[T2k]

Originally Posted by Doomtrooper

I suggest you educate yourself ....here I will bold the part that is most educational for you

http://pc.ign.com/articles/456/456054p1.html?fromint=1

ROTFLMAO...

[McDusty]

On the downside, now that Humus works for ATI, he'll have to watch his posts on these boards and others such a rage3d because his opinions could be misconstrued as those of ATI.

Large companies gernally give staff guidelines on what can / cannot be said on particular subjects.

Anyway - perhaps shader speed will increase with new driver revisions. Perhaps NV30 shader code optimisations are having a negative impact on NV40 shading abilities.

Ok, im clutching at straws here

[duncan36]

NV40= $499 retail(ouch) available in 45 days on shelves.

http://biz.yahoo.com/rc/040414/tech_nvidia_1.html

[onetwo]

...and 'slightly slower' 6800NU is $299 (according to them).

[plat]

the 299 jives with what was posted before, perhaps the 499 was a typo? i dont see how the ultra version could be $200 more and have virtually the same specs?

[mreman4k]

Originally Posted by AlphaWolf

Surfaced where? All the screenshots say 6800U and graphs say 6800 Ultra. Unless you are thinking its just a misinformation campaign.

At NVNews...

http://www.nvnews.net/vbulletin/showthread.php?t=27297

[AlphaWolf]

Originally Posted by onetwo

...and 'slightly slower' 6800NU is $299 (according to them).

Anyone wanna guess vanilla 6800 clocks? 350/500(1000) has my bet.

[demalion]

Well, I've (more than once) specifically outlined how [H] fails to provide the reader information about graphics cards, and is moreso providing information about games, and covered the issues with the approach and what information it fails to provide readers. It is encouraging at least to see that Brent recognizes it and is telling readers about it somewhere, because there is at least some hope they'll avoid conclusions their approach isn't suited to making.

However, an NV40 versus R380 review is about the perfect place for the problems for it to be most prominent, because all the AA, AF, shader, vertex processing, and CPU limitation relationship factors between the cards and on the selected games are uncharted territory. Hopefully there are 1) other results from some tool to establish these relationships usefully included (strangely enough, a successfully cheat-policed 3dmark 03 could be such a tool), 2) some other article(s) released in a timely fashion with such info (without the issues with Anandtech's article splitting), and/or 3) they have a disclaimer prominently displayed directing readers somewhere else for such info and thoroughly educating readers about what their approach doesn't inform about (might be required on every page or under every graph to be successfully communicated). The forum blurb indicates that 3) at least is quite possible, so the article could still be successfully informative in the intended way once reviews that more clearly inform about hardware comparison are available.

To me, the ideal would be to have the hardware-centric comparative analysis done, and then provide tools for users to pick comparison graphs based on their own image quality evaluations (maybe with a blind testing selection method to aid in objectivity) or for information on hardware comparison, depending on the information the reader wants to evaluate at the time. Similar to the type of comparison mechanism that has often been proposed in the past, maybe with [H]'s method thrown is as "reviewer's equivalence" settings as a guideline for a site like [H] that tries to do that for readers.

[Helevitia]

Something has got to be horribly wrong with the FarCry benchmark. The latest patch for FarCry says this:

Improved shader speed (up to two times faster) for NVidia Geforce FX cards on very high settings
- Fixed fog bug with NVidia drivers
- Added support for NVidia NV40 chipset
- Fixed realtime shadows from blended trees
- Made improvements to lighting shaders on very high settings
- Fixed bug with lake reflections
- Added support fro PS3.0 shader profile


So either the 6800U was terribly slow to begin with or there is a bug.

[plat]

if you look at the screens, farcry wasn't useing vs/ps 3.0...

[Megadeth]

Originally Posted by Helevitia

I'll save you guys the time and post what he says about his testing

The problem I see with this method is that you have no idea on how much of impact (or little for that matter) the increased settings have on the other cards. So, these measurements are pointless in my opinion. For example, let's just say the in NFS, increasing the 9800XT to 1600x1200 reduces the FPS by only 5-10 (unlikely naturally of course), how would we know? Since the reviewer is taking the liberties to decide if the fps difference is enough or not is the problem. These numbers are pointless until the ENTIRE range of equal settings are shown and then the user can decide which settings is HIS / HER ideal gameplay settings / performance for their card.

[Helevitia]

Originally Posted by plat

if you look at the screens, farcry wasn't useing vs/ps 3.0...

Right, so I would think that PS2.0 would be kicking ass right now. Don't you?

Dave

[Chalnoth]

Originally Posted by Helevitia

Something has got to be horribly wrong with the FarCry benchmark.

