NVIDIA claims top DX9 GPU marketshare spot.

RussSchultz said:
I'm saying your original statement denied that properly could include performance by agreeing with someone who proposed that it could not, and by attacking me when I said "properly can include performance" to that person.
Your responses also propose that "You weren't disagreeing with saying properly could include performance" in your reply to me, and you ignored my question about what you thought I was saying when you...disagreed with me.
If that is the bone of your contention, you have seriously wasted a lot of my time and yours.

Of course, you couldn't be the source of wasted time?

Again, in plain english for you:

Me stating an opposing conclusion with my reasoning does not equate to me denying that any other conclusion is possible.

Russ, why do you keep insisting on ignoring my specific discussion of your quoted text over and over? You either propose something I've already addressed, again, or go off on a confused tangent.

Consider the following:
Somebody else: "JoeSmith would never drive drunk."

Is this supposed to be Dave H? As close I can get in your extremely flawed analogy model, what Dave H said was more like "JoeSmith crashed due to driving drunk, and it is not a matter of opinion", after seeing a beer bottle next to the car.

You: "I think he was drunk and drove off the side of the cliff after dosing off. Look, there's a broken beer bottle in the car"
No, what I said was more like "There is a dead deer here, and you can't rule out that he hit it and ended up crashing due to that."

Me: "I never knew Joe to drink and drive. There's a dead deer here on the side of the road, I think its safe to assume he lost control after hitting the animal"
What you said to "me" was more like "No, it is safe to assume that he was driving drunk because that's what a beer bottle in the car means. And if his being drunk is a matter of opinion, then could I say I'm not drunk after I've had a half a bottle of vodka?" Except that what you actually did propose in the first part doesn't even make that much sense, because there wasn't a camera to record Joe drinking or that the beer bottle was or was not in the car before the accident and not just lying where the car crashed, but there are English rules and dictionaries to apply to an exact record of what was said in the actual conversation.

If you contradict me when I am saying to "Someone else" that it is possible that he hit the deer and then lost control, and propose a ridiculous example to disagree with the premise I propose that it is a matter of opinion whether he was drunk, you are maintaining "Someone else"'s statement.

This is a (bastardized, sloppy, ill-suited, but the best I could fit into your framework) represenation of exactly what I've proposed to you numerous times. If this is news to you, that is your fault. If you can't see it, disagreeing and taking the conversation down another erroneous path contrary to my statmements, is you wasting time, not me.

I don't recommend you argue against this example, and instead deal with...what I actually said.

When I ask you to stop proposing flawed representations of my point and ignoring the discussions I already provided, of course it is perfectly natural to ignore the discussion I already provided, and propose another flawed representation, and this doesn't display a dedication to arguing by ego at all. :oops:

Your analogy ignores exactly the same things as you've ignored before, that there is a record of what actually happened and an objective standard for interpreting it (English), and doesn't even come close to representing my actual statements even aside from that. :-? It is also in an unrelated context of an unwitnessed physical event involving limited information being discussed after the fact, in contrast to a discussion that is recorded in exact detail and is/can continue to be witnessed and referred to with specifics. Perhaps your continuing to ignore those details makes these situations seem the same to you?

Do you see me denying that it is possible that Joe, our hapless driver, was driving drunk? No. I'm simply asserting my view on the situation.

Well it is nice when you can win an argument against yourself. My actual argument remains unaddressed. :-?. Look, analogies and reinterpretations can be a valuable tool for communication, but not when used over and over to completely ignore what the other person is saying.

If you don't understand that, I don't know what I can quote or illustrate to help you. And, we'll just call this conversation over.

Does that mean you'll address the discussion I provided about PS 2.0, or has perpetuating this discussion made it convenient to ignore the questions and discussion I propose? Has Andy's commentary jogged any new thoughts with regard to it? Do you still think no competent reviewer can be in doubt about the 5200 "running DX 9 properly"?
 
If you contradict me when I am saying to "Someone else" that it is possible ... you are maintaining "Someone else"'s statement.

As I suspected. It is impossible to continue.

If you cannot grasp that my initial response was a differing conclusion than yours and the attached statement was support for that conclusion, there's no point in me continuing.

Feel glad. You won. Completely and utterly. 8 pages of back and forth for a point that was completely and utterly meaningless.

