The official GT300/Fermi Thread

[***Deimos***]

ClickMe

Fake?

I'm going to let you all in on an INSIDER SECRET.

NOBODY knows what final retail Fermi clocks/SP etc config will be. NOBODY. Not even the CEO.

You can huff and puff and really want that x Mhz goal but the reality steps in. MOST of the time, idealistic targets are not met and plans change (ie 4Ghz P4). And same is obviously true of AMD and Intel. Since clock speeds are so dependent on yield, they are often finalized just weeks before launch, based on most recent wafer quality.

Back in the spring Intel's 32nm Clarkdale was already benchmarked and overclocked on the forums. But, by the time it ships, it will be a different stepping, and different maturity of lithog-process. Even cache sizes, #SP, onboard video mem, etc can change quickly in few months before launch based on competitor's products. ie HD4850 was originally planned to be lower clocked and 256MB.

So... guess and predict all you want. But "fake?" response is always a rhetorical question - NOBODY knows how things will change.

[Chumbucket843]

I may have sold my GTX 260 Core 216 for a 4890 ( I like 8xMSAA and it kills Nvidia cards ) But I know one thing.... When Fermi finally does see the light of day these forums are going to erupt!!

I think it will be slightly faster then a 5870 and if the rumors are true, it'll trade blows with the 5890

After all the waiting and talk from Nvidia it needs to be faster although things never do live up to the hype

you know you can enable 8xCSAA on nvidia cards and its much more efficient compared to 8xMSAA? its safe to say gf100 will beat the 5870. nvidia's goal is to stay #1 even if that means lower margins or more power consumption. its simply a monster of a chip and thats why it will be faster.

[Onoff312]

Gt300 is supposedly only just over the limit for 2 6pins so it is not that power hungry at all considering the specs and leaves alot of overclocking headroom.

[YukonTrooper]

although things never do live up to the hype

I remember G80 surpassing the hype and blowing everybody through the roof. Those were glory days!

[d50man]

I love home people are dreaming of owning the GT300 hardware thats going to require a personal loan and a dedicated powersupply...

[Boulard83]

Fermi isnt gooing to hit the market before Q2 2010

[LedHed]

I love home people are dreaming of owning the GT300 hardware thats going to require a personal loan and a dedicated powersupply...

I don't see them charging anymore than the 5970, so pricing isn't really the issue here. NVIDIA knows what their products are worth in the market (remember the 295 launched at $500 the lowest price ever for the highest end part available, look back at the 8800 Ultra price) and they will adjust according to ATI.

[***Deimos***]

I don't see them charging anymore than the 5970, so pricing isn't really the issue here. NVIDIA knows what their products are worth in the market (remember the 295 launched at $500 the lowest price ever for the highest end part available, look back at the 8800 Ultra price) and they will adjust according to ATI.

Just to make it perfectly clear for everybody that's been disillusioned by the HD48xx series very low MSRP -

Each and every nVidia (and often ATI/AMD) high end product has launched at $500-$650 and even as much as $800.
AMD HD58xx is less transistors, smaller chip (better yields), simpler, only 256bit PCB, and less memory - all indications that Fermi will cost more to make. But what about price?

Regardless if Fermi is faster (or slower), since nVidia can't get game revenues like Sony to sell at a loss, IT WILL DEFINITELY BE EXPENSIVE.

[JohnZS]

The availability is going to be in early Q2 2010, this was in my nVidia news letter today
The most powerful Fermi based GPU processor is not going to reach us until Q3 2010.

[Boulard83]

The availability is going to be in early Q2 2010, this was in my nVidia news letter today
The most powerful Fermi based GPU processor is not going to reach us until Q3 2010.

Yes... ATI can be happy !

[JohnZS]

Only if they release a 2GB card otherwise ATi will not get a penny of my money. We need more than 1GB of VRAM!

[N19h7m4r3]

Only if they release a 2GB card otherwise ATi will not get a penny of my money. We need more than 1GB of VRAM!

