-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
Anandtech reviews the NV35! + Doom3 Benchmarks!!!
- Thread starter Czar
- Start date
FreshPrince
Diamond Member
sweeeeeeeeeeeeet! 😀
http://images.anandtech.com/reviews/video/nvidia/geforcefx5900/board.jpg
damn that card is big
damn that card is big
FreshPrince
Diamond Member
takes out tape measure to make sure it'll fit in shuttle sff xpc
FreshPrince
Diamond Member
lol, QIII 1024x768, 1280x1024, 1600x1200 all tops out @ 223fps! :Q that's pretty sweet.
FreshPrince
Diamond Member
whew! finished reading all 30 pages!
overall, not really impressed about the fact that nVidia used "older" technology for this card. Was really hoping it would continue to use DDRII but @ 256bit + faster GPU core. Oh well, looks like NV40/RS400 for me 🙁
overall, not really impressed about the fact that nVidia used "older" technology for this card. Was really hoping it would continue to use DDRII but @ 256bit + faster GPU core. Oh well, looks like NV40/RS400 for me 🙁
FreshPrince
Diamond Member
1st thing I did was load up all 30 pages, then I start reading 😀
FreshPrince
Diamond Member
like the other dood said, Anand might not like that. I can however tell you what it's all about.
with nothing turned on, nVidia is much faster than ATI. However, with the 'trimmings" turned on, ATI closes the gap, but still beat out by nVidia.
FreshPrince
Diamond Member
that's the thing I hate about nVidia...
they always release a new card with less than perfect specs...3 months later, they release an "ultra" version. Otho, in this case, it's already called "ultra" I wonder wut they're going to call the real ultra 😉
Random Variable
Lifer
WTF? Octuple Post?!? 😱
Fixed by the Tidy Mod.
Fixed by the Tidy Mod.
Random Variable
Lifer
The 256MB Radeon 9800 won the Comanche 4 benchmarks. But Comanche 4 is a game that relies more heavily on the CPU than the GPU.
Kyle's review at HardOCP doesn't paint as pretty of a picture as Anand's. 😕
Mem
Lifer
That`s the problem with ATi drivers,they do their best to fix the problems(all credit to them) and it breaks something else 🙁 if ATi and Nvidia ever got together we would have awesome drivers and video cards 🙂.
Back on to the NV35,looks like Nvidia have fixed the main problems and has good AA/AF performance even compared to ATi 9800.
🙂
GTaudiophile
Lifer
Just read Kyle's NV35 review. He indeed wrote a better, more-detailed review than Anand! Kyle thinks that the 9800 PRO still holds the IQ crown. When it comes to performance, he says it's a mixed-bag with the cards flip-flopping their leads. For his personal system, he says he's going to keep the Radeon 9800 PRO 256MB.
GTaudiophile
Lifer
Here is HardOCP's Conclusion for those who can't access their site:
Conclusion:
There are several very big improvements in the NV35 core over the NV30, the biggest being the inclusion of a 256-bit memory bus operating with 425MHz DDR giving it 27.2GB/sec of raw bandwidth. This is a huge progression over the 16GB/sec bandwidth of the GeForceFX 5800 Ultra and surpasses the Radeon 9800 Pro 256MB?s bandwidth of 22.4GB/sec. However, raw bandwidth alone does not a card make.
The architecture is what drives the core and determines how well that bandwidth will be used. To that it is a disappointment to see that the pipeline architecture has not changed. It is still in essence, by the understood definition, a 4 pipeline GPU. It has the bandwidth, but lacks the raw power of ATI?s 8 pipeline 9700/9800. However, they have made some very welcomed improvements as far as DX9 shaders go. The NV30 had very slow Pixel Shader 2.0 performance due to only having one shader op. The NV35 has increased this now to two shader operations. Therefore pixel shader performance has been increased a great deal. This was very apparent in Shader Mark where the scores are now comparable to the 9800.