If you didn't notice, on page 3, I decided properly wasn't worth arguing over so I stated so and couched my further response in the term that "properly" included performance and included detailed reasoning over my stance. But no, you insist on continuing to beat the dead horse. Well, sometimes it is fun to beat a dead horse so I joined you and I kept it up for another 8 pages because, quite frankly, it was fun to irritate you with the added bonus of demonstrating to the forum that you can't accept common communication mechanisms at the same time that you excel in writing essays over the smallest of topics using the most daunting of language.

Its amazing to me that you can't see the irony of you lambasting me for refusing the possibility of opinion in properly, while at the same time refusing to accept the possibility that my statement was exactly as I said it was: me stating my take on what properly meant in that context. Even with me here to clarify what I meant and intended, you were willing to continue to interject your own thoughts contrary to mine.

And, yes, I really was being honest and not faceteous at all when I say you write too much and don't get to the point. Though maybe the problem is you have too many points, or maybe you attempt to pre-emptively dismiss points and obscure the ones that are important inadvertantly in the process.

And yes, intentionally or not, your discourse style is very insulting. It comes across as intellectual slight after intellectual slight. Quite frankly, you do not write in a manner that is easy to read or follow and then turn around and lambast people for missing your very subtle points. Which even after the lambasting don't seem any more apparent then before.

So, from now on, you get a "two response" limit from me. If you don't catch my drift, or I don't catch your drift by then, its "agree to disagree" time.

As for the 2.0 discussion: I think I've shared enough information that outlines how you can and should achieve greater certainty than "probably" in "properly" statements. Look at the discussion between Andy and I (I know you read it) to see how you can get there. You're a smart guy, fill in the blanks if they're missing.

Once again (for posterity), my providing evidence that contradicts your conclusion does not mean I am rejecting it as an invalid conclusion.

At least not in my world.
 
Russ, demalion... please continue this in private. The two of you don't seem to be able to see eye-to-eye on whatever-tf you two are arguing about and I (and I presume 99% of those here) don't care to read any of your posts pertaining to your argument with each other.

This is the way I see it and I hope this is cool with the two of you.

And my sort-of-on-topic contribution to this thread :

I think anyone can make a piece of next-gen hardware that runs next-gen software at 5fps. The 5200 performs very poorly, not only with DX9 features but also in the toughest DX7 and DX8 environments. Its very existence, and the fact that it is the DX9 hardware that has shipped the most and hence likely to be in more folks' machines than other DX9 hardware, means that developers will likely have to take it into account when they create their "DX9 games" (and no, I haven't done such a survey with developers because I know I won't get too many replies due to diplomacy and politics). Hopefully there will be more developers who can afford to take the risk of long development time with their games like Valve with HL2 and games can look extremely cool with lots of DX9 effects while also being able to look decent with DX7 effects. IMO however, there aren't too many such developers.
 
RussSchultz said:
...
As for the 2.0 discussion: I think I've shared enough information that outlines how you can and should achieve greater certainty than "probably" in "properly" statements. Look at the discussion between Andy and I (I know you read it) to see how you can get there. You're a smart guy, fill in the blanks if they're missing.

But my understanding is that Andy said "probably" with regards to the 5200...where is this outline of how you can and should achieve greater certainty than "probably" in "properly" statements? This is the question I was proposing to you about 6 or 7 pages ago, and I don't see an answer.

I do see Andy's answers, and your statements and questions, but what he states doesn't seem to at all indicate what you are saying it does AFAICS. Seems to be filled with conditionals that would depend entirely on the specific game and a person's criteria for performance.

I assume, since you point to this discussion, that you are saying that what should be said by a competent reviewer is that the 5200 "certainly" can't run DX 9 "properly", since your criteria for "truly acceptably" is 60 fps or higher? That would answer one of my questions, then.
What I don't understand, since we've established that "properly" can indeed include performance criteria, is why a reviewer posting in a forum would have to be incompetent not to both have decided on a threshold and known whether games utilizing PS 2.0 that he hasn't seen yet would run above it or below it? I don't even see Andy (who I'd propose is certainly competent with regard to the technical side of reviewing) doing this anywhere.

Maybe we can focus on this, and how I'm misunderstanding (highlighted as an aid):
andypski said:
RussSchultz said:
Well, when I said "truly acceptable" I meant 60+ FPS, or 180+ if you're a quakehead, particularly for twitch type games. I doubt that even the 9800 pro could do that for every pixel being shaded with PS2.0. (Assuming the shader was something difficult enough to require PS2.0)
Fair enough - certainly you are right if every pixel is shaded with very long PS2.0 shaders, however a mix of short and long ones should give very acceptable performance, depending on resolution.

