ATI tries to downplay SLI

[DaveBaumann]

Ben,

It is clearly evident for anyone who is remotely honest.

No, Ben, the filtering was fairly close to bilinear, and much more evident to me.

No, you had it pointed out to you repeatedly and did nothing.

I had no explaination at the time - I can't look for something if I don't know what I'm looking for or where to find it.

If you went out and did something really odd, like play a few games, you may be aware of the fact that well over 90% of games don't have AF as an option. Funny you aren't aware of this as it again has been pointed out to you repeatedly for years.

4 out of the 5 titles I'm currently using for benchmarking have it, 3 of 5 had it last year.

Argue with yourself over that one.

This is what MS are telling devs - usually different to the marketting peaks they try and sell the public.

Have your Saphire X800Pro review open right now, I'm not seeing the FarCry numbers. I saw them in the 6800 review, that was the only one I noticed them in(I don't bother to read each vendors individual board review though so maybe you have had it in two or three reviews?).

First time I've used it was in the GeCbue X800 PRO review and each of the (p)reviews subsequent to that have included it.

I can't find the DooM3 numbers on any of your reviews, where are they? You have a review that was posted after D3 launched too.

http://www.beyond3d.com/previews/nvidia/nv45/

And, not being a full time site it can take some for things to pipeline through, so what was posted was complete in terms of benchmarking before D3 was available over hear. UK also recieved Doom3 two weeks after the US did, even though NVIDIA requested them from Activision directly (in fact, my Activision copy was delivered a day after I actually purchased it!).

That's not what you told us Dave, you told everyone how critical they would be a year ago(you were saying it further back then that, a year ago is when they were supposed to be paramount). The combo PR team of ATi and B3D did a fabulous job burying a lot of people's heads in the sand in no small part thanks to your help. How many R300 and R350 core chips did you help move with your PR campaign stating things that anyone with honest insight knew was wrong?

Good lord, Ben, are you really that bitter and twisted about it all?

The fact of the matter is that it was important for a number of games then and it will continue to be important now, we used one gaming title that used DX9 because we could, and we expanded effort into bringing the benchmark to a title that was readily available to the public - many other sites, including this one, also adopted that benchmark - presumably they are all part of ATI's vast PR campaign?

Of course, you are neglecting that ATI's last generation was pretty handy all round in the first place, it wasn't a one trick pony.

And everyone of them knew that the first gen parts didn't have anywhere close to enough power to make shaders a viable serious option anytime soon, can you sit there and honestly expect anyone to believe that you didn't know this?

Far Cry, Halo, TR, HL2 are not serious options on that class of hardware? Sorry, but thats not the case - sure, you're not going to be playing at 1600x1200 in most cases, but that doesn't mean it can't be used, many people happiliy played these titles with the best available options for their boards quite happily long before X800 and 6800 came along. Had you mentioned the mid and low end, I'd be agreeing heartily with you.

The one time you come out hard for a new technology is the time when it fails utterly in making any meaningful market penetration for years. Push as hard for SM 3.0 as you did for SM 2.0 and you could claim some level of objectivity, but we know that isn't going to happen.

Most technologies have a fairly long gestation period for more than scratching the surface use, and as the entire industry now has PS2.0 as a baseline, thanks to Intel and 915, then this will be the baseline for numerous titles to come - its no coincidence that UnrealEngine 3's baseline is SM2.0 upwards. As for SM3.0 we'll be using it and pushing it as we can - we'll use the Far Cry patch when its finals and we'll be looking out for other things we can use; we're also in contact with developers asking if there is anything we can get from them.

 

[DaveBaumann]

DAPUNISHER,

Doom3 was readily available for nearly 2 weeks before the article was posted, so what gives?

As I just explain in the previous post, we are not full time at B3D, we all have "normal" lives and normal jobs, so (except for specialy NDA cases) there is a pipeline of work filtering through that often has somewhat of a lag to it with whats available in comparison to when we post it.

It was also the case that Doom3 was only available over here on the 13'th of August, which I believe was a week or two after the States (and just about every other part of the world, os so it seems!).

 

[nRollo]

BFG10K:

Also first you claimed the 9700 Pro was too slow to use any AF and AA but then when you picked up the slower 5800U you started using 4xAA and 4xAF on it. No matter which way you spin it ("I'm a collector", "I'm a good family man", "I've spent $3000 on ATi hardware") it doesn't change the fact that you are a troll and an nv-fanboy. Your comments have an unbelievable, almost child-like, bias to them.

Hmmm. What a mean spirited, hostile thing to say. Not long on manners, are you BFG10K?

Oh well. They say "imitation is the sincerest form of flattery" and I notice you posted a chart of benchmarks with your new card, and I've never seen you do that before.

