The Rydermark Thread (TM)

Geo

Mostly Harmless
Legend
http://www.theinquirer.net/default.aspx?article=32809

Don't really know how to take this piece. It bothers me there is no link to any discussion somewhere with the Rydermark folks talking about what they found.

Faud never talks about _pp per se, but presumably the allegation here is that NV is using it universally rather than only when requested (personally, I find it hard to believe this could be true for the G7x lineup). With Rydermark not even being released yet, it seems unlikely it would be targeted in NV's drivers, which would lead to even more disturbing questions about just how generically this is happening.

But there clearly aren't enough facts associated with this to get too excited about it yet. Anyone know who these Rydermark people are to get some more info from them on their findings? Any of our NV people care to get a response on this issue?
 
DirectX 9 requires you to use at least 24 bit or full 32-bit Shader precision.
Incorrect.
The only problem is that 16 bit precision is below the requirements of DirectX 9, so if you use less than 24 you are not DirectX 9 compliant.
Incorrect.
If you want to do normal mapping, parallax mapping and water reflection/refraction, your Shader requires 32 bit precision.
Incorrect.
Nvidia doesn’t leave you any choice, it's claimed. You simply cannot turn 24 or 32 bit precision on, you are always locked at 16 bit.
Nvidia hardware support FX8 (32bit) precision alright... Now, if they're talking about FP24 (96bit) and FP32 (128bit), then Nvidia never did advertise the support of FP24, but did advertise FP32 support.

These folks working on this 'Rydermark' should probably sustain their accusations with something a wee bit more solid than just a public claim and a bunch of incorrect explanations. Also, why would these folks be the first and only persons working on post NV3X Nvidia hardware being able to tell that?
 
It is part of the spec, however, to provide either FP24 or FP32 *unless* _pp is used by the dev to request FP16. Isn't it?
 
FP24 is an ATI thing, nV never had it AFAICR. But FP32 is there since GF6, if I recall correctly.
 
Vysez said:
Incorrect.

Incorrect.

Incorrect.

Nvidia hardware support FX8 (32bit) precision alright... Now, if they're talking about FP24 (96bit) and FP32 (128bit), then Nvidia never did advertise the support of FP24, but did advertise FP32 support.

These folks working on this 'Rydermark' should probably sustain their accusations with something a wee bit more solid than just a public claim and a bunch of incorrect explanations. Also, why would these folks be the first and only persons working on post NV3X Nvidia hardware being able to tell that?

To be honest, we only have the Inq take on it - we dont have a link to the actual statement by the devs (or do we?)
 
kihon said:
hmm - didnt FX also "support" fp32

Well, if they've "discovered" that GFFX cards are doing this (and only GFFX), then I vote we break out the rotten fruit and save the watermelon for Fudo.
 
geo, you give the Inq far too much credit. Despite all the inaccuracies highlighted above, the fact that they were equating shader precision with texture formats was enough for me to dismiss the entire article. There's nothing in the article to lend it any credibility.

"16 bit textures only, not 32 please, breaking DX9 specs"
"DirectX 9 requires you to use at least 24 bit or full 32-bit Shader precision. "

:rolleyes:
 
_xxx_ said:
FP24 is an ATI thing, nV never had it AFAICR. But FP32 is there since GF6, if I recall correctly.
The NV30 supports FP32.
FP32 Shaders had bad performance results on the NV3X series compared to the R3XX series.
kihon said:
To be honest, we only have the Inq take on it - we dont have a link to the actual statement by the devs (or do we?)
"Anonymous developers"...
http://www.theinquirer.net/default.aspx?article=31799
 
_xxx_ said:
FP24 is an ATI thing, nV never had it AFAICR. But FP32 is there since GF6, if I recall correctly.
To attain "DX9 SM2.0" full precision you have to support FP24 or greater (which translates to FP24 for ATI (plus others) and FP32 for NVIDIA.
 
trinibwoy said:
geo, you give the Inq far too much credit. Despite all the inaccuracies highlighted above, the fact that they were equating shader precision with texture formats was enough for me to dismiss the entire article. There's nothing in the article to lend it any credibility.