I think he sounds very competent, as usual, but I don't see the greater certainty than "probably" evident in this discourse.
Perhaps you meant the mathematical calculations, but I don't see them being proposed outside the above discussion, but as examples of one possibility for the sake of illustration.
 
Hmm...didn't see your post until now, Rev. This discussion is the one back on topic I've been asking for, and it is also open for andy to clarify my question. If it heads down the same road as the other branch, I'll stop replying.
 
demalion said:
But my understanding is that Andy said "probably" with regards to the 5200...where is this outline of how you can and should achieve greater certainty than "probably" in "properly" statements? This is the question I was proposing to you about 6 or 7 pages ago, and I don't see an answer.

I don't really want to get dragged into a long debate on the DX9 merits or otherwise of particular pieces of hardware. I am hardly going to come across as unbiased in these cases whether my opinions are justified or not.

When considering environments as complex as modern 3d application software it is a bit naive to make absolutely definitive statements when predicting performance in future applications, while at the same time it is certainly possible to make general statements about performance expectations, and provide information as to why this sort of performance level is expected.

I assume, since you point to this discussion, that you are saying that what should be said by a competent reviewer is that the 5200 "certainly" can't run DX 9 "properly", since your criteria for "truly acceptably" is 60 fps or higher? That would answer one of my questions, then.

Russ further makes the distinction about the differences in acceptable frame rates between certain games, and certain game players -

RussSchultz said:
Well, when I said "truly acceptable" I meant 60+ FPS, or 180+ if you're a quakehead, particularly for twitch type games

Given these criteria (which are pretty harsh - there are game types where lower rates are generally acceptable), low end parts are going to be restricted to low resolutions like 640x480 in order to attain "truly acceptable" frame rates when running environments that are composed using complex pixel shaders. The 'twitch' framerates that he talks about will simply not be achievable without dialling down shader complexity.

I don't see the greater certainty than "probably" evident in this discourse. Perhaps you meant the mathematical calculations, but I don't see them being proposed outside the above discussion, but as examples of one possibility for the sake of illustration.

The calculations were there to support the basic performance conclusions - they give a common understanding of the baseline of performance that can be expected. By establishing a reasonable lower limit it is then possible to have specific discussions about performance.

Once you go beyond purely mathematical cases it is difficult to make accurate predictions.

However, this having been said, provided that you have an in-depth understanding of current applications and hardware and have done extensive analysis of their respective performance bottlenecks then it does become possible to make strong and well supported statements about performance expectations for future applications. Engineers at IHVs have to do this sort of thing all the time when making design decisions for future hardware in order to correctly target the appropriate performance, and then when the hardware becomes available we see whether our experience allowed us to make the right choices.
 
andypski said:
demalion said:
But my understanding is that Andy said "probably" with regards to the 5200...where is this outline of how you can and should achieve greater certainty than "probably" in "properly" statements? This is the question I was proposing to you about 6 or 7 pages ago, and I don't see an answer.

I don't really want to get dragged into a long debate on the DX9 merits or otherwise of particular pieces of hardware. I am hardly going to come across as unbiased in these cases whether my opinions are justified or not.

When considering environments as complex as modern 3d application software it is a bit naive to make absolutely definitive statements when predicting performance in future applications, while at the same time it is certainly possible to make general statements about performance expectations, and provide information as to why this sort of performance level is expected.

I understand and agree with both points, and the latter is exactly what I was pursuing with the latter half of this post, and what I do believe Jerky established directly with "probably can't run DX 9 properly" (even though he said that in a forum post, and not a review).
I certainly agree that the 5200 can be definitevely said to fail the 60 fps criteria for the PS 2.0 shaders examined, and I was maintaining that it would even fail the 30 fps criteria at 640x400, even when other DX 9 cards might run above it.

Now that we all agree on that, including Jerky (this is based on my considering "probably can't run DX 9 properly" to exactly to fit "to make general statements about performance expectations, and provide information as to why this sort of performance level is expected."), and that to satisfy Russ, Jerky was supposed to say "The 5200 definitely cannot run DX 9 properly" by Russ's criteria and the information available to Jerky from his review:

Russ, why would Jerky be incompetent to recognize other criteria than your 60 fps criteria for "twitch games", such that his behavior would have to fit andy's first description to anyone who thought differently than you? There are other game types, shaders, opinions, and he wasn't just talking to you.

Andy, somewhat related to this discussion, what is the minimum resolution for DX 9 rendering? 320x200?

