GF FX 5800 Ultra Vs. Radeon 9700 Pro

[cmdrdredd]

anyone that cares about 100% stability and uses their system for more than Games like myself. And oh...I think Microsoft finds it kind of important

 

[BenSkywalker]

anyone that cares about 100% stability

Only thing that has managed to hard lock my current build is ATi's WHQL certified 3.2 Catalyst drivers.

BFG

No, I only checked to make sure it was off. Mouse lag is usually tell-tale sign of vsync being on.

It is odd, but enabling VSync does help out quite a bit(seems to be an issue with the older Lithtech engine games).

Croteam say it's nVidia's drivers not reporting the correct available Z depth. And yes, the workaround does work as I used it on my Ti4600.

If the problem showed up in something else, I'd tend to believe Croteam. All of the OpenGL based apps I have one of them has a problem

I don't know, why don't you ask nVidia who have probably been releasing more drivers than ATi after the FX arrived?

Because they are obviously shorting out issues. What I was commenting on is why release new drivers for the GF4 when there weren't any problems? ATi still has plenty to sort out with their drivers, as does nVidia with their FX line.

The problem is that most of them were leaked betas released in a shotgun style while ATi's were tested and WHQL certified.

And the latest WHQL certified ATi driver is the only thing that has managed to crash my computer totally. Forgive me if I put little faith in to the certification process involved

 

[Rogozhin]

Ben

So you base your perception of the quality of the WHQL certification process on your own system's hardlock?

That is a logical fallacy.

Had you said "My system is locking up since I've installed these whql drivers, I think I will eliminate any variable that might be causing the problem besides my video card (run it on a fresh rig)." Then you'd be golden.

BUT

To reach the conclusion that ati's WHQL certification is meaninless is one HUGE logical fallacy-"converse accident"

Rogo

 

[Rogozhin]

The whole problem with the whql certification with nvidia is that they are trying to get the highest benchmarks possible and without using drivers that aren't whql cerified they would loose badly in all benchies.

Their floating point ops are killing them at the moment (check the latest benchmarks from Hardocp and thier visiontek 9800 review, the fx is getting thrashed).

And the fact that futuremark will accept ONLY whql drivers in their database (as valid) is only another red flag that something shady is going on with nvidia's driver department.

Why can't nvidia come out with whql certified drivers, and why haven't they in the last 3 months?

If you can answer these questions without anwers that place nvidia in some unfriendly light then good luck to you.

Rogo

 

[BenSkywalker]

Had you said "My system is locking up since I've installed these whql drivers, I think I will eliminate any variable that might be causing the problem besides my video card (run it on a fresh rig)." Then you'd be golden.

Well, I do have ATi on my side on this one. They have acknowledged that there is a bug in their 3.2 WHQL drivers that hard locks systems and it is easily repeatable.

To reach the conclusion that ati's WHQL certification is meaninless is one HUGE logical fallacy-"converse accident"

In terms of stability? Why do you think it isn't?

Their floating point ops are killing them at the moment (check the latest benchmarks from Hardocp and thier visiontek 9800 review, the fx is getting thrashed).

Using the settings they decided to use the FX isn't looking great, that's for sure. What exactly is causing this isn't perfectly clear however.

And the fact that futuremark will accept ONLY whql drivers in their database (as valid) is only another red flag that something shady is going on with nvidia's driver department.

Is it as questionable as FM using two PS 1.4 game tests for DX8?

Why can't nvidia come out with whql certified drivers, and why haven't they in the last 3 months?

Look at who cares the most about the WHQL issue. Don't you find it the least bit odd that it is those same people who are running ATi boards? How many nVidia customers have you seen shouting for WHQL drivers? Is it nV's jobs to make ATi's users happy, or their own? In other words, why should nV users care at the moment that they aren't using WHQL drivers? Since I had my first nV product, a TNT back in '98, I haven't ever been concerned with it. Actually, until the FX launched I haven't seen too many people anywhere really care about the issue.

Are the higher levels of precission currently functioning properly under DX with the FX? No. Is W buffer currently working on ATi hardware with WHQL drivers? No.

 

[nRollo]

Rogo:

"So you base your perception of the quality of the WHQL certification process on your own system's hardlock?
That is a logical fallacy."

Why didn't you trot this out when Cmdrdredd tried to fly this bit of stupidity:
"er...who has those problems? I've built 5 systems using 9700Pro and NOT ONE has that issue. "

Not only is this a logical fallacy, it spits in the face of good experimental method as well as his sample is tiny and has no power?

