ATI Catalyst 8.12 Performance Analysis

galaxy s3 covers

 

9

So last week ATI was kind enough to send me over their new Catalyst Drivers, version 8.12, for testing, not only does it include several improvements but also the new AVIVO Video Converter Program that utilizes the ATI Graphics Chip for faster encoding of videos. This version isn’t due out until the 10th, but I’ve been hard at it testing the performance in several games for comparison between version 8.11 and version 8.12.

So read on to learn what I learned, I tested with many games including FarCry2, Crysis, Crysis Warhead, Call of Duty World at War, Fallout3 and Spiderman Web of Shadows as well…

Here’s a quick overview of the new release:

Highlights of the ATI Catalyst™ 8.12 Windows driver release include:

New Features:

ATI Stream support:

*With the release of ATI Catalyst 8.12 users can unlock the potential of accelerated applications enabling ATI Stream technology for millions of ATI Radeon™ graphics processors worldwide

*ATI Stream harnesses the tremendous processing power of the graphics processing unit (GPU) for high-performance, data-intensive computations over a wide range of scientific, business and consumer applications.

*ATI Stream enables the hundreds of parallel cores inside AMD graphics processors to accelerate general purpose applications.  These capabilities will allow ATI Stream-enabled programs, for a variety of different tasks and from a growing number of software vendors, to operate with optimized performance or with new functionality.

*The CAL (Compute Abstraction Layer) is now include in the Catalyst driver package. CAL is the enabling component for ATI Stream

ATI Stream application:  ATI Avivo Video Converter

* Using ATI Stream technology, owners of ATI Radeon™ HD 4800 Series and ATI Radeon™ HD 4600 Series of graphics cards can take advantage of this video conversion tool (found in Catalyst Control Center Basic View) to achieve incredible performance levels in video transcoding.

ATI Catalyst Control Center – Noise reduction for progressive video content:

* This release of Catalyst™ introduces a new advanced video quality control:  Noise Reduction for progressive video content, which removes ghosting artifacts while preserving details of the original video.  The slider control can be found in the All Settings page of the Avivo Video tab found in the Catalyst Control Center.  

Performance improvements:

Catalyst 8.12 Preliminary Performance Notes

Crysis + 2-7%
Crysis Warhead + 2-3%
Devil May Cry 4  + 1-6%
Far Cry 2
+ noAA scores are fixed for Crossfire configs; this is worth +20% on slower cards, and as much as 70% on faster ones
+ 5-10% gains when AA is enabled

FEAR + 2-6%
Left 4 Dead + 2-4%
Lost Planet Colonies + 3-10%, mostly in Area 2
Prey + 2-5%
STALKER Clear Sky + 5-10%, primarily on Crossfire configurations

————————

Now onto my part of all of this….

First up the basics of  my system:

Intel C2Q6600 @ 2531Mhz

4gigs of OCZ Fatal1ty DDR2 memory @ 5-4-4-15

Diamond Radeon 4870 1gig

Vista Ultimate x64

of course I’ve got hard drives in there and DVDRW drives as well, but you don’t need the particulars of those.

Windows Vista is fully patched and all games have the most recent patches on them.

Ok, for most games I used FRAPS to record the framerates and then averaged them out, I played the same section three times over to try and get some semblance of equality in testing.

For some games I used the built in benchmarks, like FarCry2, Crysis and Crysis Warhead.

First up we’ll check out SiSoft Sandra 2009 Video Rendering Benchmark using DX10.1

Benchmark the graphics performance of the video adapters.

The test involves the generation of Mandelbrot Set fractals that are used to realistically describe and generate natural objects such as mountains or clouds. Such instructions are used for audio and video encoding and decoding, games and other applications where the same operation needs to be applied to large amount of data

Float Shaders (Pixels/s) – higher results are better, i.e. better single (32-bit) floating-point performance.

Double Shaders (Pixels/s) – higher results are better, i.e. better double (64-bit) floating-point performance.

sandra graph

We can see here that actually version 8.12 scored a bit lower than the 8.11 drivers…

Next, I figured why not run 3dMark06:

3dmark06

Very, slight increase, nothing to get excited about at all…

Now we have Far Cry 2:

Settings DX9:

Demo(Ranch Small), 1680×1050 (59Hz), D3D9, Fixed Time Step(No), Disable Artificial Intelligence(No), Full Screen, Anti-Aliasing(2x), VSync(No), Overall Quality(Optimal), Vegetation(Very High), Shading(Very High), Terrain(Very High), Geometry(Very High), Post FX(High), Texture(Very High), Shadow(Very High), Ambient(High), Hdr(Yes), Bloom(Yes), Fire(Very High), Physics(Very High), RealTrees(Very High)

Setting DX10:

Demo(Ranch Small), 1680×1050 (59Hz), D3D10, Fixed Time Step(No), Disable Artificial Intelligence(No), Full Screen, Anti-Aliasing(2x), VSync(No), Overall Quality(Optimal), Vegetation(Very High), Shading(Very High), Terrain(Very High), Geometry(Very High), Post FX(High), Texture(Very High), Shadow(Very High), Ambient(High), Hdr(Yes), Bloom(Yes), Fire(Very High), Physics(Very High), RealTrees(Very High)

farcry2

We see a very small increase when moving to the new driver version on the average.

