Hot Hardware QuadroFX 3000 review

[dvornik]

I wish they included 1000 and some XGL cards. And ran SpecAPC for Maya. Their conclusion - the most expensive Nvidia card is a little bit faster than the one that costs a little less.

Also, would still be nice to see max results under OGL as well. 5900 is doing pretty good with just OGL - competitive with older Quadro4s with Maxtereme I would guess.

[elvis]

yeah, i find a lot of pro 3d card reviews seem to leave out the juicy bits we really want to see.

who gives a toss about unreal tournament benchmarks on a quadro, after all?

[dvornik]

Good point. The thing is the reviews are written by either gamers or techies with no experience in supporting, let alone using, 3d applications.

Umm... There's gotta be a way to get some hardware review site based on the CGTalk franchise...

[elvis]

i'm sure if someone wanted to loan me a quadroFX 1000, 2000 and 3000 i could do a half decent review on a number of platoforms for it. :)

although, i only have access to 3dsmax and maya. an all-over benchmark with the top 10 3D applications would be a great start.

[mark_wilkins]

Actually, I'd love to see some of these boards benchmarked with both pro 3D apps and games. I'm struggling to find a pro board that will both run well with Maya and run Half-Life 2 well. :D

-- Mark

[swag]

mark the QuadroFX are great with maya but nvidia has problems with the dx9 integration the ps2.0 are very slow - here is a nice test
http://www.beyond3d.com/misc/hl2/

[mark_wilkins]

Yeah, and FireGL is bad with OpenGL!

I'm going to wait a few months and see what driver updates bring. I suspect it will be easier for Nvidia to wring out good HL2 performance than for ATI to improve their high-end drivers, but we'll see.

-- Mark

[ZrO-1]

I agree mark,
I think Nvidia will really show an appreciable speed improvement in their 50 series drivers.

I think on the gaming card front, nvidia won't be able to catch up to ATI in dx9 games...at least not till 51/52 series drivers. However, on the proGL front, I think nvidia can get enough juice out of the Quadros to make HalfLife 2 playable, if not really fast.

As far as I'm concerned, I'd rather have awesome pro speed and features with average gaming performance, than great gaming and buggy pro support.
...that's just me though ;) :p

[dvornik]

Originally posted by swag
mark the QuadroFX are great with maya

Any proof of that? The only comprehensive benchmark I've seen shows that just about ALL nvidia cards are equally good in Maya, with old Geforces Ti trailing 10% or so behind (as if anyone would notice):

http://www.3dchips.net/content/review.php?id=63&page=17
http://www.3dchips.net/content/review.php?id=63&page=18

[elvis]

the 50 series of drivers are getting nvidia into a whole load of trouble.

the whole problem is that their particular pixel shader 2.0 implementation is not exactly to spec with the official DX9 spec. therein lies the problem that any custom code nvidia writes/compiles will run well, but any DX9 code that runs to spec will aways run better on cards that adhere to the specifications fully (ie: the radeon 9500 through to 9800 pro cards).

nvidia are trying to fix this by writing application-specific code into their drivers, starting with the 50 series. this is all good and fine, but nvidia are going to be playing catch-up for the next few months on any game that is released. the other problem is their image quality will suffere greatly due to these problems.

the NV40 is supposedly a new core design, and as such should perform MUCH better under true DX9. i'm assuming nvidia will also make a professional accellerator series from this card, which should satisfy the CAD/CAM/DCC users out there who also wouldn't mind playing a game or two between modelling sessions.

as i've mentioned in other posts: if half-life 2 is the most important thing in your world right now, then rush out today and get yourself a radeon 9800pro.

if however you can handle not playing this game with all eye-candy turned on at blistering framerates (or not playing the game at all), and actually care about getting work done, stick with the cards you own now, and wait for the next generation of professional 3d accellerators.

the 50 series of drivers will be nvidias blunder coverup for the geforceFX cards and true DX9 games. it won't offer any gain for anyone else, and at this stage is looking like a bit of a white elephant.

[CgFX]

Everyone should ignore 90% of what Elvis just said.

:banghead:

[elvis]

Originally posted by CgFX
Everyone should ignore 90% of what Elvis just said.