Other improvements include a new shadow volume scheme to increase performance using stencil shadows. ATI was first to implement an optimization of the z-cache to improve stencil shadows with HyperZ III+ on their R350 VPU. NVIDIA?s advantage is to render the shadows in a single-pass and to also work with their compression schemes with Anti-Aliasing. HCT is NVIDIA?s compression scheme to work at high resolutions with a high level of AA and AF.
NVIDIA is releasing their new DetonatorFX drivers that greatly improve the filtering image quality that we have seen in the past. The control panel image quality settings are easy to understand and the highest quality setting is already selected by default. These drivers are solid and much better then the first batch we had to test with on the NV30 back in January.
ATI?s Catalyst 3.4 drivers are very solid as well. We experienced no driver problems whatsoever with these drivers. Just remember, if you are going to get a 256MB 9800 Pro, you need to run with Catalyst 3.4 or higher to utilize all the RAM on the video card.
Looking at AA Quality, we came to the conclusion that the ATI 9800 Pro 256MB card still has hands down the best AA quality. You cannot beat ATI?s rotated grid with gamma correct AA. NVIDIA still uses an ordered grid sampling pattern on the NV35. They do get marks though for implementing Multi-Sampling + Super-Sampling modes. However, even though those options exist, ATI?s AA still looks better, even though its just Multi-Sampling. The only outstanding issue with ATI and AA is that we would like to see a way for alpha textures to get Anti-Aliased in D3D and OpenGL. Even if the only way is to include a Super-Sampling mode, the extra option would be nice.
Looking at Anisotropic Quality, we have to come to the conclusion that NVIDIA?s Anisotropic ultimately is done with a better technique on the 5900 Ultra. When push comes to shove, you cannot deny what the Filter Test is showing us. Still we have yet to see where the ATI AF technique gives us a negative in real gameplay. For most people the 9800 Pro?s Anisotropic is just fine, us included.
Overall, when judging image quality, we have to give the hat to the 9800 Pro. There isn?t a huge noticeable difference in the Anisotropic quality, even though ATI?s may not be filtering certain angles "correctly". Most games have straight 90 degree angles or long hallways or ceilings or walls, etc. At 16XAF the ATI 9800 Pro 256MB looks awesome in games like Medal of Honor or Battlefield 1942. On the 9800 Pro 256MB we did not notice as many jaggies whereas on the 5900 Ultra, aliasing is quite apparent when you are given a static screen shot to gaze at. Where it is really important is duing movement in gameplay. We are going to look further into this with some following subjective opinion articles. Given those facts, the 9800 Pro still holds the IQ crown, but NVIDIA is right behind them. If they can improve their AA to match the likes of gamma correction and rotating grid or jittered grid, then the tables would most likely turn...or rotate in this case.
Looking at performance between the two cards, we found that in Unreal Tournament 2003 the GeForceFX 5900 Ultra was on top in Antalus, but the Radeon 9800 Pro 256MB led in Inferno. However, CTF-Face3 showed what all that memory bandwidth can do on the GeForceFX 5900 Ultra. In a very memory bandwidth intensive situation the 5900 Ultra will excel. Serious Sam showed the 5900 Ultra on top as well. Splinter Cell was more interesting, showing the 9800 Pro 256MB card leading and by a good margin. The demo we ran is a demanding Pixel Shader oriented benchmark, and perhaps the 9800 Pro still has the faster shader speed although Shadermark would suggest otherwise. Quite possibly in the scene we focused on, one type of shader was being used that favored the 9800.
It almost seems like in every benchmark the cards go back and forth, dancing around each other. One map will run one card faster, another map will run the other one. One resolution shows a card faster, another resolution shows the other fast. It makes deciding which is the fastest card a very hard decision. Here we have an 8 pipeline card clocked at 380MHz beating in some cases a 4 pipeline card clocked at 450MHz with a lot of bandwidth. It appears in those situations that the 9800 Pro 256MB card just has a better architecture with faster shader speed. Then on the other side we do have the GeForceFX 5900 Ultra winning in some DX8 tests as well. One thing at least is certain, both cards offer awesome performance with AA/AF at high resolutions, and the 256MB of RAM does make a difference once you overcome CPU bottlenecks. If you only play games at 1024x768 and have an aging CPU, a card of these calibers will probably not help you much. But if you run at high resolutions with AA and have a fast CPU, a 256MB card can help stabilize FPS at those resolutions.