 

[Shamrock]

Speaking of Downplaying...ATI did it AGAIN

http://theinquirer.net/?article=18119

Yeah, I know, I know, but read the article...ATI is failing the new DCT 5.2 tests. The only way they are getting WHQL is using the old test, but they can only do that for 90 days. They are on day 20 already. It's an interesting read, even for the Inquirer.

 

[Drayvn]

Originally posted by: Shamrock
Speaking of Downplaying...ATI did it AGAIN

http://theinquirer.net/?article=18119

Yeah, I know, I know, but read the article...ATI is failing the new DCT 5.2 tests. The only way they are getting WHQL is using the old test, but they can only do that for 90 days. They are on day 20 already. It's an interesting read, even for the Inquirer.

Why is that downplaying?

It just so happens that Microsoft has said that Geometry Instancing is for SM3 only, even tho an SM2 card can do it, and Microsoft wont certify it because its not SM3. And they are very strict about this, so in fact ATi is putting this as off, but has the option to be turn on.

 

[SickBeast]

Originally posted by: Shamrock
Speaking of Downplaying...ATI did it AGAIN

http://theinquirer.net/?article=18119

Yeah, I know, I know, but read the article...ATI is failing the new DCT 5.2 tests. The only way they are getting WHQL is using the old test, but they can only do that for 90 days. They are on day 20 already. It's an interesting read, even for the Inquirer.

It really doesn't seem very fair that ATi's geometry instancing does not seem to jive with Microsoft's certification program; the R9700Pro has supported GI since September 2002 whereas the first nVidia card to support the feature was the 6800U which was paper launched early this year. If Microsoft had implemented the feature in DX9 early on, people could have been enjoying GI for well over a year now, and it would have allowed game developers to work with it while creating games like HL2.

In any event, I'm wary about listening to the inquirer when it comes to the instancing feature; I read an article on their site last week which stated that only the NV40 cards support the feature. That site is all too often a sorry excuse for computer journalism.

 

[Drayvn]

Originally posted by: SickBeast

Originally posted by: Shamrock
Speaking of Downplaying...ATI did it AGAIN

http://theinquirer.net/?article=18119

Yeah, I know, I know, but read the article...ATI is failing the new DCT 5.2 tests. The only way they are getting WHQL is using the old test, but they can only do that for 90 days. They are on day 20 already. It's an interesting read, even for the Inquirer.

It really doesn't seem very fair that ATi's geometry instancing does not seem to jive with Microsoft's certification program; the R9700Pro has supported GI since September 2002 whereas the first nVidia card to support the feature was the 6800U which was paper launched early this year. If Microsoft had implemented the feature in DX9 early on, people could have been enjoying GI for well over a year now, and it would have allowed game developers to work with it while creating games like HL2.

In any event, I'm wary about listening to the inquirer when it comes to the instancing feature; I read an article on their site last week which stated that only the NV40 cards support the feature. That site is all too often a sorry excuse for computer journalism.

Lol, i didnt see that article, i cant believe they said that a week ago, since most of us techeads knew that ages ago...

 

[bunnyfubbles]

Originally posted by: Rollo
BFG10K:

Also first you claimed the 9700 Pro was too slow to use any AF and AA but then when you picked up the slower 5800U you started using 4xAA and 4xAF on it. No matter which way you spin it ("I'm a collector", "I'm a good family man", "I've spent $3000 on ATi hardware") it doesn't change the fact that you are a troll and an nv-fanboy. Your comments have an unbelievable, almost child-like, bias to them.

Hmmm. What a mean spirited, hostile thing to say. Not long on manners, are you BFG10K?

more like he's calling 'em as he sees 'em.

Oh well. They say "imitation is the sincerest form of flattery" and I notice you posted a chart of benchmarks with your new card, and I've never seen you do that before.

you were the first to do so? oh and they also say "correlation does not imply causation"

 

[trexpesto]

They should make it so you can plug 2 sli's into a third and use 4 mid range vid cards.....

 

[nRollo]

Originally posted by: SickBeast

Originally posted by: Shamrock
Speaking of Downplaying...ATI did it AGAIN

http://theinquirer.net/?article=18119

Yeah, I know, I know, but read the article...ATI is failing the new DCT 5.2 tests. The only way they are getting WHQL is using the old test, but they can only do that for 90 days. They are on day 20 already. It's an interesting read, even for the Inquirer.

It really doesn't seem very fair that ATi's geometry instancing does not seem to jive with Microsoft's certification program; the R9700Pro has supported GI since September 2002 whereas the first nVidia card to support the feature was the 6800U which was paper launched early this year. If Microsoft had implemented the feature in DX9 early on, people could have been enjoying GI for well over a year now, and it would have allowed game developers to work with it while creating games like HL2.

