Nvidia Fermi GF100 working desktop card spotted on Facebook

[JohnZS]

yeah overpriced sure when gtx 285 is nearly selling for same price and what are you gonna do with 2 gb model how many games are there demanding that much ?

Eric
If we try to find a Radeon 5870 IN STOCK for shipping 24hrs the only one I can find costs a whopping £400 odd pounds the most expensive 2GB GTX 285 is hideously over priced @ £330. Now when the Radeons which are showing (and have been showing as Pre-order) since late October/Early November finally come in stock the Radeon 5870 will cost around the £270 mark IMHO
as for the 5970... £525 is the cheapest I can find in stock (around 855USD)

As for 2GB, already I have a few games which require 2GB of VRAM for optimal performance such as "Grand Theft Auto IV, and S.T.A.L.K.E.R Complete 2009. FYI GTA IV Lost and Damned is for release next week too.

I am waiting on either a Radeon 5870 with 2GB or a Fermi

EDIT
I would hardly call the GT200 a failure, I would however say that the Radeon 4850/4870 were a bigger success than the mean green machine anticipated, however the GT200 was (and still is) a very fast GPU and the fastest SINGLE GPU* although the Radeon 5870 has changed that as that is now the fastest single GPU card
John

[Helloworld_98]

"GeForce GTX 380 will be 15% faster than dual GPU Radeon HD 5970."
"In terms of performance GTX 360 will sit between HD 5870 and dual GPU HD 5970."

Words on MSRP for GTX 380 is $499 and GTX 360 is $379, if that's true then count me in.

the idiot who made that table is just that, an idiot.

tell them to study what nvidia does with their cards before making a table.

[SamHughe]

Eric
I am waiting on either a Radeon 5870 with 2GB or a Fermi

Me too. There was a thread about THIS 2GB MODEL a while back. Hopefully we'll see some better third party cooling solutions for both cards. In any case, waiting until Fermi is released seems to be the logical option here.

[JohnZS]

Me too. There was a thread about THIS 2GB MODEL a while back. Hopefully we'll see some better third party cooling solutions for both cards. In any case, waiting until Fermi is released seems to be the logical option here.

Thanks SamHughe
Nice article

However I'd (and perhaps you too?) would like to see a Gainward Golden Sample Heat Pipe special cooler Edition of this card. Their GTX 285 2GB cards are very cool and quiet.

A Sapphire VaopourX or HIS ICECOOL Edition of this card would be good too.

I am hoping Fermi would have a decent cool and quiet cooling solution.

John

[highoctane]

Ummm... he was specifically talking about the gaming card/market, not the professional/workstation which is what was saving Nvidia for most of the G200/RV770/RV870 time period.

G200's BOM was around 2x RV770s. The gap between GF100 and RV870 is going to get even larger.

Ok so is this your own personal claims based on assumptions/opinion or is this coming from a reliable source?

I've never seen the BOM costs of either card from ati or nvidia, would you mind posting that info?

The volume of sales is by and far largely for end consumer sales which is what makes up most of the market share for the quarter with professional/workstation sales being a much smaller percentage overall of higher margin sales.

[Manicdan]

I am waiting on either a Radeon 5870 with 2GB or a Fermi

John

Me too. There was a thread about THIS 2GB MODEL a while back. Hopefully we'll see some better third party cooling solutions for both cards. In any case, waiting until Fermi is released seems to be the logical option here.

Me three
but with lemon

i hope the 5890 is going to be sick, thats probably when ill make the jump, since it should last a very long time

[InCredible]

ill probably end up waiting for atis new arch before changing anything..im all about eyefinity and the 4870x2 is good enough for anything out right now so theres really no reason to upgrade..id rather get my ssd(which is within a week or two) and do some other small upgrades to case/cooling...however id like to see what nvidia has to offer and probably wouldn't even consider them until they have a refresh of fermi..but id also need to see something like eyefinity unless the card is just straight kickin ati's arse which i dont see happening...

[JohnZS]

Me three
but with lemon

i hope the 5890 is going to be sick, thats probably when ill make the jump, since it should last a very long time

Perhaps the 5890 might even be THE 2GB 5870? (so not only 2GB of VRAM but a nice speed boost as well )

[Manicdan]

Perhaps the 5890 might even be THE 2GB 5870? (so not only 2GB of VRAM but a nice speed boost as well )

the only thing it better do is please me until im bored of bulldozer

[eric66]

Eric
If we try to find a Radeon 5870 IN STOCK for shipping 24hrs the only one I can find costs a whopping £400 odd pounds the most expensive 2GB GTX 285 is hideously over priced @ £330. Now when the Radeons which are showing (and have been showing as Pre-order) since late October/Early November finally come in stock the Radeon 5870 will cost around the £270 mark IMHO
as for the 5970... £525 is the cheapest I can find in stock (around 855USD)

As for 2GB, already I have a few games which require 2GB of VRAM for optimal performance such as "Grand Theft Auto IV, and S.T.A.L.K.E.R Complete 2009. FYI GTA IV Lost and Damned is for release next week too.

