The GT300/Fermi Thread

[annihilat0r]

1) 360 beats 5850 and 5870, in which case the 380 should beat the 5970

no. What? There's at least 40% perf. difference between 5870 and 5970. If the difference between 360 and 380 is not HUGE, 360 beating 5870 will by no means mean 380 defeating the 5970.

What I think is that, 360 will be around 5870 and 380 will be 20-30% faster than that, which places it a lot above 5870 and a few percent behind 5970.

[GoldenTiger]

no. What? There's at least 40% perf. difference between 5870 and 5970. If the difference between 360 and 380 is not HUGE, 360 beating 5870 will by no means mean 380 defeating the 5970.

What I think is that, 360 will be around 5870 and 380 will be 20-30% faster than that, which places it a lot above 5870 and a few percent behind 5970.

5970 may be 40% "faster" than a 5870 on MAX fps, it sure isn't that much in most cases with minimum FPS at high settings such as 2560x1600 with AA by a longshot, sometimes barely doing much better than the single 5870. I don't care if the max is, hypothetically, 150 when the single card gets 120 max, but maintains 30 min vs. 35 min and the same average... crossfire/SLI scaling still are very driver-dependent and have a long way to go to make any real sense to me. The gains just aren't that amazing other than pure benching. Most of the cases where it has 40%+ gains are on tests where the framerate is 8-10, boosting to 12-14, still leaves it 100% unplayable .

[annihilat0r]

I was talking about average FPS

[saaya]

well it all depends on yields... if they CAN make full blown fermis with 512sps, they will def sell them... unless even those wont beat a 5970...
assuming a 512sp full blown fermi does beat 5970, then if yields are bad, they can still decide to make the 360 a 256sp part and sell all the "trash" that way, if thats still fast enough to beat the 5850 (should be... slightly more sps than a 285, more mem bw, maybe higher clocks...)

in that case youd have a big difference between a 360 and a 380, theoretical perf difference of 100%, realistically more around 50% i guess...

[Nedjo]

What I think is that, 360 will be around 5870 and 380 will be 20-30% faster than that, which places it a lot above 5870 and a few percent behind 5970.

IMHO 380 will be 15% faster than 5870 on average, and 360 will be in line with 5870, making the 5970 still the fastest card in the world!

[saaya]

IMHO 380 will be 15% faster than 5870 on average, and 360 will be in line with 5870, making the 5970 still the fastest card in the world!

that would make the 360 a 384sp part and the 380 a 480sp part, right?
and that would explain why they need a dual gpu card, to beat the 5970... i think nvidia will price that dual fermi card WAAAy high though, as otherwise they wont make a lot of profits with it, and they have problems ramping up production, so selling cards with 2 gpus when they have problems getting good yields is not a good idea... a dual fermi has very good chances at beeing a record breaker in more than just performance benchmarks, it has very good chances to be the first consumer graphics card that reaches or even surpasses the 999$ mark

but you know what?
people are spending that on cpus for years, and a highend vga is 10x more worth it than a highend cpu imo...
actually im surprised highend vgas havent been selling for more for the past years...

[kgk]

I was talking about average FPS

The number most people should be concerned with is minimum FPS.

So what if a card has 150 FPS max and a 70FPS average if the minimum drops into the teens?

I'd rather have a card that gets 90FPS max, 50FPS average and a minimum that stays above 25-30.

[Manicdan]

The number most people should be concerned with is minimum FPS.

So what if a card has 150 FPS max and a 70FPS average if the minimum drops into the teens?

I'd rather have a card that gets 90FPS max, 50FPS average and a minimum that stays above 25-30.

because a quick spike could simply be from a HDD loading issue

i prefer charts (hardocp does that right?)

[Solus Corvus]

Did you see nVidia statement about Fermi on its websites?

Certain statements in this press release including, but not limited to, statements as to: the benefits, features, impact, performance and capabilities of NVIDIA Tesla 20-series GPUs, Fermi architecture and CUDA architecture; are forward-looking statements that are subject to risks and uncertainties that could cause results to be materially different than expectations. Important factors that could cause actual results to differ materially include: development of more efficient or faster technology; design, manufacturing or software defects; the impact of technological development and competition; changes in consumer preferences and demands; customer adoption of different standards or our competitor's products; changes in industry standards and interfaces; unexpected loss of performance of our products or technologies when integrated into systems as well as other factors detailed from time to time in the reports NVIDIA files with the Securities and Exchange Commission including its Form 10-Q for the fiscal period ended July 26, 2009. Copies of reports filed with the SEC are posted on our website and are available from NVIDIA without charge. These forward-looking statements are not guarantees of future performance and speak only as of the date hereof, and, except as required by law, NVIDIA disclaims any obligation to update these forward-looking statements to reflect future events or circumstances.

