Announcement

**Kruzin** · 24 July 1999, 13:48

From your desciption, it has to be something in your Q2 autoexec.cfg, or some script somewhere (if you use them). If you replaced the config.cfg, and it worked, once in the game, that config.cfg is re-written with the values from your autoexec. It's just a matter of finding the offending line in that startup cfg.....

**paulcs** · 24 July 1999, 14:45

I'm working from a clean install of the game. The problem emerged without an autoexec.cfg file present.

And it happens with Kingpin as well.

Something is happening during or when I leave the game.

It's all very mysterious. I guess I should make a before and after comparison of the config file, line by line, to see if any modifications are being made.

Thanks again.

**Ashley** · 25 July 1999, 22:55

Come on - enough with this "PowerStrip ate my computer" nonsense. The PowerStrip does not and cannot modify anything in your BIOS; any changes you make with it (it makes none of its own accord) are temporary and cannot survive a POST. Use a bit of common sense: if the problems you're having only occur in Quake2, check your Q2 configuration...

**Jihad** · 26 July 1999, 06:07

Hey dammit! I lost an eye to Powerstrip so kiddies remember never ever use Powerstrip without eye protection!

------------------
Celeron 366->550, 128 MB RAM, 10.1 IBM Deskstar, MX300, G400 vanilla, HP 8100i CD-RW

**Kruzin** · 26 July 1999, 06:26

Ashley - Believe what you want.
I can't specifically prove that PS affected my BIOS, but something did, and it just happened to be on the same day that I tested the very first beta version of PS that had G400 support. My clocking dividers got screwed up after trying to o/c the card too far, and I had to use my emergency bios recovery disk to get things back to normal. I still have the e-mails of that frustrating day when I reported the problem to the Matrox beta list. I did nothing else that day that could have affected my card. Others have reported similar trouble in the past.

There have been at least 3 updates to PS since then, so if it was a problem, it could very well be fixed be now. Either way, PS sux IMO, and I won't bother trying it again. There are other ways to o/c, without all the garbage PS throws in your way.

Since I know Haig sent Paulcs a bios file to reset his card already, and judging by the problem specific to this thread, I do think it's a problem with the way the Q2 is set up. This is reinforced by the success with Q3. Have you got any other OGL games you can try, like Half-Life?

**Ashley** · 27 July 1999, 00:52

Kruzin - Its not a question of me "believing what I want". I wrote the PowerStrip and I'm telling you that it does not, and cannot, modify your BIOS, period. If you don't want to take my word for it, that's fine: ask Greg or Liew if it is even theoretically possible for any software to modify your video BIOS while running under Windows.

I already explained what the PowerStrip does with the G400 clocks and divisors, when it does this, and why it does this, right here in this forum. It was never a secret, and if Liew or Greg ever release simplified, GUI versions of their utilities, they're going to impose constraints and make these kinds of decisions as well. If I had received a single message from you, Paul or anyone else asking for assistance or clarification or help on this score, you would have gotten it.

As for your repeated "I hate the PS - it sucks" comments here and elsewhere, what can I say? Its not for everyone, and I can understand why it wouldn't be to the liking of someone who was only interested in overclocking, who is only familiar with Windows 95, and who only uses one kind of graphics card. If you look at the FAQ on my site you'll see I recommend free, stand-alone utilities like Greg's and Liew's, if they are available.

Finally, just so you know - the latest PowerStrip update isn't about the G400 at all. It fixes the poke-in-the-eye problem Jihad reported...

**Ashley** · 27 July 1999, 06:15

Come on now Hellmut, please... This isn't the X-files...

Reprogram a BIOS with just "one or two typos"? As in accidentally hitting one key when you meant to hit a different one altogether?

If it really only takes a couple of accidental typos, I don't suppose you'd care to provide a little example with some premeditated, deliberate code?

And note: in one sentence you write "reprogram" and in the next you write "nil", so remember now: we want a modified, but *working* BIOS when you're done...

**Kruzin** · 27 July 1999, 06:49

Now, I'm not saying anything about your programming ability. You obviously know more than I about that. I do understand what PS does when it's o'cing the G400. It's writing to the registers of the G400 chip, and adjusting the PLL clock speeds. It's pretty basic in that respect. But let me outline what happened to me when I used the first beta PS that had basic G400 support.

I ran some baseline benchmarks with Q2 and 3DMark99Max on my G400, un-overclocked. Then I installed PS. I went to the o/c page, and moved the slider up 5MHz at a time, searching for the cards limit. Once I went past a certain point (I think is was about 180Mclk), I saw the lower slider (Gclk) drop instead of climb.

At that point, I am assuming the dividers changed. If those dividers changed because of the bios compensating, or because PS changed them, I don't know. After that, I lowered the Mclk, but the dividers did not change back.

I exited PS, rebooted the PC, and went back to PS. It appeared to show the correct clock rates. I clocked it back up (but not past the point that would make the dividers change), and ran some more benchmarks. They had all gone down about 5%. I exited PS, rebooted again, and ran the same benchmarks, un-o/c'd. Same results...5% lower than my originals.

So I ran G4set to see what it reported my clock rates as. To my dismay, they where all buggered up. In my haste, I did not write down what it was reporting, but they where wrong. I quickly popped in my BIOS recovery disk, flashed my BIOS with it, and all was back to normal. I un-installed PS, and have not tried it again since.

