The problem here is we are arguing who makes more money. We all know nVidia makes more money than ATi. There should only be concern here if ATi was making no money, but as long as both companies are making money, we should refrain from arguing about who makes more.
I don't know if nVidia will make money off of their cards so I am not going to comment but I do think that if they thought they wouldn't be able to make money, I doubt they would sell it.
Agreed, not all features appeal to everyone.
I disagree, ATi can easily raise the price of the 5850 beyond $330 and still be a better value than the GTX 285. I am saying ATi priced their card nicely because only a die-hard fan would buy a GTX 285 over a 5850. Browsing through Newegg I found the cheapest GTX 285 at $339.99:
http://www.newegg.com/Product/Produc...nd&Order=PRICE
and the cheapest ATi 5850 at $299.99:
http://www.newegg.com/Product/Produc...0Video%20Cards
Yet the ATi card is faster across the board:
http://anandtech.com/show/2848/14
That was 6 months ago before 10.3s came out. I think the performance gap should be around 20% average now.
How can you say they haven't priced their cards nicely? Also I think nicely and smartly are quite similar in this context. Smartly for ATi, nicely for us.
evolucion8
Platinum Member
- Jun 17, 2005
- 2,867
- 3
- 81
It actually mattered, look at the Steam's Hardware Survey. The HD 4800 series grabbed a nice score finally outpacing the venerable 8800 series of cards in DX10 GPU's, something that the GTX series barely missed.
DirectX 10 GPU's
GTX series - 10.59
HD 4800 series - 11.48
8800 series - 10.73
DirectX 10 Systems
GTX Series - 6.46
HD 4800 series - 6.45
8800 series - 5.18
While it wasn't a total blow to nVidia, did very well considering the previous lackluster performance of the HD 3x00/HD2x00 series compared to the nVidia's better position at that time. So considering that nVidia did so well with the 8800 series, its sucess wasn't that resounding this time with the GTX series thanks to a much better competitive ATi.
Last edited:
Aside from the obvious validity questions of using the latest valve hardware survey for anything there are some problems with data organization.
The 4850, 4830 and 4860 are included in the '4800 series' category. Those are cards competing with the G92 (read: 8800GT,8800GTS,9800GT,9800GTS,GTS250,9600 of various flavors) and not enthusiast GT200 hardware.
In other words, we really don't know how well the 4870 and 4890 sold compared to the GT200 models they were up against. For all we know NV outsold the ATI competition 100 to 1 in the higher price brackets and the '4800 series' numbers are padded with $85 4850s sold in rat horde numbers.
The 4850, 4830 and 4860 are included in the '4800 series' category. Those are cards competing with the G92 (read: 8800GT,8800GTS,9800GT,9800GTS,GTS250,9600 of various flavors) and not enthusiast GT200 hardware.
In other words, we really don't know how well the 4870 and 4890 sold compared to the GT200 models they were up against. For all we know NV outsold the ATI competition 100 to 1 in the higher price brackets and the '4800 series' numbers are padded with $85 4850s sold in rat horde numbers.
evolucion8
Platinum Member
- Jun 17, 2005
- 2,867
- 3
- 81
HD 4830 vs 9600GSO
HD 4850 vs GTS 250 512
HD 4860 vs GTS 250 1GB
HD 4870 512MB/1GB vs GTX 260/260+
HD 4890 vs GTX 275
HD 4850X2 vs GTX 280/285
HD 4870X2 vs GTX 295
I don't understand how's possible that nVidia outsold ATi's high end 100 to 1 with such competitive lineup. The Steam surveys speaks it all and the HD 4830 were and still hard to find, plus the HD 4860 was released almost 4 months ago, compared to the more than 1 year ago debut of the GTX series. If the GTX series sold that well, why companies like XFX switched to sell HD 4x00 series of cards? A pity that eVGA didn't. Plus the 9800 share is quite high, meaning that they didn't compete straight with the HD 4800 series.
It's all been covered. NV is perceived as a better product even when an identical performing ATI SKU is out there. Marketing, PhysX, 3dvision, TWIMTBP, better board partners and Linux support would have been several reasons to choose green. As a personal anecdote I can't consider any ATI card for my main machine because I use it for both work and play and work often means Linux or Solaris.
I'm not saying NV *did* outsell 10 to 1, and judging from signatures here at AT the reverse is true. What I'm saying is the 4800 series report has low end cards mixed up with high end ones. Cards which didn't compete with the GT200 in price or performance make it seem like the 4870/90 outsold the GT200 when the actual relative sales are unknown.
I'm not saying NV *did* outsell 10 to 1, and judging from signatures here at AT the reverse is true. What I'm saying is the 4800 series report has low end cards mixed up with high end ones. Cards which didn't compete with the GT200 in price or performance make it seem like the 4870/90 outsold the GT200 when the actual relative sales are unknown.
