G400: Odd OpenGL Problems Continue

[Ashley]

I'm beginning to feel like I stepped into quicksand here...

Matt - I'm not questioning anyone's "experience", just your intepretation of what happened and your conclusions. I do insist the conclusion Kruzin related from Haig was too vague to be helpful - period. I can't replicate a problem based on what you guys are saying. Only some people are having problems with the PS, not everyone. And I know very well that the normal G400 has 6ns memory, and the MAX has 5ns memory.

Kruzin and Haig - Now what you're suggesting is that I'm doing something really malicious: dumping the existing BIOS to disk, modifying it, and then reflashing it, right? There can be nothing accidental about that process, as you both know very well. The chances of writing the correct BIOS checksum, after modifying even a single byte in the BIOS, are just too remote to make it anything but deliberate, and hence malicious. That's a pretty strong accusation: I don't care who Haig is; I'm not walking away from that kind of implied accusation...

Haig - How helpful would it be to you if I reported "The G400 screwed my system". How would Matrox Tech Support go about responding to, and fixing that "bug"? As for the PS doing nothing at all as far as your HAL tools are concerned, what can I say? I use a digital frequency counter attached to the CLKIN pin - I do this with lots of chips, and its never failed me as a method of independent verification. (Of course, this only works for the actual memory clock, not the sclk or gclk, but since these are related...)

Everyone - If you have a monitor with an OSD and you want to check the accuracy of the PLL programming done by the PS, go to the Screen adjustment page and select a custom refresh rate. Then compare what the PS does with what your monitor OSD reports - programming the pclk is very similar to programming the sclk, but its even more complex.

Kruzin and SwampLady - Difference of opinion only: Greg's PINs stuff is very clever, but I don't agree about permanent overclocking. A lot of work in PS 3.0 (the Performance Tuner thing) is in fact premised on the idea that o/c'ing is worthwhile only when it makes a difference - as it clearly does not while I'm writing this message...

SwampLady (and Agallag) - I wasn't jumping all over you because of a typo. A while back someone wrote in claiming all kinds of problems using a STB Voodoo2 card. I wasted a lot of time trying to find and borrow a BlackMagic card and see what was up, only to find out they had a Voodoo3. So to your real question: I think not. Registers are supposed to be reset to their defaults by the video BIOS on power-up. Register contents are lost when you shut down or reboot. And there is nothing remotely perfect about any software I write. You're trying to be helpful. I appreciate it. (Are we on good terms again?)

Since I'm not receiving details from you guys (not a single bug report so far, with quite a few G400 users - mostly in Asia), let me try to put this back on track with a few details on my side. Maybe someone will spot something.

1. I like Matrox cards in general, and the G400 in particular, but I have no relationship to Matrox whatsoever. I buy my boards thru the retail channel just like the average user.

2. I only have one G400 card. Its a retail, boxed, single-head 16MB card. It has a Rev-02 chip - most of you, I think, have Rev-03 chips.

3. I got my card in HK in June - iXBT suggested these were not even genuine Matrox cards, but Matrox HK assured me they were. (Just thought I'd mention that.)

4. I'm working just 20 minutes down the road from Matrox's HQ in Montreal, but to this day there are no normal and certainly no MAX G400s available anywhere around here. I don't consider the G400 support in the PS done (if anything of this sort is ever "done") until I've tested with at least one MAX.

5. I always validate cards with a digital frequency counter, but that only validates the mclk, not the gclk/sclk/wclk. (Anyone with an OSD can validate the way I write the pclk.) With all due respect to Haig, a frequency counter is *vastly* superior to software , whether it be Matrox's HAL, G4Set, G200clk, or the PS. I could be mucking up the other clocks, or dozens of other things, for sure, but if so, its accidental.

If you go back to the beginning of this thread, you'll see it was started by Paul, who's having problems with Q2 and who believes it was caused by the PS messing up his BIOS. He's reflashed his BIOS, and he *still* has problems. I repeat: there are people who have problems with the G400 who have never heard of the PS.

Now Haig - head of Matrox's Tech Support - says the PS "screwed" his card completely. A pretty heavy condemnation. Don't think for even one second that I'm not extremely concerned about what is said here.

