The GT300/Fermi Thread - Part 2!

[Decami]

This stayment may be somewhat close to the truth... in current titles. As newer engines and upcomming games using liberal tesselation come about, I fully epect the 480 to pull ahead of the 5870/GTX295 by a fair margin. However where launch titles are concerened and past games, I don't think it will have much of a lead if at all. All this said, I am treating GF100 as a long(er) term invesment, not simply a day 1 upgrade. Nvidia are playing the same card AMD has in the past ( designing a gpu with more shader power vs texture power ; tradtionally they've gone for a more texture heavy approach ) Hopefully it turns out well for everyone.

I personally expect to see the 480 launch at $549 USD. As far as what vendors will actually charge for them intially... who the hell knows

well said entirely on the money as far as I am concerned.

[temporarychicke]

Forgive my ignorance, please don't RTFF me, but what exactly are these parts going to be:

GeForce GTX 470
GeForce GTX 480

Will they both be single fermi CPU cards with a few clock cycles inbetween them or will the 480 be a dual GPU (unlikely I know)

[Mad1723]

Forgive my ignorance, please don't RTFF me, but what exactly are these parts going to be:

GeForce GTX 470
GeForce GTX 480

Will they both be single fermi CPU cards with a few clock cycles inbetween them or will the 480 be a dual GPU (unlikely I know)

Single GPU with a few frequency/shader count/memory/price differences

[temporarychicke]

Single GPU with a few frequency/shader count/memory/price differences

Thanks kindly for the info.

This 480 could be the one to go for in 6 weeks then, even if it's slightly slower than the HD5970 on games which scale well with crossfire (such as Stalker series).

For most existing DX9/10 titles I predict the 480 will be the best all-rounder card, since the HD5970 is "crossfire on a card" - a technology that is inherently limited in terms of number of titles strongly benefiting.

But I've been wrong before and I am still using two HD4870 512MB cards in crossfire.

[Dimitriman]

This stayment may be somewhat close to the truth... in current titles. As newer engines and upcomming games using liberal tesselation come about, I fully epect the 480 to pull ahead of the 5870/GTX295 by a fair margin. However where launch titles are concerened and past games, I don't think it will have much of a lead if at all. All this said, I am treating GF100 as a long(er) term invesment, not simply a day 1 upgrade. Nvidia are playing the same card AMD has in the past ( designing a gpu with more shader power vs texture power ; tradtionally they've gone for a more texture heavy approach ) Hopefully it turns out well for everyone.

I personally expect to see the 480 launch at $549 USD. As far as what vendors will actually charge for them intially... who the hell knows

If you look at things that way, then perhaps its better to start looking at Fermi as a competitor to 6800 series? Since they will only be about half a generation apart and both completely new architectures.

If Fermi only considerably beats 5800 series in tesselated DX11 games , then a "smart" investor in future technology would definitely wait for the 6800 series to make his decision.

[FischOderAal]

If Fermi only considerably beats 5800 series in tesselated DX11 games , then a "smart" investor in future technology would definitely wait for the 6800 series to make his decision.

Yeah right At first everyone said "wait until Fermi is there before buying HD5k" and now you say "wait until HD6k is there before buying Fermi"? Jesus You should buy your freaking GFX whenever you need it, not when your glass globe tells you to.

[Dimitriman]

Yeah right At first everyone said "wait until Fermi is there before buying HD5k" and now you say "wait until HD6k is there before buying Fermi"? Jesus You should buy your freaking GFX whenever you need it, not when your glass globe tells you to.

That's kinda my point though, most people want a graphics card fast for now, today, so they shouldn't be thinking too much about how gaming will be late 2010 and early 2011. Just buy what's fastest for today's games, be it tesselated or not. If you start pondering too much about how well Fermi will run tesselated games and that you should buy it for such games then you might as well wait for 6800 series as well.

I understand waiting for Fermi cause it might beat 5800 series in every way for todays games, but waiting for it cause you will have a future proof product then you might as well wait longer.

