More Fermi bad news

[Freddie1980]

Maybe, but it seems ati have squeezed out Nvidia's profit, so Nvidia has cut back on RnD to save money which creates a viscous cycle that's hard to get out of.

We don't have actual numbers but we do know roughly how much a silicon wafer costs and how many chips ATI & nvidia can get from a single wafer and the actual die size and it's been mentioned numerous times before that ATI have with there smaller more efficient dies are more profitable then the larger Nvidia dies.

As for cash that's a different matter, Nvidia is loaded with surplus cash they have no loans or outstanding long term debt and are financially very stable so even in the worst case scenario that they don't have a first generation DX11 product it won't hurt them to badly as they have cash to see them through bad times. The same can't be said for AMD who are up to it's eyeballs in debt and has to fork out money to service it which could be going on R&D.

 

[Owenb]

And the big difference here is at least the 2900XT was apart of strategy shift to target the sweet spot between performance and value for money, I can't see see how Fermi is going to be anything close to good value for money given what it costs make.

The 2900XT wasn't aimed at the 'sweet spot', it's just that ATi had no other choice but you aim their top end chip at the mainstream market as it was no match for the 8800GTX. IIRC the 2900XT had even more transistor's than the G80.

And ATI are trashing Nvidia

Yet Nvidia are outselling ATi 2:1 and are making a lot more money.

 

[stulid]

Just read at xtremesystems that nvidia have posted a profit for their 4th quarter, compared with a loss for the same time last year.

As for that charlie D, whenever I see him post in the fermi discussion at beyond3D, he normally gets proved wrong by the experts there.

 

[Yamahahahahaha]

Truth and trust personified.

 

[forest4]

My! What a big forehead!

 

[elmuzzy66.]

And the big difference here is at least the 2900XT was apart of strategy shift to target the sweet spot between performance and value for money, I can't see see how Fermi is going to be anything close to good value for money given what it costs make.

I don't think they were aiming for the sweet spot, but made go down that route by how poor the 2900 was. In the end it has worked out good for ati.

 

[mmj_uk]

Didn't he post this very same article a few months ago? NVidia responded basically calling him an ATI fanboy spreading misinformation.

 

[DavyBoy]

I really don't know what to make of nVidia anymore, I'm starting to believe a lot of the bad stuff that's been said about Fermi simply because there has been next to nothing said about these cards from nVidia themselves, just dribs and drabs really.

Surely if they had a kick ass product soon to be released and it was pwning anything ATi had to offer, don't you think they'd be bragging about it by posting benches to try and stop people from buying ATi's current lineup?

 

[Rroff]

While everyone paints a picture of doom and gloom things aren't actually so bad for nVidia in reality.

The GF100 design (theoretically) should have no problem going up against the next gen ATI and comes into its own on sub 40nm processes so aslong as their foundry partner has a good quality process they are good to go. 40nm was always going to be a problem even if TSMC hadn't screwed up majorly.

Aslong as they can accomplish this before DX11 takes off proper they will have no problem retaking their market share either.

Short term not good for consumers but for nVidia its only a blip - they have the finances to weather the intermediate period.

 

[forest4]

Answer: Yes. (to Davyboy)

 

[drunkenmaster]

The 2900XT wasn't aimed at the 'sweet spot', it's just that ATi had no other choice but you aim their top end chip at the mainstream market as it was no match for the 8800GTX. IIRC the 2900XT had even more transistor's than the G80.



Yet Nvidia are outselling ATi 2:1 and are making a lot more money.

the 2900XT was built on the wrong process, as in because TSMC were so late they shifted it UP a process, not down, and it beat the 8800gtx in a few games, Bioshock being one. IT was close in several others and beating by the same margin the 280 gtx would beat the 4870 in several games, yet people don't talk about teh 280 spanking the 4870, because by and large it didn't.

Can you imagine lets say, a 280gtx on 80nm beating the 4870, or Fermi on 55nm beating the 5850, or the 5850 on 55nm beating the Fermi? No, thats the situation the 2900XT was in. Had it been on the right process, about 45% smaller due to the process, using about 60% of the power, at 2/3rd's the cost with 30% higher clocks, with REAL DX10 functionality(and really ful dx10.1) had DX10 been kept as is, frankly it would have utterly wiped the floor with the 8800gtx. TSMC basically lied, the 65nm process wasn't anywhere near to being ready, it was over a year late in the end, if it wasn't AMD had the better architecture design and functionality, all ruined by a process.


