPU & RAM & AGP): 1500uF 6.3v replaced by 2200uF 6.3v
Green (I guess for chipset): 1000uF 16v replaced by 2200uF 6.3v (!!!)
Purple (I guess for AGP & chipset): 1000uF 16V replaced by 1000uF 16v (better quality)

I am particularly concerned with the green couple (what is it for anyway?) and that's why I'm referring to the tradition of not replacing with lower attributes. I'm asking if this could be done and if it's okay...but partially, I have the answers on my hand right now:

1. It boots and runs

2. Since it is a board for PIIIs, I consider the PIII as the older CPU (correct me if wrong) which uses the 5v & 3.3v rails - there is no extra connector for 12v rail, which is used for the CPU (for example in P4 CPUs)

3. It behaves even more stable in OC'd status

I have my machine fully loaded (USB2.0, LAN, sound (PCI) and video card (AGP) inserted + HDD with DVD + some fans), 2x 1GHz @ 1245Mhz OC'd at 166MHz FSB, 41.5MHz PCI and 83MHz AGP. The SDR are Tonicom PC166 single-sided 128MB modules, so it's OK. I tried this config for about 1 hour playing Rainbow Six LV2 and no issues. The Vcore is at 1.85v and Vcc3.3 at 3.5v (but in real its 3.4v, if I raise it to the 3.6v, it's gonna be 3.5v - in Everest). Everest shows all basic rails OK but the 5v is quite lower than should be. When I boot into winxp and run everest, it shows 4.80-85v in idle, stress test lowers it to 4.61v - that doesn't seem right. The PS is Fortron Blue Storm II 500W. Any thoughts?

So what do you think? Is it ok in this board to leave it with those 6.3v instead of 16v? I suppose I'm right that PIIIs have no need to have 16v capacitors, but doesn't the board need those 16 volts? Please explain me this.

Just for fun: on another board, there are the 16v 1500uF replaced by better quality 16v 1500uF, but the system won't run stable unless I remove the LAN NIC and USB add-on card, though tested this only for 15 minutes, will try again. But the 6.3v 2200uF seem to be more stable - does this mean that higher capacitance equals to better stability?

Sorry not providing the brand of the caps, will post later.

BadCaps ment to be said that ''on their forum'' they are always recommending something, not here.

I know that this may be off topic, but it relates to the KT133a chipsets, but I have experienced these results based on the thorough search for answer and it turns out to be an electrical based issue, so I apologize for a huge amount of info regarding the capacitor stuff...😀 Anyways, KT133a rules!!! 😀

prophase_j wrote:

..I should have bought an Asus or Abit instead, as they use award.

I'm very sorry to hear about your mobo's BIOS, and its inability to "ignore" the modified L6 bridge on the XP-M CPU when it boots up. I assume that your mobo is too unstable (@ 166fsb / 14x) for you to quickly adjust the multiplier when (or if) the desktop appears?

What will you do now? Will you replace your current mobo?

retro games 100 wrote:

I'm very sorry to hear about your mobo's BIOS, and its inability to "ignore" the modified L6 bridge on the XP-M CPU when it boots up. I assume that your mobo is too unstable (@ 166fsb / 14x) for you to quickly adjust the multiplier when (or if) the desktop appears?

What will you do now? Will you replace your current mobo?

So with my current mobo (KT3 Ultra 2) and 2500+, if I want to use my chip at 166fsb, I would need to lower my max to 13x (paint L6[1]) or 13.5x (then cut L6[0]). But if I do that i'll be limited in speed if I go back to 133fsb. The other option is to set at at 17x (1 cut, 1 paint) or 16.5x (cut 2, paint 1). But what ever I set it to is going to be what it boots at, and the max setting at the same time.

prophase_j wrote:

retro games 100 wrote:

I'm very sorry to hear about your mobo's BIOS, and its inability to "ignore" the modified L6 bridge on the XP-M CPU when it boots up. I assume that your mobo is too unstable (@ 166fsb / 14x) for you to quickly adjust the multiplier when (or if) the desktop appears?

What will you do now? Will you replace your current mobo?

So with my current mobo (KT3 Ultra 2) and 2500+, if I want to use my chip at 166fsb, I would need to lower my max to 13x (paint L6[1]) or 13.5x (then cut L6[0]). But if I do that i'll be limited in speed if I go back to 133fsb. The other option is to set at at 17x (1 cut, 1 paint) or 16.5x (cut 2, paint 1). But what ever I set it to is going to be what it boots at, and the max setting at the same time.

But isnt L6 only defining the "max" multiplier, and only w/ -M CPUs? It shouldnt define the default values. That information should be stored elsewhere.