:banghead:

CgFX, once again you prove yourself to be nothing but an annoying post troll. either shit or get off the pot.

the web is full of speculation at the moment regarding the situation i talk about above. only time will tell what the true story is. if you care to comment do so, but you can leave your smart-arse throw away comments at the door.

valve are the main contributors to all of this fuss. go for a quick browse around any of the current major hardware sites (anandtech has a great artcle on it) and see what's occurring and why, and what nvidia and ati are doing about it.

[elvis]

and for those who find it difficult to search for these things themselves:

http://www.anandtech.com/video/showdoc.html?i=1862

- Valve is pissed at all of the benchmarking "optimizations" they've seen in the hardware community;
- Half-Life 2 has a special NV3x codepath that was necessary to make NVIDIA's architecture perform reasonably under the game;
- Valve recommends running geforce fx 5200 and 5600 cards in dx8 mode in order to get playable frame rates.
- even with the special NV3x codepath, ATI is the clear performance leader under Half-Life 2 with the Radeon 9800 Pro hitting around 60 fps at 10x7. The 5900 ultra is noticeably slower with the special codepath and is horrendously slower under the default dx9 codepath;
- the Radeon 9600 Pro performs very well - it is a good competitor of the 5900 ultra;
- ATI didn't need these special optimizations to perform well and Valve insists that they have not optimized the game specifically for any vendor.

[elvis]

and some more:

http://www.anandtech.com/video/showdoc.html?i=1863&p=5

i won't quote from that. you can go read it for yourselves.

and folks, i do stress the following if it's not already obvious: this entire fiasco pertains to the default microsoft directx 9 codepath particular moreso to gaming than anything else.

openGL performance, and performance within your pro 3d apps will probably not even feel the ripples generated by all of this crap.

[CgFX]

Originally posted by elvis
the web is full of speculation at the moment regarding the situation i talk about above. only time will tell what the true story is.
Bingo! You made a post that stated all of the above as fact but as you suggest it was all based on internet rumor, hearsay, and speculation. The least of which was your own speculation which is not based on any knowledge of the situation or parties involved. How you can speak for NVIDIA or valve is beyond me.

I was too tired to correct or object to your post so I banged my head on a wall instead.

if you care to comment do so, but you can leave your smart-arse throw away comments at the door.
It is interesting to see that my comments frustrate you as much as your misleading, speculative, and outright wrong comments frustrate me.

[elvis]

hearsay? the facts are:

1) the nv35 dx9 codepath is inferior to the r300/r350 dx9 codepath

2) the 50 series of detonator drivers are attempting to fix this with application-specific code, and are pissing of end users and developers world-wide by doing so.

i don't speak for valve nor nvidia, but they themselves have in numerous interviews and statements. go read them for yourselves. i post nothing but what comes from "the horse's mouth". it's hardly a case of a rumour someone heard on tomshardware, which seems to get far more attention than any reliable website these days.

if my facts are misleading and frustrate you, then post a rebuttle with proof. however, i find it difficult to post rebuttles to posts with sad one-liners that don't actually take part in a thread conversation.

rise above the commoners and do something good for the planet instead of harping on other's posts. your name comes up regularly these days on and off cgtalk as one of those who really doesn't add to conversations, but is far happier tearing people down from a distance. i'm not the only one who's noticed.

[mark_wilkins]

Just remember that the NVidia side to the story hasn't really been told yet, beyond their brief press release. It's pretty obvious that relations between Valve and NVidia had kind of broken down prior to this, so it's hard to say exactly how straightforwardly the presentation by Valve (at an ATI event) portrays the facts.

Between this and the Softimage matter, though, it certainly appears like Valve is attempting to greatly reduce their development risk by teaming up in public with anyone they can in exchange for compensation. It's a smart strategy for the short term, but in the long run, particularly on the graphics card side, it runs the risk of making Valve look bad, even if they're technically in the right about NVidia products.

