ATi Radeon 9700 Full Release Review w/ Benchmarks
Chalupa_Man writes: "ATi Technologies has officially released their new Radeon 9700 Pro today.
Real benchmark numbers and a full review can be found here. The card is
impressive for sure and should have NVIDIA on the ropes for a while, as it beats
out a GeForce 4 Ti 4600 handily, especially with Anti-Aliasing and Anisotropic
Filtering enabled. Image quality is also top notch for this new high end DX9 compliant
product from ATi." sunny_talwar adds these links to more reviews of the new high-end Radeon at AnandTech's and Tom's Hardware. Update: 08/20 03:06 GMT by T : Cp writes "Gamers Depot also has their full review up of the Radeon 9700 Pro, including nice images of the driver tabs and 6x Antialiasing performance."
Comment removed based on user account deletion
This is old news.. the Radeon 9700 has been out for a few hours already. Why do we have to wait so long for news on this site?
Well there are some floating around, but from what I hear it shouldn't be out till the end of August. I got to play on one of these cards at QuakeCon and let me tell you they are SWEET. Wolfenstien in 1280x1024, lightmap, all eye candy was usually 250-330 FPS. When it hit 400 FPS I about dropped a load.
One part of it is that it's something people are accustomed to seeing, so a score of X is more meaningful to them than one from some game without any sense of reference. People also still play Q3 a fair bit, with baseq3, Urban Terror, Reaction Quake 3, and so on all being played.
As well, when video cards come out every six months, and games like the Quake series every year or more, you're going to see the same game used for a while.
Once Doom 3, Unreal 2k3, eetc. come out. maybe those will be added to benchmarks.. Who knows.
[H]ardOCP Also has a review and benchmarks. Good stuff from the [H]ard crew.
.... um, i lost you after "0110100001101001".
Competition is a good thing. The last thing I particularly want is for nVidia to get stomped by ATI because they start getting complacent like 3Dfx did. Let's hope they keep each other on their toes.
"Destroy science and religion. Science would re-emerge exactly the same; but not religion." - Penn Jillette, paraphrased
...until John Carmack responds with his take on the card.
I'm serious. How many of us base our video card purchases on the recommendations he makes? He knows the cards in detail, knows what features they support and how well, and he sure as hell knows how well they'll perform with the next id game.
So John, is this card worthy?
Compared to some other companies *cough*NVIDIA*cough* ATI has been very helpful to linux developers. While NVIDIA only releases binaries, and only for x86, ATI actually provides developers with technical specs to aid development on other platforms (PowerPC anyone?).
From ATI's website:
While ATI does not develop Linux or XFree86 drivers for its graphics cards in house, we actively support 3rd party developers that provide driver support for the majority of ATI products with development kits and information.
Radeon drivers for Linux are in development. XFree86 and the DRI Open Source Project offer Radeon 2D support with their latest released source code. 3D support is scheduled to be released Q1 2001.
Schnapple
This is the number one reason why I stopped using ATI products once the Mach64 chips came out. Their driver support has always been slow, incomplete, and crippling to their hardware. For many products, downloading even ORIGINAL drivers was impossible, and one would have to order a $4.99 CD of the original, old, buggy, broken drivers. Some products they made (PCI TV Wonder) were left completely unsupported, and never got correct driver support for anything above Windows 98 original release.
Despite their recent excellent showings in hardware, I too refuse to buy ATI because their driver support is, at the very least, a complete insult to the sensibilities of even a modest geek. For that reason I'll continue using my NVIDIA card until it burns out (which will be as soon as the fan stops spinning), and then I'll go and buy their latest and greatest. At least their drivers are generously provided and updated, sometimes on a weekly basis.
.... um, i lost you after "0110100001101001".
Comment removed based on user account deletion
Sure, we know it runs awesome on Quake3... but will it run awesome on Quack3?
actually there's a 9000 model inbetween. I think the increments are more like 500 depending on the nature of the upgrades to the card 7000, 7500, 8000, 8500, 9000, [[9700]]. Maybe they had a 9500 but decided to tweak it a little more but not to the point where it was a 10000.
"Do not be swept up in the momentum of mediocrity." - anon
It doesn't surprise me one bit that ATI can push the envelope of 3D graphics.
They've been in business since the dawn of the x86 age. They always made solid cards.
