Graphic Cards Quake 4 High-end Graphics Shootout

[Blade_Runner]

Quake 4 High-end Graphics Shootout

FS has a good article and some surprises.

As you probably know, Quake 4 is built on id’s DOOM 3 engine. From a hardware perspective, the DOOM 3 engine has traditionally been very unforgiving to ATI’s DX9 hardware. In our RADEON X1800 XT/X1800 XL Performance Preview earlier this month, both X1800 cards were easily outgunned by the GeForce 7800 GTX/7800 GT, with ATI’s flagship X1800 XT falling behind NVIDIA’s lower-end 7800 GT offering by 7-9% in our testing. The X800 series ran into this same problem around this time last year in OpenGL titles such as DOOM 3 and Chronicles of Riddick. Eventually it was just accepted that ATI’s OpenGL performance was inferior to that of NVIDIA. This stigma has stuck with ATI all the way up to the present day.

Until now that is.

A little over a week ago, rumors began spreading that ATI was working on a new tool that delivered substantially improved performance to their recently launched X1000 cards in OpenGL titles such as DOOM 3, Quake 4, and many others. Some reports claimed ATI’s performance improved by up to 35% in these titles in 4xAA mode. Then, posts on Beyond3D’s forums and sites like Guru3D confirmed these rumors.



Altogether, these improvements are astounding – this is the type of performance improvement you expect in a new product, not a driver release. The scary part is these are only ATI’s initial changes, they’ve acknowledged that there’s still room for further tweaking.



Conclusion

ATI’s pulled off one hell of an accomplishment. As shown by our testing, with one simple driver update, ATI’s gone from last to first place in Quake 4 performance. There’s a wealth of data you can glean from these benchmarks.

For one thing the RADEON X1800 cards really don’t begin to shine until you turn on 4xAA. For an example, let’s take a look at the Quake 4 numbers at 1600x1200 in high quality mode. While the margin between the X1800 XT and GeForce 7800 GTX was merely 5% in favor of the X1800 XT 512MB without AA, the margin opens up to a whopping 18% once 4xAA is applied. Meanwhile, the X1800 XL goes from finishing behind the GeForce 7800 GT, to outperforming it by 10% at 1600x1200 with 4xAA/8xAF.



The tables turn a little once Quake 4’s ultra quality mode is turned on. The GeForce 7800 GT is able to outperform the RADEON X1800 XL, while the GeForce 7800 GTX pulls to within 16% of the X1800 XT. We have a feeling that the X1800 XT’s 512MB of memory is helping it maintain its lead a little here. That would help explain the X1800 XL’s surprisingly slower showing in Quake 4’s ultra quality mode (remember that the ultra quality mode is borrowed from DOOM 3, using uncompressed textures and full resolution normal maps, with a typical level taking up over 500MB of texture memory.)

Now that’s ATI’s demonstrated that they can deliver a faster OpenGL RADEON card, the onus is now on them to deliver boards to retail. The RADEON X1800 XL still hasn’t shown up on Price Watch. Both ZipZoomFly and Newegg carry X1800 XL boards from ATI and Sapphire (with prices starting as low as $380), but this is still woefully inadequate to compete with the avalanche of GeForce 7800 GT cards that can be found on the market. As we showed you earlier this week in the news, the GeForce 7800 GT can be easily purchased online with a free copy of Call of Duty 2 for as low as $340. ATI’s got to get more boards on the market in order for prices to fall.

Meanwhile, the RADEON X1800 XT can’t be purchased at retail yet. ATI has stated availability of November 5th. That’s about 2 weeks from now. Until ATI is actually able to deliver a product on time, we’ve got to take this date with a grain of salt. Sure, it’s possible ATI may be able to deliver on their promises this time, but given their track record it’s prudent to be skeptical. If ATI can deliver though, they’ve got an excellent chance at regaining mindshare among enthusiasts.

Clearly ATI’s got a strong product. The benchmarks we’ve presented to you today show they’ve got the fastest graphics card on the market for Quake 4. Now the question is how will NVIDIA respond? It’s certainly possible they may have a few tricks up their sleeve in the driver department to improve the performance of the GeForce 7800 GT and 7800 GTX as well.

Source

 

[KiD0M4N]

Quake 4 High-end Graphics Shootout

FS has a good article and some surprises.

