Catalyst 5.6
This month, ATI is focused on broadening compatibility and improving performance.
From this release on, ATI will be offering the lasted Catalyst software for their mobile GPUs as well as their desktop solutions. Rather than require mobile OEMs to certify and release their own drivers all the time, ATI will take the reins. Unfortunately, ATI must limit the notebook PCs on which their driver will install. OEMs that focus mainly on the business computing side of the market would rather have all their clients running exactly the same software configuration and don't want to allow end users to download and install their own drivers. In these cases, working with stability is the only important concern. The result is that, even though ATI's driver will work with any of their current generation mobility parts, they will only allow the driver to be installed on a select few systems that have opted in to the process. This will mainly include companies that focus on gaming notebooks.
Improving the CCC itself was also a goal of the 5.6 release. There have been some performance enhancements as well as some interface tweaks. The new driver loads faster than the 5.5, but there is still a noticeable pause before the window pops up. Navigating the menus and options is much smoother, and almost all the laggy behavior has disappeared. This fix has been necessary from the beginning, and we are happy to see that ATI has made progress here.
On the interface side, the cursor now changes on the display manager section to show how to use the interface. For example, a mouse with the right button flashing will appear if a right click menu is available for something. The drag-and-drop interface is easy to use, but unless people know what they need to do, it isn't very useful. ATI has also enhanced their system tray context menu to allow just about any setting to be accessed. This makes simple adjustments easy to make without loading up the entire UI. Also added is the information window, which provides software version and hardware details. The Linux install process has also been greatly simplified and automated.
From a feature set standpoint, we see a couple of new things in this release. Video settings have been added that allow real-time previews of video in a side-by-side comparison. This is helpful for adjusting the deinterlace settings. Unfortunately, only one example is given and it's difficult to show the benefits of the different methods without showing specific videos that respond better to one than the other.
It will now also be easier to view video on a secondary display in full screen. Before users were required to run in clone mode in order to watch full screen video this way, but now, it will be possible in extended desktop mode as well.
ATI has also been touting some performance enhancements this time around. We've heard that there should be a general improvement in OpenGL performance and that Doom 3 would specifically benefit. ATI showed us a list of games including Doom 3 and Call of Duty that showed performance gains. The most significant gain was in Lock On: Modern Air Combat, which benefited from Catalyst AI enabling texture compression with "no visual impact" on cloud textures. We didn't verify the results that ATI had with Lock On, but we did test a few games from our suite just to see what would happen. Here's what we ended up with:
| Radeon X800 XL @ 1600x1200 noAA/AF | ||||
| Doom 3 | Half Life 2 | Wolf: ET | Splinter Cell: CT | |
| Catalyst 5.5 | 50.2 | 104 | 95.3 | 39.9 |
| Catalyst 5.6 | 50.2 | 104.1 | 96 | 39.9 |
We also ran these numbers with 4xAA and 8xAF enabled. They were even closer in performance than what these numbers show. Needless to say, we weren't very impressed with the performance improvement (or lack thereof). It is possible that we could see a bigger performance improvement as CPU overhead becomes more of a factor (in other words, if ATI decreased the CPU load of their driver, we might see low res performance increase). We just feel that if a performance gain isn't going to show under most conditions (not the least of which is heavy load), then it doesn't have a very large impact on user experience.
Facebook
Twitter
RSS
30 Comments
View All Comments
ImJacksAmygdala - Thursday, June 9, 2005 - link
The pictures are worthless... I had to put my beer goggles on just to read the menus.I am most interested in buying a card that can run Unreal3 and is the best card for home theater HDTV applications. Whoever makes that card first gets my money.
It is nice to see the video card drivers supporting home theater HDTV, but both ATI and Nvidia could do much better. Tier 1 companies could also do much better by offering software bundles and they need to market a home theater specific card package that has bundled codecs (like Nividias codecs) and player software right out of the box. I'm talking about players like Theatertek and Zoomplayer. PowerDVD and WinDVD are fine but for the money you spend on these cards it would be nice to have some real software that a user can get the best PQ for the money they spent. It should also be said that at this point every card sold should have VGA, DVI, Svideo, and Component connections. HDTV theater packages should also include things like DVI to HDMI adapters. If I'm spending $400-$550 on a card it would be nice if all I had to do is install it, hook it up, grab FFDshow and start tweaking. I could care less about all the bundled fluffware and games I will never play.
BubbaJudge - Thursday, June 9, 2005 - link
I guess all talk, or rumor, of a Nvidia Digital Vibrance equivalent has been dropped. guess I should be thankful we still have hotkeys for gamma adjustment for CS:Sfishbits - Thursday, June 9, 2005 - link
"ATI has also been touting some performance enhancements this time around as well."Then the performace increases should have been uncovered in the testing that was done. If the increases are very specific "Only in this game, on this map, at this resolution with these settings," then IMHO ATI shouldn't be bragging about them, or should at least qualify the statement. Why disappoint customers by implying they should expect an across-the-board boost in performance, when that won't match the actual experience in most cases?
"would rather have all their clients running exactly the same software configuration and *down* want to allow end users"
Not complaining, just pointing out.
DerekWilson - Thursday, June 9, 2005 - link
My appologies for the grammar.Yes, there is someone who normally edits our articles before they go live (hi Karen). Unfortunately, it's been a busy week and I was unable to get this artile to her in time for proper editing.
Thus I was forced to employ MS Word as my editor for the article.
And no more coming up with titles at 4AM for me :-)
Thank you for pointing out the errors.
Questar - Thursday, June 9, 2005 - link
Holy Crap Batman, the commas got out of their cage and are running amok on the Internet!Can you guys hire someone who knows how to write?
gsellis - Thursday, June 9, 2005 - link
Hey Derek. Did you only test 16x12? What about at other resolutions. And yes, they are now available.probedb - Thursday, June 9, 2005 - link
This is fantastic news on doing more for HDTV. It's one area that gets left behind a lot.bbomb - Thursday, June 9, 2005 - link
And for LoneWolf just change where to were.Woodchuck2000 - Thursday, June 9, 2005 - link
Anyone fancy petitioning ATI for the option of a command-line interface? I wonder how small they could make it and still retain the same technological feature set. Just imagine;CCC/>set aa="4"
CCC/>set refresh="85"
now wouldn't that be better... We could just have a few litle batch files (.ccb I'm thinking) and execute them to configure our exact settings. It'd be handy for benchmarking too!
#10 - Well spotted! I'm glad I only skimmed through the article!
bbomb - Thursday, June 9, 2005 - link
HardOCP said that 5.5 used almost 80MB of memory and that 5.6 used only 15MB for them. Im surrised that Anandtech didnt mention what theirs was. They said in 5.5 there where 2 instances of CLI.exe, one used 35 and the other used 45. In 5.6 there were 3 instances and each used around 5Mb each.