NVIDIA claims top DX9 GPU marketshare spot.

Evildeus said:
For Nv it's the same thing.

Thats for accountancy purposes. Mercury just ask each IHV how many units ship, they don't go back to the outlets and ask what have sold.

FYI - there's an interesting discussion over DX Dev list as to whether its better to ship a lower performance currnet DX board, or a high performance previous board in the low end.

Look for Support of D3DFMT_A32B32G32R32F ?
 
Demalion: he specifically mentioned performance in his original statment, so it would be safe to assume that the second part (can't run dx9 properly) does not include performance as a criteria.

And if this is all opinion, can I therefor say that the R300 can't run Dx9 (or even dx7 or dx8) properly because the anisotropic isn't good?

Dave: share with us these hardware waivers. What are they and where did you hear about them?
 
DaveBaumann said:
Evildeus said:
For Nv it's the same thing.

Thats for accountancy purposes. Mercury just ask each IHV how many units ship, they don't go back to the outlets and ask what have sold.
For Nv the number of chip shipped depends on the number of chip sold to board makers. Actually, the number of chip sold should be > than the number of chip shipped... That doesn't mean that the boards makers have sold them.

FYI - there's an interesting discussion over DX Dev list as to whether its better to ship a lower performance currnet DX board, or a high performance previous board in the low end.

Look for Support of D3DFMT_A32B32G32R32F ?
Interesting, but a bit useless don't you think? It's not up to them to decide how the markets of VC fluctuates, but it's up to them to be as near as possible of this fluctuations. Anyway, there's no definite answer in this thread if i'm not mistaken.
 
DaveBaumann said:
A waiver is effectively an area of non-compliance but MS are willing to overlook for certain reasons. I'm hearing that the 5200 has several waivers - we know that it only runs in FP16, so this is likely to be one of them.

The WHQL process is quite convoluted, not necessarily in terms of testing for compliance but more in terms of penalties for non-complaince (i.e. for non-WHQL hardware saying windows compatible being sold, the vendor has to pay MS money for each unit).

Also, too, as has always been true, if a company can produce a set of drivers which is capable of supporting API version features in software the IHVs commonly announce that the given hardware supports that version of the API without explaining that some of the required function support occurs in software as opposed to the hardware. The idea from a commercial perspective is that a buyer sees "DXx supported" as a marketing bullet associated with the product and does not appreciate that the "DXx support" in the $150 product is not the same as found in the $300 + products. Every IHV has done this at one time or another, AFAIK.
 
RussSchultz said:
Demalion: he specifically mentioned performance in his original statment, so it would be safe to assume that the second part (can't run dx9 properly) does not include performance as a criteria.

That is not logic, that is stating "A follows from B" by depending on a presumption that someone can't "specifically mention performance in his original statement" while including performance as a consideration elsewhere as well. I think that assumption is completely nonsensical.

Now, let's discuss why I think that: for it to make sense, the wording would have to indicate that the two statements are mutually exclusive. What about the wording indicates that? His original statement highlighting performance does not preclude that he can't be refering to performance again in the same sentence. The word "which" does not indicate that something is mutually exclusive.

Let's propose a different assumption: that he is proposing that the reason the 5200 "can't run dx9 properly" is because of its poor performance. How would a person say such a thing? Well, connective words like "which" seem to me to be one way, and I propose that's exactly what Jerky did.

Now, maybe you are only paying attention to your quote selection, which deleted the connecting commentary (that would be your fault, because Jerky put it there pretty clearly). I'll quote it here for examination:

Jerky said:
...the FX5200, which we all know is an under-powered card which probably can't run DX9 properly.