I think Ben was only pointing out that WHQL doesn't guarantee system stability, you're reading too much into it. He never said ATI WHQL is meaningless, only you did.

 

[bunnyfubbles]

Seems as tho according to you his chances of not getting rolling bars on any of the 5 boards is 16.8%, seems he's gotten more 9700s to work despite the odds being against him.

Logical fallacy is assuming concluding that the 5800 Ultra is better just because amongst a sample where there practically aren't any 5800 Ultras, because none of those who don't have a 5800 Ultra will trade you for your 9700 and other bait. Let's just ignore the bias and confounding variables that exist in such an experiment and jump straight to a conclusion (a more than likely incorrect one at that)!

 

[Rogozhin]

Rollo

Here is the quote that I responded to of Ben's.

"And the latest WHQL certified ATi driver is the only thing that has managed to crash my computer totally. Forgive me if I put little faith in to the certification process involved."

It's a logical fallacy to believe that his faith in WHQL drivers (and ati's at that since he hasn't mentioned any others) is contingent upon his machine hardlocking.

Ben is generalizing from a individual occurence to a universal statement to reach this conclusion- "I put little faith in to the certification process involved."

Meaning

Ati's certification process for WHQL is BOGUS becuase my machine hardlocks with the most recent ati whql drivers-this is a logical fallacy and the name is "converse accident."

Along the same lines I put much faith into ati's whql certification process because my machine never has hardlocked with any ati whql drivers-this is also a logical fallacy.

Also

To believe that ati's WHQL certification is "faithless" or without worth one should first thouroughly test the machine these drivers are installed on, latest chipset drivers, latest OS updates, latest sound card drivers, array drivers for card, and CLEAN install. I am asking Ben here if his install is clean or if he has ANY doubt about the install of these drivers. Since he's not a real reviewer and since he didn't clarify if his machine is clean I take this comment about whql certification lightly.

Rogo

 

[Rogozhin]

Rollo

I actually skim threads for people who post well, you and Ben are two of these and since I skimmed this one I didn't run across creddmans post.

Also

Would anyone mind rating me for my posting? I would appreciate it

Rogo

 

[nRollo]

"Seems as tho according to you his chances of not getting rolling bars on any of the 5 boards is 16.8%, seems he's gotten more 9700s to work despite the odds being against him."
I see. Of course you will provide a link and a quote to where I say 16.8% of 9700s get the bars, and that the odds are against him finding 5 that work? (since I never said anything like that, I won't hold my breath)

 

[nRollo]

Rogo:
You're reading too much into his post, taking him too literally. (much like Bunny considers my stating "there are a lot of people who get the gray bars with 9700s" means "the odds are against you getting a 9700 that works")

I'm sure Ben knows some WHQL certified drivers works and has some faith left in them.

You're just being argumentative.

 

[BenSkywalker]

It's a logical fallacy to believe that his faith in WHQL drivers (and ati's at that since he hasn't mentioned any others) is contingent upon his machine hardlocking.

Considering that my initial comment regarding stability came in a reply from the laughable assertion that WHQL gave any sort of assurance upon stability it doesn't really.

Ben is generalizing from a individual occurence to a universal statement to reach this conclusion- "I put little faith in to the certification process involved."

An individual occurance that happens with one of the more popular games in the world. In terms of stability, I put no faith at all in the WHQL process, and it isn't because of that lone example.

Ati's certification process for WHQL

ATi has no WHQL certification powers, that is handled by Microsoft.

I am asking Ben here if his install is clean or if he has ANY doubt about the install of these drivers.

It is a new build. It was a new HD, needed to be partioned and formatted and then had the OS installed along with SP3(Win2K). All applicable patches were applied along with all of the latest drivers for every component(including my monitor). The hard locking issue has been admitted to by ATi, how much more proof of this issue being real do you need?

Since he's not a real reviewer and since he didn't clarify if his machine is clean I take this comment about whql certification lightly.

You mean compared to Anand who can't identify the difference between bilinear or trilinear all the time, or perhaps Kyle who thinks fill limiting a board is the best way to test the T&L unit, or maybe Tom who focuses on a benchmark that runs much lower quality settings on one piece of hardware compared to the next and fails to make note of it? I can certainly agree that I am not going to do a review of the above example's caliber. I have to admit, every time I've ever done a vid card review I have always made the huge mistake of playing as many games as possible on the hardware to detect if there are any issues with the drivers or hardware not apparent in the cookie cutter reviews most of the big tech sites offer. I also avoid running the normal five to six game benches(expanding only to include different settings of those games) and calling it a day, of course this tends to eliminate any tuning done by a particular vendor in particular titles. You want to talk about Digit-Life or Beyond3D you have a point, most vid card reviews posted by the major tech sites are seriously lacking in pretty much every aspect.