Where do you live? It must be some faraway land where you cannot download drivers...
...but wait: then how do you post here?
Last edited:
Not under the current circumstances.
Sometimes in the future when the yields are better.
"Nvidia drops their open-source driver"
http://www.linuxtoday.com/developer/2010032603035NWHW
Does that make much difference to Linux users? I don't use Linux so I don't know.
It's one thing to have drivers available for download, it's quite another thing to have functional, usable drivers for a modern desktop. ATI targets professional OpenGL apps on older, semi-proprietary distributions of Linux and embedded systems with their drivers. End users are encouraged to talk to their hardware vendors for support.
Both the closed source (fglrx) and open source drivers have a large laundry list of issues.
I'll start off with the closed source ones not supporting anything remotely modern in terms of X server or kernel (they require X server 1.6 which is not easy to downgrade to in modern distributions). They crash or hang frequently and have stunningly poor 2D performance. Support for the 5 series is very much hit and miss and non-existent for mobile chips. Needless to say, crossfire does nothing. No card older than the HD2 series is supported, including the still selling X1... series.
Open source drivers have rudimentary support for the 5 series in VGA mode. Golden oldies like the R300 are supported fairly well for 2D and 3D. 3D is very fresh and somewhat incomplete, with performance about half of the proprietary driver for supported cards. There is some rudimentary power management for some cards.
Video decode and playback acceleration is partly funcitional with some hardware with the proprietary driver.
So yes, while an ATI card will work as a basic VGA framebuffer and maybe even a little bit more (at 100% clocks and voltage the entire time for the open source version) depending on which model you have it's miles and leagues and worlds away from a Windows-like stable and full featured driver NV provides. You can even run a 4 series card from NVidia with their drivers. NV has at least a 5 year head start on ATI when it comes to Linux and it shows.
Oh yes, it's very possible to run many Windows games on Linux with an NV card. Far fewer games run with ATI hardware and the performance is awful. A GTS250 will easily equal a 4890 when gaming under Linux for native and windows-emulated games.
Both the closed source (fglrx) and open source drivers have a large laundry list of issues.
I'll start off with the closed source ones not supporting anything remotely modern in terms of X server or kernel (they require X server 1.6 which is not easy to downgrade to in modern distributions). They crash or hang frequently and have stunningly poor 2D performance. Support for the 5 series is very much hit and miss and non-existent for mobile chips. Needless to say, crossfire does nothing. No card older than the HD2 series is supported, including the still selling X1... series.
Open source drivers have rudimentary support for the 5 series in VGA mode. Golden oldies like the R300 are supported fairly well for 2D and 3D. 3D is very fresh and somewhat incomplete, with performance about half of the proprietary driver for supported cards. There is some rudimentary power management for some cards.
Video decode and playback acceleration is partly funcitional with some hardware with the proprietary driver.
So yes, while an ATI card will work as a basic VGA framebuffer and maybe even a little bit more (at 100% clocks and voltage the entire time for the open source version) depending on which model you have it's miles and leagues and worlds away from a Windows-like stable and full featured driver NV provides. You can even run a 4 series card from NVidia with their drivers. NV has at least a 5 year head start on ATI when it comes to Linux and it shows.
Oh yes, it's very possible to run many Windows games on Linux with an NV card. Far fewer games run with ATI hardware and the performance is awful. A GTS250 will easily equal a 4890 when gaming under Linux for native and windows-emulated games.
No, it doesn't. This was a 2D-only obfuscated open source driver that saw little to no usage. There are licensing issues prohibiting distributions from bundling the NV binary blob, so this open source driver let users boot into better than 800x600x60hz out of the box.
Think of it as a better than 800x600 windows safe mode driver installed by default. Now users will have to use the standard safe mode driver long enough to download and install the real NV driver -- a one or two click operation on the user friendly distributions.
I'm probably making a mistake replying to this as it's clear you have fanboi tendencies, but if Charlie's numbers are correct I can see the logic in what he's saying. As others have pointed out, if the bare package price for the chip itself is $250, once you tack in costs for the board, HS/Fan (those vapor chambers aren't cheap), misc hardware, assembly, packaging, instructions, support ($$), warranty ($$) and sale margins for 2 middle-men (minimum for the board maker is 5% and retail margin is at least 20% or it wouldn't be on the shelf and it's probably closer to 50%) that $500 is either going to be 0 profit or a loss. Even if they make a small profit they aren't going to ever recoup the R&D. The only reason IMO that nVidia is even producing Fermi is because they want to give their executives time to sell stock before the damage to earnings/revenue is announced.