Now, if you have some reasoning that interprets this as the first assumption, and not the second, it would have been helpful if you'd included it. A straight reading of English does not seem to support your interpretation unless you drop inconvenient phrases, and I'm at a loss as to how your quoting can reasonably be proposed as anything else: you dropped exactly the text that works against your assumption, and the full quote doesn't seem to really need much trimming AFAICS. :oops:

In any case, it seems to me that you took a potshot, and, based on still not being able to see how your assumption in your reply makes any sense whatsoever, it also seems to me that you have decided to conduct your conversation purely based on saying he isn't entitled to an opposing opinion.

And if this is all opinion,

Please be careful when you characterize someone's specifically targetted commentary as "all". Your description using that word is erroneous, as I'll clarify.

can I therefor say that the R300 can't run Dx9 (or even dx7 or dx8) properly because the anisotropic isn't good?

Sure you can, I'm not aware of any facts that preclude that opinion from being stated, just ones that make it seem ridiculous (IMO) once someone does so. For instance...someone might have the criteria that a "good DX 9 card" has to run something like GT 4 in 3dmark 03 at 100 fps...there are no facts that say that you cannot have that criteria, though logic can dictate that you'd have to apply it consistently and meaningful discussion would require that you have some sort of coherent reasoning behind the criteria.

However, you can't say it isn't DX 9 compliant, and you can't propose your opinion as a fact, or even more than an unsupported opinion (at least without some indication that there is some support).

I do think these are facts, though: that "properly" and "with compliance" are not the same words, and that "DX 9 compliance" is a specific set of standards associated with the co-existence of "DX 9" and forms of the word "compliant" being used together. Do you disagree? I could be mistaken (though the first seems a bit basic), but some indication of my error would be helpful if you are proposing that.

I was trying to avoid linking to the dictionary...will it be necessary? :-?
 
Sigh, I'm not going to get into a back and forth.

A) If he's a reviewer, he should know if it does or does not do DX9 properly. "Probably" shouldn't enter the equation.
B) I deleted 'underpowered' in my quote, because that definately is an opinion, and its one I generally agree with. (Though, for its pricepoint, I think it does OK)

He's entitled to an opinion, but a statement like "probably can't run DX9 properly" tells me he probably doesn't know enough about a card he's supposedly reviewed.

Or am I not entitled to my opinion and the ability to express it?

p.s. you can keep your dictionary put away. I really don't want to continue the discussion which will boil down to another semantics debate.
 
Russ, as I said earlier we know 5200/5600 and 5800 all run at FP16 precision in the WHQL certified drivers, which is not to DX9 specification - I'll wager that the proviso there was that subsequent parts would ensure FP24/32 compliancy (as 5900 does - afterall, if they could get away with FP16 with the rest you'd think they would want to with 5900 as well, as its faster). I think fog is borked as well. When I get some time I'll run DCT over it and see what crops up.

And as Jerky is a reviewer do you think its not at all possible that he's privy to more source then you are? And not necessarily competing IHV's bs'ing about each other?
 
Is it or is it not DX9 compliant? Are or are not the drivers WHQL? The responses to these are facts, anything else is opinion.
 
The specifications of DX9 call for a minimum FP support of FP24 in the pixel shader - NVIDIA's WHQL drivers for 5200 only provide support for FP16.
 
RussSchultz said:
Sigh, I'm not going to get into a back and forth.

1) Refusing to discuss is not a virtue, nor does it validate your viewpoint. Yes, I've said this to you before because...yes, you've given me cause to do so before.

A) If he's a reviewer, he should know if it does or does not do DX9 properly. "Probably" shouldn't enter the equation.

2) Changing the point of dispute does not strengthen your argument. This comment depends on my commentary on "properly" being in error without the bother of actually showing it is the case. :-?

B) I deleted 'underpowered' in my quote, because that definately is an opinion, and its one I generally agree with. (Though, for its pricepoint, I think it does OK)

3) In case you lost track: I said that "properly" can include something besides just "DX 9 compliance", and supported it with specifics. You then replied he could not be talking about performance with properly, by the expedient of ignoring my support utterly. What is amazing to me is that you agree with Jerky that the 5200 is underpowered, and continue your attack without once addressing my discussion of how "properly" could include "underperforming" perfectly validly in the text Jerky actually said.