Around the time of the stealth64 ATI lost its edge because they didn't see the potential for the consumer gaming market. (Stealth64 was the hot gaming card back in the doom days, ask thresh) Despite companies like 3dfx releasing the voodoo1 and Creative releasing the VLB 3D blaster, it was years before ATI came out with a graphics chip with even rudimentary 3D support.
Nvidia, a new company only took couple of chip revisions before they were able to match 3dfx's performance. It's no surprise that a company like ATI with years of 2D behind them would be able to quickly beat out the new top dog Nvidia.
Kudo's and good job ATI. Now if you could only price these new cards in a reasonable range, let's say less than $200, you could definetly become the new king.
I gaurentee you're using an Athlon system. My last (and I mean last) Athlon system didn't work with any of the ATI boards. I thought ATI was shite, so I bought an nVidia board. When my Athlon decided to cook itself (taking my board with onboard RAID with it, a mistake I will not make again), I decided no more - I bought a P4 system. All of the cards the previously wouldn't work in my computer now worked flawlessly, including:
- ATI All-in-Wonder Radeon
- ATI TV Tuner
- Hauppage TV Tuner
Granted, it might be partly AMD's fault, but I shouldn't have to worry about compatibility, and with Intel I don't have to. I didn't want to use nVidia because they don't have an acceptable alternative to the All-in-Wonder series.
Honestly anyone could have told you months ago the 9700 would beat the gf4... it's a new generation card.
And whats the use in getting it this month, since most games out now are still based on 5year old GFX engines that run decently on a geforce2.
and please spare me the tears of 60fps vs 200fps :)
--me
Wolfenstien in 1280x1024, lightmap, all eye candy was usually 250-330 FPS. When it hit 400 FPS I about dropped a load.
LMAO!!
Other than being glad that the architecture is advanced enough to achieve such numbers, why would you be astounded at this? I mean, its only another ~100 FPS that you only notice because you can see the actual FPS numbers, not because the quality is any better. See, I was astounded when I dumped my old TNT2 for a Radeon 7500 a month or two ago and I could actually walk through a fire fight (in any game) without the FPS dropping into the single digits (5 FPS TFC is not fun). I was astounded at that, but still not load-droppingly-astounded
Having said that, I still can't get extremely high resolutions with all the extras on to work absolutely great on my 7500, although $57 for a 64 MB DDR 7500 back in May was not that bad
The 9000 is actually the sucessor to the 7000 line. The 8500 is faster and more expensive then the 9000. There will be a 9500 as is mentioned on the Anandtech site. The numbering is confussing. Typical. Recall GForce4-MX cards from Nvidia.
Quake 3 is still being used as a benchmark because there are still games being released that use the Q3 engine.
However, Anandtech's review of the 9700 has some benchmarks that include the Unreal Tournament 2003 engine. There are also some cool CPU scaling charts in there. Epic has been providing Anandtech with build of the UT2003 engine for quite some time. All of their recent reviews include UT2003 numbers.
>Maybe they had a 9500 but decided to tweak it a little more but not to the point where it was a 10000.
The 9500 will be released in a couple of months (as in mentioned in the Tom's Hardware Guide article). It will be a scaled down 9700. It should have a lower clock speed and fewer texture units.
Its a demo/test model that I was using in the lab to verify compatibility with our applications. Yes, that is corporate speak for "I played quake for a couple hours on company time". I am payed to do that. Anyway, here it goes...
* 2D: WOW! I have been a diehard Matrox fan because of the awesome 2D on their boards. However, I think Matrox might have a challenger on their hands. Even at dizzyingly high resolutions, the fonts were crisp and clean.
* 3D: Very nice. It has been image quality than the Geforce Ti's with FSAA enabled. However, it cannot compete with the Matrox Parhelia here. The Parhelia, though it has slower framerates, has better color saturation and 16x FSAA w/o a massive performance hit.
* Drivers: so far it was worked fine under WinXP. I got the SVGA xserver running on it after mucking around with Redhat for a couple hours. I am hoping a dedicated XServer is coming out for this card since it needs one badly.
Anyone else have any luck under Linux?
Probably because they want some competent people to write some drivers for them. :-)
My beliefs do not require that you agree with them.
"Aye'm!"
"And like that
So then they're backfilling the product line. Build out the high end product then go back and fill in blank spots in the market coverage.