 

[Mem]

What issues were there with nV's drivers?

I agree with Ben on some points, in my books there`re no issues with Nvidia drivers,I`ve yet to have a problem with any Nvidia drivers or games .

ATI has to release regular driver updates to fix known problems,which is good ATI support.In the end you can argue merits both ways ie if a driver is perfect you won`t need to update it for a very long time if at all or you could say bad driver support ,on the other hand if they are updated often then you could say there was bugs,problems etc that needed to be fixed or good driver support.


Getting back to WHQL ,that really does not mean a lot in my books,sure Microsoft gives them their approval but that`s it,we all know there`s lots of good non-WHQL drivers out there .

How many nVidia customers have you seen shouting for WHQL drivers?

I for one don`t care about WHQL approval,all I care about is games working fine period.

If you want to buy ATI fine,if you to buy Nvidia fine,in the end it`s what makes you happy in games etc and for me it`s really nice to stick in any game with no issues with drivers (that`s really all I care about) and enjoy the gaming experience online with my fellow players.

The bottom line is right now ATi has the edge with hardware while Nvidia has the edge with software (drivers),so you take your pick .
I don`t want to sound like a Nvidia fan but this is my take on the subject,in the end I`m just one happy gamer.

 

[Rogozhin]

Rollo

I don't know Ben from Adam and since he is posting his beliefs in this forum and I call into question the validy of his claims this is discussion NOT argumentation. Argumentation connotes agression and ad baculum kinds of posts, this is not one of them.

Ben

"ATi has no WHQL certification powers, that is handled by Microsoft."

I know that and that is what this sentence meant-sorry for the confusion.

And thank you for posting your setup of the video card. I'm pleased that it's fresh and that you were willing to post it.

I didn't mean any disrespect about you not being a reviewer but since you aren't a professional reviewer it's difficult for you to post screens or mpegs or a complete review.

I do appreciate your imput and posts as they are informative.

rogo

 

[chizow]

All this fuss about WHQL drivers.....

About a month and a half ago, all the GF4 users were whining about how there hadn't been an official Detonator released in some time, while ATi was releasing new Cats every 4-6 weeks. If you've followed the GPU industry for any time at all, you'd know that the major factors dictating driver updates are:

A) New Hardware, which can be divided into 2 major subcategories:

• 1- A new GPU family: This is the most obvious need for updated drivers, since new feature sets and core architecture usually need revised or completely new drivers. In the past, nVidia's drivers have more or less been tweaks to optimize driver performance, since the GeForce architecture since the GF2 has been more an evolution rather than a radical revision.
  
  2- Platform, chipset, or other hardware compatibility - this is usually bug squashing that ensures the new GPU is compatible with most hardware on the market. Of course, IHVs will also test GPU compatibility to ensure their products are compatibile with the popular graphics adapters on the market, which is why I've felt for some time nVidia has an edge over ATi when it comes to hardware compatibility. Being the market leader, nVidia benefits from these efforts as IHVs will ensure hardware compatibility with their GPUs.

B) New Software, which can also be divided into 2 major subcategories:

• 1- A New OS release: The most obvious, as anyone looking to download new drivers needs to distinguish what OS they are using.
  
  2- A Game Engine or DX library that enables new feature sets - This is also pretty obvious, as we've seen in the past that many games are based on a revolutionary game engine that incorporates new feature sets. This requires new drivers to optimize performance with these titles.

Now if you look at the GF4, none of these factors have changed in the last year or so other than DX9 being released. However, GF4s are DX8 parts that will be fully compatible given DX9's backwards compatibility with older DX libraries. The GF4 core is based largely on GF3 architecture, so no one was expecting incredible performance increases a year after release, they were expecting great performance out of the box, which the GF4 provided. Patch updates were largely to fix game/hardware specific bugs and incompatibilities, but nothing major in terms of performance.

On the other hand, the NV3X cards are dealing with every factor listed above other than a new OS release. There's no doubt nVidia is hard at work on their drivers for this new generation of GPUs, which is the most obvious answer to the lack of WHQL certified drivers. Why would you bother to obtain WHQL certification if your product has not fully launched (NV31 and NV34 are just arriving on shelves), or you know that an improved, revised version is already being tested?

