Tom's Hardware Linux NVidia Benchmarks

diehard writes: "Tom's Hardware has posted a set of benchmarks of NVidia cards running under Xfree86 4. They are pretty impressive - it looks like Linux has finally become truly viable for gaming."

Comments

Encourage Tom!

[molo]

Tom has taken a big step for him in dedicating a review to Linux. He has the following to say about it:

This was my first Linux hardware review and it will certainly not be my last. The first time is always supposed to be the hardest, but the most rewarding as well. Please let me know how I performed here. Was I babbling too much about Linux? Weren't there enough facts in the review? Don't you care about 3D stuff in your Linux-box? Please give me feed back under tomslinux@tomshardware.com . I will try to live up to the expectations of the Linux community, but first I need to know what they are.

We need to encourage him! Tell him some of the things we would like reviewed with linux benchmarks. Thank him for taking a big step in dedicating a whole review to linux. Only good things can come of this!

Page flipping should not be supported.

[John Carmack]

All signs are pointing towards a future without page flipping, so adding the messy infrastructure for it now would be a mistake. Don't let benchmarking furor encourage a messy code architecture.

Points:

The benefit of page flipping is decreasing as more and more computation is done per pixel to the back buffer.

In the old days of 2D scrollers, you might barely cover the screen with one pass of writes, so page flipping could double your speed over blitting.

On a typical modern 3D game that becomes fill limited, under 25% of the performance is in the blit, and often under 10% in scenes with significant overdraw.

In upcoming games that composite 20+ layers of textures, the cost of a blit is down in the noise.

Blits add flexibility. Anti-aliasing is better done through a blit operation than with a deep front buffer. Other operations, like converting from a 64 bit work pixel to a 32 bit display pixel, or performing convolutions, are also better done with blits.

Back buffers are more optimally arranged in tiled patterns, while front buffers prefer linear scans.

Basically, our back buffers are starting to look less like raster

Page flipping doesn't apply to windowed rendering unless you butcher the X server to render all 2D to multiple buffers and clip all 3D operations. I consider that a bad thing. Making the full screen rendering more distant from windowed rendering is also a bad thing.

Every implementation of page flipping brings in a class of bugs, and obfuscates several code paths. It's not worth it.

John Carmack

Whaddya mean, "finally"?

[Skald]

it looks like Linux has finally become truly viable for gaming

I beg your pardon... Nethack has looked great on Linux for years!

What about NT?

[ink]

It would be more interesting to see how Linux fares against Windows NT/2000 rather than Windows 98se. NT and Linux have much more in common than the DOS line does; and NT shares many of the high performance gaming problems that Linux has to address as well (like protected memory, premptive multitiasking, userland libraries and such). Window 98se's job is to get out of the way of games. NT and Linux still have to have running operating systems underneath.

The wheel is turning but the hamster is dead.

Re:What about NT?

[be-fan]

As far as I recall, Windows 98 has protected memory, preemptive multi-tasking, and the whole OS is basically a user-land library. (Actually, by making the OS mainly consist of a set of DLL's loaded by the application, performance is IMPROVED, since API calls don't have to do a switch into Ring0. It also means that Win98 is in a way a pseudo-exo-kernel OS.)
NT actually performs better in Quake than Win98 does.
Windows2000 solves the whole "big OS/gaming" problem quite nicely. You see, the OS needs of games are quite limited. Thus, in effect, DirectX is really as much of an OS as most games need. Since you are rarely being productive when you're gaming, some modes of DirectX allows a program to hog the system at the expense of other applications. However, since you're not using other applications while you're gaming, this isn't a problem. Quite an elegant solution really. Although, I'm against heavy OSs in general, but within the context of the problem, Win2K's solution is a pretty good one. (Though don't get me started on the code bloat. What I want is NT4 with full DirectX. Is that too much to ask?)

...but the Radeon driver will be!

[Straker Skunk]

VA is currently working on an XFree4.0 driver for the new Radeon cards. It's being developed under contract from ATI themselves, it'll use the DRI, and it'll be open-source.

Check out the adjacent article on Tom's, about the Radeon. It performs slightly slower than the GeForce in 16-bit mode, and at lower resolutions, but at high-res 32-bit modes the Radeon edges out the GeForce.