As you probably know, Quake 4 is built on id’s DOOM 3 engine. From a hardware perspective, the DOOM 3 engine has traditionally been very unforgiving to ATI’s DX9 hardware. In our RADEON X1800 XT/X1800 XL Performance Preview earlier this month, both X1800 cards were easily outgunned by the GeForce 7800 GTX/7800 GT, with ATI’s flagship X1800 XT falling behind NVIDIA’s lower-end 7800 GT offering by 7-9% in our testing. The X800 series ran into this same problem around this time last year in OpenGL titles such as DOOM 3 and Chronicles of Riddick. Eventually it was just accepted that ATI’s OpenGL performance was inferior to that of NVIDIA. This stigma has stuck with ATI all the way up to the present day.

Until now that is.

A little over a week ago, rumors began spreading that ATI was working on a new tool that delivered substantially improved performance to their recently launched X1000 cards in OpenGL titles such as DOOM 3, Quake 4, and many others. Some reports claimed ATI’s performance improved by up to 35% in these titles in 4xAA mode. Then, posts on Beyond3D’s forums and sites like Guru3D confirmed these rumors.

Altogether, these improvements are astounding – this is the type of performance improvement you expect in a new product, not a driver release. The scary part is these are only ATI’s initial changes, they’ve acknowledged that there’s still room for further tweaking.


Conclusion
Source

omg........ help ATI if it is found out that the changes they made were overoptimizations. but i dont think these companies will commit such blunder again.

 

[Blade_Runner]

omg........ help ATI if it is found out that the changes they made were overoptimizations. but i dont think these companies will commit such blunder again.

I take it you haven't read the article yet . The "optimisations" are aimed towards hardware architectural features specifically the memory controller no software optimisations involved. Anyways good to see ATi finally correct their opengl performance.

 

[Shripad]

yep the rendering method is exactly same as before. rendering quality is also exactly same. No optimisations there.

Great to see this. Hopefully ATI improves it more

 

[Aditya]

Well. But the 7800GTX they tested on was a 256MB Version. And im Sure that 7800GTX 512MB will beat the x1800xt 512 @ Ultra Settings.

Ofcourse, its a very big accomplishment by ATi.

Hopefully this applies to my x800xl a bit too.

I wud love to see more FPS leeched outta my card.

 

[Blade_Runner]

@golden: No gains for x800xl right now but there is a opengl rewrite going on with ex-engineers from SGI doing it.

Guru3d has a similar article

http://www.guru3d.com/newsitem.php?id=3208

 

[Aces170]

Well, the optimisations affects all cards which have 256bit ring bus, what they have done is optimised the way memory controller handles the data in Opengl.

As Posted by Sireric, one of the devs in the CAT team at

http://www.beyond3d.com/forum/showthread.php?t=24628&page=2

This change is for the X1K family. The X1Ks have a new programmable memory controller and gfx subsystem mapping. A simple set of new memory controller programs gave a huge boost to memory BW limited cases, such as AA (need to test AF). We measured 36% performance improvements on D3 @ 4xAA/high res. This has nothing to do with the rendering (which is identical to before). X800's also have partially programmable MC's, so we might be able to do better there too (basically, discovering such a large jump, we want to revisit our previous decisions).

But It's still not optimal. The work space we have to optimize memory settings and gfx mappings is immense. It will take us some time to really get the performance closer to maximum. But that's why we designed a new programmable MC. We are only at the beginning of the tuning for the X1K's.

As well, we are determined to focus a lot more energy into OGL tuning in the coming year; shame on us for not doing it earlier.

Lol this is just the beggining, wonder what the NV driver team is upto.

 

[~HeadShot~]

Wow! Thiat frigging awesome news... :hap2: :yahoo:

 

[Blade_Runner]

OMG !! ATi on a roll ! :O

Some benchies with the 8.183 hotfix !!

X1800 XT - Quake 4 (Internal Demo) - Normal Rendering                    

Driver Rev      640x480   800x600   1024x768   1280x1024  1600x1200

8.173           113.1     110.6     108.0       92.8       72.8

8.173 + Mem Fix 112.7     111.6     108.2       92.5       73.0

8.183           131.1     129.5     126.2      108.3       91.6

                    

% + from 8.173  640x480   800x600   1024x768   1280x1024  1600x1200

8.173 + Mem Fix -0.40%     0.90%     0.20%      -0.30%     0.30%

8.183           15.92%    17.09%    16.85%      16.70%    25.82%

                    

                    

X1800 XT - Quake 4 (Internal Demo) - 4x FSAA & 8x AF                    

Driver Rev      640x480   800x600   1024x768   1280x1024  1600x1200

8.173           109.8     101.7      85.3       63.3       48.9

8.173 + Mem Fix 110.3     104.6      90.7       69.8       54.8

8.183           128.6     119.6     109.9       87.4       70.7

                    