In any event, I'm wary about listening to the inquirer when it comes to the instancing feature; I read an article on their site last week which stated that only the NV40 cards support the feature. That site is all too often a sorry excuse for computer journalism.

nVidia has many SM3 cards in the hands of the public SickBeast. I own two myself. To say ATI has had this since 2002 and nVidia has only paper launched it is a gross exaggeration.

Technically, ATI doesn't have it. They have hardware that can be made to do it if the developer decides to spend the time to make it work. It will be interesting to see how many do.

 

[DaveBaumann]

Technically, ATI doesn't have it. They have hardware that can be made to do it if the developer decides to spend the time to make it work. It will be interesting to see how many do.

FYI, this is the procedure the developer goes through:

- Check for presence of "INST" FOURCC format
- If present, set POINTSIZE to the "INST" FOURCC code to enable the instancing path
- Once enabled, it will stay enabled until the device is destroyed or Reset is called (same thing from the driver's point of view)

Which is about 4 lines of code - once you've established its there, its called and supported exactly as 6800 or any other SM3.0 part is called.

 

[nRollo]

BFG10K:

quote:

--------------------------------------------------------------------------------
Interesting how all the cards in the benchmarks I link perform at substantially different levels on different settings if they're all totally cpu limited, isn't it BFG?
--------------------------------------------------------------------------------


The benchmarks you produced and linked to were largely 1024x768 and you claimed AF and AA didn't matter. Then slowly you started moving to 1024x768x4x4, 1024x768x8x4 and finally to 1280x960x8x4 when it became apparent that nVidia's 6800 series wasn't any faster than the NV3x series in the CPU limited settings you had been previously pimping.

Also first you claimed the 9700 Pro was too slow to use any AF and AA but then when you picked up the slower 5800U you started using 4xAA and 4xAF on it. No matter which way you spin it ("I'm a collector", "I'm a good family man", "I've spent $3000 on ATi hardware") it doesn't change the fact that you are a troll and an nv-fanboy. Your comments have an unbelievable, almost child-like, bias to them.

Of course, you didn't answer the question at allyou called me names instead and started rambling about the resolution I liked to play UT2003 at.:roll:

I'm "child-like"? Why don't you try to stay on topic and address the issue at hand? Let me help you BFG:

I replied to your statement

Those "benchmarks" of yours were nothing more than CPU tests and to claim otherwise is an insult to the benchmarking paradigm. So yes captain obvious, thank you for pointing out that a new video card doesn't actually change the CPU in your system. Or were you expecting that it would?

by responding

Interesting how all the cards in the benchmarks I link perform at substantially different levels on different settings if they're all totally cpu limited, isn't it BFG?

.

The question put to you ADD BFG is: if my benchmarks are cpu limited as you say, why does the performance on the cards substantially decrease with each level of resolution?

No one cares about what settings I used to run UT at, or that you think I'm a nVidia fanboy. (even though I've spent a ton more money and time on ATI cards than you :roll: )

Here's the benchmarks you state

The benchmarks you produced and linked to were largely 1024x768

http://www.anandtech.com/showdoc.aspx?i=1821

Gee, right here on this site, 10X7 4X8X, 12X104X8X, 16X12 4x8x. I use these in every response I make on every board where someone says "Teh 5800 suxorz! Teh 9700Pro ownz it's sorry azz".

So I say again to you BFG "I won't answer a question":

Why are there significantly different levels of performance at each setting if the cards are cpu limited at the settings on the benchmarks I post? Why do the different cards perform at different levels if the cpu is the limiting factor? Answer the questions?

Since 10X7 is only 1/3 of the benchmarks, why do you state the benches I linked were "mostly 10X7"? Do you understand "most"? 1/3 isn't most BFG.

Now go ahead and post how I'm a fanboy and you don't like me some more and keep on ignoring the facts presented to you, like you always have. :roll:

 

[BenSkywalker]

BFG-

AFAIK Halo doesn't even support AF or AA which makes it quite a useless benchmark. A lot of ATi's extra performance came from cranking the eye candy.

SplinterCell is still being used by B3D which also can't utilize AA, although you did hit on the important part- it is hard to make ATi look as good using the bench which is an important factor of course.

The R3xx series was sometimes 50% faster than NV3x boards in shader intensive titles.

Actually, there were no shader intensive titles when Dave started his big PR push, after that one of the two shader titles we saw inside of eighteen months was significantly faster on ATi hardware- even though it still performed horribly, looked poor, and was a horrificly awful embarassment to the gaming industry(TombRaider). The next title we saw that was very shader intensive was Halo, and there was close to parity in that one.

There are games right now that will not run on cards without shaders and the Unreal 3 will require SM 2.0 to run.