"16 bit textures only, not 32 please, breaking DX9 specs"
"DirectX 9 requires you to use at least 24 bit or full 32-bit Shader precision. "

:rolleyes:

I think the Rydermark folks told Fudo something. What it was is a different question. We all know that between techies' lips and Faud's keyboard there has been many a slip, but I'd be very surprised if there isn't a kernel of something there (at least so far as the Rydermark guys bitching about NV precision, fairly or unfairly).

This Rydermark team and their "Sahara game engine" are anonymous at this point, so it's also an open question what level of expertize they are bringing to the table.

I don't like these kind of amorphous allegations. I'd much prefer to get this cleaned up sooner rather than later. Hence asking for somebody to talk to the Rydermark people, NV, or both. That's because I have faith that in this particular community there are people who can make that happen and get a response.

Edit: Inq broke the existance of this new benchmark here: http://www.theinquirer.net/default.aspx?article=31799 So I think it reasonable to assume Faud has been in contact with them for some time.
 
geo said:
I think the Rydermark folks told Fudo something. What it was is a different question.

Well that's the premise which, if true, would indicate that all of the tier1 developers and our very own eyes have been deceived for three years and we're now being rescued by these anonymous vigilantes. There's no substance in the article to point a finger at. There isn't even some scrap of evidence to start a conspiracy theory and we all know how little it takes to start one of those.
 
Sorry, I didn't see you reply at first

geo said:
It is part of the spec, however, to provide either FP24 or FP32 *unless* _pp is used by the dev to request FP16. Isn't it?
Yes, if you advertise full precision support, you need to support at least FP24 or FP32.
 
trinibwoy said:
Well that's the premise which, if true, would indicate that all of the tier1 developers and our very own eyes have been deceived for three years and we're now being rescued by these anonymous vigilantes. There's no substance in the article to point a finger at. There isn't even some scrap of evidence to start a conspiracy theory and we all know how little it takes to start one of those.

You don't make these things go away by ignoring them. You make them go away by getting responsible people in a position to know and prove them untrue to do so in public.
 
geo said:
So I think it reasonable to assume Faud has been in contact with them for some time.
I was in contact with them at last years GDCE and they have email dropped multiple editors several times since then.
 
Dave Baumann said:
I was in contact with them at last years GDCE and they have email dropped multiple editors several times since then.

Ahh. Anything interesting in those emails? Any sense of what level of expertize and prior experience they are bringing to the table?
 
geo said:
You don't make these things go away by ignoring them. You make them go away by getting responsible people in a position to know and prove them untrue to do so in public.

Heh. If responsible in-the-know people took their time to prove all of Fuad's unfounded ramblings untrue they would be extremely busy people. How about we lay the burden of proof on him instead?
 
trinibwoy said:
Heh. If responsible in-the-know people took their time to prove all of Fuad's unfounded ramblings untrue they would be extremely busy people. How about we lay the burden of proof on him instead?

If he'd asked me, I would have strenuously advised him not to publish without more than that. At least a quote, and preferably a link to a discussion somewhere by the Rydermark people.

But he did, and now it will be all over the place, because that's the power Inq has, for good or ill. Do a search on Rydermark 2006 and you'll find a great many links, almost all of them pointing at Inq's original article on Rydermark.

"more to come" says Inq.

Personally, I've never understood the theory that says wait until there's a full-throated roar before you step on unfounded rumors that will damage your reputation if allowed to grow without a response that shows they are indeed unfounded.
 
geo said:
Personally, I've never understood the theory that says wait until there's a full-throated roar before you step on unfounded rumors that will damage your reputation if allowed to grow without a response that shows they are indeed unfounded.

Yeah I can see that perspective. But would you really like to set that precedent for a publication like the Inquirer that trades in conspiracy and rumor? As a PR manager reading that Inquirer piece what would your reaction be? Would you really want to validate that drivel with a response?

In it's current form it is completely harmless and without further (any) evidence will die a natural death because there are sufficiently knowledgable people on the boards to put an end to it without Nvidia's intervention. This thread is evidence of that already. It is no where near the point where an official rebuttal is necessary. An official statement directly from the devs with some hard facts would sure stir things up though.
 
Dave Baumann said:
To attain "DX9 SM2.0" full precision you have to support FP24 or greater (which translates to FP24 for ATI (plus others) and FP32 for NVIDIA.

That's what I meant, nV did the "or greater" thingy with FP32.

Didn't know about GFFX supporting it though. I suppose it was slow as a**, maybe just somehow emulated in SW?
 
Back
Top