-- Mark

[elvis]

before i start this rant, i'd just like to say i'm no fanboy. i've been using nvidia producst for years simply because at the time i considered them the best. i still use quadro cards for pro 3d, but have turned to ati recently for gaming. i'm no brand junky.

anyways... enough of that drivel, on with different drivel:

nvidia seem to have been on the back foot since DX9 applications started filtering along. first there was nvidia's attack on futuremark for 3dmark2003 which they stated was an "unrealistic" benchmark, and then as suddenly as the whole ruckus started it was all over and done with. now we are seeing the same thing with valve and half-life 2.

granted nvidia have been top-dog for a while, and the idea that someone else can come along and virtually overnight (or in one chip revision at least) knock them off pole position (in the gaming world, at any rate) probably bruises the ego somewhat.

i can't help but be reminded of 3dfx's childish antics in their marketing and CEO shortly befor etheir demise several years ago. and even more amusing is the fact that most of 3dfx's engineering team was absorbed into nvidia, and happened to work on the geforceFX series of cards.

i agree with you mark: i think valve need to be very careful about jumping on bandwagons and what they say in an official manner. developer to hardware manufacturer relationships are important, and it would be a shame to see it get worse for the sake of some PR hunting.

[elvis]

More interesting quotes:

source (http://english.bonusweb.cz/interviews/carmackgfx.html)

Hi John,

No doubt you heard about GeForce FX fiasco in Half-Life 2. In your opinion, are these results representative for future DX9 games (including Doom III) or is it just a special case of HL2 code preferring ATI features, as NVIDIA suggests?


Unfortunately, it will probably be representative of most DX9 games. Doom has a custom back end that uses the lower precisions on the GF-FX, but when you run it with standard fragment programs just like ATI, it is a lot slower. The precision doesn't really matter to Doom, but that won't be a reasonable option in future games designed around DX9 level hardware as a minimum spec.

John Carmack

[CgFX]

Elvis,

Post quotes until you are blue in the face. The doesn't change the fact that you are wrong. All the quotes you have posted discuss that NV30's pixel shader 2.0 path is turning out slower than ATI's. That has been shown to be true in the general DX9/PS2.0/HLSL cases.

That isn't what you said.

You said "their particular pixel shader 2.0 implementation is not exactly to spec with the official DX9 spec" which is not only completely false but if you were of any importance to anyone in this industry it would probably be actionable. You went on to reinterate and suggest that nv3x is not DX9 compliant. Wrong.

[CgFX]

Here is my time wasting effort that I let sit and was hoping not to have to post:

1) Saying NV3x is not in spec with DX9 is simply false and misleading to anyone reading this site, and not even based on any facts. NV3x is a superset of DX9 and works at higher precision than the DX9 spec. DX9's HLSL was co-developed by nvidia and is identical to nvidia's Cg.

Your comments are incorrectly based on known details and comments that NV3x's architecture is not as optimal for generic DX9/HLSL code. This is, in large part, due to ATI's big win with M$ (driving the minimum DX9 spec requirements) at the time of NVIDIA's and Microsoft's nasty Xbox chips pricing arbitration.

i.e. If DX9 had required a minimum 128 bit FP pixels and 128 bit FP precision (as it should by anyone's account) or even 64 bit FP as a half-precision performance mode instead of 96 bit FP pixels then the situation would be 100% different. ATI would be multiples slower.

Nobody has ever suggested that 96 bit is where the gfx industry needs to end up, everyone says 128 bit. And it makes no sense that the half-precision mode would be 96 bit instead of 64 bit. This was simply a coup by ATI to define the minimum DX9 spec well below the precision of NV3x and where ATI needed it to be to keep their die size within reason on their older .15u process R3x0,

B) You have no idea what nVidia is doing in their rel50 driver development related to application specific tuning and they haven't said. Should they be tuning PS2.0 shaders to better fit within the flow of their architecture it would be completely valid in my book. Keep in mind that both nvidia _AND_ ATI's DX9 hardware do not generate identical screenshots to Microsoft's software DX9 test renderer. There is currently no way to suggest that ATI's final image is any more or less valid than nvidia's in Microsoft's DX9 world.

Valve can say ATI's is the valid or most accurate image but that is what you get when you hand them a reported $6+ million dollars to be the exclusive development and launch platform.

III) You have no insight into NV40 (nor do any of the webzines you read) so why even comment? Of course it will have higher perf. Why wouldn't it?

d) You have no information on what release 50 is going to provide those that are not playing HL2 or DX9 games so I don't understand how you can give advice on it? I have seen nvidia's release 50 presentation and not a single slide in the version I saw talked about games. There was a lot of stuff in there on general performance and features so there must be something there for the rest of us.