Both cards offer exceptional performance, but if I have to place my finger on which card I would choose for my primary system that I play my games on, I would have to opt for the 256MB 9800 Pro. It has better AA quality, and can play at 6XAA on a 256MB card very easily. It has also proven to be very strong in shader operations. It is not an easy decision for sure now. NVIDIA has done a great job at sizing up the issues with the NV30 and fixing those issues and making the NV35 an incredibly competitive part.
So, it all comes down to you, how much you want to spend, and what video card you need to fill your wants and desires while gaming. With the race this tight, small things might certainly tip your buying decision one way or the other. Brand loyalty, past hardware performance, and past driver support should come into play. Certainly what specific games you want to play should be weighed if possible and being able to actually purchase one is a plus as well. Both of the cards looked at here should be available for sale in quantity in 30 to 45 days.
One thing we hate to do around here is ride the fence. Currently, given the results we have here, it is hard to firmly suggest one card over the other.
Conclusion:
There are several very big improvements in the NV35 core over the NV30, the biggest being the inclusion of a 256-bit memory bus operating with 425MHz DDR giving it 27.2GB/sec of raw bandwidth. This is a huge progression over the 16GB/sec bandwidth of the GeForceFX 5800 Ultra and surpasses the Radeon 9800 Pro 256MB?s bandwidth of 22.4GB/sec. However, raw bandwidth alone does not a card make.
The architecture is what drives the core and determines how well that bandwidth will be used. To that it is a disappointment to see that the pipeline architecture has not changed. It is still in essence, by the understood definition, a 4 pipeline GPU. It has the bandwidth, but lacks the raw power of ATI?s 8 pipeline 9700/9800. However, they have made some very welcomed improvements as far as DX9 shaders go. The NV30 had very slow Pixel Shader 2.0 performance due to only having one shader op. The NV35 has increased this now to two shader operations. Therefore pixel shader performance has been increased a great deal. This was very apparent in Shader Mark where the scores are now comparable to the 9800.
Other improvements include a new shadow volume scheme to increase performance using stencil shadows. ATI was first to implement an optimization of the z-cache to improve stencil shadows with HyperZ III+ on their R350 VPU. NVIDIA?s advantage is to render the shadows in a single-pass and to also work with their compression schemes with Anti-Aliasing. HCT is NVIDIA?s compression scheme to work at high resolutions with a high level of AA and AF.
NVIDIA is releasing their new DetonatorFX drivers that greatly improve the filtering image quality that we have seen in the past. The control panel image quality settings are easy to understand and the highest quality setting is already selected by default. These drivers are solid and much better then the first batch we had to test with on the NV30 back in January.
ATI?s Catalyst 3.4 drivers are very solid as well. We experienced no driver problems whatsoever with these drivers. Just remember, if you are going to get a 256MB 9800 Pro, you need to run with Catalyst 3.4 or higher to utilize all the RAM on the video card.
Looking at AA Quality, we came to the conclusion that the ATI 9800 Pro 256MB card still has hands down the best AA quality. You cannot beat ATI?s rotated grid with gamma correct AA. NVIDIA still uses an ordered grid sampling pattern on the NV35. They do get marks though for implementing Multi-Sampling + Super-Sampling modes. However, even though those options exist, ATI?s AA still looks better, even though its just Multi-Sampling. The only outstanding issue with ATI and AA is that we would like to see a way for alpha textures to get Anti-Aliased in D3D and OpenGL. Even if the only way is to include a Super-Sampling mode, the extra option would be nice.
Looking at Anisotropic Quality, we have to come to the conclusion that NVIDIA?s Anisotropic ultimately is done with a better technique on the 5900 Ultra. When push comes to shove, you cannot deny what the Filter Test is showing us. Still we have yet to see where the ATI AF technique gives us a negative in real gameplay. For most people the 9800 Pro?s Anisotropic is just fine, us included.