Exactly! We need more memory, specially sicne ATI feature eyefinity now.

That and ATI and NV need to make cards shorter!

I'm tired of things over 9 inches, and that 5970 is longer than some chassis.

[JohnJohn]

The availability is going to be in early Q2 2010, this was in my nVidia news letter today
The most powerful Fermi based GPU processor is not going to reach us until Q3 2010.

Huh! Care to post the whole mail ?

Edit: So fermi single gpu in april, and fermi gx2 in Q3/10 ?

[BenchZowner]

I heard it won't be launched before Q6 2010

[[XC] riptide]

I heard it won't be launched before Q6 2010

lol.

[Your_Boss]

Question! If the GF100 is having problems ramping up their clocks..then what possiblity, if any, do you think they will have of any overclocking?

IF they manage a blistering 750MHz core and do pass the 5870 by this supposed 30%~40%, then what if that's all they can go? Maybe to 775 at maximum. Couldn't the 5870 be OC'd ALOT higher? I've heard of some doing over 1050MHz on air. ATI could even sell some speed binned chips that start at 950 or 1000MHz. Wouldn't this cut Nvidia GF100's lead down signifigantly? Maybe within 20% or so?

And by that time ATI's drivers have improved, like they have already, and cut the difference between them even more, maybe down to within 10%. Maybe ATI won't even need that 5890, though they probably will with a few months. 5890's may only be OC's binned chips anyway.

And IF NV doesn't quite reach their intended clocks of 750MHz and only manage 650MHz or so to start off then NV may be in bigger trouble than people think.

To be honest, if NV had these working chips at their target speeds they most likely would be out already barring some other technical problems. I don't think NV even knows the exact speed these things will be running at yet. If they knew at all..or were confident of even being close to a speed they would be flooding the internet with performance numbers(if they were 40% or more higher than the 5870). They would be doing everything to keep AMD from selling their cards and taking away their marketshare. This would keep those people that didn't want to wait..it may make them wait for Fermi with numbers like those.

Now if NV performance is about 20%~30% above the 5870 then it probably is best to not really say anything until they're released. Otherwise people will just go 'Meh'. Because we all know they won't be cheap when first released. So that would cause many to STOP waiting for Fermi and buy the competition right now. 20%~30% is not a showstopper so who will wait for that except for maybe the die hard NV fans.

Now, if the performance is 20% or LOWER than the 5870 then there is no doubt that they should not reveal any performance numbers until release...as this would be absolutely devistating and cause really noone to wait for Fermi. NV is not stupid, they KNOW ATI has something waiting to be release at Fermi launch..whether it be a 5890 or new performance drivers waiting in the wings. NV knows they need to combat ATI's show stopper as well.

All I do know is the longer the wait for the GF100 the worse for NV it gets. They need to get something out very soon. I hope for good competition and these are my opinions. Maybe others have the same views..maybe different, who knows. :-)

[zalbard]

Source: http://www.fudzilla.com/

Over the past couple of days several benchmarks pitting two Nvidia Fermi cards against ATI Evergreens have surfaced on the interweb. The results indicated a single Fermi-based Geforce GTX 380 should beat ATI's dual-GPU HD 5970, however, the slides don't appear to be kosher.

Nordichardware got in touch with Nvidia and it turns out the slides were the result of some rather simple photoshopping. Someone basically inserted a bunch of new product names on an old Nvidia presentation template. We've witnessed plenty of these alleged performance charts laying around, but we just don't bother posting them as they are about as reliable as looking at a heap of fresh chicken entrails through a crystal ball whilst eating mushrooms.

Be as it may, we're still expecting Nvidia's Fermi-based cards to have an edge over ATI's Evergreen series. At this point, however, we just don't know how fast they will actually be and whether the performance margin will be worth the wait. Bear in mind Fermi probably won't ship until late Q1, but ATI's is also facing problems of its own, as it is still unable to meet demand for HD 5800/5900 cards.