Now, is it not possible, that in that first beta, at some clock settings, PS could have sent something to the wrong register, or fed the card some bad info, and caused this anomolly? I think it's at least possible, and given what happened to me and paulcs, I suspect so even more.

I have loosley followed the PS threads, and beleive that you changed something in there with regards to the way it handles dividers (particularly on the MAX). Perhaps you could explain to us how you are doing this now, and how it compares to the initial G400 release.

On a side note, I did try Performance Tuner, but after o/c too far with it, I can not get it back to default settings, and whenever I start it up, no matter what it thinks the clock settings are, I get immidiate corruption all over my desktop. Perhaps uninstalling/reinstalling would fix this, but I haven't had the inclination to do so.

G200clk is doing the job just fine o/cing for the time being, and Powerdesk handles most of the stuff PS does (although I could see PS being useful to other cards with lesser desktop tools)...

**Scytale** · 27 July 1999, 07:46

I see PowerStrip going where PowerDesk NT as well as 'the' other utilities shine by absence. [NT PowerDesk is quite bare, try it some time.]
Until so far, PS never ate my Matrox, TNT and 3dfx cards, nor any other system component. Maybe it's because I'm not beta testing Matrox drivers

**Ashley** · 27 July 1999, 11:34

Two different questions you're asking. The first: Is it possible the PS changed something in the BIOS? No - I don't want to appear too defensive on this score, but the answer is an unqualified No.

The second question is: Did the first PS change the clock divisors. Here the answer is a qualified Yes, but only above a certain speed. As someone (maybe you?) aptly put it in this forum: the PS underclocked the core in order to overclock the memory.

The thing is, that wasn't a bug or an error on my part. It was quite deliberate.

The G400 defaults are a mclk divisor of 3:2 and a gclk divisor of 2:1. The G400 runs at a sclk of 252MHz; the mclk is 168 (252/3*2) and the gclk is 126 (252/2*1). Agreed?

(The MAX apparently runs at an sclk of 300MHz; the mclk is 200 (300/3*2) and the gclk is 150 (300/2*1) - but the MAX wasn't available then, and as far as I can see, it still isn't available now.)

Now let me ask a question: are there any limits you're aware of on the "sclk"? Liew doesn't think so: you can feed his utility a sclk of 1 or 5000, and G200clk will try to program it. Greg, on the other hand, does believe in constraints: his G4Set will reject anything below 33 and above 400, as I recall. But why 400? Why not 300 or 500?

In the first PS with G400 support, I used the figure of 305 - that's the number Matrox gives in the G400 docs - minus 5% for a margin of safety: that's gives a maximum sclk of 289 (305-5%).

The thing is, at an mclk divisor of 3:2, an sclk of 289 only gives a maximum mclk of 193MHz (289/3*2). So to go above that, you need to change the mclk divisor to 1:1. And if you do that, you don't want to leave the gclk at 2:1 because 194/2*1 only equals 97 - you want to change it to 3:2 because 194/3*2 at least gives 129.

And there you have it: the gclk has been deliberately reduced in order to increase the mclk. (Worse: once it changed your clock divisors, the first PS wouldn't change them back - you had to reboot.)

Its not necessary to do all that? Too conservative? Too low a ceiling? Sure - I'm willing to accept all of that, and have in fact: the latest PS uses a ceiling of 360; it doesn't change the divisors under any circumstances, and it doesn't auto-adjust the duty cycle.

But consider this: people run 5ns. memory at 225MHz w/o a hitch on other cards; no one seems to be able to do that with a G400, using any of the four utilities available. How come?

**Kruzin** · 27 July 1999, 11:52

Well, I can understand your desire to believe your util had nothing to do with my bios fiasco. But I can't think of any other explanation. The fact that the first release did attempt to change clock dividers at a certain point leaves a lingering doubt in my mind. No matter. It's all straight now. And I appreciate your level headed replies.

I am not aware of a limit that Mclk can be set at. Both Liew and Greg know their stuff, but on specifics like this, I usually go with Greg's opinion (he has a good line on reliable info). I'll ask Haig tomorrow if this is info I can get and pass along.

As far as Ppl not getting up to 225 with their G400s, that would be because the regular G400s don't use 5ns memory, they use 6ns. Only the MAX uses 5ns. We are all still waiting for some brave soul who has gotten a MAX to report on what it can do...

**SwAmPlAdY** · 27 July 1999, 13:54

Howdy,

here's some food for thought. With PS, I have to apply it several times before the O/Ced setting kicks in. Even then, I can't get close to how high I can go if I just O/C the G400 directly within the bios.

A few times, PS has messed up my system. I had to flash it back to the original bios settings.

I compared the bios pins between the regular bios settings and the bios settings after I used PS. I don't see any changes in the bios.

Ashley, it's quite possible that PS is screwing something up in the chip's registers or your settings are wrong since you are under the impression that the G400 is running at 5ns rather than 6ns (~183Mhz).

Does this have anything to do with the 2nd vga using a different ramdac?

SwAmPy

[This message has been edited by SwAmPlAdY (edited 07-28-99).]

**Hellmut** · 27 July 1999, 16:12

Hi Ashley,

you absolutely CAN reprogram the MGA bios under 9x!
One or two typos in your source would be enough to nil the bios.

Care,
Hellmut

**Kruzin** · 27 July 1999, 17:02

gasp
SwAmPy!
You o/c the G400 via the BIOS?!?

Announcement

G400: Odd OpenGL Problems Continue

G400: Odd OpenGL Problems Continue

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment