Nvidia agp8x works fine while Ati not!!!

[Fotis]

Reading a few reviews of geforce4 agp8x cards I can say that there has been no problem with sis648 or kt400 mobos.
Can Ati convince us that mobo makers are to blame since theirs is the only card that does not work with agp8x mobos?:rolleyes:

[SnakeEyes]

I was waiting to see when someone would point this out. I know it doesn't give absolute proof that it's the ATI cards causing the problem, since the AGP specs have some leeway for interpretation and implementation. ATI's implementation is probably either borderline for the spec or out of spec, at least in the first revisions of their cards.

[jbirney]

Check again, Xaber cards are also having issues on 8X AGP mobos as well.

Its also funny how you folks seem to forget the oringal GF launch when it had so many issues with AMD mombos. Other cards worked fine but GF cards caused lockup if they even booted at all. Turned out to be that the mobos were not following the AGP power specs and not the GF cards "fault". Funny how quick you to blame another vendor......

[Gator]

and when was the GF1 launch? how many years ago? :rolleyes: weak arguement my friend

[SnakeEyes]

I'm not surprised that Xabre cards have issues too. In fact, being released earlier than available motherboards, much like the R9700, as well as coming from the economy brand card maker :p, it's a given that their implementation is either buggy or borderline (even moreso than the conclusions to be drawn concerning the 9700).

As far as the GF1 problems with AGP specs, what has that got to do with the price of rice in China, or the current issues with AGP8x? Just because nVidia's the only company that got it right pre-launch, don't feel bad. And I'm fairly certain that the motherboard / chipset makers (other than nVidia themselves, of course) didn't get with nVida to make sure that their implementations matched nVidia's specifically (the argument used concerning incompatibilities in methods used for some game's graphics / features with non-nV cards0. :D

[vitocorleone]

"weak arguement my friend"

likewise

[jbirney]

Originally posted by Gator
and when was the GF1 launch? how many years ago? :rolleyes: weak arguement my friend

And

Originally posted by SnakeEyes
As far as the GF1 problems with AGP specs, what has that got to do with the price of rice in China, or the current issues with AGP8x? Just because nVidia's the only company that got it right pre-launch, don't feel bad.

What does it have to do with the DX8 Issue? Nothing and everything. Nothing as its a completely different issue. Everything as now instead of NV cards not working its ATI cards. So these people that are now claming that's its ATI's fault were the same ones claming it was not nV back then. Basically your being hypocritical. Besides Kyle over at HardOCP has stated that most of these issues lie with the mobos (just like they did back in the GF days). I can provide a link but doubt it will help as some folks are too busy spreading crap without knowing what the issue is...

[SnakeEyes]

Thanks for the feedback jb. But the fact that two chips released prior to hardware supporting one of their features both have the same problem, while a chip released after the same supporting hardware has no problem says that at a minimum the two earlier releases are riding the edge of the specification. The same can be said for the GF issues (ie. nV rode the bleeding edge of the power specs for those chips), but that is definitely settled now, isn't it? :D

[Joe Shmoe]

Originally posted by SnakeEyes
Thanks for the feedback jb. But the fact that two chips released prior to hardware supporting one of their features both have the same problem, while a chip released after the same supporting hardware has no problem says that at a minimum the two earlier releases are riding the edge of the specification.
It doesn't mean that at all. How do you know that the boards appearing later (i.e. the nvidia boards) don't have workarounds for flakey motherboards in the BIOS/drivers?

[SnakeEyes]

Joe Schmoe,
In general:

Release order:
ATI 8x cards early revisions / Xabre, Sis648 / KT400 chipsets, nV 8x / ATI revised 8x cards (I believe the rev number people have been mentioning is 1.03 or 1.3?) cards.

8x compatibility with mobo chipsets and unrevised mobo bioses, cards released before the chipsets available to test the video card implementations on:
ATI 8x early revisions - Sis648 - sometimes, but enough problems to become a hot topic
ATI 8x early revisions - KT400 - sometimes, again enough to make it a hot topic
Xabre - Sis648 - same as ATI / Sis 648
Xabre - KT400 - same as ATI / KT400

8x compatibility with mobo chipsets and unrevised mobo bioses, cards released after the chipsets available to test the video card implementations on:
nV 8x - Sis648 -No problems reported.
nV 8x - KT400 -No problems reported.
ATI 8x new revisions - Sis648 -No problems reported.
ATI 8x new revisions - KT400 -No problems reported.

If I stopped at the list of cards prior to the new nV18/28 based models, I'd probably stick to the idea that chipset manufacturer's implementations were flawed, and that both ATI and Xabre cards were likely suffering because of it. With the introduction of the nV chips, it becomes more unlikely, because:
1) Each motherboard chipset is based on unique code, used to implement the AGP spec. (Intel only releases specs, not proprietary implementation information). This means nVidia would have to have hacks / patches / workarounds for two different chipsets, and the workarounds would likely be different, in addition to making sure that the nV18/28 still worked properly on any chipset that isn't flawed (ie. 3 code paths for AGP3.0 spec).
2) ATI was able to change the later revision R9700's in such a way that BOTH of the problem chipsets start working fine at 8x. This again would require 3 codepaths if the issue were with the chipsets, instead of with the video card itself.

So while I suppose it's entirely possible that nV did include workarounds for faulty implementations in the motherboard chipsets for the Sis648/KT400, I find it less likely than that ATI simply revised the implementation in the 9700 to be closer to spec, which results in 1 supported code path (based on the AGP3.0 spec, not needing hacks/workarounds). Otherwise both ATI and nVidia have workarounds in their new cards.. :D

That's my thinking anyway.

[SnakeEyes]

And then there's this, which lends even more weight to it being an ATI issue:

originally quoted by jbirney in this topic: Change 9700 to AGP4x (http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=2261)
Spent over an hour with ATI Customer Service Canada and more important, the Engineering Dept. The subject: Our testing shows the Radeon 9700 will NOT work at all on the Asus P4S8X (SiS648), Gigabyte 7VAXP (KT400/8235), and VIA P4PB (P4X400/8235) which are the only 8x AGP boards on the market. Customer Service was not completely aware of this, but Engineering is.

According to ATI, they are now receiving and testing these 8X AGP boards. They are ALSO finding the 9700 does not work at all or works intermittently at best on these boards. Engineering indicated most of the boards already shipped have this problem.

Solution: A new BIOS and some board revisions are now available that fix this problem. Call ATI Customer Service for help with the issue. At present, they will ship a replacement board that WILL work in SiS and VIA 8X, but they are also considering a BIOS flash solution for problem boards.

Customer Service and Engineering deserve a 10-1/2 for their wonderful handling of this issue. They are very concerned that it be corrected quickly. They also shared that most of their 8X validation testing was done with Intel's unreleased 8X chipset and that they are just now receiving shipping versions of the SiS and VIA 8X chipsets.

[Joe Shmoe]

Originally posted by SnakeEyes
1) Each motherboard chipset is based on unique code, used to implement the AGP spec. (Intel only releases specs, not proprietary implementation information). This means nVidia would have to have hacks / patches / workarounds for two different chipsets, and the workarounds would likely be different, in addition to making sure that the nV18/28 still worked properly on any chipset that isn't flawed (ie. 3 code paths for AGP3.0 spec).
No offense intended, but you have no clue what you are talking about. I've done video driver work and we routinely put workarounds into the driver for flakey AGP implementations. That was back when AGP 4x appeared, and I have little doubt that things have changed with AGP 8x.

2) ATI was able to change the later revision R9700's in such a way that BOTH of the problem chipsets start working fine at 8x. This again would require 3 codepaths if the issue were with the chipsets, instead of with the video card itself.
They could just have increased the tolerances for certain chipsets, it's not a big change to the BIOS.

So while I suppose it's entirely possible that nV did include workarounds for faulty implementations in the motherboard chipsets for the Sis648/KT400, I find it less likely than that ATI simply revised the implementation in the 9700 to be closer to spec, which results in 1 supported code path (based on the AGP3.0 spec, not needing hacks/workarounds). Otherwise both ATI and nVidia have workarounds in their new cards.. :D
As I said, it's not hard to increase tolerances for marginal AGP implementations. There are a whole host of AGP parameters to tweak (latency, clock skew, etc.).

[SnakeEyes]

So what you're basically telling me is that by making your AGP implementation (in bios or drivers) less marginal (ie, not pushing the spec to the limits), you're able to get cards stable on marginal chipsets? :p

Besides, reread the part I quoted above, where ATI admits that the changes were to the bios AND to some board revisions to fix the issues. Sounds a bit less like simply adjusting those parameters you're talking about to me, and a bit more like fixing a hardware bug in conjunction with possibly adjusting software / bios code.

Keep in mind that even if ATI is within the specs with their implementation, I never said they weren't. If you reread my posts above, I only said that it's likely that the 9700 was riding the edge of the specification, ie., pushing tolerances so that any imperfections in motherboard chipsets would cause problems.., or possibly actually broken. I made sure to leave the fact that ATI didn't have to be out of spec to be having these problems as a possibility. However, you would expect that a hardware manufacturer would want to have as much tolerance built into their hardware, bios, and drivers that any marginal (but still correct) motherboard chipset implementations would still work.

Correct me on one more thing if you like, but haven't a number of users (most if not all of the original revision radeon 9700 owners with one of the motherboard chipsets we're discussing) have problems even posting with their cards? If so, how do drivers fix anything at all? The AGP settings are effective immediately on posting, and so take effect at boot- so maybe the card bios could help, but drivers don't touch crap at this point.

As an end-user of the afforementioned products, it matters less to me that ATI was in the spec and just had to tweak (apparently invalidating the hardware itself, in this case. Again, re: my quoted section above) the hardware, bios, and drivers, as the fact that with my system I wouldn't be able to use it. This is a fact proven already by the scores of people having issues with the 9700 in the 8x capable motherboards out there. And an option to step the motherboard down to 4x isn't a fix, it's a workaround, because you don't get the full benefit of features fully expected when you pair an 8x motherboard with an 8x video card (sigh, before you say it, even IF it doesn't result in any real improvement at the moment over 4x).

The fact that there are products available that will work with the motherboard in 8x just fine, while there are no products that will work with the particular ATI 9700's in question at 8x is a big discriminator too, no matter how easy you say it is to increase tolerance for what some are claiming is a motherboard issue. It's only made worse by the high likelihood that if I had that combination of hardware, the only fixes available would be either to try to get the card exchanged for a replacement through the manufacturer (assuming it's ATI and not some clone, which might not yet be manufacturing cards based on the fix), use a different brand of card altogether, or change motherboards (and losing 8x capability period).

[Joe Shmoe]

Originally posted by SnakeEyes
So what you're basically telling me is that by making your AGP implementation (in bios or drivers) less marginal (ie, not pushing the spec to the limits), you're able to get cards stable on marginal chipsets? :p
More like making the AGP interface more tolerant on marginal systems.
Correct me on one more thing if you like, but haven't a number of users (most if not all of the original revision radeon 9700 owners with one of the motherboard chipsets we're discussing) have problems even posting with their cards? If so, how do drivers fix anything at all? The AGP settings are effective immediately on posting, and so take effect at boot- so maybe the card bios could help, but drivers don't touch crap at this point.
Perhaps ATi is putting a new BIOS on the boards? No idea. But as far as system stability goes, you can do driver tweaks as well, once you get to Windows of course.

But some people mentioned that many of those POST problems weren't caused by AGP issues, but, instead, by powersupply problems. It must be difficult to handle power coming from two separate sources (AGP port and floppy power). I read on Rage 3D that some power supplies aren't giving enough voltage on some of the rails (5V? I dunno, I'm not a hardware person), so that could contribute to stability problems.

As an end-user of the afforementioned products, it matters less to me that ATI was in the spec and just had to tweak (apparently invalidating the hardware itself, in this case.
Of course! Things should "just work". But the PC market is a real joke, IMO. Too many motherboard, chipset, CPU, powersupply, audio card, video card, networking card, etc. etc. etc. combinations to guarantee that everything "plays nice". So if ATi is not at fault (similar to my past experience), they still have to fix it because people will blame them anyway.

[Smokey]

Originally posted by jbirney
[B
Its also funny how you folks seem to forget the oringal GF launch when it had so many issues with AMD mombos. Other cards worked fine but GF cards caused lockup if they even booted at all. Turned out to be that the mobos were not following the AGP power specs and not the GF cards "fault". Funny how quick you to blame another vendor...... [/B]


I havent forgotten :p My wife is still using my MSI motherbaord with AMD chipset with a GeforceDDR. What was the work around? Using 2x AGP and not 4X. Why? Because AMD was at fault. What do ATi9700 owners have to do if they cant use AGP 8x? Use 4X, why? Well its not the chipset companys SIS or VIA, is it the card manufacture? Well cards are being sent out now with a new bios?

I think if the card manufacture is making a fix, then it was the card manufacure at fault.

[SnakeEyes]

Good level-headed reply there Joe Schmoe. (No sarcasm) As far as the hardware and needing to update the bios, unless someone badly misworded their reply, that message that was quoted earlier seems to indicate that the change to the hardware was in addition to the new bios. Besides, the bios is flashable, so why would the card need replaced for the new one?

Smokey, I tend to agree with you. If the card manufacturer is replacing the bios and/or cards, it's likely the card's problem. Again, that's not saying I KNOW it's the card, but it sure is likely.