nVidia Fanboi's like ragging on Charlie and trying to say he's been wrong and point out his bias, but every rumor was either dead on or close enough that it indicates he was picking up the actual fluctuation in specification during production. Charlie has a source at TSMC that's telling him all the dirt, my guess would be in the Fab in either validation or testing or less likely possibly at the executive level. He pegged price, power, heat within a reasonable margin (while nVidia was still trying to peg them down themselves).
nVidia needs a major re-tape/redesign for Fermi to be competitive in both performance and power usage/heat production. The 104 isn't going to be a full new taping with a completely new design, it's at best a re-spin out of a slightly modified design. Honestly I don't think all of the blame lies on nVidia, TSMC has frankly botched up the .40 process, if .40 had been mature and nVidia hadn't had to re-spin 3 times to get to 470/480 they probably would have wiped the floor with ATI but if ATI moves to Global Foundries for Northern Islands and TSMC doesn't get their crap together it's game over for nVidia in the performance segment IMO. They will have probably 6 months before the lack of performance affects the OEM sales, but the worst damage is going to be when evergreen moves to laptops, my guess is it will devastate OEM sales at nVidia as laptops now exceed 50% of all computers.
Anyway back the point, $250 for the chip for a $500 card doesn't guarantee a profit for anyone given what margins and additional costs aren't included in that $250. If afterall things are so peachy at NVDA why did the insiders sell 3 millions shares in the last 6 months? And why did the institutional investors dump 14 million? HUANG JEN HSUN alone dumped 150K in March, 150K in Feb and 520,950 In Jan. Don't tell me those transactions are normal scheduled transactions.
http://finance.yahoo.com/q/it?s=NVDA
Last edited:
Meghan54
Lifer
- Oct 18, 2009
- 11,684
- 5,225
- 136
Actually, if you really read what's there, you'd realize that the sales were, by the officers of the company at least, automatic sales immediately preceded by options. The option was the "purchase" the officer made and the automatic sale was already in the system. Officers of a company have to file months and months ahead of time before selling stock.
What you're seeing is their payment.....their salary being paid in stock options....and how they're converting their salary of stock options into spendable cash.
Remember, Huang Jen Hsun owns over 20M shares of stock, either directly or indirectly. So his sales of stock you mentioned, even combined with all other officer sales, amounts to less than 10% of all officer owned stock. Doesn't sound like dumping to me.
And the institutional sale.....that sale you mentioned represented less than 3.3% of all institutional held stock. Again, not exactly a dumping of stock.
In my experience, ATI's drivers, both the official and open source, are much better at 2d performance and compositing than nvidia's. So for general desktop use, I find them superior, but for anything that requires an actually working 3d card, nvidia is better.
Nvidia's drivers seem more out of date with the direction of linux desktop at this point, 2d performance is horrible at the moment, even the nouveau driver outperforms the official one in 2d. And it does matter, my laptop with an intel igp runs ubuntu far more smoothly than my desktop with a 9800gtx+.
My profession is purchasing for a distributor- let's say to be kind you are way off the mark in terms of what the margins "have" to be. The PS3 as a general example has no margin, if you include labor it is a net loss to handle it. Margins on games are stellar, so it is worth handling. The GTX480 is less then a cubic foot, under 5lbs and retails for $500, it is an exceptionally friendly product on the distribution and retail level if there is reasonable demand for it. I haven't seen what the margins are on that part in particular, but I've been around long enough to know that even at a 10% margin right now that is a killer product particularly on the distribution/retail level- $50 gross margin with miniscule labor/shipping overhead is the type of thing we love to handle. On a retail/distribution end there is no more labor involved nor additional shipping overhead on a GTX480 then there is on a 5750(well, perhaps a few additional cents in shipping for the weight- 15 tons runs about $600 for us to ship and we normally make money on that too). With the demand for this product right now outstripping supply by a significant margin, inventory issues across the board combined with extremely low inventory costs the only remote hesitation you may have on some retailers parts is if they work on an interest charging inventory acquirement system; WalMart has put most of them out of business already though.
I can buy a mobo, CPU w/HSF, and 1GB of RAM for under $250 easily from NewEgg. The 5850 launched at $260- you think a magic fairy delivered the chips for free and ATi was only making $10 per sale even then? All of the associated costs for putting a graphics board together minus the chip doesn't come remotely close to being in the league of $200. This is extremely simple to observe.
The only things he had right were that it was going to be late, big and hot. Everyone who knows anything about processor design knew that well in advance.
Fermi needs a major redesign to be competitive in performance? I have to say that is the first time I have heard that. As far as power usage/heat production- they only need a 'major redesign' if the public doesn't buy it or they are losing money because of it.
