The real deal w/HL2

Balderdash · Sep 13, 2003

ATI utilizes a 8x1 pixel shader path, with one path at 8 bits. Nvidia, on the other hand, uses a 4x2 path with two paths each 4 bits wide. Currently, any game using PS 2.0 with the FX cards is only accessing shaders at 4x1, due to driver and DX9b limitations (we will see DX9c soon, mark my words) and so, the DX9 games and 45:23 driver is effectively ignoring the second PS 2.0 path.

The preview 51:75 driver alleviates this problem, enabling the full second path for use in the game as much as possible before any update to DX9 is implemented to allow true dual channels as intended by its design.

We see these HL2 benchmark results now because HL2 is seriously dependant on pixel shaders in their current form and it is singly responsible for the framerate discrepancies.

The fix coming with the Det.50 should bring the numbers in line with ATI's, and additionally, the updated DX9c from Microsoft will likely make the FX cards the winner once true dual channel shaders are implemented and dual channel benefits can be accessed.

The next incarnation of DX9 should include the ability to use simultaneous wait states for PS 2.0 textures in DX9 applications. This will greatly reduce the 'problem' shown in these 'benchmarks.' The DX9 SDK was built (without any hardware available mind you) to favor one long pipe (and thus currently favor the ATI 8x1 version) since each texture has to go through a myriad of call back and wait/check states and has a definite FIFO for all textures in the pipe the nV (4x2) pipe is crippled during these operations. With the next version of DX9 you'll see the included paired texture waits in the shader process allowing the nV to actually utilize the 4x2 pipe simultaneously instead of a defined FIFO for each.

EDITED for spelling and clarity...

coolsquirtle · Sep 13, 2003

let's just hope your right (sigh)

Proud Owner the Block Heater
120% nVidia Fanboy
PROUD OWNER OF THE GEFORCE FX 5900ULTRA <-- I wish this was me
I'd get a nVidia GeForce FX 5900Ultra... if THEY WOULD CHANGE THAT #()#@ HSF

fragglefart · Sep 13, 2003

Thats the best piece of news i have heard today...
i was thinking of sticking my 5900 Ultra on ebay and getting some Ati action...
Its ok Nvidia, i still love you, you run 3DSMax just fine...

............................................
Render times? You'll find me down the pub...

coolsquirtle · Sep 13, 2003

the thing is, nVidia still wastes all with its Quadros and nforce2/3s

Proud Owner the Block Heater
120% nVidia Fanboy
PROUD OWNER OF THE GEFORCE FX 5900ULTRA <-- I wish this was me
I'd get a nVidia GeForce FX 5900Ultra... if THEY WOULD CHANGE THAT #()#@ HSF

ufo_warviper · Sep 13, 2003

Mobo-wise for the Nforce stuff of course

;-)

My OS features preemptive multitasking, a fully interactive command line, & support for 640K of RAM!

jurians · Sep 13, 2003

hmm thats an interesting viewpoint or possible fact if its true.... thing is why didnt nvidia create drivers that would support dual channel in the first place. And it doesnt change the fact that the origonal direct x specifications did not list a dual channel configuration. I don't like the fact that microsoft and game companies have to cater to nvidia because they cant read specification. If they want to deviate from it then they should make drivers that can support the deviation from day 1 not 1 year later.

eden · Sep 13, 2003

You know, if nVidia gets such sweet treatment, ATi deserves it too, since they have not gotten a single damn optimization.

Wow, looks we have someone Dave will actually debate with!

Get yer popcorns out you freaks!

--
<A HREF="http://www.lochel.com/THGC/album.html" target="_new">Are you ugly and looking into showing your mug? Then the THGC Album is the right place for you!</A>

Balderdash · Sep 13, 2003

hmm thats an interesting viewpoint or possible fact if its true.... thing is why didnt nvidia create drivers that would support dual channel in the first place. And it doesnt change the fact that the origonal direct x specifications did not list a dual channel configuration. I don't like the fact that microsoft and game companies have to cater to nvidia because they cant read specification. If they want to deviate from it then they should make drivers that can support the deviation from day 1 not 1 year later.

I'm not going to defend nvidia's decision to implement 4x2 archetecture over a more standard single channel mode, it seems reasonable to think that their mobo chipset driver with dual channels was probably their motivation in trying it. But with that in mind, wouldn't it be a sad state of affairs if nobody ever tried to push the envelope with technilogical advances and only stuck with the tried and true? We would still be using 486's and EGA graphics cards if that were the case.

I think nvidia's driver programmers have a chance of resolving (or at least softening the blow of) this particular shortcoming before the HL2 goes to market (or shortly thereafter). We'll see, one way or the other.

phial · Sep 13, 2003

where did you get this information?

