ATI Drivers Geared For Quake 3?

alrz1 writes: "HardOCP has posted an article wherein they accuse ATI of writing drivers that are optimized for Quake 3, just Quake 3, and only Quake 3. Apparently, using a program called quackifier, which modifies the Quake3 executable by changing every "Quake" reference to "Quack" and then creating a new executable called "Quack3", they have demonstrated to some extent that the Quack3.exe benchmarks are around 15% slower than with the original Quake3.exe (same box, os, drivers, etc). The slant seems to be that there is something inherently wrong about writing game-specific optimizations into drivers, if in fact this is what ATI has done. I think this is perfectly acceptable: Quake 3 is the biggest game out there on Windows, and if ATI has invested a little extra time into pumping a few extra (meaningless) frames out of your Radeon 8500, is this really an act of treachery?"

A paralell question

[OmegaDan]

When intel optimizes adobes plugins at the expense of amd processors -- so they can use it as a benchmark -- thats ok as well?

I'd have to say yes...

[egomaniac]

This is benchmark manipulation more than trying to give customers benefits. They know perfectly well that Quake 3 is used as a benchmark, so they artificially inflated their scores.

This is nothing new, and I don't think the fact that they're catering to a real program rather than an artificial benchmark makes it any less reprehensible.

Since I never Run Quake

[Royster]

I think this is perfectly acceptable: Quake 3 is the biggest game out there on Windows, and if ATI has invested a little extra time into pumping a few extra (meaningless) frames out of your Radeon 8500, is this really an act of treachery?"

Yes it is. It's writing for the benchmark rather than writing for the user.

I'm reminded of a Richard Feynman quote "For a sucessful technology, honesty must take precedence over public relations for nature cannot be fooled."

Just a guess, but...

[Ether]

It's possible that a specific driver feature or features causes Quake 3 to run 15% faster, but at the expense of making the system unstable or unreliable in other games. It would make no sense for ATI to make the cards run as fast as possible on Quake 3 and only Quake 3.

It's also possible that the Quack-quake transfer screwed something up in Quake- I'd be interested to see how a quackified exe performed on NVidia's chipset.

See this statement at the end of the text:
John B. Challinor II APR - Director, Public Relations at ATI Technologies Inc. "ATI optimizes its drivers on many different levels, including the application level, the game engine level, the API level, and the operating system level. That is, some optimizations work only on specific games, while others work only on specific game engines or only on specific operating systems. In the case of Quake III and Quake III Arena, we were able to achieve certain optimizations specifically for that game, as we do for other popular games. "

Bah, I don't even see where the "Optimizing for Quake 3 only" comes in. The Quake series has been and still is the benchmark of 1st person, 3D FPS graphics.

That being said, it would be convenient to have a checkbox in their control panel "(X) Enable unstable 3D support. May speed up certain apps, may cause problems. Use at own risk."

This is not simply optimizing for the game.

[amohr]

Optimizing your card/drivers for the popular drawing method is the natural thing to do. Even optimizing for the way a specific benchmark draws is fine.

The problem here is that it appears ATI has gained performance by reducing the image quality -- forcing a reduced texture resolution specifically in Quake 3. Compare the screenshots shown on the site. This means comparing their benchmark scores on Quake 3 with other cards is meaningless -- their card isn't performing the same task. This was a bad decision on ATI's part.

Alex Mohr

Most reviews don't use CounterStrike for benchmark

[ergo98]

ATI knows that just about every review compares cards primarily based upon Quake 3 (looks at any of a large number of sites to see this), often under the premise that it's totally relevant because so many current and up-and-coming games are based upon the Quake 3 engine.

Re:Since I never Run Quake

[StenD]

It's not even writing for the benchmark. If it was writing for the benchmark, then functions heavily used by Quake would be more optimized, but it wouldn't matter if the program which used the functions was named "quake.exe" or "quack.exe". This sounds like a marketing payoff, where the publisher of Quake is paying ATI to slow down competing games. I can't think of any other rational explanation for the drivers to care about the name of the executables.

Given that, is it really wrong?

[Nindalf]

Different video cards give different quality video, so benchmarks have to take that into account. Some are butt-ugly or just plain wrong, and competent hardware reviewers mention that.

They always have to make trade-offs between frame rate and image quality, what makes tweaking this trade-off for certain games necessarily some kind of trickery?

Id's games have always tended to be a bit freakish, based on unusual, privately researched approaches. Maybe the standard approach isn't perceived as being as playable for Quake 3.

Ideally, they could tweak the system for every individual game, but maybe it's just a case of focusing such efforts on a particularly popular title. Others have pointed out that there are more popular high-performance games, but it would make sense that the default would be optimized for the most popular games, and exceptions coded only for those nearly as popular but different enough for the default settings to be sub-optimal.

Optimizing drivers

[Squeamish Ossifrage]

It seems that this is not necessarily optimizing for Quake3 at the expense of other software. When making design decisions, in some cases one option is simply better than others. e.g. Choosing a radix sort [uwa.edu.au] over a bubble sort is basically all benefit and no cost. However, other decisions will improve performance for some tasks while reducing it for others. Think about Amdahl's Law [duke.edu] for instance. Or consider adding an index [nec.com] to a field in a database system improves the performance of queries but reduces the performance of inserts and updates. In optimizing a system, your goal is to maximize it's performance in actual use. That means that the more you know about the expected use, the better design decisions you can make. If your database is being used to log transactions, and inserts will outnumber queries, you don't put the index on. If it's being used for census data, and will be updated seldom or never, but querried frequently, you do. That is to say, you use what you know about which operations will be requested the most often, and favor those at the expense of less-frequently used operations. The problems is that you often don't know which operations will be used most frequently at design time. If, however, some of these decisions can be made at run time, you may have more information available, and will be able to make better decisions.

