Very Odd FiringSquad GeForceFX review

[VTi2]

Russ,

It strikes me that you are so trusting of nvidia, for someone whose in the semi-con industry it strikes me as odd.

you said "its been stated that it can be benchmarked for a long period of time while overclocked with no problems." pray wheres the source of this statement?

It is obvious to me that the geffx at 500 MHz is obviously overclocked, and i would not be surprised about clock throttling. yet someone as intelligent as you tries to look the other way.

Shame on you!

 

[RussSchultz]

Shame on me? Shame on you for not reading!

We were told by NVIDIA that GeForce FX 5800 Ultra had been extensively tested in ovens, so they don’t believe it’s an issue with heat, rather a bug in the newer drivers (as they hadn’t run into these issues with 42.63). Quite honestly, we tend to believe them, as we were able to run countless tests with 3DMark 2003 while the GeForce FX 5800 Ultra board was overclocked, so we’re hoping that this will be addressed soon.

http://firingsquad.gamers.com/hardware/geforce_fx_5800_ultra/page5.asp


As for the rest of your comments:
Why should I not trust NVIDIA on this? They're on a public stage and any shenannigans will be exposed. How can they believe they'll get away with it?

Obviously overclocked? Do you even know what overclocking is? Its running it faster than the manufacturer specs it for. NVIDIA is the manufacturer, so they're not overclocking it, by the definition.

But beyond that, the manufacturer specs it based on actual test data to determine MTBF: running burnin tests, voltage characterization tests, etc. All chip designs have a performance curve that requires so much voltage at so much speed and produces so much heat. During manufacturing, each chip is tested that it is within those parameters. If the manufacturer feels they can run it at X speed, and the silicon will last for the life of the product, then they're not "overclocking" it.

 

[Joe DeFuria]

Why should I not trust NVIDIA on this? They're on a public stage and any shenannigans will be exposed. How can they believe they'll get away with it?

Since when has the risk of shenannigans being exposed actually deterred marketing from performing those shenannigans? Did you just ignore the whole 4x2 / 8x1 debate?

To be clear, I give it maybe 50/50 that the Ultras will ship with a "downclocking" problem. I'm only saying that I have no idea why you think any company (particularly nVidia, and particularly with this low volume part) would have a problem with doing something like this.

I mean, the advertise the downclocking thing as a feature, both for noise reduction and protection IIRC.

 

[RussSchultz]

You're saying most companies would have no problem shipping a product that obviously doesn't function under normal operating conditions?

 

[WaltC]

]i think if I read any more of these Nausiatingly innacurate pandering "Reviews" Im going to be sick...

*Tweaked ATis AF to make it the same IQ as the GFFX*

Is so Completely Absurd that I Am at a complete loss for words...

I wouldn't worry too much, as it doesn't appear the 500MHz nv30 chip will ever ship into the mass market. The only fair way to do it, unfortunately, is either to omit the nv30 products entirely (as some sites have done, since the products aren't shipping), or else to do it as [H] has done it and restrict all testing to the only mass-market accelerator it seems nVidia may actually ship--the non-Ultra FX 5800 (at the 400/400 clock.)

Heck, 15 years ago the print review mags were always in the pockets of the hardware companies--and worried a lot more about ad income than about their readership. Most of them are gone, now. As the dot.com boom continues to go bust the pennies get tighter and tighter so you can expect to see the whole thing deteriorate in exactly the same way that print went. Plus, you get a somewhat lower caliber of people doing the writing, too--and they simply don't know what questions to ask and what statements to make so they depend on their kindly "sponsors" to fill in the gaps. So you wind up with a lot of meaningless rubbish to pour over.

The question that I'd love to see answered is even among the ones who write in such glowing terms about the all-but-positively-cancelled 5800 Ultra, the chip-without-an-OEM, how many of these people would actually fork over $400 to *buy* one? When you consider that the R300 is approaching its seventh month of quantity production and the R350/RV350 may actually hit store shelves before nVidia's 5800 non-ultra see the light of day in a store in a meaningful quantity--it just all becomes even more a paraody of the truly absurd, and much ado about nothing. Definitely full of sound and fury...*chuckle*...and even though that does nVidia no good at the bank, perhaps some of the more clever websites can relieve nVidia of yet a little more its cash...

I agree, though--this is a definite low mark--one of the worst I've seen in years inside of 3D technology.

 

[Joe DeFuria]

You're saying most companies would have no problem shipping a product that obviously doesn't function under normal operating conditions?

That's the trick....define "normal operating conditions." How many times has the "consumer" definition of normal, not quite met the "company line" of normal. What? I can get an LCD that doesn't have 100% working pixels, and that's "normal?"

No one is accusing nVidia for possibly shipping the NV30 in a form that doesn't function. (Or are you putting words in out mouths?) "Downclocking" obviously IS a function of the chip, control panel applet and all. I see no reason to assume that in "company line terms," they would say "the occaisional slight reduction in speed is normal. This is a protection measure to ensure that no damage comes to the chip. It's temporary, and will resume speed...."

 

[RussSchultz]

Well, yeah, they might be able to SAY that, but they'd be crucified. (even worse than they're getting crucified for their 4x2/8z thing). They'd be crucified even worse because downclocking while using it in its intended application has a direct and visible impact on the user experience.