He's entitled to an opinion, but a statement like "probably can't run DX9 properly" tells me he probably doesn't know enough about a card he's supposedly reviewed.

So, you refuse to discuss anything I said, and, therefore, there isn't a problem with what you said in contradiction to what it proposes. To demonstrate, you proceed to...say what you said again.

Thank you for a productive conversation. Or not.

Or am I not entitled to my opinion and the ability to express it?
Well, I called your initial reply an unproductive pot shot, and said why. You replied that Jerky couldn't be talking about performance with "properly", and therefore seem to be maintaining that it was justified. I illustrated, clearly it seems to me still, that this assertion was counter to any reasonable interpretation of the full quote. As a reminder, the full quote (of what you edited and quoted) goes "...the FX5200, which we all know is an under-powered card which probably can't run DX9 properly".
Now, I really could spend some time discussing what "which" means in English, but you seem uninterested in irrelevancies like what things mean. :oops:

Anyways, nowhere in that did I deny you an opinion on the 5200 or Jerky. What I denied you was your apparent belief that your opinion did not require justification such that you could make a post calling Jerky incompetent and accomplish anything useful (outside of your desire to call him incompetent). I.e., I called it an unproductive pot shot. What I was denying you was a pretense for not having to support your opinion to make it more than just an insult thrown in for your personal satisfaction.

p.s. you can keep your dictionary put away.
Not when we don't appear to be speaking the same language. I know I'm calling what I'm using English, and with your usage diverging so widely it seems I might need to provide evidence of which one of us is really speaking it.
Well, if you were willing to entertain disagreement or correction to your statements, that is. :-?
I really don't want to continue the discussion which will boil down to another semantics debate.
Well, there really isn't much of a debate if one person is ignoring the meaning of words and language (in case you're confusing "semantics" with a swear word and forgot what it means), or is making up their own as they go along. There also isn't much communication when this is the case, either.
I don't think I'm the person who is ignoring "meaning", Russ. The lack of a semantics "debate" is exactly the problem...you're exhibiting only an interest in telling me the way things are without maintaining relevance to things like language. :-?
 
Evildeus said:
Is it or is it not DX9 compliant? Are or are not the drivers WHQL? The responses to these are facts, anything else is opinion.

As I said above, "DX9 compliant" can mean many things...;) It's far from opinion that although the GFFX 5200 and the R350 can both be said to be "DX9 compliant" they each offer very different hardware support for DX9, both in type and efficacy.

Come on, would you prefer the guy to say "It is not a DX9 part" when he's not sure about that, or would you prefer him to say "It probably isn't" when he's not entirely sure? I think the latter is the only honest answer if he's not sure. Last I checked it is not a crime for a reviewer to say "I'm not sure" about something he's not sure about. I'd certainly prefer that to him saying he's sure when he isn't.

Parts can easily be "DX9 compliant" without supporting the rigid specification of "DX9" in hardware. So the person who talks about "DX9 compliancy" and means rigid adherence to the spec in hardware isn't talking about the same thing as the person who talks about "DX9 compliancy" and is satisfied with less.
 
Just a hint, Demalion. If you really want people to read everything you write, write less.

I wrote a total of 1 sentence prior to your first response of 8 paragraphs showing me how I am in trying to prevent this person from holding an opinion.

I then responded with two sentences, explaining how I came to my conclusion (yes, I know, which contradicts your previous essay), again with a return volley of 10 paragraphs, explaining how a) I didn't read what you said, and b) how I am wrong again.

Then, I further clarified why I expressed what I did with a simple 4 sentences, including a plea to you to avoid another back and forth, and devolvement into a semantics debate, and once again another 9 paragraphs telling me about how avoiding discussing with you is non-productive, and again, how I can't grasp the english language, and how I'm wrong.