DooM3 is the only game I have that requires some level of shader functionality to play(and you could argue that calling register combiners shaders is stretching it)- which titles are you talking about? Unreal3 will also require FP32- R3x0 parts are out anyway.

Dave-

No, Ben, the filtering was fairly close to bilinear, and much more evident to me.

I had no explaination at the time - I can't look for something if I don't know what I'm looking for or where to find it.

You managed to track down what was going on with nV's filtering hack quite quickly without being told explicitly what was wrong- and another site did manage to track down exactly why ATi's filtering was so poor(not that you ever made mention of the increased aliasing in the first place).

4 out of the 5 titles I'm currently using for benchmarking have it, 3 of 5 had it last year.

Like I said, if you went out and did something really odd, like play some games, you may realize that the overwhelming majority of titles DO NOT have the option to enable AF. Should it surprise anyone that the devs who took the time to include benchmarks in their game are far more likely to have an option to enable AF?

This is what MS are telling devs - usually different to the marketting peaks they try and sell the public.

Ask ERP about that. You have the resources at your disposal to easily find out just how dishonest and overhyped console makers are to devs along with the public, there have been thousands of posts talking about it on your forums.

And, not being a full time site it can take some for things to pipeline through, so what was posted was complete in terms of benchmarking before D3 was available over hear.

Spend the extra five pounds and have it shipped.

Good lord, Ben, are you really that bitter and twisted about it all?

I didn't lie to people about what was coming, I just had to try and explain over and over why people were seeing the performance that they were when the titles they wanted to play hit.

The fact of the matter is that it was important for a number of games then and it will continue to be important now

Which games Dave? Here it is two years later and I can count them on one hand. That's real important? To which gamers?

The fact of the matter is that it was important for a number of games then and it will continue to be important now, we used one gaming title that used DX9 because we could, and we expanded effort into bringing the benchmark to a title that was readily available to the public - many other sites, including this one, also adopted that benchmark - presumably they are all part of ATI's vast PR campaign?

With your aid, yes; ATi was quite succesful in pulling off that PR win(and an impressive one I might add). The big difference between your take and other sites is that when a viable DX9 game came along they dropped that pathetic POS(in every sense of the word) you still use. TR:AoD was an embarassment to the gaming community Dave, it was so bad that the creators of the game lost their rites to handle the franchise ever again- it was that bad. The game runs extremely poorly on any R3x0 hardware(to put it kindly) and is easily embarassed in terms of end visuals by numerous DX7 class titles. And yet you still use it. It still runs real well on ATi hardware comparitively though- the only thing you can really say about the game.

Of course, you are neglecting that ATI's last generation was pretty handy all round in the first place, it wasn't a one trick pony.

Not really, I've been running a R9800Pro now for the last six months(unfortunately bought it @ $299 a couple of weeks before the prices dropped fast). I simply never was under the delusion that anything you said would be close to reality.

Far Cry, Halo, TR, HL2 are not serious options on that class of hardware? Sorry, but thats not the case - sure, you're not going to be playing at 1600x1200 in most cases, but that doesn't mean it can't be used, many people happiliy played these titles with the best available options for their boards quite happily long before X800 and 6800 came along.

Did you not hear that HL2 got delayed by a year or so? Not sure why you listed that one, the other three I own- they don't play very well on R3x0 hardware at all. Forget 1600x1200- they don't play well running 1024x768 with everything cranked(FC even struggles with 800x600 in many different locations).

Most technologies have a fairly long gestation period for more than scratching the surface use, and as the entire industry now has PS2.0 as a baseline

Where are all these PS2 titles you have been talking about for years? You mentioned again in your post the 'number of games'- what games? I can think of five, two years later. EMBM saw far greater support, as did Dot3, EMCM and static hard T&L- all features that you went through quite a bit of effort to downplay.

 

[nRollo]

Ben:

Which games Dave? Here it is two years later and I can count them on one hand. That's real important? To which gamers?

Where are all these PS2 titles you have been talking about for years? You mentioned again in your post the 'number of games'- what games? I can think of five, two years later.

You screwed up now Ben.

Didn't you see the post above where BFG10K implies there were lots of PS2 games out over the last two years?

He won't name them, but you can trust BFG, there were many!

Listen to BFG and Dave, Ben. PS2 was the determining factor in any 9800 vs 5900 choice last year, and continues to be in the X800 vs 6800 choice this year. Only ATIs feature set is necessary.

SLI is meaningless, you'd have to be crazy to want that additional performance. When you need it, ATI will give it to you.

All is well, buy another SM2 chip and go back to bed......


Sigh.