I am waiting on either a Radeon 5870 with 2GB or a Fermi

EDIT
I would hardly call the GT200 a failure, I would however say that the Radeon 4850/4870 were a bigger success than the mean green machine anticipated, however the GT200 was (and still is) a very fast GPU and the fastest SINGLE GPU* although the Radeon 5870 has changed that as that is now the fastest single GPU card
John

they are ripping you off thats what i am gonna say situation is totally different in newegg/us so i guess conditions change from country to country and that pound thing really f's you up sorry must say that as for 2 gb games u said gta is lazy port and stalker is bug hole but of course may be they really need that much of memory don't know ...

[ajaidev]

These are great articals on EDC and ECC. I am thinking that the AMD's EDC is based on parity bit and Nvidia ECC is based on either Hamming ECC.

http://www.eccpage.com/ "ECC"

http://graphics.stanford.edu/~seander/bithacks.html "EDC"

The end result is ECC has default EDC but the complexity varies. ECC does not always correct the errors tough resending the code may have to be used. In EDC case its the only option. So which is faster, ECC is faster but not always take a scenario where ECC tries to fix a code but could not and has to ask for a resend of the code in that case the EDC system will be faster as it will detect and request and not like ECC which first detects then tries to fix and the requests.

[mibo]

A3 stepping in production:

http://www.semiaccurate.com/2009/12/...-silicon-oven/

Let's see what frequencies the A3 chips will be able to handle.
Hopefully TSMC get their 40nm yield problem fixed. Otherwise we might end up waiting for fermi cards till summer :-(

[Manicdan]

are there any mid range cards in that A3 hot plate? what are their plans for <200$ cards? it better not be renaming dx10 cards until 2011.

[safan80]

are there any mid range cards in that A3 hot plate? what are their plans for <200$ cards? it better not be renaming dx10 cards until 2011.

no they'll make a special 200b chip for those products

[Manicdan]

does that mean its 200 series architecture with dx11?

[eric66]

well from what we saw with gt210 to gt310 transformation that is unlikely. mid parts will prolly support only dx 10 fermi arch based ones will be dx 11

[***Deimos***]

A3 stepping in production:

http://www.semiaccurate.com/2009/12/...-silicon-oven/

Let's see what frequencies the A3 chips will be able to handle.
Hopefully TSMC get their 40nm yield problem fixed. Otherwise we might end up waiting for fermi cards till summer :-(

Surprisingly, as video cards have progressed over the decade, clock speeds are becoming less and less relevent (anybody here recall the first power hungry 120Mhz GeForce 256 made on 220nm from '99.. back before DDR memory)

Despite ALL THE HIGH PRAISE OF MOORE'S LAW, I find that Amdalhs law is much more improtant

Back in '99 using higher than 1024x768 was considered needless and excessive. But with today's 3200 SP 5970, thats not enough pixels to show speedup.

We are seeing LESS AND LESS SPEEDUP WITH EACH VIDEO CARD GENERATION.

Perhaps this quote from wikipedia illustrates the point best:

For example, if 95% of the program can be parallelized, the theoretical maximum speedup using parallel computing would be 20× as shown in the diagram, no matter how many processors are used.

So next time you're overclocking that 5870 to dizzying 1Ghz stratosphere and wondering why you're only getting 1fps gain.. its no fluke... its Amdalhs law.. diminishing returns.

[LordEC911]

Ok so is this your own personal claims based on assumptions/opinion or is this coming from a reliable source?

I've never seen the BOM costs of either card from ati or nvidia, would you mind posting that info?

The volume of sales is by and far largely for end consumer sales which is what makes up most of the market share for the quarter with professional/workstation sales being a much smaller percentage overall of higher margin sales.

There was a chart circling around awhile ago that was pretty close, though that wasn't the first/main source of the info. I had heard how much both companies were selling their kits for which didn't include all components nor packaging/shipping and everything else. So the base numbers are pretty accurate but are used to extrapolate the real cost. The final numbers seemed to be pretty accurate when it leaked that the GTX260c216 cards were selling for ~$150, they were actually losing money.


As far as the gaming vs professional market, I recall a statement from nvidia saying that 70% of their profit comes from the workstation/professional market. I cannot seem to find the quote, so I might be wrong.