Overall, when judging image quality, we have to give the hat to the 9800 Pro. There isn?t a huge noticeable difference in the Anisotropic quality, even though ATI?s may not be filtering certain angles "correctly". Most games have straight 90 degree angles or long hallways or ceilings or walls, etc. At 16XAF the ATI 9800 Pro 256MB looks awesome in games like Medal of Honor or Battlefield 1942. On the 9800 Pro 256MB we did not notice as many jaggies whereas on the 5900 Ultra, aliasing is quite apparent when you are given a static screen shot to gaze at. Where it is really important is duing movement in gameplay. We are going to look further into this with some following subjective opinion articles. Given those facts, the 9800 Pro still holds the IQ crown, but NVIDIA is right behind them. If they can improve their AA to match the likes of gamma correction and rotating grid or jittered grid, then the tables would most likely turn...or rotate in this case.
Looking at performance between the two cards, we found that in Unreal Tournament 2003 the GeForceFX 5900 Ultra was on top in Antalus, but the Radeon 9800 Pro 256MB led in Inferno. However, CTF-Face3 showed what all that memory bandwidth can do on the GeForceFX 5900 Ultra. In a very memory bandwidth intensive situation the 5900 Ultra will excel. Serious Sam showed the 5900 Ultra on top as well. Splinter Cell was more interesting, showing the 9800 Pro 256MB card leading and by a good margin. The demo we ran is a demanding Pixel Shader oriented benchmark, and perhaps the 9800 Pro still has the faster shader speed although Shadermark would suggest otherwise. Quite possibly in the scene we focused on, one type of shader was being used that favored the 9800.
It almost seems like in every benchmark the cards go back and forth, dancing around each other. One map will run one card faster, another map will run the other one. One resolution shows a card faster, another resolution shows the other fast. It makes deciding which is the fastest card a very hard decision. Here we have an 8 pipeline card clocked at 380MHz beating in some cases a 4 pipeline card clocked at 450MHz with a lot of bandwidth. It appears in those situations that the 9800 Pro 256MB card just has a better architecture with faster shader speed. Then on the other side we do have the GeForceFX 5900 Ultra winning in some DX8 tests as well. One thing at least is certain, both cards offer awesome performance with AA/AF at high resolutions, and the 256MB of RAM does make a difference once you overcome CPU bottlenecks. If you only play games at 1024x768 and have an aging CPU, a card of these calibers will probably not help you much. But if you run at high resolutions with AA and have a fast CPU, a 256MB card can help stabilize FPS at those resolutions.
Both cards offer exceptional performance, but if I have to place my finger on which card I would choose for my primary system that I play my games on, I would have to opt for the 256MB 9800 Pro. It has better AA quality, and can play at 6XAA on a 256MB card very easily. It has also proven to be very strong in shader operations. It is not an easy decision for sure now. NVIDIA has done a great job at sizing up the issues with the NV30 and fixing those issues and making the NV35 an incredibly competitive part.
So, it all comes down to you, how much you want to spend, and what video card you need to fill your wants and desires while gaming. With the race this tight, small things might certainly tip your buying decision one way or the other. Brand loyalty, past hardware performance, and past driver support should come into play. Certainly what specific games you want to play should be weighed if possible and being able to actually purchase one is a plus as well. Both of the cards looked at here should be available for sale in quantity in 30 to 45 days.
One thing we hate to do around here is ride the fence. Currently, given the results we have here, it is hard to firmly suggest one card over the other.
Yeah apparantly the peformance is very close....except where the broken ATI drivers fail.
And overall quality gives ATI the edge.
I'll be keeping my 9700 PRO till the N40 and R400's come out. 🙂
I think history might end up showing the 9700 to be the best overall card ever to come out!
And overall quality gives ATI the edge.
I'll be keeping my 9700 PRO till the N40 and R400's come out. 🙂
I think history might end up showing the 9700 to be the best overall card ever to come out!
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
-
-