The truth of the matter is, your discussion style ends up in people wanting to avoid discussion with you precisely because such discussion is futile. Its no wonder nobody gets what you're trying to say, as you deconstruct any intended meaning out of sentences and bury it in semantics. There's only so many times I can restate my opinion and have it torn apart without simply getting frustrated.

/ignores oncoming 10 paragraph response on how I missed the point.
 
DaveBaumann said:
The specifications of DX9 call for a minimum FP support of FP24 in the pixel shader - NVIDIA's WHQL drivers for 5200 only provide support for FP16.
That doesn't answer my qestion. Is it or not qualified by MS to be called DX9 card (ie compliant), and are the drivers WHQL?

If they are, then they are. MS is the only one to be able to judge.
 
WaltC said:
Evildeus said:
Is it or is it not DX9 compliant? Are or are not the drivers WHQL? The responses to these are facts, anything else is opinion.

As I said above, "DX9 compliant" can mean many things...;) It's far from opinion that although the GFFX 5200 and the R350 can both be said to be "DX9 compliant" they each offer very different hardware support for DX9, both in type and efficacy.

Come on, would you prefer the guy to say "It is not a DX9 part" when he's not sure about that, or would you prefer him to say "It probably isn't" when he's not entirely sure? I think the latter is the only honest answer if he's not sure. Last I checked it is not a crime for a reviewer to say "I'm not sure" about something he's not sure about. I'd certainly prefer that to him saying he's sure when he isn't.

Parts can easily be "DX9 compliant" without supporting the rigid specification of "DX9" in hardware. So the person who talks about "DX9 compliancy" and means rigid adherence to the spec in hardware isn't talking about the same thing as the person who talks about "DX9 compliancy" and is satisfied with less.
Well sorry, but it's not what you think, or what Nv think, or even god, it's what MS think, that is important. They are or aren't DX9 compliant (labelled by MS), there's no half path. Anything else is irrelevent.
 
DaveBaumann said:
Russ, as I said earlier we know 5200/5600 and 5800 all run at FP16 precision in the WHQL certified drivers, which is not to DX9 specification - I'll wager that the proviso there was that subsequent parts would ensure FP24/32 compliancy (as 5900 does - afterall, if they could get away with FP16 with the rest you'd think they would want to with 5900 as well, as its faster). I think fog is borked as well. When I get some time I'll run DCT over it and see what crops up.
Can you run DX9 Caps Viewer and see what's available and what isn't? Are these waivers all precision-related? How is fog broken?
 
Evildeus said:
That doesn't answer my qestion. Is it or not qualified by MS to be called DX9 card (ie compliant), and are the drivers WHQL?

Well, it does answer your question since MS have stated what the specification is, and it evidently hasn't met it.

However, the 44.03 drivers are WHQL, but whether that includes a statement saying these are DX9 compliant is a separate question - if it does then it must have been achieved via the use of a waiver.
 
In regards to the quite silly debate on the meaning of the word "properly": I agree that, in certain contexts, the ability to run DX9 "properly" can be sensibly construed as requiring a certain level of performance.

However, the question whether a card belongs in the DX9 category in marketshare breakdowns is not such a context.

Having said that: If the 5200 doesn't have FP32 hardware, then I would generally not categorize it as a DX9 part. If it does, but there are no drivers which expose the functionality for DX9 (correct me if I'm wrong, but this appears to be the case here), then I'd classify it as a DX9 part without fully-DX9-compatible drivers.

It is not unusual for a part to ship with drivers that are incompatible with the spec in some way or another; this isn't typically taken to mean that the part itself isn't compatible with the spec. Of course, if Nvidia never intends to (or never will) produce a fully DX9 compatible driver for the 5200 then perhaps one might begin to question its status as a DX9 part, but I don't believe anyone is suggesting that's the case.