[saaya]

This stayment may be somewhat close to the truth... in current titles. As newer engines and upcomming games using liberal tesselation come about, I fully epect the 480 to pull ahead of the 5870/GTX295 by a fair margin. However where launch titles are concerened and past games, I don't think it will have much of a lead if at all. All this said, I am treating GF100 as a long(er) term invesment, not simply a day 1 upgrade. Nvidia are playing the same card AMD has in the past ( designing a gpu with more shader power vs texture power ; tradtionally they've gone for a more texture heavy approach ) Hopefully it turns out well for everyone.

I personally expect to see the 480 launch at $549 USD. As far as what vendors will actually charge for them intially... who the hell knows

if you have been around in the hw world for a few years you know that it makes no sense whatsoever to buy a videocard as a longterm investment, ESPECIALLY if its the first generation of a new dx standard... by the time that standard is actually used there will be MUCH faster AND cheaper cards out...

so yes, im sure it will pull ahead once tesselation is used widely... but when that will be... who knows... could be the end of 2010, could be 2012... and even if its the end of 2010, by then there should be cards with better price/perf from both camps afaik... so it really doesnt make much sense to buy a fermi betting on it performing better in the future when tesselation is used more widely...

[Metroid]

The maximum a high end graphics card may last is 3 years which must be followed by the operational system which is the most important in this role.

The first Nvidia graphics card to support DX10 was the Geforce 8 series which launched in November 2006, at that time Vista had been released 2 months later January 2007. The Radeon 5800 Series was released in September 2009 and the Windows 7 a month later October 2009, this is the key point for a successful marketing strategy, rely on the future market share of other companies in this example both Nvidia in the end 2006 with Windows Vista and now AMD's evergreen series 5800 with Windows 7. I call it a successful human combo buying habit, not to mention that AMD is doing a better job giving alternative options to its customers. Nvidia with its Geforce 8 series took 6 months to deliver mainstream based products from November 2006(8800GTX) to April 2007(GeForce 8600 GTS). AMD took 1 month releasing the 5700 series, from September 23, 2009 (5800 series) to October 13, 2009 (5700 series). It clearly shows Nvidia's hungry money appetite at that time.

I was expecting Fermi to be launched before the Windows 7 again. It looks like something got wrong somewhere. Nvidia does not have any backup plan, if something delays it delays much more than any other company that has backup plans for occasions like this. It looks like AMD graphics division took the lead and its market share only tends to grow.

I'm certain that Nvidia will launch only high end graphic cards to cash in and then after some months mainstream graphic cards.

I still think AMD is overpriced selling its 5xxx series, the 4xxx is a much wiser buy but then again the Win 7 DX11 syndrome is here and real.

[FischOderAal]

AMD is charging what they can for HD5k due to the lack of competition. Can you blame AMD for that? They need every Dollar they can get (more than many other companies).

[saaya]

AMD is charging what they can for HD5k due to the lack of competition. Can you blame AMD for that? They need every Dollar they can get (more than many other companies).

i think ati could have sold quite some additional 5800 cards if they would have reduced the price 50-100$ which would have resulted in higher net profits... but they were yield limited, so it made perfect sense not to lower the price...
now that fermi is around the corner cutting the price doesnt make much sense cause people will wait for fermi before buying a new card...

so as soon as fermi launches, 5800 prices will probably drop notably... 299$ is my guess...
ooor, if ati is smart, they will replace the 5870 with a 5890 and try to keep the same price... 350-399$

[zalbard]

i think ati could have sold quite some additional 5800 cards if they would have reduced the price 50-100$ which would have resulted in higher net profits... but they were yield limited, so it made perfect sense not to lower the price...

Yeah, exactly, yields were bad, so the cards cost more to make, and the availability was low enough to keep high prices.

[highoctane]

I was expecting Fermi to be launched before the Windows 7 again. It looks like something got wrong somewhere. Nvidia does not have any backup plan, if something delays it delays much more than any other company that has backup plans for occasions like this. It looks like AMD graphics division took the lead and its market share only tends to grow.