Russ further makes the distinction about the differences in acceptable frame rates between certain games, and certain game players -

RussSchultz said:
Well, when I said "truly acceptable" I meant 60+ FPS, or 180+ if you're a quakehead, particularly for twitch type games

Given these criteria (which are pretty harsh - there are game types where lower rates are generally acceptable), low end parts are going to be restricted to low resolutions like 640x480 in order to attain "truly acceptable" frame rates when running environments that are composed using complex pixel shaders. The 'twitch' framerates that he talks about will simply not be achievable without dialling down shader complexity.

Agreed, and thank you for getting answers for some of my questions from him.

I don't see the greater certainty than "probably" evident in this discourse. Perhaps you meant the mathematical calculations, but I don't see them being proposed outside the above discussion, but as examples of one possibility for the sake of illustration.

The calculations were there to support the basic performance conclusions - they give a common understanding of the baseline of performance that can be expected. By establishing a reasonable lower limit it is then possible to have specific discussions about performance.

Once you go beyond purely mathematical cases it is difficult to make accurate predictions.

Agreed.

However, this having been said, provided that you have an in-depth understanding of current applications and hardware and have done extensive analysis of their respective performance bottlenecks then it does become possible to make strong and well supported statements about performance expectations for future applications.

Out of curiosity, how strong a statement could you make about the 9200 or the IGP chip and the shaders they'll be running, after you've picked the minimum fps and resolution criteria as you stipulated? (EDIT: this might be unclear...what I mean is that with the capabilities of PS 1.4, and their limited transistor count, the chips seem to be "closer to the edge"...this relates to the minimum DX3D resolution issue).
And how do you represent the way things like texture usage as a bandwidth bottleneck can vary so widely in your consideration of performance in future games? I personally don't know how to narrow this down reasonably to consider how fillrate might be hindered. There doesn't seem to be a clear consideration like resolution that I know can represent the amount of texture usage per pixel as there seems to be to calculate fillrate.

Engineers at IHVs have to do this sort of thing all the time when making design decisions for future hardware in order to correctly target the appropriate performance, and then when the hardware becomes available we see whether our experience allowed us to make the right choices.

Heh, reviewers need to know as much as Engineers to be competent? I don't exactly disapprove of that idea. :p
 
demalion said:
Andy, somewhat related to this discussion, what is the minimum resolution for DX 9 rendering? 320x200?

There is no minimum resolution restriction that I am aware of in DX9. Typically the lowest screen resolution exported by drivers is 320x200, but developers routinely render into lower resolution buffers such as texture maps, and theoretically there is nothing to stop you doing this and then stretching it up to fill the screen.

Certain hardware could have restrictions on the lowest rendering resolution for surfaces.

Out of curiosity, how strong a statement could you make about the 9200 or the IGP chip and the shaders they'll be running, after you've picked the minimum fps and resolution criteria as you stipulated? (EDIT: this might be unclear...what I mean is that with the capabilities of PS 1.4, the cards seem to be "closer to the edge").

I'm not really sure what you're asking here - I guess you want to know how capable the architectures are of dealing with long shaders? I wouldn't really make strong statements about either because I have not had much direct involvement with these chips. The 8500 and its more recent cousins do have highly capable shading engines that are well balanced in terms of program length to their overall performance.

And how do you represent the way things like texture usage as a bandwidth bottleneck can vary so widely in your consideration of performance in future games? I personally don't know how to narrow this down reasonably to consider how fillrate might be hindered. There doesn't seem to be a clear consideration like resolution that I know can represent the amount of texture usage per pixel as there seems to be to calculate fillrate.

With difficulty ;). I can't give you a magic formula - we work closely with developers so that we are aware of the types of workload they are producing. Given a particular application we have to look at aspects like the ratio of texture lookups to ALU operations, size and type of textures etc. As with anything else a lot of it comes down to experience, and there's still always room for surprises when developers do things in ways we haven't necessarily considered.
 
I should have been clearer: the minimum full screen resolution supported for 3D display. If the answer is the same (EDIT: heh, that doesn't quite make sense since you mentioned screen resolution..."if my narrowing it down doesn't facilitate you adding more details"), you don't have to bother yourself to reply.

And your guess answered the other matter, thanks.
 
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.

Where did you hear that it runs only in fp16 mode?
(Obviously you don't "know" that it does since it does run fp32.)
 
above3d said:
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.

Where did you hear that it runs only in fp16 mode?
(Obviously you don't "know" that it does since it does run fp32.)
Ooooh, this should be fun! :D

BTW-I still want to know what happened to Joe...I'm getting worried. :cry:

Sayeth the married man with callouses... ;)
 
Back
Top