 

[nRollo]

Originally posted by: DaveBaumann

Technically, ATI doesn't have it. They have hardware that can be made to do it if the developer decides to spend the time to make it work. It will be interesting to see how many do.

FYI, this is the procedure the developer goes through:

- Check for presence of "INST" FOURCC format
- If present, set POINTSIZE to the "INST" FOURCC code to enable the instancing path
- Once enabled, it will stay enabled until the device is destroyed or Reset is called (same thing from the driver's point of view)

Which is about 4 lines of code - once you've established its there, its called and supported exactly as 6800 or any other SM3.0 part is called.

If that is true, it is good news for ATI owners. Source?

The only reason I ask is that articles I've read, even other than the Inquirer(cough) seem to imply that the coding to make ATIs geometry instancing work is more substantial than typing four lines and it's done.

 

[DaveBaumann]

Ben,

You managed to track down what was going on with nV's filtering hack quite quickly without being told explicitly what was wrong- and another site did manage to track down exactly why ATi's filtering was so poor(not that you ever made mention of the increased aliasing in the first place).

Again, Ben, thats because I saw it!

Like I said, if you went out and did something really odd, like play some games, you may realize that the overwhelming majority of titles DO NOT have the option to enable AF. Should it surprise anyone that the devs who took the time to include benchmarks in their game are far more likely to have an option to enable AF?

A.) As far as the benchmarking goes the majority of the titles I've been using for the past 12 months or greater have application control AF, hence I've been showing the performance with app AF in the reviews.

B.) It would appear to be the case that an increasing number of titles are using AF/AA controls in the control panel or through other settings - Far Cry, Doom3 and HL2 being examples (going from CS:S).

Ask ERP about that. You have the resources at your disposal to easily find out just how dishonest and overhyped console makers are to devs along with the public, there have been thousands of posts talking about it on your forums.

Yes, Ben, I do have the resources at my disposal, and these figures are from people that know the reality of the situation. It still doesn't alter the fact that clearly MS are expecting lots of shader use with their next console, as are Sony (although more on the Vertex end from what I hear) and as NVIDIA and ATI are preaching to PC developers.

Spend the extra five pounds and have it shipped.

Good grief, Ben, you hold grudges that bad you believe I must be biased because I didn't get Doom3 via alternative means!

I didn't lie to people about what was coming, I just had to try and explain over and over why people were seeing the performance that they were when the titles they wanted to play hit.

Neither did I Ben - I took benchmarks and posted numbers those benchmarks tested some specific elements and they also test current gaming titles, none of them were lies. The same testing procedures / format applies now and we'll continue to highlight the elements that the industry tells us its going.

And thats where the benchmarks we choose a derived from - the rest of the industry; we test what they say they are capable of which is why we use a range of tests. Both ATI and NVIDIA tell of of the importance of Shaders, so we test them (and include a wide range of shaders) they tell usof the importance of games, so we test them, we also use tests that highlight specific capabilities of pipelines such as Z culling and stencil fillrates - these tests don't favour one platform. Our tests have also been using GT2 from 3DMark for specific AF/AA testing, which given its Doom3 rendering properties favours NVIDIA hardware, surely you should be arguing that we're biased to NVIDIA for that?

The game runs extremely poorly on any R3x0 hardware(to put it kindly) and is easily embarassed in terms of end visuals by numerous DX7 class titles. And yet you still use it. It still runs real well on ATi hardware comparitively though- the only thing you can really say about the game.

Nope, what you can say is that it is still quite fill-rate limited even on the high end - there is no point in a site that is concerned primarily with 3D performance in filling up the reviews with CPU limited games. And for that matter, there hasn't been that many games that have decent bnchmarking tools in there that allow you to test without FRAPS.

And having said that, there are still sites using it it (FS for instance), presumably for the same reasons we do.

I simply never was under the delusion that anything you said would be close to reality.

Again, I ran some benchmarks and posted the numbers, thats all.

Did you not hear that HL2 got delayed by a year or so? Not sure why you listed that one, the other three I own- they don't play very well on R3x0 hardware at all. Forget 1600x1200- they don't play well running 1024x768 with everything cranked(FC even struggles with 800x600 in many different locations).

Well, HL2 is soon to be released, given the Steam preload, and the Stress Test indicates that 9800 will be well in the range of playability. As for the other titles - these have been available for a fair while before 6800's or X800's were available and I'm sure manay people managed a reasonable gaming experience before then.

You mentioned again in your post the 'number of games'- what games?

Not every games that uses it is immedatly obvious - things like Colin McRea Rally and Tiger Woods make reasonable use of shaders and they aren't ones that are necessarily commonly thought of. Any future titles that are based off the Source engine, Unreal Engine 3 or even Doom3 engine will almost certainly be using PS2.0 or greater effects - do you think all of these engines will be unpopular in the future?