Basically, both Nvidia and ATI played Grinch and stole Christmas 2009 from FPS hungry gamers. No wonder gamers are turning to consoles in droves.

More here.

Not surprising, but I thought I'd mention it.

[Weevil]

Question! If the GF100 is having problems ramping up their clocks..then what possiblity, if any, do you think they will have of any overclocking?

IF they manage a blistering 750MHz core and do pass the 5870 by this supposed 30%~40%, then what if that's all they can go? Maybe to 775 at maximum. Couldn't the 5870 be OC'd ALOT higher? I've heard of some doing over 1050MHz on air. ATI could even sell some speed binned chips that start at 950 or 1000MHz. Wouldn't this cut Nvidia GF100's lead down signifigantly? Maybe within 20% or so?

And by that time ATI's drivers have improved, like they have already, and cut the difference between them even more, maybe down to within 10%. Maybe ATI won't even need that 5890, though they probably will with a few months. 5890's may only be OC's binned chips anyway.

And IF NV doesn't quite reach their intended clocks of 750MHz and only manage 650MHz or so to start off then NV may be in bigger trouble than people think.

To be honest, if NV had these working chips at their target speeds they most likely would be out already barring some other technical problems. I don't think NV even knows the exact speed these things will be running at yet. If they knew at all..or were confident of even being close to a speed they would be flooding the internet with performance numbers(if they were 40% or more higher than the 5870). They would be doing everything to keep AMD from selling their cards and taking away their marketshare. This would keep those people that didn't want to wait..it may make them wait for Fermi with numbers like those.

Now if NV performance is about 20%~30% above the 5870 then it probably is best to not really say anything until they're released. Otherwise people will just go 'Meh'. Because we all know they won't be cheap when first released. So that would cause many to STOP waiting for Fermi and buy the competition right now. 20%~30% is not a showstopper so who will wait for that except for maybe the die hard NV fans.

Now, if the performance is 20% or LOWER than the 5870 then there is no doubt that they should not reveal any performance numbers until release...as this would be absolutely devistating and cause really noone to wait for Fermi. NV is not stupid, they KNOW ATI has something waiting to be release at Fermi launch..whether it be a 5890 or new performance drivers waiting in the wings. NV knows they need to combat ATI's show stopper as well.

All I do know is the longer the wait for the GF100 the worse for NV it gets. They need to get something out very soon. I hope for good competition and these are my opinions. Maybe others have the same views..maybe different, who knows. :-)

You said "IF" 9 times, that's a lot of if's, pure conjecture. Why don't we just wait till the chip arrives at a reputable review site for our opinions.


BTW, my wife and I like your avatar.

[***Deimos***]

Schematic diagrams all info on Fermi is out!!
http://wiki.opengraphics.org/tiki-index.php

To those proposing that if Fermi could reach 750Mhz it would be out already.

You are right. A 3B transistor chip, complex PCB, and millions of lines long driver are simple and easy and SHOULD be here now.

Afterall, if Toyota can design,test and mass produce the 2009 (extra cup holder) Corolla in a week, and Boeing can roll-out fully tested and certified 787 in a couple days, then nVidia is obviously lazy - right!?

[Your_Boss]

Schematic diagrams all info on Fermi is out!!
http://wiki.opengraphics.org/tiki-index.php

To those proposing that if Fermi could reach 750Mhz it would be out already.

You are right. A 3B transistor chip, complex PCB, and millions of lines long driver are simple and easy and SHOULD be here now.

Afterall, if Toyota can design,test and mass produce the 2009 (extra cup holder) Corolla in a week, and Boeing can roll-out fully tested and certified 787 in a couple days, then nVidia is obviously lazy - right!?

What my original question really was is..

IF NV manages to get to a 750MHz core, how much further do you think it can overclock? As in, what are the ranges of clock speeds for chips normally of this complexity?