-------

<A HREF="http://www.teenirc.net/chat/tomshardware.html" target="_new"> come to THGC chat~ NOW. dont make me get up ffs @#@$</A>

Balderdash · Sep 13, 2003

You know, if nVidia gets such sweet treatment, ATi deserves it too, since they have not gotten a single damn optimization.
<snip>

You know as well as I that, that statement is really not true. All hardware manufacturers get help from Microsoft when a already released and widely distributed item ends up with a glitch or bug that can be resolved with a simple driver or OS or DX update. How many games have ATI cards had problems with that weren't eventually resolved?

And this is one of the main reasons that ATI has traditionally had problems with openGL and Linux applications, because ATI had to deal with these issues without a big conglomerate corporation like Microsoft to help them out...

Balderdash · Sep 13, 2003

where did you get this information?

Paying real close attention to white-papers and tech notes, you can find out a lot of things. Having a few friends tell you their problems is another

p.s., anandtech is a very good place to read about 'whys' and 'why nots,' and he's much more eloquent as well.

Willamette_sucks · Sep 13, 2003

I'm not sure if this is true or not but I'll have to see some proof. If something this big and obvious was holding nVidia back I'm sure nVidia or someone important (not you) would've stated it by now. We've already seen 50.xx beta drivers in action and performance was not increased much. I have no idea where you get 51.75 as I have heard nothing about them.
We'll see.

"Mice eat cheese." - Modest Mouse

"Every Day is the Right Day." -Pink Floyd

speeduk · Sep 13, 2003

If something this big and obvious was holding nVidia back I'm sure nVidia or someone important (not you) would've stated it by now

I second that!

<A HREF="http://service.futuremark.com/compare?2k1=6988331" target="_new"> 3D-2001 </A>
<A HREF="http://service.futuremark.com/compare?2k3=1284380" target="_new"> 3D-03 </A>
 120% overclocker (cheapskate)

Ion · Sep 13, 2003

Just curious are you the bladerdash from hardocp forum?

Just wondering why you show up here...maybe your posts got bashed badly there :tongue: .

Balderdash · Sep 13, 2003

There's only one Balderdash...

I go all over the place. Sometimes I get bashed, sometimes I deserve it and sometimes I don't. My words fend for themselves. If they turn out to be right, nobody remembers where they heard it first, if they turn out to be wrong, then they are remembered!

Balderdash · Sep 13, 2003

If something this big and obvious was holding nVidia back I'm sure nVidia or someone important (not you) would've stated it by now

I second that!

I didn't just make this shader business up. It's been out for weeks (if not months) all the way back to the rumors (on http://www.theinquirer.net) that some unidentified nvidia employee admitting to having problems with the ps 2.0 back in july. It's the same thing now.

The driver to effectively run the dual 4x2 archetecture is still giving them problems today, I'm NOT trying to pretend that it's an easy thing to fix, I'm only saying that it is fixable with the current cards.

It hasn't been a secret, just not something nvidia likes to make public either. And I'm not the first person to reveal it, that's just silly to say that I am.

speeduk · Sep 13, 2003

A quick look into this reveals this is the case. The new drivers will fix a problem relating to the 4x2 mode. If you read the statements by nvidea, they drop hints about this problem, but don't exactly say what the problem is.
This will increase the HL2 perf and pix2 perf in all cases, and when they have worked on it some more, it should get better....

<A HREF="http://service.futuremark.com/compare?2k1=6988331" target="_new"> 3D-2001 </A>
<A HREF="http://service.futuremark.com/compare?2k3=1284380" target="_new"> 3D-03 </A>
 120% overclocker (cheapskate)

Balderdash · Sep 13, 2003

Thank you speeduk,

I appreciate the fact that you actually checked into it, thanks again.

kmolazz · Sep 13, 2003

"ATI utilizes a 8x1 pixel shader path, with one path at 8 bits. Nvidia, on the other hand, uses a 4x2 path with two paths each 4 bits wide. Currently, any game using PS 2.0 with the FX cards is only accessing shaders at 4x1, due to driver and DX9b limitations (we will see DX9c soon, mark my words) and so, the DX9 games and 45:23 driver is effectively ignoring the second PS 2.0 path.

The preview 51:75 driver alleviates this problem, enabling the full second path for use in the game as much as possible before any update to DX9 is implemented to allow true dual channels as intended by its design."

I sorry but, and i'm no expert in the matter, what yousaid doesn't make much sense. "dual channel... 4 bits wide".. i mean i've nerver heard of that..

Read this article about NVFX and CineFX architecture..
http://www.3dcenter.org/artikel/cinefx/index_e.php

I really don't think your comment made any sense, in regards to the architecture aspects..

And benchmarks with those newer drivers are already floating aroud, without any noteworthy increase in speed... some increase, but nothing special.

Balderdash · Sep 13, 2003

<snip>
Read this article about NVFX and CineFX architecture..
http://www.3dcenter.org/artikel/cinefx/index_e.php

I really don't think your comment made any sense, in regards to the architecture aspects..