I am not convinced the PS is doing anything that can't be fixed by a reboot. Still, I've already decided to disable G400 clock support in the next PS release, at least for the time being.

I think eveyone will agree that this is the only responsible thing to do, especially since - unlike Matrox - I don't have to support the G400 at all...

[Kruzin]

I have no idea where you get the impression I (or Haig for that matter) is accusing you of doing anything malicious. That was never implied. The PINs dumping and modifying was something I said WE do to our G200, not that PS does to the G400. The only thing I have claimed is that I had a problem, and believe it started with PS. Flashing the BIOS (not just rebooting) was the only way to set things straight. I never said it was intentional on your part.

Get a grip. We're just trying to find the answer, same as you.

[rickT]

Ashley: explain this to me. I was using g200clk with the command line

c:\g200clk.exe /1 297. that set my core at 148.5, memory at 198. I never had any problem.

After using the latest powerstrip, I am able to set 148.5/198 in powerstrip, but whenever I run g200clk now my screen gets a bunch of vertical lines. g200clk no longer works anymore. Care to comment on that?

Also, if someone could send me the original bios file so that I may see if g200clk works again with an original bios flash. 32MB DH G400. send to rnay1@tampabay.rr.com

thanks

[Haig]

Hi Ashley,

"Now what you're suggesting is that I'm doing something really malicious: dumping the existing BIOS to disk, modifying it, and then reflashing it, right?"

Read my posts carefully before talking nonsense.

"Haig - How helpful would it be to you if I reported "The G400 screwed my system". "

I am not reporting anything to you. If I wanted to help you fix your problem, I would have done so privately.

I won't waste my time on some 3rd party overclocking utility having problems with our card especially when we even won't support it.

"With all due respect to Haig, a frequency counter is *vastly* superior to software , whether it be Matrox's HAL, G4Set, G200clk, or the PS."

??? The HAL Tools was used to see what registers have been changed in our chip!

I'm not about to start reading pinouts of our chip for this utility.

"I could be mucking up the other clocks, or dozens of other things, for sure, but if so, its accidental."

At least you are starting to admit something.

Haig

[JamesA]

Hi,

This is such an interesting thread that I think the title should be modified to "Debugging PowerStrip/G400". Heck, I am suppose to be working now, but I am writing this. ;P

Ashley: I don't think anybody want to intentionally malign/slander your product. Perhaps you could outline a procedure whereby a person could follow to prove or disprove their thesis that "PowerStrip made me blind!" or "PowerStrip whacked my DualHead!!".

Everyone Else: I am quite impress by the level of civility in this thread! Keep it up! This got to be the nicest forum around.

Just my el cheapo $0.02.

JamesA @ Work.


------------------
System Hardware:
www.cis.ohio-state.edu/~ang/demerzel.html

[Ashley]

Sinking deeper and deeper...

Let me try to get back on topic here. The thread was started by someone saying I modified their BIOS. I said that's not possible and I don't do it. A couple of people said it is possible. I repeated: it is not possible, and I don't do it.

Now comes the twist: the conversation changes from one about *modifying* a BIOS to *flashing* one. To me, that's not the same thing at all. You could, in fact, modify a BIOS if you reflashed it. No one's suggesting I do that? Good.

This leaves me, not modifying the BIOS, but screwing the card up in such a way that reflashing the BIOS is necessary to set things right again. No one thinks I'm do this deliberately - it is by accident. Is that put fairly?

Then I'm suggesting, not that it isn't possible, but that it should not be possible. The PowerStrip does not do this sort of thing to any other card, and it should not be able to do this to the G400. After all, I'm doing this accidentally, right? Therefore other programs can do it too. And if it can be done accidentally, it can obviously be done on purpose.

Kruzin - If you really wanted to help find an answer, then you could supply some "before" and "after" specifics. All I have seen here is "before it was OK, and after it was screwed up". What was screwed up? Paul says QuakeII. You say you lost 5% on benchmarks. Others say the BIOS, the card, or even the whole system was mucked up. I have asked again and again for specifics. I haven't gotten any.

rickT - I'm on the road right now using a notebook, so I can't even comment on, much less answer that one. There is more than one way to program a PLL to generate an sclk value, and it would be perfectly possible (probable even) for 3 different programs to arrive at the same sclk using 3 different parameters.

