GF100 evaluation thread

[eastmen]

Haha, what a cop out. You don't have much tolerance for opinions different from yours, eh.

Huh ? I presented evidence of why many of doubt a gtx 495 or whatever the sli gtx 480 would be on a single card.

It uses a crap ton of power in its single gpu varient. I showed the gtx 470 using slightly less than a dual cypress gpu board as more proof.


I am certian that both ati and nvidia can produce cooler less power hungry chips as time goes on using the same process node. I just don't see it happening for nvidia for awhile. The dual cypress is already at gtx 470 power draw and a gtx 480 is already pushing 100c at full load. I'm not sure where anyone sees an mgpu gtx coming in the future. Mabye a 2 slower clocked gtx 470s might be possible on a single board with binned parts. But at that point it may still loose to the 5970 because of lower clocks

 

[compres]

Ah that must be why you used the word exactly then.



Tssk tssk, there you go again.



Heh, I honestly don't care what you think about the value of this card. You seem really distraught about the fact that not everyone agrees with you though, to the point that you start accusing people of being unreasonable and biased. Do you experience these feelings of insecurity more often?

Just provide me with any hard facts that make the 480 a good buy.

And keep the personal attacks coming, like I care.

 

[Florin]

Huh ? I presented evidence of why many of doubt a gtx 495 or whatever the sli gtx 480 would be on a single card.

It uses a crap ton of power in its single gpu varient. I showed the gtx 470 using slightly less than a dual cypress gpu board as more proof.


I am certian that both ati and nvidia can produce cooler less power hungry chips as time goes on using the same process node. I just don't see it happening for nvidia for awhile. The dual cypress is already at gtx 470 power draw and a gtx 480 is already pushing 100c at full load. I'm not sure where anyone sees an mgpu gtx coming in the future. Mabye a 2 slower clocked gtx 470s might be possible on a single board with binned parts. But at that point it may still loose to the 5970 because of lower clocks

Oh I completely agree that the dual GTX is not looking very likely at all at the moment, at least not with < 300 Watt TDP. But to put trini on ignore for considering the possibility, c'mon now..

 

[neliz]

Cheer up guys, GT300 beat Larrabee, time for a party!

 

[eastmen]

Oh I completely agree that the dual GTX is not looking very likely at all at the moment, at least not with < 300 Watt TDP. But to put trini on ignore for considering the possibility, c'mon now..

Sometimes instead of bickering with a person its better to ignore them. Sometimes its hard to keep reading thier comments and so having the posts auto ignore makes it easier.

 

[nutball]

This thread needs pruning. Hard.

 

[trinibwoy]

Haha, what a cop out. You don't have much tolerance for opinions different from yours, eh.

Well several pages ago I made a point about Fermi being a bigger gain over GT200 both feature and performance wise than Cypress was over RV770. After several futile attempts to deflect from the issue because there wasn't any evidence to the contrary the best option was probably to give it up. :smile:

 

[compres]

Well several pages ago I made a point about Fermi being a bigger gain over GT200 both feature and performance wise than Cypress was over RV770. After several futile attempts to deflect from the issue because there wasn't any evidence to the contrary the best option was probably to give it up. :smile:

Cypress consumes less power than rv770. Compare power on gf100 to gt200.

 

[KimB]

Cypress consumes less power than rv770. Compare power on gf100 to gt200.

While that is a true statement, it doesn't contradict what trini said there.

 

[Arty]

While that is a true statement, it doesn't contradict what trini said there.

Like I said numerous occasions earlier, I can find reviews that have a different suite of benchmarks where GF100 doesnt scale as good. Do you honestly think this is impossible? If so, then who is the one cherry picking there.

Besides, if no one cared about the power draw or heat issues then I'm sure AMD would have also clocked Cypress at 1.1GHz and wiped the floor. The minority, same group of people are suggesting we look at the GF100 in a vacuum, this can never be done. A product is a package.

 

[eastmen]

While that is a true statement, it doesn't contradict what trini said there.

some might say power and heat are parts of performance.

So while the gtx 480 might be slightly faster than the 5870 it uses alot more power. When you factor in heat , power usage and absoulte power which one does better ?

Some might also say time to market is important also. The gtx 4x0 series will be almost 7 months after the 58x0 series launched. What would peformance for fermi be in sept of 2009. What would a cypress part launched in the high end of april 2010 yield over a sept 2009 varient.

 

[Malo]

While that is a true statement, it doesn't contradict what trini said there.