Well on the matter of backup plan, I don't think any of the companies playing with the silicon have much of a backup plan. Look at ATI with the r600 and its delays, look at Intel with the P4, look at AMD with Phenom 1.

Sometimes issues result in product delays, less than expect performance or both but sometimes things are very successful from the start.

Then you have Athlon 64, Phenom2, Core2, G80, & RV770 which would be arguably very successful hardware from launch.

[bfar]

Well on the matter of backup plan, I don't think any of the companies playing with the silicon have much of a backup plan. Look at ATI with the r600 and its delays, look at Intel with the P4, look at AMD with Phenom 1.

Sometimes issues result in product delays, less than expect performance or both but sometimes things are very successful from the start.

Then you have Athlon 64, Phenom2, Core2, G80, & RV770 which would be arguably very successful hardware from launch.

+1

One really bad one comes to mind - Nvidias FX series effectively allowed ATI to swallow up a big chunk of market share with their 9700 series. Up until then Nvidia were looking like they had it wrapped up. History doesn't always repeat itself - there's no way to know how this gen will pan out until we see the products.

I personally think this is a bad time for graphics card makers - there just aren't enough games out there that need the power. I used to get very excited by new gens of graphics cards, but these days, what I have in my rig is already more than enough.

[saaya]

there usually are backup plans... we just never notice it cause the decision to go for either one is made way before release...
for example, xenos was originally planned as a desktop part but was canned because of performance shortcommings, and instead they just doubled up on their previous desktop gpu...

and look at nvidia, gt200 is nothing but an (almost) double pumped G92... im sure they didnt plan that originally, but many times doubling up a current design gives you roughly the same performance as a new design, and its known to work so its less risky. and nvidia actually DID have a backup plan afaik, 40nm G92 and G200, its just that those didnt work out... it was a poor backup plan cause shrinking from 55 to 40 is very difficult from what i heard.

amd K9 never made it either... intel tejas was canned... intels celeron SOC in the late nineties was canned... and once again in the early 2000s they tried it again and cancelled it... intel using imagination technologies gpus in their chipset is OBVIOUSLY a backup plan because their own gpu solutions didnt work out as planned to cover some segments...

[SamHughe]

Isn't this thing hitting the markets in March? Why we don't see any decent benchmarks and reviews?

[highoctane]

there usually are backup plans...

I think the better term with the hardware industry would be damage control!

[Metroid]

...

Let's take my example for this matter when I said Nvidia does not have backup plans. If you were the Nvidia's CEO what would you do? just remember that in that position there are many secrets that only higher ups know, it is like playing cards hand by hand, you don't know everything but you know many things. You know quite well that Windows 7 is coming and will use DX11 (which I think is pretty basic since most of us knew that) and also that the competition will use that as an advantage to take the lead. Will you let it go or will you do something about it? This situation occurred months before the launch of the Windows 7. You as a CEO had a choice right? it is not like you did not have one.

I as a CEO, If fermi was not ready for that month then I would play their game and do the same, release something with DX11 support and or something extra even if that is not as powerful as the 5800 series this new model would compete with the mainstream 5700 series or would give customers the DX11 feeling that they wanted and personally I think that this would not have weaken the brand as it is.

What about a G92 with DX11 only to compete? That is my point, don't let the competition go on a rampage.

[saaya]

What about a G92 with DX11 only to compete? That is my point, don't let the competition go on a rampage.

thats kinda what the 40nm 10.1 parts are... just that they are 10.1 only and not 11

and yes, nvidia played poker... they bet that tsmc can mfc a big gpu on 40nm in late 2009 and they were wrong...

[zerazax]

thats kinda what the 40nm 10.1 parts are... just that they are 10.1 only and not 11

and yes, nvidia played poker... they bet that tsmc can mfc a big gpu on 40nm in late 2009 and they were wrong...

Exactly, although ATI had the right idea by first testing out 40nm on the RV740 and going for a smaller design, which made their transition much smoother (relatively)

[highoctane]

What about a G92 with DX11 only to compete? That is my point, don't let the competition go on a rampage.