Which gets us back to the issue at hand. I don't know anything about the inner workings of the Radeon driver, but there are probably a number of similar tradeoffs involved in its design. The most reasonable interpretation is *not* that Radeon has optimized for Quake 3 at the expense of other programs. If that were true, it would run at the same rate whatever it were named. The better explanation is that when the driver knows what program is being run (such as Quake 3) it optmizes itself to the known characteristics of that program, and when a program which the driver knows nothing about (such as "Quack") is run, it uses default settings.

Thus, it's not necessarily favoring Quake 3 over other applications, but is instead using optimizations for for known programs which are not available for unknown ones. There's nothing in this article to indicate that similar optimizations haven't been made for Counter Strike, Half Life, or any other popular 3D programs.

ATI repeats itself

[D3]

Many years ago I bought an ATI expert@play card because it had good Quake benchmark numbers. Guess what, they had written the drivers for that benchmark and the card itself couldn't actually play games at the speed the benchmark indicated. Nothing new here. This is also why I stopped buying ATI cards.

My simple test case...

[Cinnamon]

ANY company, ESPECIALLY a computer hardware company, will trumpet any 'optimizations' to anyone who'll listen in multiple press releases. Stupid little minutiae that make no difference in frame rate for the average user (AGP, anyone?) are announced as if they were the next generation in computer 3d graphics.

So. If ATI didn't think they were doing anything wrong, there'd be something like a sort of freaky stepchild of an iD/ATI agreement where ATI would plaster "Quake 3 optimized!" all over their boxes and take underhanded swipes at Nvidia et al. in their press releases about it being an exclusive.

They didn't. So it's clear, to me, that regardless of what the Slashdot/HardOCP/etc. community thinks, ATI thought it was scummy enough to keep it under wraps, AND make a non-statement regarding it once they'd been caught.
Pretty damning, I'd say. I can't wait to hear what Tom has to say about this. (Or has he spoken up already?)

Yes, it is.

[supabeast!]

"...and if ATI has invested a little extra time into pumping a few extra (meaningless) frames out of your Radeon 8500, is this really an act of treachery?""

Quake ]|[ is THE standard for PC game benchmarks. John Carmack's engines are generally regarded as the best and fastest in the industry, and test overall performance of a system without getting bogged down on the CPU like other engines do. The Quake X engines also tend to support just about every performance enhancing feature they can (Even if the games themselves may not take advantage of it.). Quake X engines also tend to be the most OpenGL compliant engines around - something that figures greatly into why ATI would do this.

By focusing driver development on Quake ]|[, ATI is able to produce a card that will perform very well on the standard PC benchmark. Honestly, I would rather have a card that performs well on any system out there. ATI has always had horrible problems with OpenGL performance caused by weak drivers, and this has long been one of the biggest criticism of their cards. By rewriting the driver to show a great amount of Quake ]|[ performance, ATI is able to convince potential buyers that they have been fixing the OpenGL code; which if Kyle's speculation is correct, is probably one of the sleaziest things in the history of computer hardware.

I will be keeping a close eye on this one in the next few days. If this is true, I will be changing my plans to buy a new Radeon to buying a new nVidia card - because nVidia has never given me such a reason to distrust them. On top of that, nVidia drivers are custom hacked for specific cards by other vendors, so if nVidia did try this, people would leak the truth.

This has the potential to really harm ATI. If ATI loses the faith of gamers, OEMs will continue to abandon ATI for nVidia. At a time when the global economy is already faltering, ATI does not need any lost sales, and if they look weak they could lose the support of companies like Dell and Apple that are already moving to nVidia.

Re:No way

[egomaniac]

This goes for everybody else who responded to me as well:

READ THE ARTICLE BEFORE POSTING.

Image quality is worse in Q3 due to the optimizations. If you do the Quack3 rename, the image quality gets better and the frame rate gets worse.

You're telling me that you're okay with a graphics card manufacturer deliberating reducing the image quality of Q3A in order to get better frame rates, when it just-so-happens that Q3A framerates are an important benchmark? And not giving you any indication (other than reduced quality) that this is happening, nor any way to change it?

I stand by my original view.

Re:Wha??

[Shostykovich]

Remember, they run a LOT of benchmarks on video cards. Q3 is a common benchmark program... lots of people buy cards based in part or in whole on Q3 performance, under the assumption that Q3 performance is fairly representative of the card's performance in other games. So if ATI is skewing results only for Q3... well that's not "wrong", but testers and buyers NEED TO KNOW THIS that so that they can interpret Q3 benchmarks accordingly. I applaud HardOCP for raising this important issue.

The only reason potential customers have for paying attention to Quake3 benchmarks is because thats what they plan on running. For the typical non-gaming home buyer, Quake3 benchmarks are just as meaningless as theoretical memory bandwidth. For gamers, Quake3 benchmarks matter because thats what they'll play. People who refer to specific benchmarks are intentionally doing so, and the benchmarks they pay attention to are chosen entirely because they are relevant to their plans for usage.

Of course, some people are always impressed by "big numbers", in which case your agument applies to every single hype-word on the boxes, in which case the Quake3 benchmark is only one of a huge number of fabrications that marketing will place on a box.