The Fermi Thread - Part 3

[flippin_waffles]

Really, just stop trolling. It was you who started the argument, bye. /Ignore

LOL, yeah just plug your ears and yell, "lalalala I can't hear you!" really loud. I'm just pointing out how you completely ignore anything you don't want to hear/read, and yell to the top of your lungs when it's something you do want to hear/read. I got no problem with somebody who discounts a source, but when you go back to the same well when it's something you agree with, well that's just plain ridiculous. But whatever floats your boat, carry on!!

[Luka_Aveiro]

gtx 480 clocks are great, like i said! tdp is good, its far away from 298!

If this is true, you will have my respect, my friend.

I hope it is.

[Picao84]

gtx 480 clocks are great, like i said! tdp is good, its far away from 298!

Well, far away can be 275W like Charlie said

[zalbard]

Well, far away can be 275W like Charlie said

It's obvious that it's not going to be power efficient by now. All I am hoping to see is very good performance and pricing.

[Dimitriman]

gtx 480 clocks are great, like i said! tdp is good, its far away from 298!

I might get flamed, and I don't mean to insult by any means, but from what I saw...

Charlie 10x0 vs your claims when it comes to accuracy. (looking at all latest leaks, which can't be far from the truth)

So unless you have more then words, gonna be hard to believe that statement. From what I see, you're the only person around trying to convince us that Fermi is perfect, clocks super high, runs super cool, and has plenty availability at launch.

Flame away

[trinibwoy]

Come on trinib'

Of course that clocks, 700+ was originaly aimed by nVidia if I remember well.

Hey, it wouldn't be the worst thing said about Fermi I dont know if base clock was targeted at 700, but the scheduler/TMU clocks were probably targeted at ~750 for a shader clock of 1500.

It's obvious that it's not going to be power efficient by now. All I am hoping to see is very good performance and pricing.

How can it not be power efficient yet have very good performance at the same time? Those two things are mutually exclusive.

[annihilat0r]

I might get flamed, but from what I saw...

Charlie 10x0 vs your claims when it comes to accuracy.

So unless you have more then words, gonna be hard to believe that statement.

Flame away

A GTX 480 that uses 280W, is 5% faster than HD5870, is 550mm2, is 448 SP's (later he reverted to 512), idles at 70C, has around 600mhz clockspeeds.

Will be paper launch, 3.000 to 8.000 cards will be available at launch, you are not going to be able to buy one. Wide availability will be much, much later (Q4 2010. He said Q4 2011 but he was being sarcastic.) The chip is hot, huge, broken and unmanufacturable.

Those are Charlie's claims. Let's wait some time and see how many of those claims were correct. As far as I can remember, the only thing that he was correct about Fermi was the launch date up to now.

[Dimitriman]

A GTX 480 that uses 280W, is 5% faster than HD5870, is 550mm2, is 448 SP's (later he reverted to 512), idles at 70C, has around 600mhz clockspeeds.

Will be paper launch, 3.000 to 8.000 cards will be available at launch, you are not going to be able to buy one. Wide availability will be much, much later (Q4 2010. He said Q4 2011 but he was being sarcastic.) The chip is hot, huge, broken and unmanufacturable.

Those are Charlie's claims. Let's wait some time and see how many of those claims were correct. As far as I can remember, the only thing that he was correct about Fermi was the launch date up to now.

He claimed 480 would have around 480 cores not, 448, that's 470. He mentioned it would not be much faster than 5870 but I don't recall a 5% figure anywhere. If 480 does have 512 cores, and the 470 has 225W, can we really expect less than 275W for 480?
As for the "broken and unavailable" claim, when was that written? more than 1 month ago? - yet it's still nearly 1 month to go before Fermi even launches. I suppose if Fermi was peanuts to manufacture, we would have it already. His claim for Q4 availability was for mid range/low end Fermis, not 480/470

[zalbard]

How can it not be power efficient yet have very good performance at the same time?