Now I recently got Spiderman Web of Shadows for review, and I figured I could bench it while trying it out.

There’s really no settings for Spiderman at all though.

1680×1050 Resolution

Shadows Enabled

Particle Effects Enabled

Special Effects Enabled

V-Sync Disabled

spiderman

We see a large jump in frame rates when going from the 8.11 to the 8.12 drivers with Spiderman. With Spiderman it’s hard to make the game exactly the same every time you play it as things constantly change, I tried to do everything exactly the same each time I ran though it.

Next up is Crysis:

1680×1050 – High Settings – 2X AA

crysis

We see a slight increase but not much really here.

Now we have Crysis Warhead:

1680×1050 – Gamer Mode – 2X AA

crysis warhead

Not much of a real difference at all here either.

Next up we’ve got Call of Duty World at War, I used FRAPS of course and I played from the Beach to the Beach Landing 3 times each.. fun-fun

1680×1050 Resolution
Aspect Ratio – Auto
4x AA
Sync Every Frame – No
Specular Map -Yes
Ocean Simulation – Yes
Dynamic Foliage – Yes
Bullet Impacts – Yes
Number of Corpses – Medium
Texture Filtering – Automatic
Texture Quality – Automatic

cod ww

Here we see the averages do go up using the new 8.12 drivers, not sure what’s up with the minimum dropping like that though, I tried to do the exact same thing each time, but I may have done something a bit differently. Still though what we want to look at is the Averages, and they do go up quite a bit using the new Catalyst drivers.

Finally I’ve got Fallout 3

1680×1050 16:10 Widescreen Resolution
4X AA
15X AF
HDR is on
Vertical Sync is on
Texture Quality – high
Radial Blur Quality – high
Depth of Field is on
Transparency Multisampling is on
Decal Cap 10

Water, everything is set On and High
Shadows, everything is High and Max
View Distance, everything is Max
Distance LOD, evertinhg is Max

fallout3

We see a big difference here between the versions of the drivers, pushing the average way up

Of course the most important thing here is this new version of the Catalyst Drivers is the Video Converter, and honestly it’s hit and miss with it, sure it’s fast, very fast to convert movies. Only problem is though I ran into problems with it, sometimes my video would be converted with no sound, sometimes there would be sound but no picture and other times it would be all scrambled, the video I mean would be.

What I did was try several things, .AVI to .MPEG, .AVI to .WMV and .WMV to .MPEG.

What I wanted to do was give you some nice graphs with times and all that, but I can’t because it truly didn’t work correctly every time for me, so that would be wrong to give you results based on those.

Originally I used three different video converting programs and was going to compare their times to AVIVO times, sure it’s fast, like going from  20+ minutes to convert a 615mb WMV file to MPEG using Daniusoft Convertor to doing the same thing in under 5 minutes using Avivo Video Converter. That worked perfectly though, so great.

Next I tried converting a 350mb .AVI file to .WMV, using Arcosft Media Convertor and it took a little over 7 minutes, while using Avivo it took under one minute but I had audio and no video.

Yes, ok it’s new, so there’s bound to be bugs and I’m sure we’ll be seeing a hotfix for it soon for any of the problems, but still it would be nice if it worked somewhat.

Anyway, here’s some screenshots for you of how the whole process looks and works, you’ve got to be in Basic Mode of Catalyst Control Center though to access the Avivo Program.

1 2

 3 4 56

7 8

Overall the drivers themselves do give you a bit of a performance gain, and the Avivo is a very cool feature, but I would like to see it have options for changing sizes, bitrates etc for the end result videos. Yes there are choices for Sony PSP and iPod Video so they should be the correct size, but I would still like to have full control over the final output.

 

 

 

 

 

This driver will be posted live on AMD.COM December 10th, 2008

  • Pingback: Wednesday Reviews Roundup .:. 10-12-2008 « Xtremegears()

  • DAN

    Good job!

  • Riedwaan

    Im really impressed with the Call of Duty WaW scores, because I know it’s optimized for nVidia and a 10fps jump is quite significant for us ATI boys.
    I just wanted to know if i’ll be able to see such increases in framerate with my HD 3870

  • 4870 512mb User.

