ATi's New All-In-Wonder Radeon 8500 128MB

KillaBee writes "ATi has taken the wraps off their latest addition to their 'All In Wonder' product line of graphics cards with TV and video editing functionality. The All In Wonder Radeon 8500 128MB card, reviewed here, has ATi's fastest Radeon 8500 core along with a full 128MB of 300MHz DDR SDRAM (600MHz DDR). This is ATi's 'Swiss Army Knife' card that brings with it very competitive 3D graphics performance as well."

Graham

[Anonymous Coward]

ATI have this market all wrapped up at the moment, and the prices are good for both the AIW products and the standard Radeon cards. Drivers are a lot better as well. Nothing that is as good as a GeForce 4 Ti4600 though, but that is darn expensive :)

whee

[Anonymous Coward]

whee

Re:PCI?

[NoMoreNicksLeft]

Dude, for the price of such a card, you could buy a tivo. With some money left over for the 100baseT card for it. I wasn't required to subscribe when I bought mine, and I don't plan on it. A dedicated solution, a cheap solution.. no dropped frames or segfaults. If you want to build your own, to say that you've done such a thing, good luck. But if you just want a kickass linux PVR... someone already makes it.

To be honest, I can think of many early PCI TV tuner cards you might buy, but without checking I'd think the performance on those would be horrible. Everything that is current, is high end, for professional use. $700 and up.

How is the Linux support?

[j09824]

Are all the features available from Linux? Are the drivers open source, or are they semi-closed, like nVidia's? How good is OpenGL performance on Linux?

Re:Review skimps on the video recording features

[56ker]

Articles are generally based on press releases which emphasise the good points of a product - for an unbiased opinion you need to ask someone who actually has one before buying.

Re:What about a Digital broadcast TV Tuner?

[-tji]

I agree.. This NTSC tuner merits a big yawn for me.

DTV makes more sense for PVR functions. The data is already compressed digital. All you have to do it save it to disk. And, the quality is leaps and bounds better than our 50 year old NTSC standard.

It's about time they got on board with DTV. I would be the first in line to buy one.

Re:WHY are ATIs drivers so bad?

[Forkenhoppen]

I have an 8500 too, and I must agree that the drivers are a tad flaky. (I've had several "spontaneous" reboots so far, though they've been fairly far between.) Just a few observations about this card and it's drivers..

- 2D transparency of windows and stuff in 2K and XP is done using the 3D acceleration parts of the card. Sometimes, after running software that makes rather heavy use of pixel shaders, I'll end up with anything transparent suddenly being mono-color. I suspect they had a state-saving problem in that particular version of the drivers... suffice it to say, the latest driver version fixes this issue.

- As a developer, I've been using this card to write vertex and pixel shaders, and let me tell you, this thing does not react well to incorrect values. As an example, I once accidentally fed a mangled pixel shader pointer value to the SetPixelShader call in DirectX, and the following render call I made caused the computer to reboot. Ditto happens if you specify an incorrect specification for vertex information. It's a shame they don't check for obvious errors like this, something nVidia does. (Although I should point out that part of me is extremely thankful that the card does react badly to these problems. Otherwise, I probably never would've discovered the problem in the first place.)

- The OpenGL texture-loading-into-memory issue---which I really don't know much about--is not yet fixed in an official driver release, as I understand it. So most people will still be experiencing the texture memory chug in Quake III, which appears to be part of what this review is based on. I'm not sure if the other tests are OpenGL or DirectX, but maybe this'll shine a little light on why there's a bit of that discrepancy. (Was the texture shuffle thing an issue in DirectX too? Anyone know?)

- Windowed 3D rendered contexts that are rendering slow can end up feeling like they're lagging by a bit. Compared to a Geforce3, it can seem like the Radeon8500 is a slow mule, but I think it's just from being triple-buffered instead of double-buffered. (Incidentally, this might also be responsible for another issue I've seen crop up while moving from a Geforce3 to a Radeon8500; the base memory footprint, graphics memory-wise, tends to be larger on the Radeon8500. This is more of a feeling than a documented fact, but I suspect that when you're working on a Radeon8500, you actually have less texture memory to play with than on a Geforce3.. even when they both have the same on-card memory and AGP aperature sizes. I think this would actually make for an interesting comparison sometime, if someone would actually make a benchmark that compared the amount of stuffs you can stuff into each of these cards.)

All in all, I'm happy with my purchase. This is probably the most stable set of drivers I have seen come out of ATi ever. Granted, I'm not running multihead, so I don't know how much added complexity that throws into the equation, but.. hey, it works, and a hundred times better than the Rage Pro and Rage 128 drivers did. For instance, this one calculates clipped vertex coordinates correctly, something the Rage Pro had issues with in OpenGL. And I had an issue that bugged me about the Rage 128 too, but I seem to've forgotten it.. : )

Still, I have one issue that's been bugging the daylights out of me with the Radeon8500, more because I can't logically figure out why it would be happening rather than because it's annoying. I've been playing this old game called Oni, and while it runs faster than ever with the new card, and looks simply amazing, I've begun to notice that.. well.. the texture coordinates on the level geometry actually jump around ever so slightly. It's really quite bizarre to watch... : )