FX5200 and FX5800 from Inno3D. DirectX9 now also by nVidia
Introduction
After asked nVidia since late summer last year visitors to their website, whether they are prepared (for the GeForce FX) and a presentation at great expense to the official launch date with a handful of samples was raised in November, the first maps of the new nVidia generation we have arrived these days at last.
These are, as already the title to refer to a FX5200, so the smallest possible model (a version with 64 MB RAM times apart from) and one FX5800, so the model directly under the FX5800 ultra. Our test object only by the lower bar differs from this, all other features and hardware features are the same.
In our small preview, we want to briefly introduce the two cards, and give a first impression of the class in which they play.
We save on detailed comparisons in terms of image quality and a thorough check of performance with diverse FSAA/AF options us for the full review in a few days following.
Cards and boxes
Starting want we with the small FX5200 from Inno3D, which from the packaging manufacturing only differs by the lower weight and the number on the box. Otherwise, delivery as well as other facilities of the two cards are identical.
The delivery is next to the map from an adapter of DVI to operate also a normal CRT monitor to the second monitor output. There are a sufficiently long TV-out cable, must provide not the machines behind the TV to allow the output on the big screen. Computer and TV should, but then but not each other, be far away.
The software package that has put together Inno3D has failed this time unusually rich, what a little surprising especially in the case of very inexpensive FX5200. But complaining we do not want us on by.
Besides the compulsory driver CD, that includes also the WDM driver for ViVo capabilities DirectX in version 9.0 and the drivers for the release of 43.03, 3DMark03 in the downloadable, limited demo version is included in. A current game in full version there with Novalogics Comanche4 also included, such as a two-channel version of interest videos WinDVD 4.0. But the highlight is produced which probably also from InterVideo WinDVD Creator, with the movies on the PC to edit and ban on DVD can be.
The cards themselves are what is however already rather different, of course. The FX5200 is commendably because with a passive cooling which ideally suitable for the Office, and Internet-PC where at times the one or the other games to run, because by their very nature no noise.
Otherwise, however, the FX5800. This required a massive heat sink that is filled with air by a centrifugal fan to the dissipation of the heat produced by it. The about in the first test of FXFLow by nVIDIA is not used here, but the solution from Inno3D needed nevertheless two slots, so a PCI slot next to the AGP port and can be called also not just running. The video card fan is even with closed housing (under the desk) still clearly herauszuhören an average noisy PC if he is in full swing in the 3D.
The connections of both models offer everything what is required of a modern video card with a 15-pin VGA connector, a DVI and a TV-out also.
Also the back of tornado GeForce FX5800 needs still a cooler of the caliber completely dispenses with the FX5200. The cooler while covered her also the memory chips from the surface, but has gewollt’s mass no contact to same as the DDR memory taktende with 200 MHz requires no cooling. The FX5800 has a very good contact of the memory chips to the cooler and is generally very solid worked like also the FX5200. Latter seems, robbed their cooler behind the power supply in the first third of the show shallow as a flounder. An indication of a design very easy and inexpensive to manufacturing.
Performance data in comparison
To get a direct comparison of the cards to their competitors, is of course nothing better suited as a classical table, aufgeteil for low end and high end cards.