For those that aren't completely aware of how WHQL certification works, Microsoft doesn't do anything other than collect a fee and wave a magic wand granting a WHQL key. Each driver submitted for WHQL certification runs through a series of tests determined by MS, and the results are submitted for WHQL certification. Now, imagine doing that every 2 weeks while you are in driver development for a new family of video cards. Not only would it take an incredible amount of resources (imagine running benchmark suites similar to those found in reviews for ALL of your GPUs on multiple platforms and OSes), but it would also be a burden financially.

I personally think WHQL certification does more harm than good. Its just another revenue stream for MS and a way to control software and hardware development on the PC desktop. Many smaller firms simply don't have the resources to go through this Q&A, which is why you see parts off the beaten path ship without WHQL drivers or you see an initial set of WHQL drivers with few, if any, further driver updates. This leads to a competitive disadvantage, since many people have the false impression that WHQL certification means "quality". When deciding between two similar products, the uninformed consumer may be inclined to purchase the product with WHQL certification. MS has implemented a similar control on software sold in retail outlets; many major retailers announced they would no longer carry software that did not have "compatible with Windows XP" emblazoned on the packaging. Would the product still be compatible with XP? Sure would. What would you need to get that sticker? A check to MS.

Anyways, back to the topic at hand. It makes no sense for nVidia to go through WHQL certification at this point, as they're going to be releasing updates frequently. There will be a mature build released that is WHQL certified in Det. 50, which will no doubt take the best of all builds leading up to it. All this talk about non-WHQL certification b/c of FP24 is ridiculous since nVidia's drivers (unlike ATi's) actually has a UDA in theory and in practice. There are plenty of cards supported in that UDA that neither require or are capable of FP24 precision. I've never seen MS's WHQL testing suite listed for public viewing, but I highly doubt 3DMark2K3 or any other specific game settings are used. My guess is a generic suite similar to SPECviewperf is used. /rambling over.

Chiz

 

[gururu]

Nice stuff Chizow.
I myself have a few things to say about WHQL certification and MS interference.
1) Why would someone NOT want the manufacturer of their operating system to test other
manufacturers drivers for compatibility?
I myself like to have some reassurance (even if not certain) from an independent source (other than vendor) that the driver isn't gonna trash my registry.

2) Accountability.
If a driver does trash my registry, someone is to blame (if not me). Going through MS helps the card vendors alleviate some of the responsibility. Maybe its not the driver, maybe its DirectX. In such a case, MS is forced to examine the problem. If a driver doesn't go through MS, do you really think MS is gonna give a flyin' frek if it crashes computers?

3) Standards.
There has got to be a common set of standards that new drivers must meet. One might say that competition would make vendors be the best that they can be. Not true. Vendor cheat. There needs to be a third party source to set some kind of standard.

I'm not saying we need WHQL. I'm saying, we need third party testing to test drivers for function, OS compatibility, and we need the manufacturer of the OS to be available to help vendors troubleshoot problems.

 

[chizow]

gururu, I agree with a need for standards and compatibility, but realistically, WHQL certification can't be much different from the QA a product goes through in-house. Strong QA controls through development mitigate future overhead and expenses in the future, such as customer support and RMAs/returns b/c of hardware incompatibility or driver issues.

As for WHQL and drivers, if you have an existing WHQL build, and you released a patch for a specific bug fix, would that justify putting that new driver through WHQL certification (and its associated fee) again? I would argue no, as long as it didn't break anything else since its a marginal change in the driver.

I'm not saying that QA isn't important, but in the case of either ATi or nVidia (or Intel or AMD for instance) where the overwhelming majority of PCs will be supporting your product, high QA standards are a given. Its always going to be a very complicated intersection when it comes to compatibility, however, in duopolys, monopolys, and oligopolys for required components (GPU, CPU, chipsets, RAM etc.), makers of those components will always have the edge b/c of the extensive QA done from all sides of that intersection (everyone is required to test for compatibility with your parts if they want to avoid a tech support nightmare).

Edit: An example of this is game's that have the "nVidia, the way its meant...." logo. They typically run better on nVidia cards relative to the respective part from the competition. Why's that? Because the developers used those cards as target hardware and used them throughout development to optimize performance and game code to take advantage of specific features for that part. Any bugs would be squashed through the course of development. Again, this is why I feel that nVidia has better drivers than ATi. I think the gap has closed significantly for game support, but I still don't think IHVs are fully testing for ATi boards yet. I haven't had a problem with any recently purchased games (6 months) on my 9700pro other than PlanetSide (which is still Beta, so no biggy), but I have had some issues with the hardware. They've all been corrected over the last 5 months, but still shows ATi isn't the priority when testing for hardware compatibility IMO.