Such as..if the fastest chip they get from the fab can reach say 775MHz the slowest one will still be able to reach what? (from the same spin of course) Is there a normal percentage range?

Originally Posted by Weevil
BTW, my wife and I like your avatar.

Thanks, photoshopped Clinton face inside the outline of Obama.

[BenchZowner]

What my original question really was is..

IF NV manages to get to a 750MHz core, how much further do you think it can overclock? As in, what are the ranges of clock speeds for chips normally of this complexity?

Such as..if the fastest chip they get from the fab can reach say 775MHz the slowest one will still be able to reach what? (from the same spin of course) Is there a normal percentage range?

Let's add another IF and say "IF you raise the voltage, will it overclock more ?"

You guys need to get a drink and go watch a good movie with your girl/wife/family.

[Your_Boss]

Let's add another IF and say "IF you raise the voltage, will it overclock more ?"

You guys need to get a drink and go watch a good movie with your girl/wife/family.

Not raising the voltage..SAME VOLTAGE! What's the range?

And I was watching a good movie to come back to a condesending post.


I try to rephrase my question..

What percentage of chips would have to hit their 'target speed' before they say, "Yes, we've hit our target speed"? 10%, 20%, 50%? And AFTER they've reached this target speed...what's the RANGE of the lowest possible speed from this respin of that aren't defective? (let's say top tier)

I'm just looking for a possible range for a chip of this complexity. Or is there NO way of knowing?

Let's say they settled on a voltage for the top speed chips and only 1 out of 10,000 will hit 800MHz at this 'stock voltage'...obviously this would be the absolute MAX speed for this chip at this voltage and they of course would never advertise it at this speed. My question is... with a perfect non-defect chip what would the lowest speed be for a perfect chip? 600MHz? less? more? Would ALL perfect chip be able to make it over at least a certain speed? There is a 'range' just wondering what an estimated range would be. Thanks.

[BenchZowner]

1PHz.

The card's not out yet.
Nobody really knows where it'll be clocked, and nothing about its performance per clock, and you're looking for its overclocking range on default voltage ?
Talk about addiction!

By the way, have you ever seen a borderline clocked high end card ?

Can't you guys just freaking wait ?

Oh no, it's so amazing, so amusing, so interesting to talk about something you know nothing about

Let's see... 1) Good movie with friends/girl/family, 2) Go to a club/restaurant/bar/beach, 3) Have sex ( feel free to pick your partner on your own, I'm not judging you ), 4) Speculate, overreact, fight over something we know nothing about and isn't a life-changing factor...

Hmm... I'll pick the movie now

[Your_Boss]

1PHz.

The card's not out yet.
Nobody really knows where it'll be clocked, and nothing about its performance per clock, and you're looking for its overclocking range on default voltage ?
Talk about addiction!

By the way, have you ever seen a borderline clocked high end card ?

Can't you guys just freaking wait ?

Oh no, it's so amazing, so amusing, so interesting to talk about something you know nothing about

Let's see... 1) Good movie with friends/girl/family, 2) Go to a club/restaurant/bar/beach, 3) Have sex ( feel free to pick your partner on your own, I'm not judging you ), 4) Speculate, overreact, fight over something we know nothing about and isn't a life-changing factor...

Hmm... I'll pick the movie now

What do you think this forum is here for? Hmmm, let's see, why don't we grab a sex partner(pick your pref) and they why don't we just grab a nice warm blanket..maybe some snuggies.(see infomercials) Grab some popcorn..put a nice movie on the TV ..then.....prop ourselves up in a couple of chairs in front of the computer and read the Xtremesystem forums.

Come on..I'm just looking for some speculation.

Here's what happens when GOOD THREADS GO BAD!

Let's not make this thread go bad.

[***Deimos***]

MMmmm...

that picture has got me all hot and bothered...

I got my hands in my pants, and its giving me a Fermi.