Keep in mind that the 6800 Ultra is running at higher settings than the other cards in the benchmark. Because the reviewer didn't test at the same settings, we don't know what the difference between those settings is.

[Megadeth]

Originally Posted by McDusty

On the downside, now that Humus works for ATI, he'll have to watch his posts on these boards and others such a rage3d because his opinions could be misconstrued as those of ATI.

Large companies gernally give staff guidelines on what can / cannot be said on particular subjects.

Anyway - perhaps shader speed will increase with new driver revisions. Perhaps NV30 shader code optimisations are having a negative impact on NV40 shading abilities.

Ok, im clutching at straws here

I think the only concern would be people offering opinions as opposed to facts. ATI has programmers who frequent the board and they only talk to the topics they know. I don't see why any company would have an issue with that (at least any normal company) if it doesn't expose any IP.

[ark_angel]

i'm confused about something http://mbnet.fi/elixir/NV40/10817474...eutS_1_9_l.jpg
shows the board as version a1 yet there were a bunch of pictures on the web of board a2 and the board vegetto-ex posted about was also version a2. would a1 be a more recent board?
http://www.beyond3d.com/forum/viewtopic.php?t=10946&highlight=a2[/url]

[Chalnoth]

No, A2 would be the more recent board, but unless there were major bugs in the A1 revision, I would expect that most reviewers would still get the A1 rev boards. HardOCP may have had their NV40 for a while.

[plat]

Originally Posted by Helevitia

Right, so I would think that PS2.0 would be kicking ass right now. Don't you?

Dave

actually from the talk on this board 3.0 is supposed to make things more efficient i believe.

i just pointed it out because the patch log says added nv40 support and ps 3.0 and the game is detecting it as nv3x hardware.

[Chalnoth]

Originally Posted by plat

actually from the talk on this board 3.0 is supposed to make things more efficient i believe.

The key is that it can make things more efficient. There will be no change if the shaders in question don't use any of the added efficiency improvements in PS 3.0. There may be a speed drop if the programmer doesn't write a PS 3.0 shader properly (i.e. using dynamic branching for deciding whether or not to do a texture fetch, or just having a very small number of instructions to choose between in the branch).

[Mintmaster]

Originally Posted by Chalnoth

Keep in mind that the 6800 Ultra is running at higher settings than the other cards in the benchmark. Because the reviewer didn't test at the same settings, we don't know what the difference between those settings is.

True, but most of the time ATI takes a far greater hit from NoAA to 2xAA than from 2xAA to 4xAA. I don't think ATI will drop in score very much going to 4xAA.

I think we need to somehow get FarCry to do the same shaders on both cards. NV40 fixes a lot of image problems of NV3x, but some are still there.

Personally, I'm starving to see Shadermark scores, or maybe even some framerates from some of Humus' demos.

[Rev. Night]

I want all nv3x,4x and R3xx,4xxx tested against each other in one giant arena of Quake 1 engine.

[Bjorn]

Originally Posted by Mintmaster

Personally, I'm starving to see Shadermark scores, or maybe even some framerates from some of Humus' demos.

That and Half Life 2

[saf1]

Wonder why no one has run or shown NVShaderPerf stats? Also - does this only run on Nvidia hardware???

http://developer.nvidia.com/object/n...perf_home.html

[rendezvous]

Last time i checked it reported performance only for the chips in the NV3X series.

[ninelven]

One can get a few apples to apples comparisons from the IQ pics, of course only for one frame meh.

CoD:
NV40 - 214fps
9800XT - 114fps

Splinter Cell:
NV40 - 63fps
9800xT -35fps

???:
NV40 - 42fps
9800XT - 33fps

NV40 - 53fps
9800XT - 40fps

Anyway, I'm sure there will be much better information in the morning .

[Pete]

1. Why is 6800U still running in NV3x fuglified mode in Far Cry? Note the 9800XT, in addition to not showing banding, also appears to show brighter brights (to coin a phrase). Puzzling, considering NV40's supposedly huge shader performance gains.

2. The majority of [H]'s tests appear to be CPU or bandwidth-limited. Why don't they use a synthetic that's known to stress the GPU, even if they have to acknowledge potential cheats? There was a big flap when Ace's showed how GPU-limited 3DM03 was, yet now it appears that same limitation makes it superior to bench these new cards.

3. Interesting to note the core is only 400MHz. NV has decreased GPU clock speed for each of their last three high-end cards, and all are 130nm. I don't remember that happening to them ever, though I'm sure they're hitting more process limitations and dealing with far greater architectural complexity now. I'm expecting ATi to debut with much faster clocks, but I wonder if that'll matter, given the frankly uninspiring performance leap of the 6800U (per Brent's review, though I suspect the CoD improvement may be more in line with what we'll see from more conventional reviews comparing cards at the same settings).