5870 already has better performance/watt ratio, even if GTX480 manages to fit into 250W.
I don't see the issue.

[annihilat0r]

He claimed 480 would have around 480 cores not, 448, that's 470. He mentioned it would not be much faster than 5870 but I don't recall a 5% figure anywhere. If 480 does have 512 cores, and the 470 has 225W, can we really expect less than 275W for 480?
As for the "broken and unavailable" claim, when was that written? more than 1 month ago? - yet it's still nearly 1 month to go before Fermi even launches. I suppose if Fermi was peanuts to manufacture, we would have it already. His claim for Q4 availability was for mid range/low end Fermis, not 480/470

He claimed 448 for the top end Fermi, as you can read from this article:

http://www.semiaccurate.com/2009/12/...-fermi-448sps/

He also claimed a %5 number in this article. He also says that the top Fermi will be 512SP, and idles at 70C on idle clocks with fan set to 70%. Also he says that Fermi clocks will be 600.

http://www.semiaccurate.com/2010/02/...gtx480-scores/

His claim of Q4 "real" availability is for GF100, which is 470 and 480:

http://www.semiaccurate.com/2010/03/...pes-out-gf108/

"A more realistic view is late July or early August before you will be able to get bottom feeder DX11 parts from Nvidia, but that will be much sooner than real GF100 quantities will arrive. At this rate, Nvidia will have a full line of parts, in quantity, but not until Q4".

[Chumbucket843]

LOL, yeah just plug your ears and yell, "lalalala I can't hear you!" really loud. I'm just pointing out how you completely ignore anything you don't want to hear/read, and yell to the top of your lungs when it's something you do want to hear/read. I got no problem with somebody who discounts a source, but when you go back to the same well when it's something you agree with, well that's just plain ridiculous. But whatever floats your boat, carry on!!

oh please, like we are going to trust charlie just because he gets anything negative right. he is biased and will never have a good thing to say about nvidia's upcoming products, even successful ones like g80. its kind of sad because shilling a company is not what journalism is about.

[Dimitriman]

He claimed 448 for the top end Fermi, as you can read from this article:

http://www.semiaccurate.com/2009/12/...-fermi-448sps/

This is an obvious misinterpretation of a slide referring to Tesla. That is his own fault and indeed a mistake, but when it comes to 470, that info seems nearly spot on wouldn't you say?

He also claimed a %5 number in this article. He also says that the top Fermi will be 512SP, and idles at 70C on idle clocks with fan set to 70%. Also he says that Fermi clocks will be 600.

http://www.semiaccurate.com/2010/02/...gtx480-scores/

written in February 20, 2010, early tests, early drivers, obviously performance will be better. And quite plausible that a chip with 512 shaders will not pull much higher clocks than the 625 Mhz from its 470 sibling, not unless it will push 300W.

His claim of Q4 "real" availability is for GF100, which is 470 and 480:

http://www.semiaccurate.com/2010/03/...pes-out-gf108/

"A more realistic view is late July or early August before you will be able to get bottom feeder DX11 parts from Nvidia, but that will be much sooner than real GF100 quantities will arrive. At this rate, Nvidia will have a full line of parts, in quantity, but not until Q4".

He claims you will be able to GET midrange Fermi in summer, but you will be able to get GF100 before that, just not in "quantities" which in Charlie language, means full availability.
His statement claims: low availability of GF100 until Q4, launch of midrange Fermi in summer. Full availability of both only in Q4.

[highoctane]

This is not pointed to you personally, but to the whole buggus in question.

"Windows 32bit" will only see and utilize no more than 4 gig (just a little under), but all the other programs installed will use eveything available. So when playing a game, it is not windows 32 managing the memory, but the game itself and the hardware in question (which is sort of a program on its own).

So basically you guys are half right and confuse.

Everything will be in use, but windows 32 will only use what it can, 2 different thing.