So nVidia say it could lie about Fermi and get away with it ? (Example 512 cuda cores - ups we couldn't deliver that... you can have 448 tops) What you think?

It sounds like a normal disclaimer written in legalese to CYA. They probably realize they have a penchant for talking big and that could ultimately get them in trouble with the shareholders if things don't work out perfectly.

But more then that, it's also the truth. Sometimes things don't work out as expected. Sometimes something looks great on paper but doesn't measure up when produced. Sometimes external factors conspire to ruin a good product. Et cetera. It's easy for us, being so cynical, to look at those statements and think that they might just be using those statements as a cover for lying. But the reality is that they probably do think/hope that it will be awesome but have to admit that there might be problems - don't ding them for being honest, even if it's the legal department making them do it.


As for what those bolded statements mean: The first is probably a reference to airflow (ie. it's hot so if you put it in a case without enough airflow it will throttle). The second is probably a reference to not reaching target clocks or shaders.

[SnipingWaste]

If there going with 256 CC then that be it self shows there in bad shape. Think about it, that's half the CC turned off on a die that's very large and that cost alot to get. The only reason I can think of for them doing this is trying to find a way not to make money but trying to find a way to break even and not be in the red. They will be selling cards with a 256 CC at a loss but its better then trashing them. Even a 320 CC will be hard to sell to break even on with the low yields and the large size of the die.

[neliz]

2x256 > 5970?

2x320 > 5970 i could imagine... but 2x256?

Just some examples, could be 448 for my part, I honestly have no clue.

btw, launch in march = launch, or retail availability?

I thought nvidia didn't do hard-launches anymore?

[Zed_X]

Deep Dive meeting has over, and we already know how powerful GF100 is, and IT IS! Performance is better then expected, most peoples here will search in own words and feel to shy for craps in thread. We can only laugh now, because Nvidia did it! Perf in Unigine DX11 benchmark is spectacular, Radeon HD 5870 is there like low-end toy for kids against GF100.

PS. Yeah, GF100 has 512 CC, and GPUs are in massproduction now, plenty of manufacturers have first GF100 inhouse.

[Musho]

Deep Dive meeting has over, and we already know how powerful GF100 is, and IT IS! Performance is better then expected, most peoples here will search in own words and feel to shy for craps in thread. We can only laugh now, because Nvidia did it! Perf in Unigine DX11 benchmark is spectacular, Radeon HD 5870 is there like low-end toy for kids against GF100.

PS. Yeah, GF100 has 512 CC, and GPUs are in massproduction now, plenty of manufacturers have first GF100 inhouse.

Sounds interesting. Do you happen to have a source? Or were you at the deep dive meeting yourself?

[N19h7m4r3]

Deep Dive meeting has over, and we already know how powerful GF100 is, and IT IS! Performance is better then expected, most peoples here will search in own words and feel to shy for craps in thread. We can only laugh now, because Nvidia did it! Perf in Unigine DX11 benchmark is spectacular, Radeon HD 5870 is there like low-end toy for kids against GF100.

PS. Yeah, GF100 has 512 CC, and GPUs are in massproduction now, plenty of manufacturers have first GF100 inhouse.

If this is true I am really happy!

Competition will only make ATI bring out something better later and drive down prices.

If the cards aren't to big and power hungry they'll be perfect for what I mean to do.

[Dimitriman]

Deep Dive meeting has over, and we already know how powerful GF100 is, and IT IS! Performance is better then expected, most peoples here will search in own words and feel to shy for craps in thread. We can only laugh now, because Nvidia did it! Perf in Unigine DX11 benchmark is spectacular, Radeon HD 5870 is there like low-end toy for kids against GF100.

PS. Yeah, GF100 has 512 CC, and GPUs are in massproduction now, plenty of manufacturers have first GF100 inhouse.

I feel a few things here,

1- First, you're probably overexcited about nvidia "tweaked for fairness" benchmarks
2- We're talking about Unigine here, what about other benchmarks? No games?
3- We need some settings+actual numbers otherwise your statement just sounds like more hype.

PS: Do you have info on the lower range card (aka 360)?

[zerazax]

Deep Dive meeting has over, and we already know how powerful GF100 is, and IT IS! Performance is better then expected, most peoples here will search in own words and feel to shy for craps in thread. We can only laugh now, because Nvidia did it! Perf in Unigine DX11 benchmark is spectacular, Radeon HD 5870 is there like low-end toy for kids against GF100.