But even if AMI bios is made in such a way as you describe, it should still load up 133FSB by default - which shouldnt be a problem. So after the initial startup - go to the BIOS, lower the multiplier by hand and increase the FSB. No need to paint anything - cutting should suffice 😉

5u3 wrote:

Anyone willing to sell me a capacitor-plague-ridden, but otherwise undamaged KT133A board?

Well I have a "quasi" 1.3 version Abit KT7a non-raid. I say quasi because I got it of the ebay as 1.3V and the installed BIOS was indeed one of the version 1.3 (B4), but markings on the board itself (bar-code) says V1.0. I've used this board for a couple of months until I got the "real" 1.3V RAID. I havent tried the mobile CPU on this one; I've only tested a normal tbred-b 2400+ which worked at 20x100 Mhz. I could not get it to run at 133 Mhz, as the only confirmed settings with 133Mhz and high multiplier on abit kt7a is 9x(17x) which was too much for this CPU (it wakes up, but hangs the system afterwards).

If you are willing to pay 5e + postage fees (from Germany) and wait until end of holidays (15th of January), I can send it to you. For 1e more I will include the 2400+ 😀

elfuego wrote:

prophase_j wrote:

So with my current mobo (KT3 Ultra 2) and 2500+, if I want to use my chip at 166fsb, I would need to lower my max to 13x (paint L6[1]) or 13.5x (then cut L6[0]). But if I do that i'll be limited in speed if I go back to 133fsb. The other option is to set at at 17x (1 cut, 1 paint) or 16.5x (cut 2, paint 1). But what ever I set it to is going to be what it boots at, and the max setting at the same time.

But isnt L6 only defining the "max" multiplier, and only w/ -M CPUs? It shouldnt define the default values. That information should be stored elsewhere.

But even if AMI bios is made in such a way as you describe, it should still load up 133FSB by default - which shouldnt be a problem. So after the initial startup - go to the BIOS, lower the multiplier by hand and increase the FSB. No need to paint anything - cutting should suffice 😉

I know man.... it's stupid. But yeah even with my mobile chip on these AMI strapped boards I can't change the multi that way. Always going to start at L6. I couldn't figure out why until a couple day ago when I was reading the
FAB51 site, where they brought this up.

1[ Type A ]The products are started by L3 multiplier, even when Mobile CPU is used.
2
3At the Looked Athlon XP, it always starts in a rated multiplier also by the case of Mobile operation.
4In these mother boards, it always starts by L3 multiplier or the fixed starting multiplier.
5All are solved by closing L5 [1] and L5 [2].
6According to our test, the BIOS vender applicable to it is Award.
7
8[ Type B ]The products are started by L6 multiplier , when Mobile CPU is used.
9
10Also by the case of the Loked Athlon, it can start by L6 by making it operate by Mobile.
11At these motherboards, when L5 [2] is closed, it starts in L6 multiplier. Since it is going to start by 24X if you close L5 [1], cautions are required of these products. With the following products, change of the maximum multiplier is based on processing of the L6 bridge. And we recommend you for L5 [1] to leave as it is.
12According to our test, the BIOS vender applicable to it is AMI.

On another note, I had KT7a from a few months back that I pulled out yesterday. When I first got it, it wouldn't power on for more than a second, this was when I was using a dinky 200 something watt power supply. Anyway, connected up with the Antec True 480 and 2500+ as soon as I pressed the power button

POP!!!

Actually 2 right in a row. I wasn't looking at the board but I did see two poofs of dust/smoke come out the side with the cover off. I pulled the plug so quickly nothing much else happend... Kinda fun really.

prophase_j wrote:

On another note, I had KT7a from a few months back that I pulled out yesterday. When I first got it, it wouldn't power on for more than a second, this was when I was using a dinky 200 something watt power supply.

I've noticed something interesting about Abit KT7A boards, especially older revisions. If you have either removed (then replaced) the button battery, and/or cleared the CMOS jumper, you must connect a really fast fan (minimum 4000 rpm) on to the CPU #1 header. If you don't, you will see that behaviour - ie, the mobo just won't start. Once you satisfy the "cpu temperature guard thing" which is enabled inside the BIOS with a fast fan, you can then get access to the BIOS set up area, and you can disable that particular feature, and then you can plug any slower fan, or no fan at all, on to the CPU fan header #1.

prophase_j wrote:

Anyway, connected up with the Antec True 480 and 2500+ as soon as I pressed the power button

POP!!!

Actually 2 right in a row. I wasn't looking at the board but I did see two poofs of dust/smoke come out the side with the cover off. I pulled the plug so quickly nothing much else happend... Kinda fun really.

Christmas day sounds fun at chez prophase_j's! 😉 Now that the dust has settled 😉 , do you know what happened? I wonder if you fried the CPU, because the mobo was anticipating a previously installed CPU?