 

[Joe DeFuria]

That's my point Russ...they'll weigh the consequences of possible "crucifiction" of shipping such a part, vs. some other crucificion for some alternative....for say, not releasing the part at all, or lowering top speed to 450 Mhz, etc.

I never said they didn't run the risk of crucifiction. Only that it doesn't mean they won't do it. The alternatives could be worse. It's a judgement call.

 

[RussSchultz]

I'd say it was highly highly highly unlikely. None of their OEM partners would put up with the returns generated.

 

[Joe DeFuria]

What...return for a maxium of 5 cards that have been produced?

 

[nelg]

I have been putting the question of the down clocking of the Ultra to Dave here and Brent at [H] for a while. Perhaps no one will comment on this until they test an actual retail card ? I do wonder though, what are the legal ramifications of listing 500/500 speeds (unless marked max, sometimes, etc.) when it appears they are not substainable. If this card can not run at its advertised rate all the time then all of its benchmark scores are purely academic.

 

[Hyp-X]

It's not just the Doom3 engine, for example, Command and Conquer Generals uses either shadow volumes or buffers to achieve self-shadowed lighting everywhere.

It uses shadow volumes.

You can tell because the trees don't have proper shadows.

 

[kyleb]

Why should I not trust NVIDIA on this? They're on a public stage and any shenannigans will be exposed. How can they believe they'll get away with it?

well how could they not get away with it if everyone takes that attitude?


btw, what "semi-con industry" is VTi2 refering to anyway?

 

[RussSchultz]

Semiconductor industry. I work for a company that provides the chips for most MP3 players these days (plus a lot of AC97 codecs).

And Christ, Kyleb. What attitude am I taking that will allow them to get away with anything? That I believe that there is enough scrutiny of everybody's products that hiding something as blatant as not running at full speed half the time will get exposed in no time flat? The fact that even casual use of the product would expose this issue?

*boggle*

 

[kyleb]

argueing a comment like "if everyone takes that attitude (of just assumeing it will all be allright)" with "I believe that there is enough scrutiny" is blatent rhetoric. i am extreemly offened at the way you treat me with such disrespect as to even try such things, is this intentional or simply part of your nature?

 

[RussSchultz]

I'm sorry, Kyleb. Respect is a commodity. You have to give it to recieve it.

You've been going negative in my bank account by consistantly sniping me with comments like "if everybody had that attitude" or "perhaps truth would be a better policy", when you weren't even involved in the conversation and incorrect with your assessment about who was right to boot. Even in your last statement you refer to my reasoning as rhetoric, which it clearly wasn't.

You're welcome to your opinions, but don't expect me be respectful to you if you continue to snipe at me.

 

[Dave H]

I think Russ is taking for granted his knowledge of where the line gets drawn between marketing and engineering in the industry. To someone who doesn't understand this, it may seem natural that a company would be just as likely to ship a product that can't operate according to spec as they are to call a 4x2/8z organization "8 pixel pipelines".

But Russ is right. If the 5800 Ultra were really overheating under normal operating conditions, it would likely be recalled. But much more likely is that it wouldn't have shipped if the problem is as common as it appears to be in the review samples. Instead it's almost certainly either fixed in the final boards, or fixed or at least identified if it's a driver issue.

 

[demalion]

I think Russ is taking for granted his knowledge of where the line gets drawn between marketing and engineering in the industry. To someone who doesn't understand this, it may seem natural that a company would be just as likely to ship a product that can't operate according to spec as they are to call a 4x2/8z organization "8 pixel pipelines".

But Russ is right. If the 5800 Ultra were really overheating under normal operating conditions, it would likely be recalled. But much more likely is that it wouldn't have shipped if the problem is as common as it appears to be in the review samples. Instead it's almost certainly either fixed in the final boards, or fixed or at least identified if it's a driver issue.

Well...the Ultra sort of did get recalled, as someone indicated above I think.

To me, it indicates two things about the Ultra: 1) if it was widely available in the retail channel, it would have issues that "prove Russ wrong", 2) it isn't widely available in the retail channel, so it "proves Russ right".

I think a blurry line between those two things is facilitating disagreement where I don't think one really exists.

 

[chavvdarrr]

quality of AF: ATi vs NV ?

Strange things happen these days...
pro-Ati people say that R300's AF is much better than FX's. that even 8xAF performance is better/on par with FX 8xAF balanced
pro-NV people say that Ati's "ripmap"Af is bad. So bad that FXs 8xAF balanced is on par with Ati's 16xAF quality.
Who's right? Can someone PLEASE, give some hints what to look for in images when comparing AF quality.... I for myself think that:
GF4 AF is better and slower than Ati's AF in R300. Much better than R200's AF
FX balanced is faster and a bit worse than GF4's AF
R200 AF is bad. Not always, but sometimes is really bad (no AF at all at some angles.)
R300 AF is better than R200s. But on 22.5/45 etc. degrees is bad too
So..... who sees what he wants to see and who sees what he sees...

Any thoughts? 10x

 

[Nebuchadnezzar]

Check out Leadtek's cooling solution :