AFAIK all AIC's that have physical ram installed have to reserve address space with the OS and with 32bit windows the total is 4gb shared address space. There are supposedly specific apps that utilize PAE to utilize more then 4gb but not anything the avg consumer is probably using.

Don't take it the wrong way but I'm not really sure what position you are in to call others half right and confused considering this issue is clearly documented....

http://support.microsoft.com/kb/929605/

Various devices in a typical computer require memory-mapped access. This is known as memory-mapped I/O (MMIO). For the MMIO space to be available to 32-bit operating systems, the MMIO space must reside within the first 4 GB of address space.

For example, if you have a video card that has 256 MB of onboard memory, that memory must be mapped within the first 4 GB of address space. If 4 GB of system memory is already installed, part of that address space must be reserved by the graphics memory mapping. Graphics memory mapping overwrites a part of the system memory. These conditions reduce the total amount of system memory that is available to the operating system.

The reduction in available system memory depends on the devices that are installed in the computer. However, to avoid potential driver compatibility issues, the 32-bit versions of Windows Vista limit the total available memory to 3.12 GB. See the "More information" section for information about potential driver compatibility issues.

If a computer has many installed devices, the available memory may be reduced to 3 GB or less. However, the maximum memory available in 32-bit versions of Windows Vista is typically 3.12 GB.

[annihilat0r]

This is an obvious misinterpretation of a slide referring to 470. That is his own fault and indeed a mistake, but when it comes to 470, that info seems nearly spot on wouldn't you say?

Yeah it's pretty much spot on, except that the lower high end Fermi would have 448SP was pretty much common knowledge anyway, and after all he was mistaken.

written in February 20, 2010, early tests, early drivers, obviously performance will be better. And quite plausible that a chip with 512 shaders will not pull much higher clocks than the 625 Mhz from its 470 sibling, not unless it will push 300W.

Meh. So we can talk about numbers for an unfinished product like they will belong to the finished product? That way several months ago he could have written an article and say "wow there is no Fermi at the moment so there will never be any Fermi's". This is now how journalism works.

He claims you will be able to GET midrange Fermi in summer, but you will be able to get GF100 before that, just not in "quantities" which in Charlie language, means full availability.
His statement claims: low availability of GF100 until Q4, launch of midrange Fermi in summer. Full availability of both only in Q4.

His statement clearly says that this summer time will be "much sooner" than real GF100 availability, which means pretty much what I have said before.

[NKrader]

i heard from this guy they were gona be like OVER 9000$

[annihilat0r]

i heard from this guy they were gona be like OVER 9000$

I can't even imagine their power level, then!!!

[DosDuoNo]

This is not pointed to you personally, but to the whole buggus in question.

"Windows 32bit" will only see and utilize no more than 4 gig (just a little under), but all the other programs installed will use eveything available. So when playing a game, it is not windows 32 managing the memory, but the game itself and the hardware in question (which is sort of a program on its own).

So basically you guys are half right and confuse.

Everything will be in use, but windows 32 will only use what it can, 2 different thing.

this is wrong, and shows a basic misunderstanding and common wrong about 32 bit OS systems. it also shows you dont really understand 32 bit address at all.

AFAIK all AIC's that have physical ram installed have to reserve address space with the OS and with 32bit windows the total is 4gb shared address space. There are supposedly specific apps that utilize PAE to utilize more then 4gb but not anything the avg consumer is probably using.

Don't take it the wrong way but I'm not really sure what position you are in to call others half right and confused considering this issue is clearly documented....

http://support.microsoft.com/kb/929605/

this is of course correct, but i will put it simply for you

32 bit OS can only address just under 4 gig, regardless of whats sitting on top, ie: games programs etc.

now heres the basic math

OS system total availible address space - just under 4 gig

system memory - 2 gig
Other devices (ie, Device manager list) - 0.8 gig
Soundblaster Sound Card - 64 MB
Video card - 512 MB
Total = 3.31 gig

now everthing is great here, no problems, but lets change a couple of things,

system memory - 3 gig
Other devices (ie, Device manager list) - 0.8 gig
Soundblaster Sound Card - 64 MB
Video card -1 gig MB
Total = 4.81 gig

housten we have a problem , windows will now 'reduce' memory address alocation. and Memory will disapear, because? the grand total cannot be more that just under 4 gig.

it really is that simple

[Leeghoofd]

so conclusion then : use a 64bit OS.. and ditch XP... I didn't know 32bit OS also took into account videomemory...

