More bad news rumors for Fermi

[GaiaHunter]

This is absolutely hilarious. NO ONE said this about ATI when the GTX 280 and 260 came out and ATI had nothing. Then when the 4870 and 4850 came out EVERYONE loved it because it was competitive on price.

The point is being competitive on price is the ONLY thing that matters for the majority of non-ultra high end enthusiasts.

16th June 2008
NVIDIA's 1.4 Billion Transistor GPU: GT200 Arrives as the GeForce GTX 280 & 260
http://www.anandtech.com/showdoc.aspx?i=3334

25th June 2008
The Radeon HD 4850 & 4870: AMD Wins at $199 and $299
http://www.anandtech.com/showdoc.aspx?i=3341

The 2nd part is true but not the first.

 

[SlowSpyder]

Please explain how approximately ~5870 level performance for less than 5870 price is bad for consumers.

Well, considering it doesn't yet exist in retail it's not exactly helping the consumer...

 

[BenSkywalker]

I didn't see or remember that thread.

You were the third poster in that thread-

http://forums.techpowerup.com/showthread.php?t=117000

That's the link to the results. To me it's all rumors, but if we are giving an analysis that all of these rumors are accurate, then it would seem nV would be blowing ATi away on a price/performance basis. I take all of it as normal pre launch BS, when we see reviews I'll start paying a lot closer interest

However, do you really see NV getting into a price war with AMD?

If they want to, sure, why not? I think people seriously underestimate how much raw profit nV makes on each part they sell, they have a huge amount of room to move before they come close to taking a loss. For last quarter, with nV having nothing to compete with ATi at any price point over $120 nV's gross profits were greater then ATi's total revenue. If we were to say that nVidia was going to sell 100K GF100s and each one of them cost nV $300 to make and they gave them all away it would have consumed about 25% of their profits last quarter(IIRC, think they made roughly $120Million if memory serves). Obviously nV isn't going to come anywhere close to giving these things away, just trying to put into perspective how much room they have before a price competitive part at the high end is going to seriously impace their bottom line. Right now ATi is being smart and taking consumers to the cleaners(odd I don't hear the same people who were outraged back during the GT200's run outraged again, but I digress) fattening up their margins and helping the bottom line for AMD as much as possible.

If all of the rumors we heard are true so far, the 470 will launch at $30-$80 less then the 5870 did and will be faster overall. Right now the 5870 is marked up considerably, the actual price swing would be in the $80-$130 range compared to the least expensive 5870. I would think, no matter what color video card you like to buy, we all could agree that if these things do come to pass it would be a win for consumers. Clearly ATi will need to drop their prices to compete, so no matter board you are after it would be a good situation to have happen. Not saying I believe any of it though, just since we are in the rumor/speculation/analysis mode I'd figure I'd chime in

Basically Nvidia's new core runneth over with hype, so it see it 6 months late, huge, hot, power hungry and marginally better than ATI's recycled Tera Scale Arch provides quite a gulf between perception, expectation and (seemingly) reality.

The hype was about the refined way in which the chip works, if it truly is a huge leap will be proven moreso by PhysX/Badaboom/RayTracing demos then by games built for a rasterizer. They made huge changes to the way the compute end of the chip works, not much to the rasterization side of the equation(even if you read all the hype, that is what it focuses on).

 

[Lonyo]

What sort of Physics does Battlefield Bad Company 2 use?

Software.

http://www.pcgameshardware.com/aid,...y-2-x-New-tech-info-about-the-DX11-game/News/

Anders Gyllenberg: Havok is a good base for us and we have built many layers of our own on top of it to support destruction and our large-scale multiplayer worlds. One interesting feature and quite unique feature is the destructible terrain - it actually affects gameplay and allows the player to take cover in the craters created. We are currently performing all our physics computations on the CPU cores in parallel.

Oh shi-

 

[Lonyo]

This is absolutely hilarious. NO ONE said this about ATI when the GTX 280 and 260 came out and ATI had nothing. Then when the 4870 and 4850 came out EVERYONE loved it because it was competitive on price.

The point is being competitive on price is the ONLY thing that matters for the majority of non-ultra high end enthusiasts.

16th June 2008
NVIDIA's 1.4 Billion Transistor GPU: GT200 Arrives as the GeForce GTX 280 & 260
http://www.anandtech.com/showdoc.aspx?i=3334

25th June 2008
The Radeon HD 4850 & 4870: AMD Wins at $199 and $299
http://www.anandtech.com/showdoc.aspx?i=3341

The 2nd part is true but not the first.

http://www.anandtech.com/video/showdoc.aspx?i=3338
3 days later the first real stuff leaked.
And ATI were only claiming the HD4870 would compete with the 9800GTX.

So yeah, it's not even close, since there was a whole THREE DAYS between the NV launch and ATI leaked full real proper benchmarks and pricing and cards, and then another 6 days until proper full tech stuff.

 

[SHAQ]

Now we have to wait until April 6. Nice.

 

[SlowSpyder]

This is absolutely hilarious. NO ONE said this about ATI when the GTX 280 and 260 came out and ATI had nothing. Then when the 4870 and 4850 came out EVERYONE loved it because it was competitive on price.

The point is being competitive on price is the ONLY thing that matters for the majority of non-ultra high end enthusiasts.

The 4870 gave you 90% of the performance of a GTX280 for under 50% of the price. AMD launched their parts the week after Nvidia launched the GTX2x0 parts. If it took AMD a further 6 months after Nvidia's GTX2x0 launch, and the GTX280 was launched at and selling for under $400 I assure you the reaction would have been pretty underwhelming as well.

It's been pointed out quite a few times that Nvidia has a lot of money in the bank. That Nvidia spends more on R&D than AMD by a large amount. Both AMD and Nvidia use the same company (and in this case the same process) for their graphics chips. Yet this is the best Nvidia managed to do? Show up 6 months later (and it will be difficult to find Fermi for a further few months) with a part that matches performance to is slightly faster that uses more power and puts out more heat (asuming rumors are true, if we believe them all). Sorry, it just seems so very underwhelming at this point.

I guess we'll wait and see if the rumors end up true or not, but so far things aren't looking all that rosey for Nvidia from an enthusiast point of view.

 

[bunnyfubbles]

This is absolutely hilarious. NO ONE said this about ATI when the GTX 280 and 260 came out and ATI had nothing. Then when the 4870 and 4850 came out EVERYONE loved it because it was competitive on price.

The point is being competitive on price is the ONLY thing that matters for the majority of non-ultra high end enthusiasts.

Except the 4800s were only a few weeks late to the game (not half a year+), and ATI wasn't in trouble because those parts were specifically designed not to take the competition on through sheer performance, but through cost-effective price/performance. Fermi is anything but cost-effective, its a part clearly designed to cost a lot of money but cannot (apparently) offer the performance to match when compared to the 5800s.

Sure, as long as nVidia offers competitive pricing, we consumers will win for the short term, but that still doesn't mean what I said is wrong, the situation certainly isn't ideal for nVidia when they have to sell their product for prices that hurt them.

 

[v8envy]

If they want to, sure, why not? I think people seriously underestimate how much raw profit nV makes on each part they sell, they have a huge amount of room to move before they come close to taking a loss. For last quarter, with nV having nothing to compete with ATi at any price point over $120 nV's gross profits were greater then ATi's total revenue.

NV being a very profitable growth company is exactly why I don't expect a price war. There is no growth of any parameter NV stockholders like that can be increased by selling the new chips at a loss. The high end of the last generation was bad enough that way -- which is why I suspect NV EOLed the G200 instead of trying to price war with the 58xx series.

ATI only had limited quantities of competitive product at $350+ (read: barely registers saleswise) and at $160. The first was too expensive to count and the second performed worse than SKUs already available at that pricepoint for the previous year. I'm shocked they sold ANY cards, really.

The price war simply can't happen with limited supplies of 40nm parts. Even if NV gives away the hardware and board partners decide to do the same you'd see a $0 MSRP 470 fetching at least $410 on newegg so long as the 5870 sells for over $410.