The GeForce numbers may be impressive, but me, I'm going to wait until the Radeon driver is ready. And then I'm going to show ATI just how much I appreciate an open solution.

linux 3d is weak

[small_dick]

Tom's reviews are always hard to read; mostly because of the bad grammer, spelling and logic, (like "....linux is securer than windows 2000...")

but, he is honest and doesn't hold back. that makes up for it.

Linux 3d is being handled very poorly. Several years ago, when Linux was struggling through issues with 2D, a lot of people wrote HOWTO's and webpages explaining how to cobble things together. XFree was honest about what did (and did not) work.

But now, anyone who visits the 3dfx newsgroups or the XFree site, then attempts to get their card working in 3d mode, will notice two things:

1) There are no HOWTO's or webpages dedicated to your card. When you ask a simple question, you will immediately be bombarded by three morons who say "why didn't you read through the last 712 messages! everything you need is there!"

2) XFree 4.0.1 documentation claims all sorts of miracles. But the truth is, the product does very little that is new unless you are willing to read the the 712 aforementioned messages and figure out how to disregard the 213 of then that are now outdated or innaccurate.

Many cards require a CVS download and rebuild of everything, plus a 2.3 series kernel, and many files from other sites. After you build everything and and apply all the hacks, don't be suprised if the your system locks after 20 seconds or so, like mine did.

Hate to say it, but XFree is really going out on a limb by calling this beast a "3D enabled release". It's a immature, undocumented 3D release being supported by a bunch of monkeys who refuse to properly document the 3D workarounds. The later 3.X releases were better documented, both by XFree and the card vendors, and more reliable in 3D mode.

The 2D side of 4.0.1 is stable, and the new server configuration stuff is wonderful. But this is not a quality 3D release by any means.

Re:Gaming will only undermine Linux's reputation

[91degrees]

Absolutely. Text based games are the way to go. In fact, I recenty spent $179.99 on a BSD Star Trek accelerator. Quite an investment, but you should see how fast it goes. Especially if I reduce the res to 40 columns. And Hangman is incredible. Even some of the serious apps in /usr/games/bin like morse have increased their speed considerably.

The mouse STILL SUCKS

[ywwg]

I don't care how many articles are written about how great gaming on linux is, it is still _unplayable_ without decent mouse support. I have the latest drivers from nvidia that work with xfree 4.0.1. I've tried everything from Option SampleRate to tuneps2.

In short, NOTHING WORKS. the response in Q3A is slow and jerky, unless you are standing stock still. If you turn m_filter off in windows, the mouse is still smooth because the ps/2 rate acn be cranked so high. In linux with m_filter off, it is sucky sucky sucky.

Somebody has to get Loki and XFree in a room together and FIX THIS. Check the newsgroups. check slashdot. _everyone_ complains about this, but nothing gets done. I'm tired loki saying "oh, we support DGA, it should work," and XFree saying, "we fixed it in 4.0.1." Can somebody _please_ address this problem, once and for all?

</rant> sorry.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Bzzt! Wrong.

[MostlyHarmless]

Linux didn't used to be about money ... While many people in the Linux community, including myself, didn't welcome RedHat's IPO, it shows how valuable Linux has become

Why, oh why, do people always think that open-source or software liber implies that nobody can make money off of it? Contrary to many peoples' opinions, RMS does not mind if you make money. Why is it bad that RedHat went IPO? What can they possibly do to the community? The very nature of the GPL is that no one person can run off with the code. There is no chance that they would make the kernel non-free or anything like that, like many people were claiming would happen. Instead, what has happened to the free software community as a result of the IPO?

They've hired programmers to work on the kernel
They've hired programmers to work on GNOME
They've helped bring the ideas of the open-source community to the masses.

The last is the most important of all. Hiring programmers has been a great help to ensuring the success of critical parts of the GNU/Linux system, but that by itself would have meant very little without their support of the free software community. Without freedom, it would have just been another software project. Instead, their help of the GNOME project as well as the kernel has helped prepare the platform for their most important aid to our community. That aid was in spreading our ideas to the world. Yes, they are taking more of an open-source pragmatist approach as opposed to the free software idealistic approach. But even exposure to the former will help, and eventually some of the users will open their minds towards the ethics of free software, not just the business of open-source.

Sorry if this turned out to be a rant.
--