Finally, on the slightly related debate about whether it is better for a cheap card to be extra-slow with an up-to-date featureset, or only rather slow and a generation behind featureset: It depends. Both types of cards have a place in the market. Personally, were I forced to spend ~$80 on an Nvidia card today, I'd probably go for a Ti4200 or GF3 over a 5200. Although I can imagine cases where a 5200 would fill the consumer's needs better.

The assertion (in the thread Wavey referenced) that the up-to-date featureset doesn't matter anyways, because such a card (e.g. the 5200) will be so slow as to require its own path anyways, is a straw man, however. It may in fact be the case that such a card will require its own path. But it is not necessarily true that the best way to make such a path is to include absolutely no DX9-only functionality and include the full workload of DX7/8 functionality. Perhaps the best IQ bang for the performance buck would be found by including (perhaps sparingly) some of the DX9 shaders, or at least cut-down (but still PS/VS 2.0) versions thereof, while easing the workload in other ways--cutting down on geometry or texture size, for instance.

Moreover, not all games have similar characteristics to the first-person shooter types that typically get benchmarked. There are plenty of games where GPU performance is not critical and yet where DX9 effects might be made use of.

Or, in a couple years time, non-game applications, for that matter. Yes, it does sound terribly implausible, but it just might come to pass. Oh, here's one such application now: Microsoft Windows. Ever heard of it?
 
Dave H said:
In regards to the quite silly debate on the meaning of the word "properly": I agree that, in certain contexts, the ability to run DX9 "properly" can be sensibly construed as requiring a certain level of performance.

Then the "debate" wasn't "silly", since your initial commentary excluded recognition of that. That was the point of my reply.
However, the question whether a card belongs in the DX9 category in marketshare breakdowns is not such a context.

Well, if you recognize the above, how are you arriving at the conclusion that this was the context in which Jerky's post should be discussed? I ask, because he did not say the 5200 didn't belong in the DX 9 category, nor did he say it was not DX 9 compliant...in fact, I'd propose that usage such as "...even though NVIDIA dominates the DX9 market by share..." serves to indicate the exact opposite of such. What he did say was that it probably didn't run DX 9 properly, as you were just agreeing was sensible.

I'm not saying he has to use exact words, I'm saying I don't see a reasonable interpretation of his words that indicates other than the opposite (that his usage of "properly" which was quoted in your initial post included performance, and not DX 9 compliance, as you originally asserted, or market segment, as you assert now).

Having said that: If the 5200 doesn't have FP32 hardware, then I would generally not categorize it as a DX9 part.

Actually, if the waiver system Wavey is describing allows the label "DX 9 compliance" for partial "non compliance", according to Microsoft, then this is "not a matter of opinion", as you proposed. That phrase is a specific marketing phrase, AFAIK, which is the way I thought you were proposing it.
We'd still have such phrases as "running DX 9 properly" available in case that happens, though recognition of that is hindered by labelling pointing such things out as "quite silly". :-?
Whether "DX 9 compliance" remains meaningful outside of marketing, if "DX 9 compliance" diverged from the technical DX 9 minimum as (hypothetically, AFAIK) proposed above, is another discussion entirely.

...skipping various commentary that I do not take issue with...

On another topic of discussion...
(As always, if you feel my quote selection distorts, indicate what key points I omitted and I will modify my address).
...
The assertion (in the thread Wavey referenced) that the up-to-date featureset doesn't matter anyways, because such a card (e.g. the 5200) will be so slow as to require its own path anyways, is a straw man, however.

A straw man is used to preclude discussion away from a topic based on a false premise of equivalency. Did it really propose a false premise? Read on for a moment.