< div >
| GF FX5200 | GF4 MX440 | Radeon 9000 | Xabre 400 | |
|---|---|---|---|---|
| Chip | nV34 | nV17 | RV250 | Xabre400 |
| Manufacturing process | 0, 15µ | 0, 15µ | 0, 15µ | 0, 15µ |
| Core clock | 250 MHz | 270MHz | 250 MHz | 275MHz |
| real DDR memory clock | 200 MHz | 200 MHz | 200 MHz | 275MHz |
| Rendering pipelines | 4 | 2 | 4 | 4 |
| TMUs per pipeline | 1 | 2 | 1 | 2 |
| Pixel fill rate | 1000 Generation / s | 540 Generation / s | 1000 Generation / s | 1100 Generation / s |
| Texelfuellrate | 1000 MTex / s | 1080 MTex / s | 1000 MTex / s | 2200 MTex / s |
| Memory bandwidth | 6400 MB / s | 6400 MB / s | 6400 MB / s | 8800 MB / s |
| Texture lookups per pass | 16 (OpenGL: 4!)* | 2 | 6 | 4 (OpenGL: 2!)* |
| Shader hardware | PS 2.0 + / VS 2.0 + | TnL HW after DX7 | PS1. 4 / VS1. 1 | No PS / vs1. 1 |
| Vertex throughput | ~ 62MVert / s | ~34MVert/s | ~39MVert/s | k. specifying |
| * Although the reason is so far unknown to announce both the current nVIDIA detonator 43.45; at the same time only a smaller number than the drivers of the Xabre series under OpenGL Active textures. We hope this at least for the GeForce FX up to our detailed To be able to determine review. |
| GF FX5800 | GF4 Ti4600 | Radeon9700 Pro | |
|---|---|---|---|
| Chip | nV30 | nV25 | R300 |
| Manufacturing process | 0, 13µ | 0, 15µ | 0, 15µ |
| Core clock | 400 MHz | 300 MHz | 325 MHz |
| real DDR memory clock | 400 MHz | 325 MHz | 310MHz |
| Rendering pipelines | 4 (8) *. | 4 | 8 |
| TMUs per pipeline | 2 | 2 | 1 |
| Pixel fill rate | 1600 Generation / s | 1200 Generation / s | 2600 Generation / s |
| Texelfuellrate | 3200 MTex / s | 2400 MTex / s | 2600 MTex / s |
| Memory bandwidth | 12800 MB / s | 11200 MB / s | 19840 MB / s |
| Texture lookups per pass | 16 (OpenGL: 4!)* | 4 | 16 (OpenGL: |
| Shader hardware | PS2. 0 + / VS2. 0 +. | PS1. 3 / VS1. 1 | PS2. 0 / VS2. 0 |
| Vertex throughput | 160 MVert / s | 136 MVert / s | 325 MVert / s |
| * Although the reason is so far unknown to announce both the current nVIDIA detonator 43.45; at the same time only a smaller number than the drivers of the Xabre series under OpenGL Active textures. We hope this at least for the GeForce FX up to our detailed To be able to determine review. |
* nVIDIA announced the introduction of the GeForce FX although you can edit up to 8 pixels at the same time which was taken over by many Radeon9700 of à la also immediately as confirmation for a 8-pipeline design. To the explicit confirmation, which have nV30 eight separate pipelines you expressed around but again and again at nVIDIA and pointed out that in the age of advanced importance abnehmende Rendertechniken of specified units such as pipelines, separate Vertexshadern and TMUs.
As it turned out, is the nV30 built the classic 4 x 2 (pipelines x TMUs) principle and has powered get additional units for accelerated processing of Z-buffer and stencil operations. When these two types of tasks, the nV30 actually behaves like a real 8 pipeline design, which is useful for example in games, based on the Doom3 engine in development.
On the features we will enter into only opposite to the DirectX9 standard of partly significantly advanced shaders of the nV3x series in the review at ührlichen.
Test system
To test a Pentium4 operated with 133MHz-FSB on an i845D motherboard with 512 MB PC2100 RAM was used, which will provide the basis for any video cards testing ComputerBase.de in the future. We test under WindowsXP, because it now on the one way or another in the home realm has prevailed and has become the first choice for many now have to play.
To noise background of the Board of the onboard sound was used based support on the C-Media motherboard with six-channel.
- CPU:
- Pentium IV 2,53GHz (133 MHz FSB QDR)
- Motherboard:
- ASUS P4B266-E
- Memory:
- 2 * 256 MB Nanya/Apacer PC2100 DDR RAM Cl2 with sharpest settings
- Video card:
- Inno3D GeForce FX5200 128 MB (driver official detonator 43.45) tornado
- Inno3D GeForce FX5800 128 MB (driver official detonator 43.45) tornado
- Connect3D Radeon9700 Pro 128 MB (driver of official catalyst 6166)
- Connect3D Radeon9500 64 MB (driver of official catalyst 6307)
- Peripherals
- IBM DTLA 307015
- Pioneer DVD A03
- Onboard 10/100Mbps NIC
- Other
- A4 tech double wheel mouse
- Software
- WindowsXP Professional
- DirectX 9.0
In order to keep the preview as possible up to date, we have used partially older, determined by us in previous reviews benchmarks why are some versions of the driver not to date. For the detailed review of course current drivers for all Combattanden are on the program.
We only used the “application” mode, which likewise should guarantee the best Bidlqualität TI to the GeForce3/4 in particular in the use of anisotropic filtering due to the still not finally clarified optimization possibilities with the recently unnamed quality and performance settings. We will compare the listed modes later even against each other in the full review.
