Fermi benchmarked? [Fake]

[thilanliyan]

http://www.guru3d.com/news/geforce-gtx-360-and-380-benchmarks/

originally posted at:
http://www.xtremesystems.org/forums/showthread.php?t=240958

😱 if true!!

BUT the 5870 had amazing performance on PR slides too and that didn't fully pan out. Who knows...maybe someone at NV/a fanboy made them up to get people to wait for Fermi. I was going to wait for Fermi anyway and then decide whether to get a 5XXX or one of the Fermi cards.

EDIT: It's been mentioned my Keysplayr that these are definitely NOT from nV and others at XS have said things to the same effect.

So to a mod, could you please lock this thread? THanks.


Locked at OP's request, because the results are fake, and because T2k can't stop trolling.

Video Mod BFG10K.

 

[T2k]

These are PR slides form Nvidia, NOT BENCHMARK results.
They simply took some numbers and interpolated old results.

 

[T2k]

PS: there's not even a single working final chip yet, forget cards. It's some PR BS, someone threw up at NV and a rather lame at that - heck, even 58xx results are off...

 

[thilanliyan]

THe Clear SKy results are not too far off what they're getting here:
http://www.bit-tech.net/hardware/graphics/2009/11/18/amd-ati-radeon-hd-5970-review/7

Resident Evil 5 (these numbers are actually lower than what's shown on the slide):
http://www.legitreviews.com/article/1141/6/

Farcry 2 (these numbers are lower on the 5970 and 295):
http://www.guru3d.com/article/radeon-hd-5970-review-test/14

 

[Cookie Monster]

PS: there's not even a single working final chip yet, forget cards. It's some PR BS, someone threw up at NV and a rather lame at that - heck, even 58xx results are off...

These benchmarks are most likely false, but to say there isnt even a single working final chip is quite humorous.

 

[Wreckage]

Geez the 360 will be a match for the 5970. A dual fermi will be an absolute monster.

These numbers do seem in line with other rumors out there.

This is shaping up to be the next G80. (Which means I would be able to afford one till the refresh).

 

[T2k]

These benchmarks are most likely false, but to say there isnt even a single working final chip is quite humorous.

Perthaps for some who does not know what "working" means - there's a reason why they had to do another respin... final chip will be in their hands in early Jan, best case.

 

[Voo]

These benchmarks are most likely false, but to say there isnt even a single working final chip is quite humorous.

According to all the rumours the A3 stepping went in the oven around 1 Dec. and Nvidia will get them back around Jan 1 - so either these are benchmarks from the A2 stepping (which probably isn't true, as A2 had some rather large problems) or someone just made up some numbers..


I'm not sure how anyone but wreckage could take benchmarks seriously if the hardware on which it should be run isn't even in Nvidia's hands

 

[Nemesis 1]

Man this PR shit has to stop . Next thing well see is ATI 6000 series leaks beating fermi to a pulp . Than what. The ATI people say wait and by the 6000. its only 5 months out from fermi release . Than NV boys argue the oppsit argument they are using now . Its a dog chasing its own tell .

 

[Wreckage]

According to all the rumours the A3 stepping went in the oven around 1 Dec. and Nvidia will get them back around Jan 1 - so either these are benchmarks from the A2 stepping (which probably isn't true, as A2 had some rather large problems) or someone just made up some numbers..


I'm not sure how anyone but wreckage could take benchmarks seriously if the hardware on which it should be run isn't even in Nvidia's hands

You are making a lot of statements as if they are facts. I'm curious as to how you know for a fact all these details about the respins, the problems, the dates, etc. 😕

 

[T2k]

According to all the rumours the A3 stepping went in the oven around 1 Dec. and Nvidia will get them back around Jan 1 - so either these are benchmarks from the A2 stepping (which probably isn't true, as A2 had some rather large problems) or someone just made up some numbers..


I'm not sure how anyone but wreckage could take benchmarks seriously if the hardware on which it should be run isn't even in Nvidia's hands

Exactly.

 

[TheMeanestGuest]

PR slides are rarely if ever accurate. From AMD or from NV. I don't think anyone takes them seriously anyways.

I expect any lead GF100 (or whatever it's called today) has won't be quite so great as presented, if it does lead.

 

[Voo]

You are making a lot of statements as if they are facts. I'm curious as to how you know for a fact all these details about the respins, the problems, the dates, etc. 😕

I think I'll just quote myself: "According to all the rumours" - yep I see your problem, I see. And if the A2 stepping was perfectly well, exactly why would they need another stepping?


And I'd assume that keeping the date at which TSMC got the A3 stepping secret, would be rather troublesome, I mean that surely involves a LOT of people.

 

[ShadowOfMyself]

Geez the 360 will be a match for the 5970. A dual fermi will be an absolute monster.

These numbers do seem in line with other rumors out there.

What rumors? There are no performance rumors from Fermi whatsoever LOL

This is shaping up to be the next G80. (Which means I would be able to afford one till the refresh).

Suuuuure... Oh wait, G80 wasnt 6 months late

 

[Hyperlite]

Doesn't NV usually do three spins? And i think BFG's predictions about ATi drivers hold some water...as soon as fermi drops, 5xxx cards are going to magically get a sizable performance boost...