% + from 8.173  640x480   800x600   1024x768   1280x1024  1600x1200

8.173 + Mem Fix  0.50%     2.90%     6.30%     10.30%     12.10%

8.183           17.12%    17.60%    28.84%     38.07%     44.58%

X1800 XT - Doom 3 (Turkey Baster) - Normal Rendering                    

Driver Rev      640x480   800x600   1024x768   1280x1024  1600x1200

8.173           140.2     141.6     140.7      121.2       95.9

8.173 + Mem Fix 139.8     142.1     139.9      118.9       93.7

8.183           148.0     152.2     148.2      122.7       95.3

% + from 8.173  640x480   800x600   1024x768   1280x1024  1600x1200

8.173 + Mem Fix -0.3%      0.4%     -0.6%      -1.9%      -2.3%

8.183            5.6%      7.5%      5.3%       1.2%      -0.6%

X1800 XT - Doom 3 (Turkey Baster) - 4x FSAA & 8x AF                    

Driver Rev      640x480   800x600   1024x768   1280x1024  1600x1200

8.173           137.7     124.8     101.8       72.2       52.7

8.173 + Mem Fix 127.5     123.8     105.0       77.9       59.0

8.183           149.1     140.1     123.4       92.3       69.1

% + from 8.173  640x480   800x600   1024x768   1280x1024  1600x1200

8.173 + Mem Fix  -7.4%    -0.8%      3.1%       7.9%      12.0%

8.183             8.3%    12.3%     21.2%      27.8%      31.1%

Performance increase without AA as well :O

 

[Aces170]

LOL any comparisions with the 7800 GTX on a similar rig ?

The optimisations is very deep API rooted. The memory controller can be tweaked a lot in D3d too...

 

[saumilsingh]

Any improvements for r300 where it's really needed?

 

[Aces170]

Any improvements for r300 where it's really needed?

Nope these improvements are in respect with the memory controllers which are programmable specific to R5xx series, though if X1300 support the optimisations is yet to be seen, as the 512 bit Ring bus is not present in it.

As quoted by Sireric from

http://www.beyond3d.com/forum/showthread.php?t=24628&page=2

There's some slides given to the press that explain some of what we do. Our new MC has a view of all the requests for all the clients over time. The "longer" the time view, the greater the latency the clients see but the higher the BW is (due to more efficient requests re-ordering). The MC also looks at the DRAM activity and settings, and since it can "look" into the future for all clients, it can be told different algorithms and parameters to help it decide how to best make use of the available BW. As well, the MC gets direct feedback from all clients as to their "urgency" level (which refers to different things for different client, but, simplifying, tells the MC how well they are doing and how much they need their data back), and adjusts things dynamically (following programmed algorithms) to deal with this. Get feedback from the DRAM interface to see how well it's doing too.

We are able to download new parameters and new programs to tell the MC how to service the requests, which clients's urgency is more important, basically how to arbitrate between over 50 clients to the dram requests. The amount of programming available is very high, and it will take us some time to tune things. In fact, we can see that per application (or groups of applications), we might want different algorithms and parameters. We can change these all in the driver updates. The idea is that we, generally, want to maximize BW from the DRAM and maximize shader usage. If we find an app that does not do that, we can change things.

You can imagine that AA, for example, changes significantly the pattern of access and the type of requests that the different clients make (for example, Z requests jump up drastically, so do rops). We need to re-tune for different configs. In this case, the OGL was just not tuning AA performance well at all. We did a simple fix (it's just a registry change) to improve this significantly. In future drivers, we will do a much more proper job.

For a translation in English either go through the thread at beyond3d, or wait for someone like Chaos to give an explanation.

R4xx has a partially programmable memory controller, but enabling these fixes on them de-grades the performance, but the Catalyst team has promised there will be some improvements for R4xx to with CAT 6 release.

R300 is obsolete tech now

 

[Aditya]

HUH!! :O :O :O...

Now i gotta Hand it to ATi.

They Rock!!!

But god damnit! Do something for the x800's!

They said that the x800 has partially programmable memory controllers didnt they ...

But What FPS is that man :O :O :O :O :O!!!!!!!

 

[Blade_Runner]

Any improvements for r300 where it's really needed?

No these are only for the r520 series. However, unlike Nv who often forget old generation cards a Opengl rewrite is on its way for older ATi card users and r520.