Benchmarks
For a first small demonstration, we use the following programs:
Benchmarks
- 3DMark2001 – the classic, although its value greatly depends on the CPU and the engine is used nowhere
- 3DMark2003 – the controversial, still violently is argued over whose meaningfulness, less CPU-dependent than its predecessor
- Villagemark – fill rates test with gigantic Overdraw
- Fablemark – a program that benefits strongly from Stencilperformance and has massively transparent Overdraw
- Comanche4 – alone already, because it included the two cards and is a good Directx8.0 test with large textures and many polygons
- Because DoD flyby – very current, much gespielter benchmark with great demands on the graphics card
- Serious Sam SE – OpenGL version, also a classic among the benchmark-enabled games
Synthetic benchmarks
3DMark2001 SE v3. 30
Because of the 3DMark 2001 2000 is a purely synthetic benchmark such as the well-known 3DMark, he can be a good guide for future games as he fully based on DirectX 8 and this slowly to enforce begins after one and a half years in the play area. Among other things due to the nature test, which makes extensive use of pixel shader effects, is the 3DMark2001 SE not only of CPU and graphics card RAM depending on, but also very from the core clock of the video card. The current version of 330 was of course.
3DMark2001 SE v3. 30
Information on points
|
What restrains the FX5800 here in 1024 x 768, is still unclear, particularly since she can compete settings again very well on the quality-I (again to the memory of 2xFSAA and 4xAF). Actually the distance should increase more, because the Radeon9700 Pro has nearly of the double, very important for FSAA memory bandwidth.
The FX5200 considers himself a low end product according to the bottom and can remain after all in the same performance area like their predecessors, the GeForce4 MX. As it has various benefits this to, as for example shader hardware plays for DirectX9 and also in the same price range, not exhilarating performance is can.
3DMark03
The recently published 3DMark03 is controversial like no other. After looked in first test the cards from nVIDIA not as good as that suggested by arch rival ATi, hefitge controversy to this benchmark, which went so far as that nVIDIA extra a driver on the 3DMark03 wrote optimized was to show how vulnerable is this benchmark for specially adapted driver. This was also impressive with a performance gain of just under 50% compared to the regular drivers.
Nevertheless, also the 3DMark03 will play an important role in most tests so that we also us not quite want to lock the.
3DMark03
Information on points
|
Nanu, a benchmark in 800 x 600?
Yes, read correctly. Also if 3DMark03 allegedly based futures market on the latest technology, the most of all tasks is still of Directx8.0 are to deal with. Despite all this, performance in the basement will with the points, as in any new version of the 3DMarks, so here we had to stiefeln to obtain easy-with the (resolution) basement reasonably acceptable frame rates and thus points.
This may be an indication how the performance of future games on current maps will develop, but in the face of little performance optimized programming of game test 2 as “The Battle of Proxicon” is still hope.
Want to make even if the points here believe something else: Halbwegs liquid the more demanding three tests only for the FX5800 and Radeon9700 Pro and which ran only, if FSAA and AF remained disabled. The FX5200 is here not competitive, the technically advanced shaders will hardly be enough look at for more than a few demo effects in DirectX9.
Villagemark D3D v1. 20
This benchmark was developed by PowerVR and was used to illustrate the benefits of Kyro2, because in the VillageMark of the Overdraw by a factor of up to 10 is particularly large. Many, especially older video cards calculate also the surfaces that are hidden by others and mean therefore only wasted bandwidth and fill rate, making this graphically actually not very expensive benchmark than you first thought but more often a stumbling block. That is why it is of the utmost importance in this benchmark, a well-functioning technique to remove hidden surfaces hidden surface removal (HSR) own.
Villagemark D3D v1. 20
Information in frames per second (FPS)
|
The Radeon9700 Pro includes component along with their HyperZ III the largest umgesetze fill rate, although the FX5800 quite close to her even though the application performance-eating AF was used.
The FX5200 truncates weaker than one would expect the pure fill rate after. Also their memory is probably a bit too modest to clocked, with 200 MHz with one, but twice as expensive, to be able to keep up Radeon9500.
FableMark D3D v1. 0
The Fablemark was also developed by PowerVR only is used despite a very high share in Overdraw of the display of the strengths of the KYRO chip regarding the stencil buffer.
Of course the performance of the stencil calls also on all other cards strongly so that this test can be an indication of upcoming games, insert a Z / stencil-only pass prior to actual rendering, to prevent advance any Overdraw. John Carmack of id Software programmed just such an engine.
FableMark D3D v1. 0
Information in frames per second (FPS)
|
The stencil-performance of the FX5800 is strong, the additional 4 pixels per clock step that can process them exclusively in Stencil-and Z operations help you here.
