NV34-U review -- what do you want to see?

[Reverend]

I don't normally do this (ask for ideas for a review) but I just received an Albatron GeForceFX 5200 Ultra and I'll admit that I have been slightly out of the loop wrt NV3x things due to too much work.... so ideas are welcomed but don't expect all to be included in the review.

I should also mention that this Albatron review will appear after Dave's review of a NVIDIA NV34Ultra reference board (Dave, when do you expect that out?).

 

[DadUM]

For the 5200U, I would be interested to see some CPU scaling results. As this a "budget" board, I would like to see how it performs with "budget" CPUs.

As for specific tests, I like to see SS:SE as well as some more non-traditional games.

 

[Sunday]

As for specific tests, I like to see SS:SE as well as some more non-traditional games.

Why SS:SE!? Nobody us that engine! I can understand Q3 and UT2K3 tests even Max Payn ('cos of sequel)!
But SS!? Its B, or C game with no licensed title of other developera. It has some nice features, but they aren't widespread, or implemented in the same way in other titles that use same effects!
it's pointless!
G! Then why not using some sort of NOLF benchmark or Enter The Matrix, or...you choose!?
Rev stuck up FX5200Ultra against R9500 (noPro)!?

 

[SsP45]

It would be interesting to see some less-known games that probably wouldn't be too "optimized." I don't see too many benchmarks of Medal of Honour or BattleField 1942, yet these two games are very popular.
________
Easy Vape Vaporizer

 

[gkar1]

Its better to test with lesser known engines since they are not targetted for optimization by the driver teams. Even though the SS engine is pretty well known for its technical merits. Nvidia's recent actions with "optimizations" and "bugs" in benchmark programs being exposed prove my point.

 

[Reverend]

This will be a review, so I will only be focussing solely on the GFFX 5200U. Comparisons with other competing cards will be in a separate article but don't expect this anytime soon as I currently don't have any competing cards.

 

[KimB]

Well, we know about the anisotropic filtering quality, and the AA quality (cursory tests to ensure they're what are to be expected may be good).

What may be interesting is an investigation of the shader capabilities. There was a recent thread that did this for the 5800 Ultra, but it would also be interesting to see how the other cards of the familiy stack up.

Update: found the thread:
http://www.beyond3d.com/forum/viewtopic.php?t=5150&start=0

 

[DemoCoder]

Try to get the HL2 demo from Valve and bench it on it.

 

[Reverend]

Try to get the HL2 demo from Valve and bench it on it.

Done!

Hehehe.

 

[Calavaro]

Lets see now. 1st of all, what everyone have pointed out, bench NON-standard games.
If you bench Q3A, JK2, UT2k3 etc, record your own demo and bench on that. Make the demo available for download for us to do a comparison ourselves.

IQ IQ IQ. After playing with 2xAA and 16xAF on the R9700pro I would be hard pressed to go back to a non-AA and a non-AF scenario. I know the 5200U is targeted for budget people, so maybe limit to 4xAF and 2xAA. Further, there is no point in benching anything except the "application" setting on the nv30-nv34.

A forgotten point nowadays is 2D filtering quality. How does it look on a 21" monitor @ 1600*1200, 85Hz? (i.e).

 

[Sage]

i want to see the 5200-U vs GF2-GTS and V5 5500 SERIOUSLY!

edit: oh, and Xabre and Parhelia as well

 

[Killer-Kris]

I really have only one recommendation. And that's to use the Quake1 Tenebrae modification as a benchmark.

http://tenebrae.sourceforge.net/

Just like the PowerVR FableMark it uses the stencil buffer quite heavily and I think would certainly count as a more obscure title which I doubt will have been specifically optimized for. It 'might' also give some insight into whether or not the nvidia cards really are all that when it comes to stencil operations like the Doom3 benchmarks would have us believe.

 

[demalion]

I'm interested in some behavior characteristics for some advanced functionality...I think conditonal branching and looping vertex shader benchmarks flexing the new functionality for NV3x might be a good exploration, and also for other cards in the NV3x family. There were rumors that the NV34 had the CPU handling some workload that the other NV3x cards did onboard, and I'm thinking a CPU scaling for such "2.0+" VS code might be informative as it seems a logical place for some cutbacks to have been made.