The curious thing is, this little poinless tangent is in a thread about SLI - heavily shader bound games are going to be among the ones that will gain the most benefit from SLI. Its no coincidence that NVIDIA were posting the large performance increase numbers from UE3.

EMBM saw far greater support, as did Dot3, EMCM and static hard T&L- all features that you went through quite a bit of effort to downplay.

Sorry, Ben, you'll have to look elsewhere for downplaying of them. However, all their support came over time, as will support and requirements for shaders. Arguably, thanks to HLSL, the uptake of shaders in some quarters has been faster than other technologies introduce.

Having said that, there's really little point in continuing with this since, as I learn't time and again there is no concpet of a "discussion" with you - you have your opinions and they are immovable, regardless of whether there is reason for actions taken, nothing will change your opinion.

 

[DaveBaumann]

Rollo,

Listen to BFG and Dave, Ben. PS2 was the determining factor in any 9800 vs 5900 choice last year, and continues to be in the X800 vs 6800 choice this year. Only ATIs feature set is necessary.

Please, kindly do not put words in my mouth as I have said no such thing.

If that is true, it is good news for ATI owners. Source?

The only reason I ask is that articles I've read, even other than the Inquirer(cough) seem to imply that the coding to make ATIs geometry instancing work is more substantial than typing four lines and it's done.

The source is ATI's Dev rel. The inquirer doesn't know the details of how it is actually used - when I mailed Charlie and told him how it was activated his reply was "Thanks, I didn't know that".

 

[nRollo]

Originally posted by: DaveBaumann
Rollo,

<blockquote>quote:
<hr>Listen to BFG and Dave, Ben. PS2 was the determining factor in any 9800 vs 5900 choice last year, and continues to be in the X800 vs 6800 choice this year. Only ATIs feature set is necessary.<hr></blockquote>

Please, kindly do not put words in my mouth as I have said no such thing.

<blockquote>quote:
<hr>If that is true, it is good news for ATI owners. Source?

The only reason I ask is that articles I've read, even other than the Inquirer(cough) seem to imply that the coding to make ATIs geometry instancing work is more substantial than typing four lines and it's done.<hr></blockquote>

The source is ATI's Dev rel. The inquirer doesn't know the details of how it is actually used - when I mailed Charlie and told him how it was activated his reply was "Thanks, I didn't know that".

.

You have been a pox on the gaming community since 3dfx gamers Wavey! I'll get you!



Hmmm.

It does seem odd to me that the Far Cry 1.2 patch got recalled and Far Cry didn't have geometry instancing all along if it was as simple as adding 4 lines of code to the game?

Also, what you say seems to make it seem like no one should ever bother with SM3 at all, at least for geometry instancing, when all they have to do is add 4 lines of code to the game and every ATI card R300 up has geometry instancing?

It took Crytek a long time to get that 1.2 patch out, your statement makes it seem like SM3 is much harder to code for, rather than the "much easier" it's supposed to be?

Any thoughts Dave?

 

[SickBeast]

Originally posted by: Rollo
<blockquote>quote:
<hr><i>Originally posted by: <b>SickBeast</b></i><BR><blockquote>quote:
<hr><i>Originally posted by: <b>Shamrock</b></i><BR>Speaking of Downplaying...ATI did it AGAIN<BR><BR><a target=_blank class=ftalternatingbarlinklarge href="http://theinquirer.net/?articl...;BR><BR>Yeah, I know, I know, but read the article...ATI is failing the new DCT 5.2 tests. The only way they are getting WHQL is using the old test, but they can only do that for 90 days. They are on day 20 already. It's an interesting read, even for the Inquirer.<hr></blockquote><BR><BR>It really doesn't seem very fair that ATi's geometry instancing does not seem to jive with Microsoft's certification program; the R9700Pro has supported GI since September 2002 whereas the first nVidia card to support the feature was the 6800U which was paper launched early this year. If Microsoft had implemented the feature in DX9 early on, people could have been enjoying GI for well over a year now, and it would have allowed game developers to work with it while creating games like HL2.<BR><BR>In any event, I'm wary about listening to the inquirer when it comes to the instancing feature; I read an article on their site last week which stated that only the NV40 cards support the feature. That site is all too often a sorry excuse for computer journalism.<hr></blockquote><BR><BR>nVidia has many SM3 cards in the hands of the public SickBeast. I own two myself. To say ATI has had this since 2002 and nVidia has only paper launched it is a gross exaggeration.<BR><BR>Technically, ATI <b>doesn't</b> have it. They have hardware that can be made to do it if the developer decides to spend the time to make it work. It will be interesting to see how many do.