It may in fact be the case that such a card will require its own path. But it is not necessarily true that the best way to make such a path is to include absolutely no DX9-only functionality and include the full workload of DX7/8 functionality.
The way I understood what was being proposed was that having to create a unique "limited" PS 2.0 shader for the 5200 means that there is a separate "full" PS 2.0 shader being used that is not inclusive of the 5200. What seems reasonable (but not an absolute that can be proposed to exclude that 5200 could be using PS 2.0 level functionality) is that this short PS 2.0 shader might not offer an advantage over PS 1.3 or PS 1.4 on the 5200, and therefore might not exist. This proposal cannot ignore details like what shader instructions and capabilities are common or unique amongst them.

I don't see how this (if it is what was said) is a straw man.

Another way of stating this: that this divergence existing can free up "PS 2.0" to evolve independently of "PS 2.0 that runs well on the 5200". Since you agree with this divergence in your discussion here, I am presuming you are saying something else is said in what you refer to...if so, a quick pointer as to what I missed or misread would likely have me agreeing with you concerning that flaw in it.

...other discussion that depends simply on the 5200 being DX 9 compliant, which I also tend to agree with...
 
RussSchultz said:
Just a hint, Demalion. If you really want people to read everything you write, write less.

Just a hint, Russ. If you are going to write something, say something coherent with the words.

Are the "obligatory" put-downs out of the way, now?

I wrote a total of 1 sentence prior to your first response of 8 paragraphs showing me how I am in trying to prevent this person from holding an opinion.

Oh, you used less words than me. I guess you were right after all.

This would be a straw man, if an example was needed, BTW.

...Russ ignores every point I make, and counts words, while spending a lot of (countable) words doing so, which seems sort of contradictory to even the straw man "point" being put forth...

The truth of the matter is, your discussion style ends up in people wanting to avoid discussion with you precisely because such discussion is futile.

Here's a thought Russ...maybe you are wrong. Do you think if you don't honestly admit that, that you can't be? Maybe your frustration in not being able to respond to my points is the product of your own conversation, and not mine. Despite your complaints of word count in my posts, those words are absolutely rife with specific items that can be logically addressed to tackle any point I raise. My being able to discuss something wrong with what you stated in such detail might simply be because...there is something wrong with what you stated.

As long as you deny that possibility in a discussion, or alternatively, as an example, propose that your admitting that you are wrong is only an insencere statement of appeasement for someone rude enough not to let you have your way with being right when you want to be...discussions with you are fruitless.
But how is that my fault?

Its no wonder nobody gets what you're trying to say, as you deconstruct any intended meaning out of sentences and bury it in semantics.

Russ, some day you will stop saying things fall "up", and I can stop discussing the definition of up and down with you. Request: please stop saying up is down.

Since I'm not literally proposing you said up is down, hopefully I don't have to specify the exact conditions under discussion for this example as a product of your "throwing" your perception of what semantics are for "back at me"?

There's only so many times I can restate my opinion and have it torn apart without simply getting frustrated.

Russ, I'm sorry, but there seems to be no way to get around this: AFAICS, your opinion deserves to be torn apart. Not simply because I disagree with it, but because it doesn't stand. I disagree with it because it doesn't stand...if I am wrong in that assertion, hold a discussion to show it instead of spending ALL your time complaining about my disagreement and simply restating it because you don't like that disagreement! :-?

You are perfectly free to establish otherwise, but you do have to establish it. Those words you complain about are my recognition that I have to do the same, and I do. Here is a revolutianary thought: if you don't see an error in my reasoning and my description of flaws in yours, perhaps it is because there is not such an error...if you do see such an error, discuss it instead of what you are doing now.

I end up spending an extreme amount of words explaining things multiple times due to your simply seeking a way to avoid doing this. An ever expanding list of things you tack on to avoid discussing what I've already proposed.

3 paragraphs of something I've said before, and seems pretty glaringly obvious to me. Simply dismissing them as wrong or useless, as you've done before by word or deed, does not make them so...try addressing them in a different way, maybe even by taking them to heart.

/ignores oncoming 10 paragraph response on how I missed the point.

"Plug your ears" and go 'nyah, nyah, nyah', eh? :-?
 
Back
Top