 

[thilanliyan]

http://www.shanebaxtor.com/2010/03/16/gtx-400-series-date-pushed-back/

Booo!! :twisted:

 

[vss1980]

Can't wait to see the price listing for a Quadro card running all 512 SPs. If demand is tight then we've got to be looking at a a price maybe 10x higher than the GTX 470....

 

[Daedalus685]

Can't wait to see the price listing for a Quadro card running all 512 SPs. If demand is tight then we've got to be looking at a a price maybe 10x higher than the GTX 470....

The Fx5800 is $3000 on the egg, I'd expect a true 512 quadro to be running on the order of 5k at release.

Question though... I'm a bit confused as to who the quadro line is currently aimed at. I have used them in the past for Cuda work on some labview racks, will they be selling "tesla/fermi" branded cards for that now or will that name be reserved for the rackmount computing boxes?

 

[Janooo]

...
If they want to, sure, why not? I think people seriously underestimate how much raw profit nV makes on each part they sell, they have a huge amount of room to move before they come close to taking a loss.
...

Well, it seems it's not going to last for long. NV does not have mid and low end DX11 cards. OEM's are jumping to AMD.

 

[Sylvanas]

So for an extra 1 billion transistors the performance we get vs a 5870 is.....wait for it....about the same. I'll reserve judgement till launch day but it's not looking good for Nv.

 

[shaynoa]

sorry about the spelling in earler post but i was in a hurry , i can't understand why Nvidia would be doing things in this way ,i think it states to us all that there 2 videocards are not going to be anywhere near as good as a lot of people think,there is no other reason for the lack of advertisement,
i have been in business for myself for 25 years and have never heard of a company pushing people away like is going on here ,something isn't right

 

[BenSkywalker]

NV being a very profitable growth company is exactly why I don't expect a price war.

I don't expect them too, but it is certainly easily within their capability if they so choose is what I was saying. Even if we assume all of the yield rumors we have heard are true, if this latest article is also true that would indicate that nV is going with a Cell style built in redundancy on functional units which can sharply increase yields(it could move it from 1% to 50%; not saying it will but it could). They also have the option to release a '460' with say 320 shader units and perhaps a 256bit bus that would likely put their useable dies into a very comfortable position as far as yields go giving them the option to focus on the higher end parts as straight margin builders while reducing the risk of yields hurting the bottom line by a hefty amount. If they placed such a part opposite the 5830 it would put them in a rather secure position to be able to compete in a price war without a serious financial risk.

There is no growth of any parameter NV stockholders like that can be increased by selling the new chips at a loss.

This is true, but they don't necessarily need to be sold at a loss, or anything close to it, if nV plays their hand a certain way. nV paying $5K per wafer with a possible yield of ~110 parts per if they can only sell a total of ~20% between the 480 and 470 it quickly becomes extremely expensive. If they can manage to sell ~80% of them as 460s the cost of the 470s and 480s drops in an enormous fashion which could very well allow them to get into a price war, in the segment they want to, with ATi while actually improving their bottom line situation. It all depends on how they want to play it.

The price war simply can't happen with limited supplies of 40nm parts. Even if NV gives away the hardware and board partners decide to do the same you'd see a $0 MSRP 470 fetching at least $410 on newegg so long as the 5870 sells for over $410.

This is true to a point, but cross vendor pressure should help keep things within reason. nVidia is going to ramp more production then ATi on 40nm(or already have)- what percentage of it is for GF100 based parts is going to determine how much availability we have(no matter what the yields, running twice as many wafers should yield almost exactly twice as many parts) so the situation could markedly improve by the end of Q2. Again, all speculation and the assumption that he rumors we have heard are all true, which I'm certainly not sold on at the moment

So for an extra 1 billion transistors the performance we get vs a 5870 is.....wait for it....about the same. I'll reserve judgement till launch day but it's not looking good for Nv.

Fermi was very clearly aimed at the HPC market from early in the design phase. If the GF brand holds even and the HPC market situation improves markedly it would be a huge win for nV.