As for outselling AMD 2:1, they aren't, as for making a lot more money, again, they aren't. Every single segment AMD makes a higher profit in, they are closing in on overall market share, the only thing lacking right now is the long term dominance and cash for R&D to penetrate the uber profit markets of things like medical equipment and the "Quadro" market. However, as Nvidia go backwards, AMD have all the funding in the world, will likely push heavily into those markets in the next couple years which will screw Nvidia pretty badly.

 

[Rroff]

I really don't know what to make of nVidia anymore, I'm starting to believe a lot of the bad stuff that's been said about Fermi simply because there has been next to nothing said about these cards from nVidia themselves, just dribs and drabs really.

Surely if they had a kick ass product soon to be released and it was pwning anything ATi had to offer, don't you think they'd be bragging about it by posting benches to try and stop people from buying ATi's current lineup?

I've heard absolutely nothing lately... absolute dead quiet and thats not usual for nVidia... by this time before launch previously there was lots of snippets of information flying around from partners, etc.

 

[drunkenmaster]

While everyone paints a picture of doom and gloom things aren't actually so bad for nVidia in reality.

The GF100 design (theoretically) should have no problem going up against the next gen ATI and comes into its own on sub 40nm processes so aslong as their foundry partner has a good quality process they are good to go. 40nm was always going to be a problem even if TSMC hadn't screwed up majorly.

Aslong as they can accomplish this before DX11 takes off proper they will have no problem retaking their market share either.

Short term not good for consumers but for nVidia its only a blip - they have the finances to weather the intermediate period.

Utter tripe, again, you're ignoring the blantantly obvious, that performance has had NOTHING even remotely to do with Nvidia's market share loss, meaning having a higher performance card, AGAIN, won't win them back market share.

The lost market share THROUGHOUT 2009, not just this last quarter, while they had the fastest dual gpu card, and the fastest single GPU card. THe problem was due to their fundamental big core design they couldn't compete on cost, or price/performance in any segment. THAT is why they lost market share, nothing to do with AMD's better performance, because they never had a better card. Well ok, a VERY low sales part in the 4870x2 had a significant lead for a decent amount of time before the 295gtx came out, It didn't make the slightest difference in market share.

Even if the GF100 utterly spanks the 5870, even if its 100% faster, unless it is better in price/performance ratio and has a derivative in EVERY market segment thats also competitive on price/performance then they won't gain any market share.

The GF100 is a fundamentally awful design in terms of price/performance, a drop to 28nm will do it no favours at all as AMD will have a BIGGER drop in size from 28nm, with a bigger increase in performance and a larger power drop. Because GF's 28nm is simply a better more complex process that wasn't skimped on like TSMC who have again gone gate first with trying to kill leakage, which is EXACTLY the reason 40nm sucks.

GF100's future high end versions for the next gen will have a larger problem in price performance compared to AMD< not better. Its also very likely that GF are bringing their 28nm months before TSMC, with better yields on a far better process.

As for talking about their finances being able to weather the blip, lol 40nm is Nvidia's smallest problem for the next two years. AMD have been cursed with the same 40nm process. Next gen, as said AMD are moving to GF with a hugely better quality 28nm, it only gets worse. In the next 18-24 months Nvidia is also losing the massive majority of its low end market.

Now keep in mind Nvidia's increase in profits/revenue in this quarter, the massive majority of those sales were, yes, GT210's, or gt310's, or whatever the hell they are calling them. In 2 years, the same segment of card won't be sold by Nvidia. Thats by far and away Nvidia's biggest problem and it most certainly is not short term, it hasn't happened yet.

40nm is a short term blip, 28nm will be worse, losing the low end will be almost a complete disaster. T

The last 6 months will be Nvidia's easiest 6 months in the next 3 years.

The increase in revenue shouldn't really be looked at as a positive thing for Nvidia, Intel, AMD and Nvidia were down massively last year during the market crash. AMD and Intel have increased revenue's up till now by OVER 100% as the market recovered, Nvidia are up, but the last numbers put the revenue closer to 50% up on last year vs this year. With this quarter vs the last quarter showing the biggest improvement because last years 4th quarter was a disaster the likes we haven't seen for decades.

 

[Rroff]

Utter tripe, again, you're ignoring the blantantly obvious, that performance has had NOTHING even remotely to do with Nvidia's market share loss, meaning having a higher performance card, AGAIN, won't win them back market share.

The lost market share THROUGHOUT 2009, not just this last quarter, while they had the fastest dual gpu card, and the fastest single GPU card. THe problem was due to their fundamental big core design they couldn't compete on cost, or price/performance in any segment. THAT is why they lost market share, nothing to do with AMD's better performance, because they never had a better card. Well ok, a VERY low sales part in the 4870x2 had a significant lead for a decent amount of time before the 295gtx came out, It didn't make the slightest difference in market share.

Even if the GF100 utterly spanks the 5870, even if its 100% faster, unless it is better in price/performance ratio and has a derivative in EVERY market segment thats also competitive on price/performance then they won't gain any market share.

The GF100 is a fundamentally awful design in terms of price/performance, a drop to 28nm will do it no favours at all as AMD will have a BIGGER drop in size from 28nm, with a bigger increase in performance and a larger power drop. Because GF's 28nm is simply a better more complex process that wasn't skimped on like TSMC who have again gone gate first with trying to kill leakage, which is EXACTLY the reason 40nm sucks.

GF100's future high end versions for the next gen will have a larger problem in price performance compared to AMD< not better. Its also very likely that GF are bringing their 28nm months before TSMC, with better yields on a far better process.

As for talking about their finances being able to weather the blip, lol 40nm is Nvidia's smallest problem for the next two years. AMD have been cursed with the same 40nm process. Next gen, as said AMD are moving to GF with a hugely better quality 28nm, it only gets worse. In the next 18-24 months Nvidia is also losing the massive majority of its low end market.

Now keep in mind Nvidia's increase in profits/revenue in this quarter, the massive majority of those sales were, yes, GT210's, or gt310's, or whatever the hell they are calling them. In 2 years, the same segment of card won't be sold by Nvidia. Thats by far and away Nvidia's biggest problem and it most certainly is not short term, it hasn't happened yet.

40nm is a short term blip, 28nm will be worse, losing the low end will be almost a complete disaster. The last 6 months will be Nvidia's easiest 6 months in the next 3 years.

I thought it was obvious we are talking in context of the high end performance parts market not the overall market share.

You also seem to be forgetting that AMD are running into the end of the line for how far they can scale their current design before they hit the threshold on diminishing returns performance wise... their next generation will be far more like nVidias design than anything they have done to date...

 

[DrBombcrater]

Yet Nvidia are outselling ATi 2:1 and are making a lot more money.

It's very possible for a company to be profitable and still be in serious trouble, computer industry history is littered with examples. Making money now is nice, taking the decisions that mean they can make money 2, 3 or 5 years from now is critical. And that's where NVidia is in trouble.

One by one their profit centres are drying up or being significantly reduced. Their chipset division is dead, low-end graphics cards (where they get the bulk of their revenue) is a market that will go into terminal decline with the rise of on-CPU graphics, the high-end isn't making any money because NV's enormous GPUs mean ATI always has a cost advantage and can just cut prices until NVidia's margins go underwater - this will get much worse next year if Bulldozer is as good as predicted and ATI can depend on the CPU division to make cash while they engage NVidia in a bloodbath price-war.

The only area where NV is still healthy is professional graphics cards, but that market can't stand alone. It's only profitable so long as gamers buying high-end GeForces are paying for the R&D on the GPU. 3DLabs were to last company to try and live on the professional market alone and they were crushed by the sheer amount of cash NVidia could throw at R&D thanks to the gaming market. ATI would be only too happy to repeat that trick on the green team.

NVidia isn't doomed even if Fermi is the disaster most of us suspect. They have the cash to survive a barren spell while they go through the kind of painful refocusing that ATI did when the R600 crashed and burned, that AMD did after the Phenom I. But there's not much sign management actually understands that business as it's always been isn't going to work anymore, something made all the more dangerous by AMD/ATI having become very adept at long-term planning.

 

[Rroff]

It's very possible for a company to be profitable and still be in serious trouble, computer industry history is littered with examples. Making money now is nice, taking the decisions that mean they can make money 2, 3 or 5 years from now is critical. And that's where NVidia is in trouble.

One by one their profit centres are drying up or being significantly reduced. Their chipset division is dead, low-end graphics cards (where they get the bulk of their revenue) is a market that will go into terminal decline with the rise of on-CPU graphics, the high-end isn't making any money because NV's enormous GPUs mean ATI always has a cost advantage and can just cut prices until NVidia's margins go underwater - this will get much worse next year if Bulldozer is as good as predicted and ATI can depend on the CPU division to make cash while they engage NVidia in a bloodbath price-war.