How was what I said a "gross exaggeration"? I don't know how to define the 6800U's launch date in any other way but to state that it was a paper launch this past spring. The 9700Pro was available in reasonable quantities immediately after launch IIRC. The fact of the matter is that the two cards were released approximately a year and a half from each other. This means that the R300 supported geometry instancing for a year and a half before nVidia had a single product which supported this feature. As far as I'm concerned this issue is cut and dry.

My point was that developers shouldn't have to spend time to make ATi's geometry instancing work; ATi came out with it first, and logic states that their implementation should be the one used in DirectX.

When you say ATi "doesn't have it", what exactly do you mean? You should be well aware that they support it fully in hardware, and their latest drivers make the feature available. The sole bottleneck seems to be Microsoft. If you read the article, the ATi rep stated that they intend to make GI a user selectable switch in the drivers. I can't see game developers leaving half of their user base in the dark when it will require very minimal work to implement ATi's GI feature.

I know for a fact that there are people out there playing Far Cry on 9700 cards with GI enabled. Sorry Rollo, it appears that for the time being ATi *does* have it.

 

[DaveBaumann]

It does seem odd to me that the Far Cry 1.2 patch got recalled and Far Cry didn't have geometry instancing all along if it was as simple as adding 4 lines of code to the game?

The Far Cry 1.2 (second attempt) patch had more than just Geometry instancing changes to it, I doubt the recall was related to that.

The First attempt at the 1.2 patch added Geometry instancing and a single pass lighting shader for SM3.0, but at that point ATI hadn't got a method for exposing instancing at that time - the patch was actually sent out to reviewers prior to DX9.0c being available anyway. The second patch almost certainly uses the same geometry instancing calls that it does for SM3.0/6800 but all that changes in the actual detection for whether you use it.

Under normal circumstances when you test for Geometry Instancing being present you would just ask if the device is VS3.0 capable as it is arbitrarily tied to VS3.0. If you were to do that on a 6800, obviously it return "Yes" hence you can use GI, however with an 9500-X800 it would say "no". What ATI have done is added an alternative for that test so the developer just test for VS3.0 or the "INST" FOURCC mode and then you can enable it - the calls in the rest of the code from then on can just be exactly the same as you would call it for any SM3.0 device.

A similar situation has previously occurred - you may remember some issues that cropped up with Valve and FSAA support on modern cards. What solved that was another feature, Centriod Sampling, that ATI's 9500-X800 range can support, but Centriod Sampling was ties to PS3.0. However, in the case of Centriod Sampling MS removed that particular restriction in DX9.0c and gave it it's own call, but haven't for Geometry Instancing (probably because ATI didn't lobby MS to move it out in time).

Also, what you say seems to make it seem like no one should ever bother with SM3 at all, at least for geometry instancing, when all they have to do is add 4 lines of code to the game and every ATI card R300 up has geometry instancing?

OK, here is what Microsoft tells developers:

http://msdn.microsoft.com/libr...MultipleInstances.asp

This technique requires a device that supports the 3_0 vertex shader model. This technique works with any programmable shader but not with the fixed function pipeline.

Geometry Instancing works on any programmable vertex shader (VS1.1, VS2.0, VS2.x, VS3.0) program - the only thing they have done is arbitrarily tied it to Vertex Shader 3.0, and thats what MS want the developer to test for when checking it. None of the vertex shader code needs to be VS3.0 to use instancing, they are just saying "we expect this to be enabled on SM3.0 devices", but that?s not the case with ATI's parts.

And, yes, this only pertains to Geometry Instancing.

It took Crytek a long time to get that 1.2 patch out, your statement makes it seem like SM3 is much harder to code for, rather than the "much easier" it's supposed to be?

First, we are only talking about how it is detected, you also have to add code actually use GI - the first version of the patch did that work for the 6800's, however once that was in it was only a matter changing the detection to also look for ATI's version, once that was in it just uses the same code was they already programmed for 6800's Geometry instancing.

Second - in both cases it wasn't just GI that was added. They made changes to their lighting shader for PS3.0, PS2b in the second patch, and I believe further optimisations for PS2.0 in general. I'm sure they made other changes as well.

 

[DaveBaumann]

Originally posted by: SickBeast
My point was that developers shouldn't have to spend time to make ATi's geometry instancing work; ATi came out with it first, and logic states that their implementation should be the one used in DirectX.

No offense, but I wouldn't say thats the most stable logic!

Whil, technically, they could have supported it from the off, they didn't implement it and push for it when they could have so they have this rather pants detection system. Having said that, the actual call, to use it should be no different between ATI and any other SM3.0 device.

 

[SickBeast]

Originally posted by: DaveBaumann

Originally posted by: SickBeast
My point was that developers shouldn't have to spend time to make ATi's geometry instancing work; ATi came out with it first, and logic states that their implementation should be the one used in DirectX.