PS. Yeah, GF100 has 512 CC, and GPUs are in massproduction now, plenty of manufacturers have first GF100 inhouse.

Are you basing this on Nvidia's PR slides?

Cause if so, we all know that they're overstated (this is true of ATI too)

Seeing as how you've posted that actual cards won't be sent out til February, I'm going to assume you got your numbers from PR slides - so how about game performance, and not benchmark performance? And what you're saying - that it's better than everyone expected - seems to go counter against what everyone else has been hearing, and that includes hints from some Nvidia folks that the design was incredibly hard to achieve

Sorry if I'm not exactly trusting of a new poster from China who's only here to post on this thread, and from where I've heard more than a few rumors turn out false

[okorop]

Deep Dive meeting has over, and we already know how powerful GF100 is, and IT IS! Performance is better then expected, most peoples here will search in own words and feel to shy for craps in thread. We can only laugh now, because Nvidia did it! Perf in Unigine DX11 benchmark is spectacular, Radeon HD 5870 is there like low-end toy for kids against GF100.

PS. Yeah, GF100 has 512 CC, and GPUs are in massproduction now, plenty of manufacturers have first GF100 inhouse.

it seems like you invented all the things you write, ces was a free event and nobody talked like this about fermi, if nvidia does not post any benchmark of fermi cards, and if it does not talk about it with certain things we can talk about it all day long......still wainting for the cards from september........let's wait without saing uncertain things.....

[marten_larsson]

Are you basing this on Nvidia's PR slides?

Cause if so, we all know that they're overstated (this is true of ATI too)

Seeing as how you've posted that actual cards won't be sent out til February, I'm going to assume you got your numbers from PR slides - so how about game performance, and not benchmark performance? And what you're saying - that it's better than everyone expected - seems to go counter against what everyone else has been hearing, and that includes hints from some Nvidia folks that the design was incredibly hard to achieve

Sorry if I'm not exactly trusting of a new poster from China who's only here to post on this thread, and from where I've heard more than a few rumors turn out false

Actually, all he has posted has turned out true so far. (Which is 3DFinity or whatever, the deep dive, the NDA date.) And seeing as how he said the power consumtion for GTX380 was also huge, it's probably in the ~250W or else it wouldn't have been all that much. A 250W card should beat a 188W card.

[okorop]

Actually, all he has posted has turned out true so far. (Which is 3DFinity or whatever, the deep dive, the NDA date.) And seeing as how he said the power consumtion for GTX380 was also huge, it's probably in the ~250W or else it wouldn't have been all that much. A 250W card should beat a 188W card.

let's wait and see if it is true or false, I can only say that nvidia is from septembre that is promising but they did not show anything.....

[Dimitriman]

I haven't read the whole thread but does NVidia itself expect Fermi 380 only in September 2010?

Thats basically on par with an ATI refresh for 5 series. Should be interesting if true.

[okorop]

I haven't read the whole thread but does NVidia itself expect Fermi 380 only in September 2010?

Thats basically on par with an ATI refresh for 5 series. Should be interesting if true.

probably yes, gtx380 will be expected only in september 2010

[trinibwoy]

Are you basing this on Nvidia's PR slides?

I believe him when he reports these things but the original source seems to be Nvidia so adjust your expectations to suit.

[eric66]

Deep Dive meeting has over, and we already know how powerful GF100 is, and IT IS! Performance is better then expected, most peoples here will search in own words and feel to shy for craps in thread. We can only laugh now, because Nvidia did it! Perf in Unigine DX11 benchmark is spectacular, Radeon HD 5870 is there like low-end toy for kids against GF100.

PS. Yeah, GF100 has 512 CC, and GPUs are in massproduction now, plenty of manufacturers have first GF100 inhouse.

r600 was great in 3dmark too lol who cares how about real games

[GoldenTiger]

because a quick spike could simply be from a HDD loading issue

i prefer charts (hardocp does that right?)

SSD's ftw? Yes, hardocp does charts. You can still have half the time in the teens half the time in the 80's, resulting in a 50fps average with a miserable experience .

[SamHughe]

Deep Dive meeting has over, and we already know how powerful GF100 is, and IT IS! Performance is better then expected, most peoples here will search in own words and feel to shy for craps in thread. We can only laugh now, because Nvidia did it! Perf in Unigine DX11 benchmark is spectacular, Radeon HD 5870 is there like low-end toy for kids against GF100.

PS. Yeah, GF100 has 512 CC, and GPUs are in massproduction now, plenty of manufacturers have first GF100 inhouse.

Bold words. Hope they are true because if not, you'll be the one feeling shy for spreading BS.
As I told in another thread, sometimes I feel people register to XS forum just to troll in this Fermi thread.