I gotcha, it really does seem like they fell asleep at the wheel in regards to getting any dx11 hardware out the door. Both ATI and Nvidia seem to release the latest on the highend part first before releasing the lower parts so with gt300 delay's you would have thought some midrange to lower end parts would have been scheduled to release by now.

[tajoh111]

If you look at things that way, then perhaps its better to start looking at Fermi as a competitor to 6800 series? Since they will only be about half a generation apart and both completely new architectures.

If Fermi only considerably beats 5800 series in tesselated DX11 games , then a "smart" investor in future technology would definitely wait for the 6800 series to make his decision.

Why would that be the case? It seems AMD is dead set on using 28nm on its next graphic tech, which means at the minimum its going to be a 2011 part. That's a substantially longer wait. The sketchyness of 28nm just makes it actually timeline even less reliable.

Even if the gtx 480 is as fast as its hyped up to be, then AMD can still make money selling it for a price lower than it is at the moment. E.g 5870@ 299 5850@229-249 and the 5970 at $549 or $499. Both All of AMD parts right now are small chips and thus can be sold at much lower costs and still make a profit.

AMD is in no rush, the 299 and under market is quite safe for them and its actually quite a profitable area.

But in terms of upgrading for the sake of playing directx 11 games, waiting is actually not a bad idea simply because the lack of games at the moment. Dirt 2 really isn't that good and alien vs predator seems sketchy.

[xVeinx]

The impression I've gotten is that ATI is going for a refresh of 40nm parts (ie, same architecture) in the second half of this year, while intending to release the new architecture on 28nm (though likely to release some low end parts first on 28nm to test out the process) in the first half of next year.

To be on topic however, Fermi won't be up against anything but the current architecture when released. An ATI refresh might yield some improvements, but it's anyone's guess where things will end up, though I imagine they'll be working on some tessellation improvements to shore up things between the 5800 series and Fermi.

[LinkinParkBoy]

Alex Sakhartchouk, Demo Engineer at NVIDIA, explains the technologies behind NVIDIAs new Rocket Sled demo running on the new Fermi architecture

http://www.youtube.com/watch?v=HjIzo...layer_embedded

[LordEC911]

It looks like something got wrong somewhere. Nvidia does not have any backup plan, if something delays it delays much more than any other company that has backup plans for occasions like this.

Jensen doesn't believe in backup plans. (Paraphrasing a quote)

Yeah, exactly, yields were bad, so the cards cost more to make, and the availability was low enough to keep high prices.

I wouldn't say yields are bad, TSMC didn't have enough capacity at the time. Simple economics, low supply and high demand = high prices.

Isn't this thing hitting the markets in March? Why we don't see any decent benchmarks and reviews?

AIBs are just getting their cards... Give it another week or two.

I as a CEO, If fermi was not ready for that month then I would play their game and do the same, release something with DX11 support and or something extra even if that is not as powerful as the 5800 series this new model would compete with the mainstream 5700 series or would give customers the DX11 feeling that they wanted and personally I think that this would not have weaken the brand as it is.

Not that simple. These designs are being worked on for 3-4 years before they are released. This isn't a simple, hey let's throw DX11 into the architecture and get it out next month since GF100 is having problems. There isn't much you can do as far as a backup plan in this industry the R&D costs are just way too high as it is.

I gotcha, it really does seem like they fell asleep at the wheel in regards to getting any dx11 hardware out the door. Both ATI and Nvidia seem to release the latest on the highend part first before releasing the lower parts so with gt300 delay's you would have thought some midrange to lower end parts would have been scheduled to release by now.

Why would that be the case? It seems AMD is dead set on using 28nm on its next graphic tech, which means at the minimum its going to be a 2011 part. That's a substantially longer wait. The sketchyness of 28nm just makes it actually timeline even less reliable.

Sketchy? I think not, at least for GF. TSMC, on the other hand, god help them.
http://www.globalfoundries.com/pdf/G...ldBrochure.pdf
Please note proof#5 on page 2.