And benchmarks with those newer drivers are already floating aroud, without any noteworthy increase in speed... some increase, but nothing special.

From the very nice website you linked to (i've read it before, it's well done), I will quote:

nVidia mentions at multiple instances that they doubled the floating point performance compared to its predecessor. The "NVIDIA GEFORCE FX 5900 PRODUCT OVERVIEW" also claims that che chip is able to execute 12 instructions per clock. This leads to the conclusion that nVidia added a second FPU that can perform four instructions per clock. Together with the eight texture instructions, this adds up to the claimed 12 operations.

The additional FPU was most probably placed at the combiner stage. The five million additional transistors do not suffice for an FPU of this complexity. nVidia had to remove something to fit the FPU on the chip. Because the new FPU is able to handle the tasks of the integer ALUs, we can assume that those units were removed. Tests with NV35 show only minimal performance losses in PS1.1 to 1.3, so the FPU can perform almost all operations at the same rate as the integer ALUs do.

In some exceptional cases this doesn't hold true and the FPU needs more cycles for the same task. Then there is the question of which data formats the new FPU can handle. We can be sure that that the FPU is able to handle fp32 (s23e8) numbers. The mantissa of 23 bits requires the FPU to have 23 bit adders and multipliers. Extending those to 24 bits and allowing them to split yields two 12 bit adders and multipliers. Exactly what is neccessary to replace the integer ALUs of NV30. This seems more logical than replacing only one of the integer ALUs because it needs much less transistors in total.

The question whether the FPU can be split into two fp16 units will be left unanswered. Tests have not definitely shown whether performance increases can be accounted to a higher calculating power or to the smaller register usage footprint. The marketing department surely would have mentioned it if fp16 allowed for 16 instructions per clock.

Also,

But nVidia can't expect an application to always deliver such code. At this point we can only preach that nVidia has to put instruction reordering capabilities into their drivers, but without changing the function of the code. The method the current driver uses nothing more than a stop-gap solution that is acceptable as long as the number of applications containing unfavorable shader code is small. But wiht a growing number of titles using shader technology, nVidia can't expect GeForceFX users to wait for new drivers to enjoy higher performance through replacement shaders.

And that, sir, is exactly what I was trying to say...

speeduk · Sep 13, 2003

Balder I think its best to let time do the talking so to speak. When HL2 is benched with the new dets and checked for cheats etc, it will show that the performance is more what we expected in the first place. But I don't expect miracles from them either. Probably 10-20% perf increase at most.

<A HREF="http://service.futuremark.com/compare?2k1=6988331" target="_new"> 3D-2001 </A>
<A HREF="http://service.futuremark.com/compare?2k3=1284380" target="_new"> 3D-03 </A>
 120% overclocker (cheapskate)

kmolazz · Sep 13, 2003

Thanks for the reply, now i know what you were tring to say... although i still think the dual channel "stuff" you mentioned doesn't make sense..

Still, now it's more clear what you were tring to say.

I think your last quote says it all, in respect to what we can expect from new drivers. Lets just hope, for the sake of all nvidia users, they get it to work without "changing the function of the code". But i don't think the results will be amazing, because the hardware is inferior (well, different), but they can still give it a boost though.

Balderdash · Sep 13, 2003

I promise not to use the dual channel / ps 2.0 4x2 analogy anymore, I thought it would help people visualize it better... but maybe not

Balderdash · Sep 13, 2003

Balder I think its best to let time do the talking so to speak. When HL2 is benched with the new dets and checked for cheats etc, it will show that the performance is more what we expected in the first place. But I don't expect miracles from them either. Probably 10-20% perf increase at most.

That sounds about right. I've seen some results that show a 20% gain in ps 2.0 already, but then they didn't contribute anything to the 3dmark03 GT4 score at all ... so go figure?!?

But yes, we will have to wait and see, and that was a point I have been trying to make all along ... that the valve presentation and HL2 benchies now out will not be the last word in regards to the nvidia and ati cards currently available (not to mention the unknowns of the soon to be released cards)...

spud · Sep 13, 2003

Nice to know someone still beleives in the FX, thx dude your kind and gentle words have soothed me. I stopped makeing my pipe bomb I was going to send to Nvidia and decided to buy a new motherboard and some memory to make me feel better.

BTW I thought that was fishy as well the 4:2 aspect since in most cases the FX was 50% slower than the R3xxx series. I dont personally think my FX will surpase the R3xxx's but it sure makes me feel a lot bette with my purchase if im only 15% to 10% slower if this is all true.

-Jeremy

:evil:

<A HREF="http://service.futuremark.com/compare?2k1=6940439" target="_new">Busting Sh@t Up!!!</A> :evil:

<A HREF="http://service.futuremark.com/compare?2k3=1228088" target="_new">Busting More Sh@t Up!!!</A> :evil:

The real deal w/HL2

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Champion

Distinguished

Splendid

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Share this page