The only area where NV is still healthy is professional graphics cards, but that market can't stand alone. It's only profitable so long as gamers buying high-end GeForces are paying for the R&D on the GPU. 3DLabs were to last company to try and live on the professional market alone and they were crushed by the sheer amount of cash NVidia could throw at R&D thanks to the gaming market. ATI would be only too happy to repeat that trick on the green team.

NVidia isn't doomed even if Fermi is the disaster most of us suspect. They have the cash to survive a barren spell while they go through the kind of painful refocusing that ATI did when the R600 crashed and burned, that AMD did after the Phenom I. But there's not much sign management actually understands that business as it's always been isn't going to work anymore, something made all the more dangerous by AMD/ATI having become very adept at long-term planning.

Unfortunatly professional and gpgpu market aren't going to net nVidia much cash... at its best it only looks like bringing in 10-15% of their current income (in the foreseeable future).

I don't think ATI could repeat the trick on nVidia tho - nVidia already have a strong reputation within these fields and generally are good with documentation and support which isn't something AMD/ATI have exactly excelled at in the past in these markets.

 

[Owenb]

the 2900XT was built on the wrong process, as in because TSMC were so late they shifted it UP a process, not down, and it beat the 8800gtx in a few games, Bioshock being one. IT was close in several others and beating by the same margin the 280 gtx would beat the 4870 in several games, yet people don't talk about teh 280 spanking the 4870, because by and large it didn't.

Can you imagine lets say, a 280gtx on 80nm beating the 4870, or Fermi on 55nm beating the 5850, or the 5850 on 55nm beating the Fermi? No, thats the situation the 2900XT was in. Had it been on the right process, about 45% smaller due to the process, using about 60% of the power, at 2/3rd's the cost with 30% higher clocks, with REAL DX10 functionality(and really ful dx10.1) had DX10 been kept as is, frankly it would have utterly wiped the floor with the 8800gtx. TSMC basically lied, the 65nm process wasn't anywhere near to being ready, it was over a year late in the end, if it wasn't AMD had the better architecture design and functionality, all ruined by a process.


As for outselling AMD 2:1, they aren't, as for making a lot more money, again, they aren't. Every single segment AMD makes a higher profit in, they are closing in on overall market share, the only thing lacking right now is the long term dominance and cash for R&D to penetrate the uber profit markets of things like medical equipment and the "Quadro" market. However, as Nvidia go backwards, AMD have all the funding in the world, will likely push heavily into those markets in the next couple years which will screw Nvidia pretty badly.

You can't compare the 2900XT vs 8800GTX to 280GTX vs 4870. The 2900XT/8800GTX both had ~700m transistor's, while the 280GTX had ~50% MORE than the 4870(956m vs 1.4bn). Bioshock- I think your info is well off on that.

Today they posted $130m profit and you are trying to tell me that they aren't making money? They have had a few bad quarters but then again so have most companies over the last few year.

"It's very possible for a company to be profitable and still be in serious trouble, computer industry history is littered with examples. Making money now is nice, taking the decisions that mean they can make money 2, 3 or 5 years from now is critical. And that's where NVidia is in trouble."

True, Microsoft are a possible example of that- losing control of mobile OS's, search, browser's etc. However I don't see that with Nvidia. They still are the largest GFX chip maker. They having one of the best mobile SoC that has the potential to make them billions. They also are about to release a GPGPU that could help them corner a market where they don't have any competition- i.e a potential gold mine.
However it does seem clear that they need to have more realistic goals as their last few GPU's have just been too big.

 

[paradigm]

They still are the largest GFX chip maker.

Course they are... if you are a blinded fanboy.

I think you'll find that title goes to Intel, by quite some margin.

 

[Rroff]

At high res/quality settings the 8800GTX annihilates the 2900XT in most things and most earlier DX10 comparisions are flawed due to the 2900XT not properly implementing a number of DX10 features.

ATI went completely the wrong direction with the design, no amount of moving to a lower process, etc. would save it.

 

[Rroff]

Course they are... if you are a blinded fanboy.

I think you'll find that title goes to Intel, by quite some margin.

I'd love to see intel shake up the enthusiast market... they have the tech and financial clout - they just need to hire a few good key people already experienced in the particular industry and give them discretion over the actual implementation of features/software.