Haig - You have me at a distinct disadvantage. You were able to go and freely d/l my software, but I have no idea what HAL is. I didn't drag you into this discussion, and I fully understand why you don't want to discuss 3rd party software. However, if you ran the program under NT and compared it to Matrox's own PowerDesk, you might not be so quick to dismiss it as an "overclock utility". In any case, I've already said I'll disable G400 support in the next release, so you'll only have to contend with whatever Greg or Liew release...

JamesA - Guilty until proved innocent, eh? At the risk of stirring things up further: the PowerStrip only whacks both sides of a SingleHead. That's just not the same thing as whacking a DualHead...

[Haig]

HAL = Hardware Abstraction Layer
HAL tool = Hardware Abstraction Layer tool.

It's a utility that we have which tells us
everything we need to know about every single register in our chip. It also allows us to modify any register on our chip and also pop up warning messages anytime some other party tries to access our registers, among other things.

If you didn't know what this was, then don't crititicize it by telling me that your "frequency counter is *vastly* superior" than this tool.

"However, if you ran the program under NT and compared it to Matrox's own PowerDesk, you might not be so quick to dismiss it as an
"overclock utility"."

The only reason why I downloaded your utility was to answer a question for Kruzin. I have no interest in 3rd party utilities that we don't support even if part of the utility is not used to O/C our card.

I'm not picking on your utility about my non-interest because the same goes for all other 3rd party utils.

Haig

[JamesA]

Ashley,

No, no. I meant to say that perhaps you should provide a list of criteria that a person with problems should look into, and the relevant information that you need to troubleshoot the problem.

Just by asking for specifics isn't too helpful either you know. Perhaps you should be provide a version of PowerStrip that produces debug output. You know, your own limited version of what Haig has at his disposal - a HAL tool. Then, let the overclocker troubleshoot and provide the output for you to digest.

Sounds fair, no? Most programs/shareware has a debug option, no? (I am assuming here. Still new to the world of WinXX shareware. Still an old OS/2 fan.)

JamesA


------------------
System Hardware:
www.cis.ohio-state.edu/~ang/demerzel.html

[Kruzin]

Kruzin - If you really wanted to help find an answer, then you could supply some "before" and "after" specifics. All I have seen here is "before it was OK, and after it was screwed up". What was screwed up? Paul says QuakeII. You say you lost 5% on benchmarks. Others say the BIOS, the card, or even the whole system was mucked up. I have asked again and again for specifics. I haven't gotten any.

I gave a pretty specific description of how I produced the problem, and what that problem was. Excuse me for not writing down the exact incorrect dividers and clock rates I was left with.

It seems to me, that if you start up PS, then change clock settings enough times in that same session, eventually, it's gonna happen. Maybe not to many others have experienced the problem because they're not aware of how to check that there is a problem even if it is there, or they have not used it enough to produce the problem at all on their system.

So far as helping you find your problem, I'm more interested in keeping my fellow users aware of the situation. I have given you more input than time I spent (or likely ever will) actually using the program. If it's input you don't want to hear, or choose not to beleive, fine. It's you're software...you fix it...or dont...but I'll relay my personal experiences to the bothers and sisters of M either way

I told you how to reproduce the problem. Plug in your G400 and try it for yourself. Just remember to make a BIOS recovery disk before you start

[Scytale]

Haig, maybe your customers find an interest in running other software but the Matrox drivers.

And I still wonder where the NT beta testers are hiding...

[The PIT]

Mmm why don't matrox do a overclocking prog.

[Hellmut]

Hi Ashley,

my 2c here, You all propose to be the MGA specialists and after weeks, noone has noticed there is a unified bios on the matrox page. This contain the tools to flash the card underbios AND windows. This tells me, there is a way to write into the card under running windoze. You poke into certain adresses of the card, but where do you know from what any bit in the card does? If you set the wrong flag, your code might open "doors" in the card that are not supposed to.