No offense, but I wouldn't say thats the most stable logic!

I suppose one could argue that since nVidia went the full nine yards and supported SM3, their solution should be the one that is directly supported by Microsoft. It just doesn't seem fair to me. It would be like SSE working on the Athlon XP but not on the Pentium 3 simply because Microsoft decided to choose one over the other.

In any event, this will probably end up being a non-issue. As you said, the call to use the feature in software should be universal, as it appears to be in Far Cry with the 1.2 patch.

I'm quite certain that the inquirer is once again spreading FUD and we will see instancing support from all cards which support the feature, regardless of the manufacturer.

 

[jim1976]

Originally posted by: Rollo
Ben:<BR><blockquote>quote:
<hr>Which games Dave? Here it is two years later and I can count them on one hand. That's real important? To which gamers?<hr></blockquote><BR><BR><blockquote>quote:
<hr>Where are all these PS2 titles you have been talking about for years? You mentioned again in your post the 'number of games'- what games? I can think of five, two years later.<hr></blockquote><BR><BR>You screwed up now Ben. <BR><BR>Didn't you see the post above where BFG10K implies there were lots of PS2 games out over the last two years?<BR><BR>He won't name them, but you can trust BFG, there were <b>many</b>!<BR><BR>Listen to BFG and Dave, Ben. PS2 was the determining factor in any 9800 vs 5900 choice last year, and continues to be in the X800 vs 6800 choice this year. Only ATIs feature set is necessary. <BR><BR>SLI is meaningless, you'd have to be <b>crazy</b> to want that additional performance. When you need it, ATI will give it to you.<BR><BR>All is well, buy another SM2 chip and go back to bed......<BR><BR><BR>Sigh.

So let me get this straight.
Should we define "good" hardware everything that just falls in the category of "handling the job" ?
Fact : NV30 supported SM2.0 but was unable to perform anywhere near ATI did. But there weren't many SM2.0 games so far to utilize the competitive advantage of ATI.
Should we underestimate this advantage just because the gaming companies didn't respond immediately? I wouldn't think so.
Good hardware is always good hardware and everyone who's denying that confronting the "so what?" behavior is suspicious of biased opinions.
Example. The vast majority of gamers don't change video cards like shirts.
So one guy who chose 9700pro instead of a Ti4600 or a 5800 should be able to play HL2 in DX9 enviroment. That can be the case to some other DX9 games this year. He may not have the "great" framerates that a current gpu can offer him but he certainly has the competitive advantage to a NV3x user.
Bottom line. Even if this is not the case, distorting the truth is not the way to go IMO.
ATI had the competitive advantage with the R3xx line.
As Nvidia has it now with the NV4x line and the SM3.0 support. That's the reason I chose Nvidia this time over ATI.
And for some ppl that want the best product out there because they plan to use it for some additional time than the hardcore gamer does they may find it wiser to go the right way.
Speaking from the luxurious and comfortable seat of $$$ gamer you might forget that sometimes it's the little things that make the difference...

P.S: Some guys may argue that by the time these games will be here that give competitve advantage to an older gpu the framerates will be unplayable. Guess what? The gamer knows that already!! He just tries to do what's best for his interest.

 

[DaveBaumann]

Originally posted by: SickBeast
I'm quite certain that the inquirer is once again spreading FUD and we will see instancing support from all cards which support the feature, regardless of the manufacturer.

The main thrust of the Inquirers article is correct - there is now a specific DCT test that causes ATI issues with its instancing support. There certainly was some lack of understanding of some of the technilcal details, but given the Inq is somewhat of the Online Tech equivelent of Tabloid journalism, thats kinda what you'd expect.

In the long term, though, I would expect that we'll see it, or something similar, have specific support in DX9.0d if they do another update in 6 months time. In the meantime, it seems like ATI will provide a workable solution that will provide end users the facility of using it (if their dev rel is good enough to get developers to use it) whilst effectively circumventing the DCT issues.

 

[nRollo]

What's with the quotes, and this forum, today?

Jim:

For someone like me who considers the usable life of a card one year, the nV30/R300 choice made no difference.

For someone who keeps them two+, the 9700Pro was probably the better choice. It wasn't much of an advantage though, and still isn't.

Let's say I still had the 9700Pro and 5800Ultra and had to choose.

When I look over at my games, I see one clear victory 9700P (Far Cry). (although Painkiller put the hurts on the 5800U too, don't know how the 9700P would have fared.

So I have 1, maybe 2 games, that I would have noticed a big difference on. Woot?

So much text has been devoted to this, it's got to stop. It's like arguing the GTS vs V5 all over again. Let's stick to nV40/R420?