As for changing the AGP settings to keep using the card, I have two objections (one is specific to this situation at the moment though):
1) If you own one of the 8x motherboards, as well as the 9700, you're probably counting on that feature. Not using it is like getting slighted, though it probably makes about .1fps difference in most games. :)
2) This is the specific objection. From what I've read, the current bios on these motherboards doesn't allow it to be set back to 4x mode. They just automagically set themselves to the highest level the cards indicate they can handle. This should work fine if the cards work right (and the motherboards too- NOD in the direction of the people still on the motherboard side of the fence.). However, it still would be better if the mobo bios allowed this to be overriden manually (like most older boards I've ever seen do), and then if someone really needed or wanted to run at 4x for whatever reason (overclocking?) they could.

It really is going to become a moot point though. I suspect that the cards having problems will be modified/replaced (avoiding the use of the word fixed this time- again, NOD to the mobo's at fault side of the debate), and the motherboard bioses will most likely end up with manual overrides anyway.

[Mono]

Intresting considering I havn't seen any reviews/previews on these cards let alone user feedback. I saw pictures of Gainward ones but that's about it. Care to give me some linkage?

I think the problem is with the motherboard manufactures and ATi. It seems the 9700 is pushing the spec real close to borderline but some motherboards work almost all the time *Soyo KT400 Ultra Platinum, i've seen all positive on this* so i'd say some of the motherboard manufactures are being lax on following the specs intel has given. With that said, I think ATi and the motherboard manufactures should do revisions to see it dosn't happen, but of course the blame will get put mostly on ATi along with the corrections most likely.

[[Corporal Dan]]

Originally posted by Gator
and when was the GF1 launch? how many years ago? :rolleyes: weak arguement my friend

Exactly.

Every nvidia card has been flawless (from functionality) standpoint ever since.

And besides, the GF256 wasn't the ONLY incompatibility back then. The first AMD chipset was remarkably sloppy, and the motherboards were all of pretty low quality.

[[Corporal Dan]]

Originally posted by jbirney


And



What does it have to do with the DX8 Issue? Nothing and everything. Nothing as its a completely different issue. Everything as now instead of NV cards not working its ATI cards. So these people that are now claming that's its ATI's fault were the same ones claming it was not nV back then. Basically your being hypocritical. Besides Kyle over at HardOCP has stated that most of these issues lie with the mobos (just like they did back in the GF days). I can provide a link but doubt it will help as some folks are too busy spreading crap without knowing what the issue is...

Oh please.

Let's count how many years ago that was: 3-4?

That's plenty of time for ATI to get their act together.

Everytime there is a new ati card, if it's not one problem, it's another.

R9700 was launched a little prematurely.

[PreservedSwine]

Originally posted by Smokey




I think if the card manufacture is making a fix, then it was the card manufacure at fault.

Not at all, both ATI and Nvidia have to do many workarounds in their drivers and BIOS to cover up problems that other vendors (both hardeware and software) refuse to fix. This happens constantly.

[PreservedSwine]

Originally posted by [Corporal Dan]


Exactly.

Every nvidia card has been flawless (from functionality) standpoint ever since.

And besides, the GF256 wasn't the ONLY incompatibility back then. The first AMD chipset was remarkably sloppy, and the motherboards were all of pretty low quality.

Just out of curiosisty, what is the color of the sky in your world?

Remember when the Geforce2 outperformed the Geforce3?

[[Corporal Dan]]

Originally posted by PreservedSwine


Just out of curiosisty, what is the color of the sky in your world?

Remember when the Geforce2 outperformed the Geforce3?

Purple, why?

A gf2 doesnt outperform a gf3..?!?

Was it shoddy release drivers?

if so, it musta been fixed within like a week

[StealthHawk]

Originally posted by PreservedSwine
Remember when the Geforce2 outperformed the Geforce3?

we've already been over this :rolleyes: anyway, what's so unbelievable about a card with the same memory bandwidth and more fillrate outperforming a card with less fillrate and same bandwidth in stock situations ;)

[thcdru2k]

when the geforce3 was just released, the geforce2 ultra, was able to beat it in certain situations like 640x480x16 because it had more fillrate. that's of course not the case anymore after a couple of driver releases.

http://www.digit-life.com/articles/gf3/q3-16-anis.gif

http://www.digit-life.com/articles/gf3/q3-16-quaver.gif

[netviper13]

I went from a GeForce2 Ultra to my current GF3 Ti200, and I have noticed a huge increase in the performance.

I remember that maybe the first two weeks, the GF2 Ultra could outperform the GF3 in CERTAIN tests...but at the very least all of the GF3s features worked.