[CgFX]

Originally posted by elvis
your name comes up regularly these days on and off cgtalk as one of those who really doesn't add to conversations, but is far happier tearing people down from a distance. i'm not the only one who's noticed.

I am flattered but you shouldn't put me on such a pedestal.

[elvis]

Originally posted by CgFX
You said "their particular pixel shader 2.0 implementation is not exactly to spec with the official DX9 spec" which is not only completely false but if you were of any importance to anyone in this industry it would probably be actionable. You went on to reinterate and suggest that nv3x is not DX9 compliant. Wrong.

ok, my bad. too many copies of the word "spec".

the problem that valve have indicated as far as nvidia's poor performance with directx9 is that their hardware does not compile the PS 2.0 code in the same way as ati's hardware. is that better?

don't you love people who pick on your posts for minor definitions? (http://www.cgtalk.com/showthread.php?s=&postid=838104#post838104) :p

[CgFX]

Originally posted by elvis
their hardware does not compile the PS 2.0 code in the same way as ati's hardware. is that better?

Nope.

First, there is no such thing as PS 2.0 code. PS2.0 is a spec and a profile. HLSL and Cg code get compiled to machine language that runs on hardware that is in spec with the PS2.0 profile.

Second, Not only is that not what Valve is saying but why wouldn't you expect the code to compile differently? They are completely different hardware architectures. That is what the compiler is for. If they ran the same compiled code then you wouldn't need to compile them seperately or have runtime compiling. It would be more like AMD vs. Intel with IA32. You would compile once and it would run on all GPUs.

What they are saying is that generic DX9/HLSL code (after it is compiled for each GPU) is optimal for ATI's hardware and not for NV3x.

[stephen2002]

arrrgggg.... it's all a big fat mess! I just can't wait untile the game (or the benchmark) comes out so that we can all have fun playing with it instead of reading endless speculation. :banghead: Hurry up and release the games :D

[cg_fan_2003]

Bingo! You made a post that stated all of the above as fact but as you suggest it was all based on internet rumor, hearsay, and speculation. The least of which was your own speculation which is not based on any knowledge of the situation or parties involved. How you can speak for NVIDIA or valve is beyond me.

thankyou

don't believe everything you read there was also something posted about valve doing some sort of auction as to which company would be umm how u say better supported and nvidia pulled out at $8,000,000.

The most likely cause of this is simply valve letting off some steam. and or a dissagreement with nvidia or a bribe off ati. no one can comment what has happened because both sides stories haven't came out. truth is there isn't much of a story just some air swinning.

[endless]

[quote]
don't believe everything you read there was also something posted about valve doing some sort of auction as to which company would be umm how u say better supported and nvidia pulled out at $8,000,000.
[quote]

Intresting, because the story I read (off the inq, no less) was that both nvidia and ati ponied up 6 million, and it was up to valve to choose which they thought would showcase the game better.

[CgFX]

The marketing deal between ATI and Valve is a fact. Bidding went beyond $6M with nVidia pulling out. The rumors were wide spread but it was confirmed by a leaked email from nVidia's S.VP of Marketing.

I don't think that has much affect on the technical situation but I do think it affects how Valve colors the facts and with how much vigor. e.g. You don't see Carmack beating up nVidia for any architectural differences. He just adjusts ("custom [render] backend") and moves on.

[loop29]

I read about the differences within PS2.0 before when comparing performance between ATI and nv, the article that lighten it up for me can be found here:

http://www.3dcenter.org/artikel/cinefx/index_e.php

CgFX: I guess it´s not the consumers problem if nvidia is going higher than DX9 spec, at least it shows up as being slower in the end. I assume it´s time for nvidia stop stating excuses and tell the truth and make it better on nv40. I would be impressed if they can get back in one GPU cycle.

regards

[CgFX]

Originally posted by loop29
CgFX: I guess it´s not the consumers problem if nvidia is going higher than DX9 spec, at least it shows up as being slower in the end.
Totally agree, although if the app uses that full precision mode they end up with greater color and lighting range/precision (dramatic for the shadowed environment of Doom3) and higher sub pixel precision (avoid rendering errors).

No idea on NV40.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.