"Do not be swept up in the momentum of mediocrity." - anon
Comment removed based on user account deletion
Digit 1 - DirectX version
Digit 2 - Performance relative to others in the same series
Digits 3 and 4 - meaningless
What would Lemmy do?
The big difference between NVidia and ATI is that ATI releases lots of information about their chips, while NVidia releases binary-only drivers for their cards in Linux and keeps mum about details on their chipsets. What's better, a binary driver that will break with the next version of XFree, or truly free drivers that can be updated as XFree evolves? I'd say ATI is the more supportive of Linux of the two companies.
Knowledge is power. Knowledge shared is power multiplied.
I gave up ATI after my eXpert@play board. I bought the board, installed it and the RAM was bad. Called ATI tech support IN CANADA!!! No 1-800 number! At first they wanted me to spend my own money to send them the bad card and get a new one. I said, "If I have to spend my own money the card goes back to the place I bought it and I buy 3Dfx." So ATI sends me a card and a UPS shipping label. I still spent $20 to call in the first place!
Next, I found out that the benchmarks I had looked at so long were for a tweaked set of drivers that ATI had released to get better scores on Quake and the card sucked for anything else and wasn't as good for Quake as I thought! This was one week after I'd bought the card.
I'll NEVER trust them again.
Do really dense people warp space more than others?
actually, ATI did a press release. You can search for it on [H]ard OCP, I'm too lazy. But basically the first digit tells the chip's generation relative to each other so:
7xxx first gen radeon
8xxx second gen
9xxx third gen
The Radeon 9000 is not a DX 9 compatible chip, its mroe or less a tweaked 8500. meaning it gives aproxatmetly the same performance as the 8500. Its actually a little less cause the 900 can't do single pass texturing or something.
11 was a racehorse
12 was 12
1111 Race
12112
>and never got correct driver support for anything above Windows 98 original release.
Never had an ATI ISA TV card, did you?
They couldn't even get it working well with windows 95 (I know, I tried every windows version I could get my hands on).
Blech. But at least they're stepping into the open source movement, so perhaps this won't be such a problem in the future (at least on Linux).
If you could be told what you can see or read, then it follows that you could be told what to say or think - BoC
It is unlikely you will see an effective Xserver for this card any time soon. While nVidia may only provide closed-source drivers (save for the barest minimum source-level shim to allow their drivers to work with a few different kernels), at least nVidia pays programmers to support their cards under !MSWindows.
ATI will provide some documentation to selected members of the XFree development team, but they do not release all the programming information to the world, nor do they pay anybody to support their cards.
Perhaps that might change if enough people make it clear to ATI that Free Software drivers for XFree, source on the CD that comes with the card and pre-compiled binary modules for the current releases of XFree will sell more cards.
Of course, the odds of this happening any time soon are roughly 2-to-the-9421 power, and falling...
www.eFax.com are spammers
Wolfenstien in 1280x1024
My GeForce2 already runs Wolfenstien at like 800 fps. How did you manage to get it into 1280X1024 mode though? I didn't think there was a VESA mode that high in DOS.
I've had enough abrasive sigs. Kittens are cute and fuzzy.
...but if you want decent Linux support buy a Matrox. Matrox may not be the fastest in 3D but it's no dog either, and you get unbeatable image quality. They also give you full programming manuals and source code for the Linux drivers.
"I have opinions of my own, strong opinions, but I don't always agree with them." -- George H. W. Bush
This is the number one reason why I stopped using ATI products once the Mach64 chips came out. Their driver support has always been slow, incomplete, and crippling to their hardware.
ATI products were crap back in the days of Mach64 and the like--both hardware and drivers. This all changed with the introduction of the Radeon series, however. I've had no problem with their latest cards and Windoze drivers. Far more importantly, ATI products have better support in Linux because ATI, unlike NVidia, actually documents their hardware and plays friendly with Open Source developers. And it seems to me Radeon boards still have the GF4 beat hands down in 2D image (ie. analog signal) quality at high resolutions. Somebody with a high bandwidth oscilloscope want to do some S/N analysis?
bullshit- absolute and utter bullshit. I for the longest time, was an ATI supporter. They have by far, the most kickass integrated tv-tuner cards I've seen.
I supported them when everyone else laughed at me. I supported them- until I bought neverwinter nights last weekend. They have NO DRIVER SUPPORT. they're response to 'your card won't work and continues to crash' was 'suck our balls. if you want to play, you have to use the 2-versions-past drivers.' Don't believe me? look up the all-in-wonder-radeon drivers on their site and look at the known issues section. That isn't acceptable to me.
