Massive Graphics Card Review

Brian Tonka writes to tell us that rojakpot has posted a pretty comprehensive graphics card review including over 240 different desktop graphics cards. With each of the vendors given their own section and using 15 different points of comparison this should be quite a starting reference for the enthusiast and casual buyer alike.

Review?

[compm375]

This might be helpful to some people, but it can hardly be called a review. It is just a list of specs. It doesn't even have benchmarks.

Oops

[phorm]

Replying to myself, I didn't notice the "NVidia" link hidden at the bottom (I looked, I swear!), which leads to XGI, etc.

Re:simple: open source drivers?

[hamfactorial]

The best so far would be the radeon 9250, which is the most recent card supported by the current (open source) x.org "radeon" driver, and has EXA acceleration in the just-released 6.9/7.0 version.

Re:Hmm. Radeon, radeon, ati, ati

[JanusFury]

There are other brands on the following pages of the article; it's just really hard to find the 'next page' link because the site's layout SUCKS.

MTexels/s

[BrookHarty]

I'm not sure if mega-texels shows true performance. I have a ATI 9700 Pro and Geforce 7800 GT, both can run games at high resolutions at the same speed, but the 7800 can run with AA/AF enabled without a performance hit.

It is nice to see where GFX cards rate in games, and Toms hardware has the best link per game. Thats why I picked a GT over a GTX for 200 dollars less.

http://www.tomshardware.com/2005/12/02/vga_charts_ viii/page18.html [tomshardware.com]
and
http://www.tomshardware.com/2005/07/05/vga_charts_ vii/page4.html [tomshardware.com]

The original article is right.

[ShaolinTiger]

It's titled COMPARISON, not REVIEW, whoever posted it to /. got it wrong, not the adsense crazy Rojakpot.

This is really cool

[fraktus]

Ok seems I am the only one to see this usefull.
My application requiere shaders v2.0 and it's really boring to always type radeon radada in google to hunt for the specs to reply to questions from customers.
Also even if it will not tell you for sure that your engine will run faster on this one or this one it will at least give you a hint.
Having the OpenGL version supported from the driver would also have been nice.

August wasn't *that* long ago...

[twicesliced]

http://hardware.slashdot.org/article.pl?sid=05/08/ 20/1912221&tid=126&tid=137 [slashdot.org] Yeah, thought it looked familiar...

Re:What I need to know is...

[poofyhairguy82]

I went through the trouble of logging in so I could give advice. I read about Linux graphics all day, and I can tell you which card has the best open driver: the ATI 9250.

Nothing else is close. Its the most powerful card on the market with open specs!

Re:simple: open source drivers?

[odie_q]

I admin a score of machines using Sapphire 9200 cards, all running with the Xorg driver. The machines are used daily (I am posting this from one of them) and I have yet to see a single problem with the driver. Granted, OpenGL is mostly used for screen savers on these boxes, but still. In my experience the drivers are rock solid.

Re:This isn't a review

[drinkypoo]

Actually, you are mostly right but made a glaring error. AGP8x allows multiple AGP slots (2 only, I think) in a single machine. Granted, that's only half as many video cards as you can have with PCIe... But dual-slot was the main draw of AGP8x, which rarely provides any performance improvement over AGP4x with sideband and fast write.

AGP vs PCI

[HalWasRight]

I have a hunch that AGP would show barely an improvement over PCI in a typical system (where the PCI bus is mostly idle), and PCI-E would only show improvement in a dual-card config.

There are a couple things you may not have considered with your hunch. First, if you are doing 3D textured graphics, then transfer speed to texture and vertex memory is key to performance, and PCI is many times slower than AGP. 10x is not "barely an improvement" in the real-time 3d graphics world. Secondly, there typically isn't just one bus in a system, and that PCI bus is typically on the other side of more than one bridge relative to the CPU, where AGP is typically only one bridge away.

Finally I just don't understand the obsessiveness of your argument. Who cares about PCI? Do you think it costs that much more to manufacture an AGP card? The $$$ are in the GPU and memory, not in the bus interface. A PCI card wouldn't save you $$$ other than being not in demand and therefore cheaper because no one wants them. Are there really mainstream motherboards with no AGP slots? Haven't seen one in years.