[STEvil]

Surprisingly, as video cards have progressed over the decade, clock speeds are becoming less and less relevent (anybody here recall the first power hungry 120Mhz GeForce 256 made on 220nm from '99.. back before DDR memory)

Despite ALL THE HIGH PRAISE OF MOORE'S LAW, I find that Amdalhs law is much more improtant

Back in '99 using higher than 1024x768 was considered needless and excessive. But with today's 3200 SP 5970, thats not enough pixels to show speedup.

We are seeing LESS AND LESS SPEEDUP WITH EACH VIDEO CARD GENERATION.

Perhaps this quote from wikipedia illustrates the point best:


So next time you're overclocking that 5870 to dizzying 1Ghz stratosphere and wondering why you're only getting 1fps gain.. its no fluke... its Amdalhs law.. diminishing returns.

The law has a flaw.

When a non-parallelized instruction is used, queue more of them at the same period to make use of idle resources (oh hey, reminds me of tile buffering with the Kyro series. which works EXTREMELY well).

That aside, Mhz does still mean a lot or we'd be using 120mhz Radeon 5870's.

[tajoh111]

Surprisingly, as video cards have progressed over the decade, clock speeds are becoming less and less relevent (anybody here recall the first power hungry 120Mhz GeForce 256 made on 220nm from '99.. back before DDR memory)

Despite ALL THE HIGH PRAISE OF MOORE'S LAW, I find that Amdalhs law is much more improtant

Back in '99 using higher than 1024x768 was considered needless and excessive. But with today's 3200 SP 5970, thats not enough pixels to show speedup.

We are seeing LESS AND LESS SPEEDUP WITH EACH VIDEO CARD GENERATION.



Perhaps this quote from wikipedia illustrates the point best:


So next time you're overclocking that 5870 to dizzying 1Ghz stratosphere and wondering why you're only getting 1fps gain.. its no fluke... its Amdalhs law.. diminishing returns.

What holding back gaming returns is primarily software development.

The more you design a game for high end hardware, the more customers you alienate and the smaller your sales.

Not to mention console gaming. Almost anything that not a MMO RPG has to be developed for consoles in mind, as a result, games are being program with being able to run on a console.

IF pc gaming was stronger and people had better hardware on average we would see a larger gap between generation as programs would likely be more specialized for stronger hardware.

How correct is that chart?

I am surprised the cost of DDR5 is not really anymore expensive than DDR3.

Also the price of the GPU seems really low. If high end GPU were so cheap, I wonder why consoles are sold for a loss when they come out considering they come out at 500 dollars.

[highoctane]

There was a chart circling around awhile ago that was pretty close, though that wasn't the first/main source of the info. I had heard how much both companies were selling their kits for which didn't include all components nor packaging/shipping and everything else. So the base numbers are pretty accurate but are used to extrapolate the real cost. The final numbers seemed to be pretty accurate when it leaked that the GTX260c216 cards were selling for ~$150, they were actually losing money.

Interesting breakdown but the ram pricing seems screwy to me, ddr5 seems a rather bargain compared to ddr3 with ddr5 512 at $13.30 & doubled to 1024 for $19.95 vs ddr3 at $13.98 for 512 & $33.83 for 1024?

Whats up with the est cost of the 285 gpu in relation to the 275 or bigger 65nm 280, the gpu cost seems screwy even if you where to factor in a skewing for extra validation for binning for the 285.

Seems like there is a little bit too much extrapolation & manipulation without enough solid verifiable information, I mean do you think the 4890 gpu cost is really the same as a big gtx275?

Shipping costs shouldn't be a bid deal with cost distributed across the volume of these things moved at a time.

Anyways interesting chart but I have my reservations about its accuracy overall.

EDIT: There was in interview where it was mentioned, the claim on more profit from the quadro market but the sales volume was in gaming/retail. You don't stay in business selling your products at a loss, things might have been skinny with 65nm but a fundamental of business is sell it for more than you paid for it. Even with higher margins in the professional market I highly doubt there is enough volume even with higher margins in that market alone to support Nvidia at its current size.

[highoctane]

Just some more cannon fodder:


http://vr-zone.com/forums/519373/gf1...s-coming-.html

[Manicdan]

Just some more cannon fodder:


http://vr-zone.com/forums/519373/gf1...s-coming-.html

lolz at them, they seem overly excited like its really news. fermi better be the fastest gpu, its much larger than ATI's

and for the duel gpu thing, im still wondering how badly they will have to downclock it to keep under 300W. Perf per watt will be extremely interesting this spring/summer

[Tim]

Just some more cannon fodder:


http://vr-zone.com/forums/519373/gf1...s-coming-.html

no spesific launch date

Ack that suck no spesific launch date.

[highoctane]

Well from what I can get from the translation it will be at CES 2010 but who knows...

http://translate.google.com/translat...tm&sl=tr&tl=en