This would need to be an uncommon or custom made benchmark test to be clearly useful, I think.

 

[DadUM]

Why SS:SE!?

Four quick reasons:
1. I have SS:SE and know exactly how it compares to my system. Not numbers for games I have no intention of dropping $50 on. (Yes its a selfish reason, but I can still ask)
2. SS:SE supports both OpenGL and D3D renders. By investigating both, we can see if drivers have been tweaked more for one path.
3. SS:SE supports some "advanced" benchmarking features like frame rate graphs and mip-map shading. Other IQ factors are also clearly adjust in the game. How about comparing enabling AF via control panel vs application. Does that make a difference?
4. The Serious engine was mentioned as having shown visual artifacts that could be simliar to the "clipping plane" issues of 3DMark2k3. It appears that changing a .ini setting should be able to clear these up. It would be nice to see if the visual errors are present, if the solution works and if the solution effects speed or render quality.

 

[Sabastian]

Yeah ....... I am not horribly interested in the GeforceFX5200 "ultra" or non "ultra" really. I would like to see a review of the non ultra simply to see just how poorly the GeforceFX5200 ("DX9 for $79!") actually works. Particularly on DoomIII. If you ever compare the card to anything I think the Radeon 9600 Pro would be the price range of the GefarceFX5200 "Ultra"? From what I have read about the NV34 core it really isn't that impressive. I would like to see what kind of frame rates the non ultra gets on DoomIII even the "ultra" would be interesting though.

 

[Sage]

I think you may have the 5200 series confused with the 5600 series. The 5200 is comparable to a GeForce4 MX... although the MX does manage to be a bit faster

 

[T2k]

As for specific tests, I like to see SS:SE as well as some more non-traditional games.

Why SS:SE!? Nobody us that engine!

And? Still one of the best engine for testing the SAME GAME in OpenGL AND D3D.

I can understand Q3

I can't: old piece of crap, shitty textures. Get lost!

Use JKII instead of Q3 - it has much better/bigger textures.

and UT2K3 tests

That's cool.

even Max Payn ('cos of sequel)!

Why? I don't understand...

But SS!? Its B, or C game with no licensed title of other developera. It has some nice features, but they aren't widespread, or implemented in the same way in other titles that use same effects!
it's pointless!

Perhaps only for you, my friend. Read my notes above.

G! Then why not using some sort of NOLF benchmark

? Maybe NOLF2, right? NOLF1 is old and "Its B, or C game with no licensed title of other developera"...

or Enter The Matrix,

This is the most stupid idea. EtM? What? Did you check that game at least? Go to the Games Talk and check relevant topic! Enter the Matrix IS A low-level, cheap piece of crap.

or...you choose!?

Rev stuck up FX5200Ultra against R9500 (noPro)!?

Yeah, it's interesting. But don't forget the R9000/9200-family or even the 8500 - there are a lot of them out there.

 

[KimB]

I really have only one recommendation. And that's to use the Quake1 Tenebrae modification as a benchmark.

http://tenebrae.sourceforge.net/

Just like the PowerVR FableMark it uses the stencil buffer quite heavily and I think would certainly count as a more obscure title which I doubt will have been specifically optimized for. It 'might' also give some insight into whether or not the nvidia cards really are all that when it comes to stencil operations like the Doom3 benchmarks would have us believe.

Ah, yes, that reminds me. I'd like to see not only how the stencil buffer usage affects the performance of the card, but also how it affects the performance of FSAA on the card.

 

[Doomtrooper]

Interesting to see it's DX9 performance on this high dynamic range demo, includes a frame counter with many different shader options...

http://www.daionet.gr.jp/~masa/rthdribl/index.html

 

[Pete]

I'd like to third the Tenebrae recommendation. Even if it's not a shipping game, you can at least get an idea of performance on different cards (if not across companies, then maybe at least across generations--GF4 vs. GF FX).

Use it wisely before the driver "optimizers" get word of it!