Look at SoftFSB, it can switch your AGP to 133MHz with one button under windows.

The CIH+ erased mobobios while under Windows, so how can you say your program is harmless?
I think that you communicate via certain registers with the card, are you sure about every single bit's function? Who says poking a 216.3 mclk does not mean for the card that the next byte coming is to be written into the BIOS?

I'm sure that though programming PS you have piled up loads of knowledge about the weirdest things, but then you should know that one bit set wrong can decide over the further life of components.

peace,
Hellmut

[Haig]

Scytale - Don't understand.

Pit - Do you think it would be a good idea if we would do this but keep it as un-supported?

Haig

[Ashley]

Haig - Now who's not reading? I said I use a digital frequency counter to validate the memory clock, and that it is superior to any piece of software, regardless of who wrote it - including Matrox.

HELLO! - Is there an engineer in the house? Anyone anywhere who wants to contest this simple truism? I didn't criticize your HAL tool; I said I didn't know WTF it was, as it's not publicly available. You respond by telling me what "HAL" stands for... Thanks!

JamesA - I was just joshing... There are actually a lot (but not enough) of both documented and undocumented switches in the PS that might help sort this problem out. For example, the PS can dump your video BIOS to disk (you can use this feature to compare the BIOS before and after using the clock control). You can also turn off the PS's clock support altogether, disable all BIOS accesses (they are read-only anyway), set the max and min values for the clock slider, and on and on. As for what I need to know, an answer to the following questions about the PS and the G400 would be nice:

1. does the DDC work?
2. does the power management work?
3. do the gamma controls work?
4. do the custom refresh rates work?

All of these things are far more complicated, code-wise, than the memory clock - so much so that Matrox itself doesn't provide all these features under NT. More important: these are things the average person can actually *verify*, in a way that you cannot verify clock speeds and divisors.

Kruzin - you say you don't have time for this discussion anymore? People who haven't had your problem really have had your problem? They just don't know it "because they're not aware of how to check that there is a problem even if it is there"... Very strange...

You say I can replicate everything you experienced by just using the PowerStrip on my card? You're seriously suggesting that I've never run the software I write? Come on...

If you want me to believe something has really gone wrong with your card you're going to have to recipricate, and believe that nothing has gone wrong with mine.

So why would this happen? Could it be because I have a Rev02 chip and my card was obtained in HK? Because I have a SingleHead? Because I have a 16MB board? Maybe. That's why I mentioned all these things earlier. But I am the one with the retail card, not you. I paid for my card; you didn't. Mine was over the counter; yours came from Matrox. For all I know, the anomolies could afflict beta-testers only...

Hellmut - The BIOS is read-only, and Haig has specifically said the BIOS is NOT modified by the PowerStrip. Sure you can reflash a flash BIOS, but you cannot modify just one byte. Even if you could, you'd also have to update the BIOS checksum or the card would never boot again - that would imply deliberation, and worse. In any case, the PowerStrip only writes to memory-mapped addresses - when you shut down your system, the changes it makes are gone.

I continue to resist this whole "PS mucks up your card" thing, with - I think - good reason. I absolutely insist that this sort of thing should not be possible, and no one here wants to address this issue. Haig says it happens, but not *why it can happen*. In fact, he specifically says the BIOS is not altered and the clock registers are not changed by the PowerStrip. Therefore it is the *process* of reflashing, not the content, that fixes whatever has gone wrong.

That is just *too* strange, and - like I said - I'll be disabling G400 support in the next release. You guys will be relieved to know I'm off to destroy some Oxygen/VX1 cards now...

[chuck]

Haig,
Sure, do it like Power Toys for windows. "Here's some cool stuff to play with, but we don't have the resources to support officialy".
You would hear vast rounds of applause in here.
chuck


------------------
(erstwhile cjolley) celery 333@500mz 2.2V, 128meg@cas2 gh, 10gig IBM 7200rpm UltraStar, Princton EO75, $15 sound card(!),V3 2000 PCI (still waiting on my MAX!)





[This message has been edited by chuck (edited 07-30-99).]