 

[dguy6789]

Seems kind of silly that the GTX 295 is so close to the 5970 in those slides, yet every review site shows it being 60-90% faster than the GTX 295.

Conclusion: Benchmark slides of a card that doesn't exist comparing to one that has been released for a while now: Nvidia false marketing again

The GTX 380 will probably be as fast as the GTX 295 and the GTX 360 will be a bit slower. Fermi is the next NV30.

 

[MrK6]

A little too ambiguous to really glean anything from, I think. Being as late as they are, if the Fermi based cards did perform this well, they could definitely sell well. Of course, looking at the 5870's overclocking capabilities, I'm sure a 1GHz "5890" part could easily be out the door the minute Fermi is released.

According to all the rumours the A3 stepping went in the oven around 1 Dec. and Nvidia will get them back around Jan 1 - so either these are benchmarks from the A2 stepping (which probably isn't true, as A2 had some rather large problems) or someone just made up some numbers..

I'm not sure how anyone but wreckage could take benchmarks seriously if the hardware on which it should be run isn't even in Nvidia's hands

Good logic, that's where I am too. There's no confirmed release date, and most reports point to March-ish; yet they have a working part and can run games well with some seriously beta drivers on a brand new architecture? Sorry, bullshit intake overload.

 

[BenSkywalker]

There's no confirmed release date, and most reports point to March-ish; yet they have a working part and can run games well with some seriously beta drivers on a brand new architecture?

Obviously these benches can't be taken seriously at all, but I can assure you that nV has working Fermi chips with drivers ready to roll. Since A0 the respins have been about improving yields, they have had working chips for a long while now and even if they didn't driver development for new parts starts off on simulators- if it didn't every new architecture would need to be done long before parts were able to ship(which isn't the case). This is not nV specific, ATi does the exact same thing for driver development on new architectures.

 

[tviceman]

PS: there's not even a single working final chip yet, forget cards. It's some PR BS, someone threw up at NV and a rather lame at that - heck, even 58xx results are off...

Actually those radeon 5x00 results on this mystery slide are nearly spot on with Anandtech's 5970 review results.




Charlie bashed Fermi's A2 silicon, saying that it's clock speeds were no where near where they wanted them to be. Well..... as we all know there are extreme cooling solutions and who's not to say that nvidia had a fully feature-functional A2 fermi that when applied with an extreme cooling solution, can get close to, if not equal, their expected clock speeds with A3 silicon.

I'm not saying these slides are for real, but everything I just stated is highly possible.

 

[MrK6]

Obviously these benches can't be taken seriously at all, but I can assure you that nV has working Fermi chips with drivers ready to roll. Since A0 the respins have been about improving yields, they have had working chips for a long while now and even if they didn't driver development for new parts starts off on simulators- if it didn't every new architecture would need to be done long before parts were able to ship(which isn't the case). This is not nV specific, ATi does the exact same thing for driver development on new architectures.

I'm sure they do have drivers for it, but how optimized are they? We continually see improvements and gains even after a cards release, and I imagine there's a mad dash before release to get the drivers as ready as possible for release - so what's the state three months before release? My point was that it's a little extreme to see such numbers on such early revisions - over the next three months with typical driver enhancements/improvements taken into account, I'd expect the performance increase over the previous generation to be ludicrous at launch. Hence why I can't see these being real in the slightest.

 

[StinkyPinky]

I'm sorry, but if Nvidia had cards THAT damn fast, theyd be inviting hardware review sites to come and have hands on experience and to let them benchmark.

 

[BenSkywalker]

We continually see improvements and gains even after a cards release, and I imagine there's a mad dash before release to get the drivers as ready as possible for release - so what's the state three months before release?

A lot of that has to do with newer games and better optimizations hitting for those games. If you look at the games listed, they all have been out for quite a while, it wouldn't be shocking if Fermi had highly optimized drivers for those titles right at launch(again, I think the chart itself is total imaginary BS, just stating high profile games used in almost everyone's bench charts are going to be what any smart GPU maker optimizes for first).

 

[MrK6]

I'm sorry, but if Nvidia had cards THAT damn fast, theyd be inviting hardware review sites to come and have hands on experience and to let them benchmark.

Excellent point as well.

A lot of that has to do with newer games and better optimizations hitting for those games. If you look at the games listed, they all have been out for quite a while, it wouldn't be shocking if Fermi had highly optimized drivers for those titles right at launch(again, I think the chart itself is total imaginary BS, just stating high profile games used in almost everyone's bench charts are going to be what any smart GPU maker optimizes for first).

Interesting, thanks, I didn't know finalized optimizations could occur that early.

 

[Keysplayr]

These slides aren't real. And they are not from Nvidia.
Seems "pre-mass hysteria" want's to rear it's ugly head.

Probably just some fan having fun with photoshop.

 

[ronnn]

Nvidia false marketing again

Possible, but just as likely amd guerrilla marketing building up false expectations. At least nice for nvidia to actually get noticed... Outside of the odd photo of a dumb looking pr guy with a maybe card or answering questions on forums, the nvidia pr has not had much focus lately.