Chiz

 

[Rogozhin]

Chizow

Thanks for the informative posts and discussion.

So you don't think that nvidia is dodging whql certification because they aren't able to comply with dx 9.0 and the floating point ops?

That's what I thought you said earlier, just wanted to clear it up

Rogo

 

[touchmyichi]

damn chizow, we're u on speech and debate or something? lol

 

[BFG10K]

It is odd, but enabling VSync does help out quite a bit(seems to be an issue with the older Lithtech engine games).

Yes, I'll be sure to test it out sometime. That's probably about the only thing I didn't try.

If the problem showed up in something else, I'd tend to believe Croteam.

But the problem doesn't show up on the Ti500 for reasons that are possible to deduce based on Croteam's explanation of the issue.

And the latest WHQL certified ATi driver is the only thing that has managed to crash my computer totally.

And? I've had nVidia drivers crash my system before too.

Forgive me if I put little faith in to the certification process involved

The WHQL process is definitely better than a leaked driver release.

but realistically, WHQL certification can't be much different from the QA a product goes through in-house.

It is different because the rules are laid down by Microsoft, not by the manufacturer's QA. If you look at the some of the requirements for WHQL you'll see that they're looking for very different things that the manufacturer tests.

 

[Rogozhin]

Chizow

I'm interested in a response to my last post if you don't mind

rogo

 

[OpStar]

I like my FX Ultra. How many ppl that bash them HAVE one? Not many. Can barely hear the fan over mp3s on my z560s at low volume (not at all in games with my Senn's on).

Its hot, that I hate, oh well.

I had a 9700 pro, wouldn't play GTA 3, too many CS problems, hella hard to try to get everything to mesh, and I have a e702 as well.

I like nVdia, haven't had one issue with this card yet.

and yes, its fast

6158 2k3

not to mention that Splinter Cell 10x7 4xaa and 8x ansio I get 35~40fps average.

I'd talk to ppl who actually OWN the card and have experience with it, and not ppl that read reviews and coat tail on others opinions.

 

[kylebisme]

you could take it from nvidia themselves who says the product failed. here is another funny story about how to cook an fx ultra.

 

[chsh1ca]

Originally posted by: OpStar
I like my FX Ultra. How many ppl that bash them HAVE one? Not many.

How many people have them? According to the others on the forum who have owned them, a high percentage of the people who have purchased them (according to various posts and polls) have returned them because they were both too noisy and too slow for the money they laid out.

Can barely hear the fan over mp3s on my z560s at low volume (not at all in games with my Senn's on).

The fact that you can hear it over MP3s is a bad thing for quite a few people who don't want a lot of noise. I personally can't hear my machine running even during high-CD access speeds, and wouldn't change that at all.

6158 2k3

Wow, now you're quoting a known-bad benchmark to prove your point? Now that is funny. Guess you haven't heard anything about nVidia's drivers that intentionally render things differently in order to boost their higher 3dMark2003 score then, have you?

I'd talk to ppl who actually OWN the card and have experience with it, and not ppl that read reviews and coat tail on others opinions.

The point is that people here who participate in the forums HAVE had it and didn't like it. Are you suggesting that the merits of a piece of hardware someone doesn't own but has a lot of information on can't be discussed simply because you don't like what they're saying? There are a lot of facts to support the various arguments being discussed.

What I find most amusing is that you as well compare the FX5800 to the R9700Pro, instead of the 9800Pro. It seems that even people who like the card recognize it's slower than ATI's flagship.

 

[gururu]

I'm glad Opstar provided us with his take, because he raises a good point. That point is,
the FX is a fast video card. It is arguably competitive with the 9700pro and 9800pro.
I personally haven't read any posts from people that have owned the card. I'm not surprised either
since production level was very low for the ultras. All I can say, is that for those that own them, I hope they
enjoy them, while being able to put up with all the trash talk condemning their purchase. No sense in making people feel stupid or ignorant over things that are largely based on bias, opinion, suggestion, ignorance, peer pressure, mob mentality, educated guessing, or EVEN hands-on experience.

Opstar, my 9700pro has never come close to that 3DMARK score. I'm very impressed.