Also the FX5200 can say here quite well, as long as no FSAA and above all no AF in the game come. In the absence of here almost all textures to the direct view, the AF tweaking of the Radeon grab here 99% and they hardly lose performance by addition sequencing the anisotropic filtering. The GeForce FX is so of course plans.
Games benchmarks
Comanche 4 demo
Comanche 4 is an also pretty new game, which makes excessive use of pixel rags, as well as a large amount of polygons to show the detailed landscape, as well as beautifully animated water and reflective surfaces. On average 30 fps the polygon throughput is round 6 million per second, a feast for the TnL units. The benchmark allows for unfortunately no 1152 x 864, so that this resolution here must account for.
Comanche4 demo
Information in frames per second (FPS)
|
In Comanche4, the game included in the two GeForces, can contact the Radeon very well in the scene. No stitch is here to pick up the cards with nVIDIA chip, and in particular the FX5200 disappointed here with relatively weak performance in this benchmark is CPU limited.
DoD demo – flyby
Only the demo of popular ego-shooters available, which has the advantage stood us that everyone without purchase of the full price product can understand the benchmarks for this benchmark.
Although released only in the last year and graphically quite impressive, tournament draws the DirectX 7 registers unreal 2003 just once. It is certainly true that PixelShader of DX8 compatible graphics cards be used, however this is done solely for the performance increase. The game should look just as on a DirectX 7 Hardware TnL enabled map, as it does on DirectX 8 cards.
The benchmark integrated in the demo consists of two, actually more four parts. There will be one after the other, with each in the meantime erfolgendem engine restart the outdoor levels Antalus and the level of indoor asbestos durchflogen first without enemy action in the so-called Timedemo mode, so as soon as possible, and the average from this educated delivers the “FlyBy”-score.
DoD demo FlyBy
Information in frames per second (FPS)
|
In this test, none of the two GeForce can prevail in the settings chosen by us. The FX5200 must fear even to be obsolete, without themselves having switched FSAA or AF by Radeon9500 with active quality settings.
As DoD requires a strong vertex has the Radeon9500 without any doubt, one of the bottlenecks could be here, should however, according to rumors, slimmed the vertex at the nV34 quite down was be.
Serious Sam SE OGL
Serious Sam is well known already from the last video card tests. But also there the frame rates are shooting now in dizzying heights, and this, we have decided on the successor’s, also conveniently allows the direct comparison of OpenGL and Direct3D which despite the very demanding quality settings, because he after a patch both supported, what we ourselves have also the same advantage.
Serious Sam SE OGL v1. 07
Information in frames per second (FPS)
|
Serious Sam in second incarnation see the FX5800 in front, only in 1280 with FSAA and there is no way to Radeon9700 Pro passes AF. The FX5200 beats here for a map, only around 100€ costs surprisingly well, as long as you dispense FSAA and AF in this game.
Comment
Without wanting to convey more than a first impression now one can say that nVIDIA with its CineFX cards first generation no longer so much Crow can, as in the past. A year ago still you could almost be sure that a new card from nVIDIA also performance perspective won 95% of the benchmarks.
Of these 95% dominance you will need to goodbye, the Canadian Red Giant is awakened and has repelled impressively with the R300 series from last year. So vehemently that also the GeForce FX roofed with linear advance praise such as with Hähme may overtake only ascending but barely.
We are looking forward, which details the detailed examination that will promote.
Additional Computer Articles from "Graphics cards"
- Inno3D GeForce 2 tornado ultra.
- NVIDIA GeForce GTX of 560. MSIs twin Frozr II and Zotacs AMP!
- Versteckspiele at ATis texture filter. What ColorMipMaps do not reveal
- ATi catalyst 10.8. The August version is here
- ATi catalyst 8.11. surprise, a new driver!
- nVidia GeForce 8 m for notebooks. DirectX 10 and PureVideo HD
- Beta-detonator version 40.41. nVidias last rebellion?
- GeForce 8800 GTS 320 OC RoundUp. Which manufacturer makes the race?
- Inno3D tornado GeForce2 MX400.
- 3 x NVIDIA GeForce GTS of 450th EVGA, MSI, and ZOTAC in the ring
Recent Computer Articles
Wood and plasma tv stands: the epitome of style and sustainability in the scenario
Continue Reading | 13 views
Collaboration between Konica Minolta and Komori Corporation for commercial printing market
Continue Reading | 14 views
Income tax is obligatory and pre-registration
Continue Reading | 17 views
Full live interview strategy linking the
Continue Reading | 20 views
Insurance that if divorce becomes a single mother?
Continue Reading | 14 views