I bought a geforce4 mx440 yesterday, and it works great. First non-ati card I've bought. I hear that each time nvidia releases new detonator drivers, it improves ALL of their cards, including the older ones. so yeah, I felt the need to rant on that.
mod me down if it gives you your jollies, but just keep in mind your supporting a company that doesn't support you.
Looking for Book Reviews? Check out Literary Escapism.
Uh, except 3-400FPS in RTCW in a given resolution means you can double the resolution and get.. 1-200FPS. Handy when the next generation of monitors come out that let you use resolutions of 4000x4000 or whatever. It also means you can make more detailed maps, and have busier scenes; instead of being limited to 2-3 characters in view at any one time, you can do a Doom/Serious Sam with the same detail level. Or you can tone up the LOD and have higher resolution textures further away and reduce mipmap artifacts. Or you can seriously concider adding bumpmaps everywhere, or more detailed volumetric self shadowing and *still* not drop below monitor refresh for the most detailed scenes.
And even IF you've still got tonnes of power left, this is what multitasking is for; if my system can push RTCW along at 400FPS, I can leave some expensive background task running and still have perfectly smooth gameplay.
NVIDIA's binary drivers don't break between XFree86 versions. They support all XFree86 versions from 4.0.1 through top of tree XFree86 CVS. And they do have open source versions of the 2D-only drivers. The open source drivers support all NVIDIA cards. I've seen more complaints on XFree86 mailing lists about newer ATI cards not working that there are about new NVIDIA cards not working. I'd take vendor support over specs any day.
Doesn't wash.
I would presume that if nVidia is that worried about their GPUs, then they're patented as well as closed source. Hardware can be reverse-engineered, but it can be a pain in the neck crawling around SEMs and trying to turn it back into a schematic, and then trying to turn that back into functional blocks so you can walk up the hierarchy and comprehend the whole. I know, I've done it. Supplying Open Source-style documentation would make it easier to reverse-engineer the hardware.
On The Other Hand...
IMHO a big part of the reason for closed source drivers is that it can take a lot of work to release proper documentation. Closed source drivers can be done by poor documentation plus the fact that the programmers may well sit down the aisle from the hardware guys. They talk daily, and that fills the gaps in the documents. Painful for both, but frequently cheaper and less painful than doing a good job of documentation.
On The Gripping Hand...
One of the harder aspects of patenting something can be detection of violation. If nVidia were to release their documentation and let this stuff work its way into the Open Source community, then they could watch the software concepts flow, and know where to start looking for hardware infringement. Presumably the nVidia driver model is most useful for nVidia hardware. If the nVidia driver model began being used against upstart JoeVideo cards, then they'd have good reason to take an SEM to JoeVideo chips, the the Open Source drivers would have pointed the way for them.
Whether Open Source wants to be in a position of assisting with patent prosecution is a different question.
The living have better things to do than to continue hating the dead.
Hmmm, I had problems capturing with the All-in-Wonder on Windows 2000 until I did some registry hacking. Then it worked fine.
On Windows XP, however, it worked right away. Perhaps you could try XP? What capture software are you running? I recommend VirtualDub.
> Either that or the motherboard was the problem.
Perhaps, but in my purchasing experience:
- 3 out of the 5 AMD-based systems I've owned (one K6/100/?, one K6-2/300/VIA, one K6-3/?/VIA, an Athlon/700/VIA and an Athlon/1.2/ALI) were unstable and/or had compatibility issues.
- 0 out of the 4 Intel-based systems I've owned were unstable (a 386/16, a 486/33, a P2/233, and recently a P4/1300, all using Intel chipsets)
I just have bad luck with AMD.
Wrong, doubling the resolution quarters the framerate, so 3-400 would become 75-100, still playable though.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Holy polygons, would you just quit the hype already? I *just* ordered a dual-867 Power Mac with GF4 Ti, and I spent a pretty penny for that upgrade - can't a guy bask in hardware glory without some bithead like you going and raining on his GPU parade? Sheesh.
Was that out loud?
Hmm. I got the "first digit = dx version" thing direct from some friends at ATI. I suspect that what happened with the 9000 was the marketing idiots messing things up in a similar way to what happened with the GF4-MX.