    I hope that it can make a big jump in performance while playing GTAIV! :)

  • Bojcha

    wtf .. where did you find 8.12 Catalysts .. ? or you tested a 8.12rc3 !?

  • kristofer

    The answer to your question is in the beginning of the article.. ATI gave them to me before the release date specifically for testing and review…

  • Pingback: CATALYST 8.12 ????? - IOPaneL()

  • Jools

    ATI 8.12 official drivers now available from ATI webpage.

  • Pingback: Catalyst 8.12 Performance - Xtreme CPU()

  • spdhax1

    be nice if you did some crossfire testings, since the scaling was claimed by ati to be at 57% (i think) for far cry 2.

  • A

    Vsync ON on Fallout 3 test? Best way to screw a test and give unrepresentative results.

  • han

    This CPU is too slow to test this card. C2D over 3.4ghz would scale better with r4870. Card is blocked by cpu so results are not good.

  • kristofer

    Which C2d runs at 3.4ghz without overclocking it? Not everyone has the top of the line CPUs, and considering the PCU I’m running is one of the most popular out there, I think it scales very well with ‘common’ systems

  • P4 + 1950pro agp

    So I still have an old P4 northwood @ 3.55Ghz with 1.5GB OCZ ram on an Abit Max3 with a 1950pro.
    I haven’t really benched these much but from what I can tell you is that these drivers lowered my 3dmark06 score from by about 40pts but where it counts is in FarCry 2… the score went up from an average of about 24fps to about 31fps. Nice jump for an old system like this.

  • cd

    pretty lame performance increases.

    And that AVIVO video transcoder has been around for ages now!! looks exactly the same as before, and seems to be next to useless as before aswell (hardly any formats actually come thru working after transcoding.)

    Come on ATI i know you can do better than this.

  • capricorn

    Hi Kristo

    I’ve noted you only paid lip service to actually ‘testing’ AVIVO’s supposedly new and improved functionality due to Stream Computing: you only showed screens of the process. A proper review should include a comparison between different version of AVIVO to SHOW the actual claimed improvement.

    To whet your interest, I did a comparison myself between the AVIVO versions in catalyst 8.10 and 8.12 RC3, and the officially released catalyst 8.12.

    Hardware:
    AMD Phenom 9600+ BlacK Edition (stock clocks)
    DFI 790FX M2R
    Sapphire HD 4870 X2 2GB
    4GB Crucial Ballistix Tracers
    Vista Home Premium SP1

    I encoded a 348MB xvid video file using the Portable Media profile in AVIVO, with quality setting set to High.
    These were my results:
    cat 8.10 2m13s
    cat 8.12 RC3 2m00s
    cat 8.12 2m16s

    WTH?

    Also of interest, is the fact that CPU utilisation was EXACTLY the same for all three tests. Previously. AVIVO only used CPU power, and since Cat 8.12 had exactly the same CPU utilisation profile as previous AVIVO versions, the only conclusion I can make at this stage is that Stream Computing via the Graphics card IS NOT WORKING in catalyst 8.12 (might be a function of my specific hardware setup, but I have my doubts…), and is STILL running ONLY off the CPU.

    I’ve logged a fault with ATI, but have yet to receive any communication with them on the phenomena i’ve observed. I am hoping that independent reviewers as yourself might be able to forthcoming with a quicker response.
    I look forward to that response.

  • kristofer

    Hi Capricorn, Yes I only mentioned I tested it, and I also said that I didn’t go into any real testing because I couldn’t get consistent results. I wanted to test it, I actually used three other programs before testing the AVIVO as I was going to use those results for comparison, but I couldn’t get Avivo to give me consistent results every time, it basically didn’t work for me every time, so I figured why post results that just weren’t able to be achieved consistently by the end user.

    I’m also in touch with ATI, and am waiting for a response about that and a couple other small things I found. Once I get a response I’ll post it here.

  • capricorn

    thanks a much. will be amusing to see who gets a response first ;)

    but was your source video, and to which format were u trying to transcode to? Also, which quality setting you use?

  • capricorn

    Hi Kristo: any joy running additional benchmarks? I know a lot of ppl are not really interested in this, so I will understand if you didnt give it further thought. I even contacted Fudzilla, and Fuad’s response was typical: he simply ran a single benchmark comparing AVIVO to competing solutions, which in no way tests AMD claims that video transcoding should now be faster because it’s being done on the graphics card. You hAVE to compare different avivo versions to test that claim, yet no one has done it, or seems interested in doing it.

    Also, I’ve raised that ticket with ATI support, but no joy just yet. Still holding out they will respond.

