NVIDIA Prepping dual-GPU GF104 card

[el Kuarlos]

ROFL Typical nvidia. I love it. Can just see them in the boardroom.

CEO- 'We need a dual GPU fermi!!!'

Techies- 'We can't, it will melt, we are already way over what we should be for TDP'

CEO- 'But the 480 works?'

Techies- 'In a manner of speaking, yes.'

CEO- <narrows eyes> 'Get me my wood screws...'

Techies- <shuffle and look scared> 'Why?'

CEO- 'There is a reason why I am the boss round here you ninnies...we shall take a 480...and cut it in half!!!'

 

[Rroff]

For some reason theres been several reviews showing the GTX4x0 not doing so well in Bad Company 2 - but there results are wrong (or old) - anyone with an SLI setup will tell you that they get far better scaling than indicated in those reviews.

Overclocked GTX470 SLI is currently the best value and performance solution IMO, despite the fact that a single GTX470 doesn't represent well in either.

 

[RavenXXX2]

Latest drivers....

http://www.hardocp.com/article/2010/05/10/galaxy_geforce_gtx_470_gc_sli_review/1

 

[Flanno]

So is there a placebo effect in play here because I have read lots of user reviews on the forums on xs, hardforums, guru3d. Everyone is saying at 2560 res that gtx 480's in sli offer a much smoother gaming experience then crossfire and in some cases trifired 5870's. I guess they are talking about micro stuttering here and lack of it on nvidia sli setups.

I'm gonna bench a load of games I've installed on my noisy gtx 480 setup, and swap back in my 5870's and see for myself.

 

[Rroff]

Latest drivers....

http://www.hardocp.com/article/2010/05/10/galaxy_geforce_gtx_470_gc_sli_review/1

snip...

While I've never held much by hardocp reviews... what that graph doesn't tell you... you can double the AA on the nVidia setup over the highest possible settings the ATI setup is comfortable at... personally I think that graph is ******** as the graphs above show a completely different picture and its different to what users - including people who have owned both cards - have said.

 

[gurusan]

wtf at that graph. Those are the kinds of numbers I was getting with my 260s in SLI. ...only difference was I was running 4xAA

Still... :/

 

[jigger]

Dual Fermi = twice the failure.

 

[RavenXXX2]

wtf at that graph. Those are the kinds of numbers I was getting with my 260s in SLI. ...only difference was I was running 4xAA

Still... :/

And DX10

So is there a placebo effect in play here because I have read lots of user reviews on the forums on xs, hardforums, guru3d. Everyone is saying at 2560 res that gtx 480's in sli offer a much smoother gaming experience then crossfire and in some cases trifired 5870's. I guess they are talking about micro stuttering here and lack of it on nvidia sli setups.

I'm gonna bench a load of games I've installed on my noisy gtx 480 setup, and swap back in my 5870's and see for myself.

The 480's have the advantage at max rez with the extra 500mb of vram they have, the 1GB ATI's will stutter and lag at max rez in the most demanding games with max IQ, that's why if you game at that rez the 480's are the best option, if you hate nvidia then the 2GB ATI cards.

 

[Rroff]

wtf at that graph. Those are the kinds of numbers I was getting with my 260s in SLI. ...only difference was I was running 4xAA

Still... :/

I've always considered ***** reviews, to borrow a term from jigger, to be somewhat fail - a lot of the non apples-to-apples stuff is subjective and even slight changes to settings can completely upset the um applecart, meaning the results are pretty meaningless... and even the apples-to-apples results seem to be handpicked for whatever point they are going after.

 

[jigger]

I never say fail. If you want good benches techpowerup have got to be one of the best.

 

[RavenXXX2]

They are meaningless to you Rroff because they don't give the results you want to see, if it was 470SLI creaming the 5850's there would not be a problem on your end. I have seen SLI results from 480/470 SLI users on youtube and they are not all that impressive, yes they scale better than ATI in some games and not so well in others but they do not demolish ATI at all.

 

[Rroff]

I never say fail. If you want good benches techpowerup have got to be one of the best.

They would be... if they actually used proper driver releases instead of weird revisions like 196.01 that no ones actually ever used.

 

[RavenXXX2]

I like techpowerup but he needs to update the games he benchmarks ASAP, we do not need to see how our DX11 beasts do in farcry, quake 4 and prey.

 

[jigger]

Rroff they state the driver revision on every review. They use the latest available across the benchmarking run.

They run fair and genuine tests.

 

[Rroff]

Rroff they state the driver revision on every review. They use the latest available across the benchmarking run.

They use the latest, now, with the GTX4x0 cards (which are the only ones they bother to benchmark, the rest are just pulled from old results database) with the latest drivers but thats probably because they have no choice... with other cards the results are more often than not from very weird choices of driver version:

NVIDIA: 195.62
GTX 470 & 480: 197.41
ATI: Catalyst 10.3

Who on earth has ever used 195.62? why even bother adding old nVidia card results, based on a driver that everyone skipped because they were rubbish, in amongst a test of ATI cards rebenched in the case of old ATI cards apparently as they don't list an old ATI driver release on the latest drivers? unless they are just slightly bias...

 

[Rapido!]

Dual Fermi = twice the failure.

You add so much to this forum.

Apparently Nvidia have shipped over 400,000 Fermi-based cards to date. This very site is replete with 480/470 stock, which is interesting because I remember you and others repeatedly stating that they would never make it to retail in any real numbers.

I mean, you really, really went on about it. Ad nauseum. Broken GPU! Cannot be made! You won't be able to buy it after day one!

Oops.

In light of your track record of Fermi-related wrongness, jigs, what exactly makes you think anyone is ever going to listen to you again?

 

[Flanno]

And DX10



The 480's have the advantage at max rez with the extra 500mb of vram they have, the 1GB ATI's will stutter and lag at max rez in the most demanding games with max IQ, that's why if you game at that rez the 480's are the best option, if you hate nvidia then the 2GB ATI cards.

Hmmm.....then why did the reviews at 2560x1600 show no difference hardly between 1gig 5870 CF and 2gig 5870 CF. Guess they didn't cover microstuttering. Anyway, it's irrelevant for me as my 5870's are the toxic 2gig ones. I know I was insane to buy 2 gtx 480's 3 weeks later. And probably mad to buy those 2gig 5870's too. Now I can't make up my mind which is faster.

 

[RavenXXX2]

LoL do you expect him to go back and bench every other " older " video card he has in the test with latest drivers...

This is why he needs to streamline his reviews to only the most recent cards.

 

[Rroff]

You add so much to this forum.

Apparently Nvidia have shipped over 400,000 Fermi-based cards to date. This very site is replete with 480/470 stock, which is interesting because I remember you and others repeatedly stating that they would never make it to retail in any real numbers.

I mean, you really, really went on about it. Ad nauseum. Broken GPU! Cannot be made! You won't be able to buy it after day one!

Oops.

In light of your track record of Fermi-related wrongness, jigs, what exactly makes you think anyone is ever going to listen to you again?

I'd love to quote back to certain people their comments about Fermi launch availability compared to ATI - as I said at the time give it til may and stock will pickup - unlike ATI where it wasn't available for months after the initial launch volume.

 

[jigger]

If that's whats out at the start of the tests, then that is the driver that should be used. When the next round of test are done they just use latest driver for that run. They are not pulling from a data base as the results change with time.