True, but trini should also not discount that power and heat come into play as well. Regarding the specific discussion, Cypress is at least as good as or better than RV770 regarding power and heat. GF100 is actually worse than GT200 for both.

 

[homerdog]

Originally Posted by Bouncing Zabaglione Bros.
1. flawed product
2. cut-down
3. over-volted
4. late to the party
5. expensive
6. hot
7. noisy,
8. yet we have all these people saying how wonderful it is because all the previous ridculousness have given it a 0-15 percent benchmark lead on ATI's second (not first) tier product.

1. Are you a GPU designer? If so, in what way is it flawed.
2. Yep, I think they threw in the towel too
3. How do you know it is over volted? See 1 and supply info where it should have been lower than 1v.
4. Yep
5. 470 is about right, but the 480 is over priced
6. Yep
7. only relavent to the end user. I have been using a nice set of stereo noise reduction head phones for about 5 years now, so I dont hear my case with its 2x GTX260s and fans set too 100% and 6 120mm fans.
8. The overall average lead is 15% with launch(are the review drivers going to be shipping with it) drivers compared to a much more mature driver set for the 5870. I'd call this not to shabby in my book.

I agree, and noise would also not be an issue for the hearing impaired. Therefore it is not valid to criticize GF100 for being noisy.

 

[KimB]

some might say power and heat are parts of performance.

So while the gtx 480 might be slightly faster than the 5870 it uses alot more power. When you factor in heat , power usage and absoulte power which one does better ?

Depends on the person. There is no non-arbitrary way to weigh performance against power consumption.

 

[Malo]

I agree, and noise would also not be an issue for the hearing impaired. Therefore it is not valid to criticize GF100 for being noisy.

The deaf can't hear the fan therefore it is not noisy for anyone. Classic.

 

[RecessionCone]

Geometry

I'd be interested in hearing people's perspective on Nvidia's claims that the future of real-time graphics is more in geometry than pixel shaders. Of course, it's hard to predict the future, but it does seem to be the case that film-quality 3d rendering makes beautiful pictures by rasterizing huge numbers of tiny polygons. If we believe that the future of real-time graphics is to approach film rendering, than GF100's emphasis on geometry over pixel shaders seems justified.

However, if we think that it's premature to push for larger geometric complexity, and instead what we really need in a new graphics architecture is more pixel shader power, than GF100 has really missed the boat, since one of its main strengths will never be practically useful.

The reason I bring this up is that I think much of today's discussion on GF100 is too conservative & backward-looking. We're judging the architecture as a failure because it's not just more of the old familiar pixel-pushing we've been used to. I sympathize with this discussion, since we have to judge things by what we can measure, and we can't measure tomorrow's workloads. But if gaming workloads really do shift towards geometric complexity, physical simulation, computer vision (Project Natal) and software rendering (a la Intel's vision), AMD's architecture will start looking much less efficient/W or mm^2 than it does now. And GF100 will look fairly prescient as an architecture, albeit flawed as a product.

Maybe I would still buy a 5870 if I were on the market for a card today, since it's a better card in terms of perf/W or perf/$ on today's games. But that doesn't mean GF100 is a failed architecture - I can still appreciate the risks and vision behind it.

 

[no-X]

The overall average lead is 15% with launch(are the review drivers going to be shipping with it) drivers compared to a much more mature driver set for the 5870. I'd call this not to shabby in my book.

Are you a software engineer? If so, what exactly is more mature on Evergreen drivers?

Anyway, nVidia had A1 in labs since fall. Even if it ran at lower clocks, they could work on drivers all the time, couldn't they? I don't believe nVidia started to work on Fermi drivers a week before launch

 

[Farid]

I guess, people care if they get that worked up

This thread needs pruning. Hard.

Yeah, but where to start?

Seriously, folks, try to respect your fellow B3Der, enough with the ad-hominem and the condescending nonesense some of you throw a each others.

 

[compres]

I'd be interested in hearing people's perspective on Nvidia's claims that the future of real-time graphics is more in geometry than pixel shaders. Of course, it's hard to predict the future, but it does seem to be the case that film-quality 3d rendering makes beautiful pictures by rasterizing huge numbers of tiny polygons. If we believe that the future of real-time graphics is to approach film rendering, than GF100's emphasis on geometry over pixel shaders seems justified.

IMO it is very justified.

The debating is because some claim that the ridiculous power and heat levels are acceptable.

 

[neliz]

IMO it is very justified.

The debating is because some claim that the ridiculous power and heat levels are acceptable.

I'll continue with this Cinema 2.0 train of thought

Gaiking