  • kristofer

    Hey, personally I’m a relatively new ATI guy… been running nvidia SLI for a couple years now, this is the first time in a while I’ve run ATI in my main rig…. and I’ll be honest with you I don’t have the time to install old drivers and re-run benches, I’m swamped with stuff going on…but it’s an idea that maybe I do do next month when things slow down.. but I can’t promise anything

  • kristofer

    here’s something that just popped into my inbox:

    We recieved a lot of feedback in regards to features, quality and performance that we wanted to address.

    For answers to our most common questions, we are providing the below Q&A.

    Q&A

    Q: What are the specific issues users of ATI Video Converter may encounter?

    A: Users running 32-bit software on 64-bit systems are encountering issues getting quality output from the encoder and 32-bit users have also experienced some anomalies in the playback. It was not released for support on 64-bit operating systems. Our engineering teams have been alerted of the 32-bit issues and are currently working to address them. We have a defined roadmap for the application and plan to add new baseline functionality, ease-of-use and stability in coming versions.

    Q: Why are reviewers seeing so little GPU processing during transcoding?

    A: The ATI Video Converter uses the GPU for only a portion of the video encoding. Specifically, the GPU currently offloads only the motion estimation portion of the encoding pipeline which is the most compute intensive part of the encoder. The GPU is particularly well-suited for this task. Given that a fixed workload is being offloaded to the GPU the load on the GPU may be relatively low or high based on the specific model of GPU.

    Q: When will the stability and quality of the ATI Video Converter improve in a new release?

    A: We have a defined roadmap for the application and plan to add new baseline functionality, ease-of-use and stability in coming versions. We are evaluating now when we can most quickly offer those improvements, and will update you as soon as we have more information.

    Q: Why did AMD release such a buggy software application into the market?

    A: The ATI Video Converter is a basic utility that introduces users of Catalyst to the capabilities of Stream processing on 32-bit operating systems. While it isn’t perfect, The level of functionality and quality is appropriate for maximizing the conversion speed on the most widely used video formats. Improvements are already being made and users can expect to have a much higher quality experience with future revisions.

    Q: Why does ATI Video Converter not work on 64-bit operating systems when so many PCs are moving in that direction?

    A: While the move to 64-bit is well underway, 32-bit OS’s satisfy the mainstream consumer and the ATI Video Converter was to fill that segment first. We are evaluating our roadmap for future improvements, including the possibility of offering a 64-bit solution.

  • Rick

    Glad to see the performance increase for the 4870. I think I saw an improvement with my MSI HD4850 running at 685/1100 clocks. Mostly testing with Mass Effect and Grid, but I think I’m still CPU limited but running at 1920×1200 with maxed detail settings. I may need to go back and test the older driver for qualitative comparison.

  • Ntreik

    I would like to ask, how was it possible to install dx10.1??

  • Arne

    Will this driver work on my ati mobility readeon x700 for laptop?

  • Noel

    And this is the main reason I love http://www.technogog.com. Fascinating post.

Divoom Launches Airbeat-10 Bluetooth Speaker for the Shower and Biking

Divoom Launches Airbeat-10 Bluetooth Speaker for t…

Kristofer Brozio
Well here’s what I thought was an odd combination, a Bluetooth speaker that can go from the shower t…
Read More
5 days agoDivoom Launches Airbeat-10 Bluetooth Speaker for t…
Boreas Announces Echo and Monterey Waterproof Series Bags

Boreas Announces Echo and Monterey Waterproof Seri…

Kristofer Brozio
I’ve always had a thing for good bags, maybe it’s because most of my life I carried a back pack arou…
Read More
3 days agoBoreas Announces Echo and Monterey Waterproof Seri…
FTC Says AT&T Has Misled Consumers with 'Unlimited' Data Promises

FTC Says AT&T Has Misled Consumers with ‘Unlim…

Kristofer Brozio
Wow, so they’re just realizing this now after how many years that AT&T has been throttling us?&n…
Read More
4 days agoFTC Says AT&T Has Misled Consumers with ‘Unlim…
Turtle Beach Stealth 500X Fully Wireless Gaming Headset for the Xbox One Now Shipping

Turtle Beach Stealth 500X Fully Wireless Gaming He…

Kristofer Brozio
Turtle Beach has just announced the Stealth 500X which is the first fully wireless headset for the X…
Read More
5 days agoTurtle Beach Stealth 500X Fully Wireless Gaming He…
WWE 2K15 Now Available on PlayStation 3 and Xbox 360

WWE 2K15 Now Available on PlayStation 3 and Xbox 3…

Kristofer Brozio
I didn’t think wrestling was still popular, then again I just don’t pay attention to that stuff anym…
Read More
3 days agoWWE 2K15 Now Available on PlayStation 3 and Xbox 3…