Anyway, thanks for the correction/update.
What would Lemmy do?
I just picked up an 8500LE, which though although claimed to be a lower end part came with same specs and s+-video out. 87 bucks at newegg.com. I flashed it to retail bios and I got a regualr 8500. Noi biggie. And it is an awsome card, 2d and 3d. For 109 dollars you can get the 128 meg version, makes sure it says le.
I should of spent the money and got the 128 meg version.
But for 87 bucks I got something that kicks ass.
Go ATI.
Puto
The Revolution Will Not Be Televised
No, not really. You've got a lot more than just pushing pixels around to worry about for it to be that simple.
4x the pixels doesn't mean 4x the geometry, or 4x the AI, or 4x the bandwidth requirements (ok, the final image is 4x bigger, but the textures and geometry you're pushing aren't also 4x bigger).
Look at the benchmarks for the 9700 for Q3; 1024*768 = 203FPS. 1600*1200 = 180.6FPS. The nearly three times as many pixels of the higher resolution resulted in a loss of just 10% of the framerate.
I'm sure you know this and I'm not really sure why I responded, but there it is.
What's a QD then? I actually have one of these... the vendor (who has since disappeared - don't you just love the net?) claimed it was an 8500 - certainly I've never seen it slow down on any game I've thrown at it - but I'm not so sure.
Not that I'm worried, it's far better than the geforce it replaced (having supported drivers rather than that binary junk that crashes every 20 minutes is a great plus) & the 2D performance is quite good... not as good as a Matrox, but usable.
I will not buy any new 3D video card. Regardless of what's demoing it now, it won't be the top choice when Doom 3 comes out.
Was a Voodoo1 the top choice for Quake 2 when it came out?
--
Internet Explorer (n): Another bug -- that is, a feature that can't be turned off -- in Windows.
Dedicated Xserver?
.o files to be loaded automatically based on the config file.
I'd really like to hear your information if it pertained to the XFree86 4.x tree. XFree86 3.x and its separate Xserver binaries for each card disapeared a long time ago. XFree86 4.x has an ABI which allows driver
Now, if there was a way that per-user accounts could have an XFree86 override and there were easy tools for both CLI and GUI configuration, and these were all the default settings in distributions, and the changes made in a session were stateful (IE: if I changed the res down a notch and restarted X, it'd be at that res, even if I had many modes defined), we'd finally be close to where Windows / MacOS is in terms of easy-GUI configuration.
Setting up X is still too much black magic.
--
Internet Explorer (n): Another bug -- that is, a feature that can't be turned off -- in Windows.
It was the last time I saw Linux Radeon drivers.
Apparently if you have a really fast Rage 128 games like Q3 will run fast. But who needs a fast Rage 128...we need drivers that treat an N-generation card as such, not an (N-1) generation card.
So my true questions are: do the _current_ drivers support
1. hardware T&L?
2. vertex shaders?
3. pixel shaders?
4. FSAA / SmoothVision?
and last but not least,
5. TV-out / Multiple monitor / Video-in?
Get to work, or it's back to default poll option for you!
Best Slashdot Co
From ati's website:
John Carmack
"The R300 is an ideal rendering target for the DOOM engine, it can do both our highly complex pixel shaders for light surface interactions and can very rapidly render all the stencil shadow volumes which deal with all our dynamic masking of way light operations"
"3D accelerators are all about performance, quality and flexibility and the R300 breaks new ground over anything thats come before it in all three areas."
> P.S: As long as you have a FAN on that Athlon, it'll be fine!
I had a fan on it. It died and the CPU fried itself. P4's underclock themselves to prevent this from happening.
The rest of you benchmarkers: fuck off. If I have to read another masturbatory "analysis" of how one card's Quake3 framerate is 4 times the refresh rate of the best available monitors, while an inferior card can only do 3 times, I will have to write you email to see whether you also spend a lot of time wondering how many angels can dance on the head of a pin.
Meanwhile, I'm seriously starting to wonder whether there is some payola behind the scenes of these uncannily similar choices of games to benchmark.
No 3D support at *all*.
You're the one full of crap.
From the DRI on BSD page:
However, in the good news, a couple of users have reported success with r200-0-1-branch of DRI CVS with Radeon 8500s on FreeBSD.
Maybe you should do a little research before posting blatant lies.
Dinivin