 

[PlasmaBomb]

That's exactly my point. At $300 the 470 sits between the 5850 and 5870 in both price and performance -- or at least it would, if it got into a time machine and went back six months. It's relative good deal now simply because prices of the 5 series have increased due to only one source being unable to cope with the high end product demand combined with etail gouging.

That, and I simply don't believe the $300 GTX470 for a moment.

So this rumor you believe, but not the benches from the other thread where the 470 is faster then the 5870 60% of the time.

The fail is strong in this one...

 

[PlasmaBomb]

what's to stop AMD from undercutting them with price cuts? The 5850 back at its original $259 MSRP or lower as well as the 5870 with plenty of room to drop would be a major blow for nVidia.

The question is -

Did AMD price the HD5800 line at a MSRP and leave themselves room to cut the prices when fermi arrived? (should it be as good as claimed).

I think the answer is yes, and that AMD and it board partners have been happy for the last 6 months...

 

[Lonbjerg]

Software.

http://www.pcgameshardware.com/aid,...y-2-x-New-tech-info-about-the-DX11-game/News/



Oh shi-

That is why all the physics is scripted...hence HL2 had better physics...oh shi...

 

[BFG10K]

I spotted this just now: http://translate.google.fr/translat...nfirmation-du-retard&sl=fr&tl=en&hl=&ie=UTF-8

According to our reliable source, it will not be before April 6 that should be available in-store cards. The reason for this delay is simple, Nvidia has still not provided the frequency of its new graphics cards to its partners.

Apparently, the photos Inno3D leaked through Guru3D were just empty boxes made to create buz. This also worked very well.

Equally disturbing, partners today still do not know how many cards they may have for the day of the official release.

 

[dug777]

Price and performance is all very well, but if it's not actually available in any meaningful volume then MSRP for fermi will meaningless as the basic mechanics of supply and demand kicks in, and that is not going to drive DAAMIT pricing down.

Given JHH's analyst conference call comments about Fermi ramping in Q2 of FY 11 (so May to July?) it seems more likely than not that there will be availability issues for some time after launch (will you actually be able to buy the cards at launch, or did I read somewhere that was delayed also?).

That they will launch the cards at a competitive price/performance bracket I have no doubt (that seems elementary logic). That's about as much use as a chocolate teapot if you can't actually get hold of one at that price

EDIT: doh! BFG answered my question, and at this stage the answer appears to be no (although that is not by any means confirmed at this stage, right?)

 

[Kuzi]

Nvidia should do fine if they can release Fermi with reasonable quantities (+100K), and maybe within the next few months release mid-range/lower end Fermi derivatives. Of course the pricing has to be right too.

If they can't do that, then they will have a tough year ahead of them.

 

[dank69]

...

EDIT: doh! BFG answered my question, and at this stage the answer appears to be no (although that is not by any means confirmed at this stage, right?)

Now we have to wait until April 6. Nice.

http://www.shanebaxtor.com/2010/03/16/gtx-400-series-date-pushed-back/

Booo!! :twisted:

I spotted this just now: http://translate.google.fr/translat...nfirmation-du-retard&sl=fr&tl=en&hl=&ie=UTF-8

Third post claiming April 6th.

 

[Will Robinson]

sorry about the spelling in earler post but i was in a hurry , i can't understand why Nvidia would be doing things in this way ,i think it states to us all that there 2 videocards are not going to be anywhere near as good as a lot of people think,there is no other reason for the lack of advertisement,
i have been in business for myself for 25 years and have never heard of a company pushing people away like is going on here ,something isn't right

No problemo.....capital "I"s would be nice though.
Also it's "earlier".^_^

 

[Headfoot]

Fermi is anything but cost-effective, its a part clearly designed to cost a lot of money but cannot (apparently) offer the performance to match when compared to the 5800s.

Both incorrect according to current rumor. Rumor says 5870 performance for ~$350 (rumor may be wrong of course).

My point is that people care about price, the example is clearly different as this is 6 months late rather than a few weeks, but all im trying to say is that PRICE is most important, as long as performance is in the same ballpark.