[MrDiaz]

9 days left for reviews... feels like an eternity

[Sn0wm@n]

I might get flamed, and I don't mean to insult by any means, but from what I saw...

Charlie 10x0 vs your claims when it comes to accuracy. (looking at all latest leaks, which can't be far from the truth)

So unless you have more then words, gonna be hard to believe that statement. From what I see, you're the only person around trying to convince us that Fermi is perfect, clocks super high, runs super cool, and has plenty availability at launch.

Flame away

i endorse this statement!!!!


ohh btw i have a flame suit if you want one

oh please, like we are going to trust charlie just because he gets anything negative right. he is biased and will never have a good thing to say about nvidia's upcoming products, even successful ones like g80. its kind of sad because shilling a company is not what journalism is about.

yes it is ... journalism these days is all about ad revenu .... even the ones in the papers need to please their boss with their article and their boss need add revenue to live .... so if charlie does it to please someone so he could get more cash then he wins ....

[DosDuoNo]

so conclusion then : use a 64bit OS.. and ditch XP... I didn't know 32bit OS also took into account videomemory...

yeah the minute 4 gig of memory became the 'norm' of 2 gig this problem arose for a lot of people. because windows will give priority to 'Devices' including video cards, most people ended up with just over 3 gig system memory. hence why a lot of OEM's still ship systems with 3 gig of memory. as anything more would be a waste

[flippin_waffles]

Yeah it's pretty much spot on, except that the lower high end Fermi would have 448SP was pretty much common knowledge anyway, and after all he was mistaken.



Meh. So we can talk about numbers for an unfinished product like they will belong to the finished product? That way several months ago he could have written an article and say "wow there is no Fermi at the moment so there will never be any Fermi's". This is now how journalism works.




His statement clearly says that this summer time will be "much sooner" than real GF100 availability, which means pretty much what I have said before.

I guess we could use nv's numbers if you prefer. Availability in august, nope november, nope december, nope january, nope march, nope april.
80% faster than 5870, nope 50% nope 30% nope depends on 'application'. dx11 doesn't matter, nope it does matter.
fermi will allow home built supercomputers, nope we'll castrate DP to 1/8 full hardware specs.
fermi will be in top supercomputers, ORNL says not anymore. maybe later. more?

[Tim]

Can anybody actually show me the source/rumour of where it says that the GTX 480 will have 448SP?

Anything I have seen so far are rumours, still it's already being preached as the truth.

[DosDuoNo]

Give a Monkey enough balls, it will eventually hit a goal, it doesnt mean it can play football

and to be fair charlie has had enough balls over the last 8 months.

i think we also need to seperate rumour from Fact from the Nvidia information, some of what you have there waffles was rumour, this is not Nvidias fault

[annihilat0r]

I guess we could use nv's numbers if you prefer. Availability in august, nope november, nope december, nope january, nope march, nope april.
80% faster than 5870, nope 50% nope 30% nope depends on 'application'. dx11 doesn't matter, nope it does matter.
fermi will allow home built supercomputers, nope we'll castrate DP to 1/8 full hardware specs.
fermi will be in top supercomputers, ORNL says not anymore. maybe later. more?

what the hell do Nvidia's statements have to do with what I said? I just laid out Charlie's claims, and to you guys who say he's a reliable source, let's wait and see how true his claims were.

You guys are just so used to seeing things from "sides" like political parties do. If I say something bad about Charlie this must mean that I have said something good about Nvidia which makes you say bad things about Nvidia so that it will counter the bad things I said for Charlie. ??