Take the first part released from any new line from nVidia and rip the heatsink of. A3 is actually pretty normal for them hitting initial retail. The only difference this time is a blogger was talking about how badly things were going due to them following standard practices.
Never had a job where they offered you stock options? You schedule an automatic sale of your stocks prior to options on buying in at your fixed price, when you are replacing the stocks you just sold for 1/4 of what you sold them for, you are making millions of dollars. These things need to be filed months some time in advance with the FTC and are standard operating procedure at companies who have lucrative stock option bonuses.
I won't debate whether you are right or not on most of what you said. But did you seriously just say the 5850 gpu costs $250 to make? I hope you realize that a 480 is a tad more expensive to produce than a 5850...
GodisanAtheist
Diamond Member
- Nov 16, 2006
- 8,031
- 9,290
- 136
-Actually I think he's saying the opposite. He can get all sorts of crazy shit for less than $200, it probably costs a fraction of that for a GPU alone, even at 5800's original MSRP.
AMD was probably making a fair amount on each card sold, now they're making that much more.
Wrong. He reported about memory controller issues, yields, fake Fermi, ...
Much more than you would give him credit for.
Absolutely not, what I'm saying is that the highest reported costs for the GTX4xx GPUs we have seen is $250 and it is reported that they are taking a loss on these parts. The GTX480 sells for $500- if it is true that the GTX480 is selling for a loss then that means there is over $250 in cost to produce the PCB, buy the memory, package ship and distribute the boards. The 5850 is comparable on all of those costs except for the GPU. What I was saying is that if a magic fairy delivered 5850 GPUs to ATi for free, anyone that believes that nV is taking a loss on the GTX480 would believe that ATi was making at most ~$10 on the 5850 seeing that it launched at $260. That is a laughable assertion to be sure, I am positive ATi had far higher margins on the 5850 then that even though they actually made the GPU themselves and didn't have a magic fairy zapping them into existance for free. The idea that getting a PCB/mem/assemble/package/distribute on a sub 1 cube sub 5lb part is close to $250 is absurd.
Yields were a given for anyone who knows anything about building a processor. They have 3billion transistors on an inmature 40nm build process, there is no way yields were going to be good. ATi's chips were around 40% and they had parts that were built to be high yield, not pushing the upper limits of what was possible. I could start writing a bunch of posts now about how summer is going to be hotter then winter, what would you think of people who used that as evidence of insider information? It is the same thing.
What was it exactly he said about the memory controller? Given that the 480 seems to scale almost perfectly with core clocks without touching the memory at all seems to indicate that they aren't having any issues with bandwidth, but honestly I don't remember exactly what he had to say about it. Fake Fermi- they used a mock up at a press event- rather standard practice across a huge variety of different industries. Yes, he was right about that, much like I could say Toyota was only showing mock ups of certain cars at the last auto show. I'd be right, but the only people who would care at all would be those with an axe to grind.
You know what, something else I forgot to give Charlie credit for, so far from everything I have seen he posts the correct date on all of his articles too.
But the PCB/memory/etc. for the 5850 isn't the same. It isn't the same amount of ram, it isn't nearly as complex, and the cooler isn't nearly as strong. All of that costs money. And since margins are percentage rather than absolute numbers, x% margin on a higher cost product is more money than the same x% on a lower cost one. You simply can't compare it that way.
Not saying anything about the rest, but you can't really compare a console to a GPU. It is well known that console makers sold them at a loss for a long time, by subsidizing the console and hoping for enough game sales to make a win. I don't think the aftersale market for GPUs brings Nvidia much money
The RAM amount isn't the same, but the 5850 actually uses faster RAM. The PCB complexity is certainly comparable(although obviously not identical) and the coolers are different. The PCB/mem/heatsink costs for the 480 and 5850 are *much* closer then the GPU costs, by a long shot. To get an idea of what PCB costs are in relation to each other, check out the cost on mobos, they are far more complex then the most complex consumer graphics card ever released. Check out the pricing in that segment.
We are talking losing money versus not losing money, we are talking about absolute numbers.
Heh, no, I meant distributors and retailers don't make any money on the PS3. We have no margin on the product. If Sony is making $245 per console or losing $900, we(distributors) don't make anything.
Last edited:
No. The GTX 480 uses the same RAM (6Gbps bin) but Nvidia's memory controller bugs means it can only run at the reduced speed. But they need that bin to achieve those speeds.
It's in one of Charlie's forum posts which I can't find the link to (he posts a lot) but I remember seeing in the last month.
http://www.xbitlabs.com/articles/video/display/geforce-gtx-480_2.html
5000MHz GDDR5 chips from Samsung.
Same speed as the HD5870 chips IIRC.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 24K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 20K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter