The Fermi Thread - Part 3

[SoF]

does anybody already has any watercooled impressions with these cards (avoiding the bad f-word - beware of dave ). It's a last hope for me that there is a way to have fun with these cards - if watercooling works good and you can overclock it a bit with it, that might be a last chance. for all-day it was out for me when I saw the power consumption - I'll prefer to stick with my 285 GTX system which has in total 275W peak consumption (entire system, sound and monitor while folding). I would double that up easily with .... -

[QuadDamage]

I see a lot of unhappy folk mostly ATi boys/nVidia haters. I've got the card and i can tell you it doesn't get that noisy unless you run Furmark (90C) but under normal load ie while playing games the cars usually hovers between 70-75C and the fan is defo a lot quieter than the 4870X2/5970 or even the 4890.

My reference card can do 825/2000 easy on stock cooling and vgpu without any extra fan blowing over it in my Cosmos S.

default 06 run at 825/2000.

[kaktus1907]

I see a lot of unhappy folk mostly ATi boys/nVidia haters. I've got the card and i can tell you it doesn't get that noisy unless you run Furmark (90C) but under normal load ie while playing games the cars usually hovers between 70-75C and the fan is defo a lot quieter than the 4870X2/5970 or even the 4890.

My reference card can do 825/2000 easy on stock cooling and vgpu without any extra fan blowing over it in my Cosmos S.

default 06 run at 825/2000.

well said.. can you try to run higher clocks(i heard it has 0.4 ns chips same as 5870s) for memories while gpu clocks stays stock speed ?

[QuadDamage]

the card crashes at 875Mhz almost instantly in '06, at 850Mhz it crashes in the second test but it should do 835/840Mhz core. I haven't pushed the memory yet. I'll do that now.

[Praz]

A shame that for benchmarking it is really no faster then Crossfired 4870s.

I see a lot of unhappy folk mostly ATi boys/nVidia haters. I've got the card and i can tell you it doesn't get that noisy unless you run Furmark (90C) but under normal load ie while playing games the cars usually hovers between 70-75C and the fan is defo a lot quieter than the 4870X2/5970 or even the 4890.

My reference card can do 825/2000 easy on stock cooling and vgpu without any extra fan blowing over it in my Cosmos S.

default 06 run at 825/2000.

[japamd]

I see a lot of unhappy folk mostly ATi boys/nVidia haters. I've got the card and i can tell you it doesn't get that noisy unless you run Furmark (90C) but under normal load ie while playing games the cars usually hovers between 70-75C and the fan is defo a lot quieter than the 4870X2/5970 or even the 4890.

My reference card can do 825/2000 easy on stock cooling and vgpu without any extra fan blowing over it in my Cosmos S.

default 06 run at 825/2000.

Using those clocks please run Vantage Extreme or Crysis Warhead AA4x at high resolution.

If you can measure power consumption and temperatures that would be nice tho.

[p2501]

the card crashes at 875Mhz almost instantly in '06, at 850Mhz it crashes in the second test but it should do 835/840Mhz core. I haven't pushed the memory yet. I'll do that now.

What brand did you get?

[QuadDamage]

nVidia reference board.

EDIT : Here's the stock untouched GTX480 BIOS. Nibitor doesn't seem to work but maybe someone can figure out the way to mod it.

http://www.megaupload.com/?d=1SIHTCTP

edit2:

Vantage default at 825/2000 with the highest recorded temps.

[zalbard]

IMO Fermi's scalability potential is upwards, not downwards.

Should be decent both ways. And we won't see an example of scalability upwards for a looooong time...

I thought the best review was Anand's.

Same. No offence to anybody else.

Fermi reminds me of something, the reborn of FX 5800

Hahaha! I've never seen that, hilarious!

[japamd]

nVidia reference board.

EDIT : Here's the stock untouched GTX480 BIOS. Nibitor doesn't seem to work but maybe someone can figure out the way to mod it.

http://www.megaupload.com/?d=1SIHTCTP

edit2:

Vantage default at 825/2000 with the highest recorded temps.

Nice! Cheers!

Another thing, can you post a screenshot of device manager showing which IRQ number your GTX 480 is using, like the one below:



I'm curious if finally NVIDIA programmed the driver to use Message-signaling interrupts instead of pin-based interrupts.

[AKM]

Grillforce GTX480 FTW!

[Sn0wm@n]

epic post .... keep those fermi joke comming i love me to death

[GoldenTiger]

[*IMG]http://i42.tinypic.com/15ez287.jpg[/IMG]

Awesome...

I'm gonna love mine! Looks tasty and well-done!! (ooh bad pun!)

[ajaidev]

[tajoh111]

This is nice manipulation with facts, you know? If you say Fermi is completely new, than you must say Evergreen is completely new architeture, too. AMD did the major changes with RV7x0 generation (memory hub and other stuff), NVIDIA with GF10x (cache and other stuff).

But the truth is, the shaders and other parts which have direct impact on 3D performance are the same till R600/G80. So this talk about "hard to make drivers especially for GF100" is total nonsense.

If they want to make special driver, nobody is holding 'em. It is so easy to name the DLL libraries differently and copy the different ones if GF100 is detected, isn't it?

Evergreen is completely new? You can't be serious, even AMD says their completely new generation of cards comes after r8xx. I don't think anyone here will agree with you that the changes from g200 to fermi are vastly greater than the changes from r7xx to r8xx.

Evergreen is more of a testiment to the scalability of r6xx architecture. They use similar ways to obtain performance, hence, why AMD dropped support for new driver of anything pre r600 last year. The recent improvements that came with drivers 10.3 also improved r7xx performance as well show this.

Fermi shaders are far different than g80 tech. Its their first fold into directx11 from mostly a directx 10. The jump from direct x 10.1 to 11 is alot less steep compared to 10 to 11.

So you are saying Nvidia had working GF100 silicon in May? I'm a Charlie supporter when his sources/rumors seem to support info I heard from a different source. Nothing more or less. If you want to look back, I actually disagreed with Charlie quite a few times over the last year or so.

You also missed what I am trying to say... Evergreen is a new architecture, in the sense of drivers, the same Fermi is. Did AMD/ATi have working silicon before their Sept launch, more than likely, did Nvidia have working silicon before the end of 2009, more than likely, I never said any different. The point is, to say that AMD/ATi should have already optimized the performance from Evergreen before it's launch, or shortly therefore after, is wrong. It will take months to correct and optimize, sort of like how it took +6 months for G80 to get stable drivers in certain games(though not the same).

I think the SLI driver performance shows how much time Nvidia really had to prepare GF100. The performance scaling is quite good and somewhat surprising, though that might be only my opinion.

So Nvidia dropping supporting for pre-G200 is of no consequence to you? We will see...

Man you didn't get my point and your original point did not address my first post at all, hence my explanation.

I will make it as simple as possible because we are honestly agreeing on the same thing. You are saying that it can take alot of work to get optimizations and performance out of a new card and it is not easy. Right? R8xx has substantial changes that might take months to take advantage of, hence the recent driver improvements which bring up performance.

I am saying the same thing. It takes time to bring out performance from new hardware.

So I will makes make this as clear and as streamlined as possible.

R8xx is a new card with some architecture changes to make it directx 11. The most important part of directx 11, a tesselation engine part of their previous architecture was already there. This is not a new architecture however, something AMD has already admitted.

But other changes that make it directx 11 requires a few changes, hence, the added work needed for driver to allow the card to perform at optimum levels. Hence the 6 months of work that has payed off to get performance to current levels.

Lets look at NV situation.

gf100 is completely new, working silicon was made in late 2009 and has vast changes over g80-g200 based architecture. This is officially NV new architecture.

Making drivers that perform well with this new architecture requires working silicon, hence NV has only had 4 month to work on this project.

The starting point to get to that final point(optimized drivers) is alot farther back because the current drivers they have are for previous architecture that was for something alot different. As a result, it will take more time and effort compared to the AMD situation.

This leads to my point towards adaivjev
It will take longer than 4 months to get good drivers out of fermi, if it takes 6 months to get optimized drivers from evergreen.

Especially when you take into consider the following points.
Fermi is a totally new architecture, evergreen is not.
AMD likely had working silicon longer than NV for their latest and greatest.
Driver optimizations towards fermi are likely to be more dentrimental to pre fermi hardware, which NV can't ignore because it is the vast majoritity of its current base.

Additionally to clarify my objective and alliances, I do not plan on getting a fermi based card at the present time. To justify the power consumption, they need atleast 15 percent more performance than what we currently see. This might come with later drivers, but I do not make purchases based on potential future driver performance. However the likelyhood of these improvements are high given the newness of fermi and the time given to prepare the current drivers.

I did not get a 5xxx card because 40-50 percent improvement over the previous generation is not enough to justify an upgrade in cards just over a year old, especially since I don't game that much. Additionally, I am not a graphic snob with a superiority complex over consoles. With my current solution, I would need to purchase dual 5870s as even a 5970 really doesn't add much to the performance equation(although likely more consistency) which doesn't make sense considering the cost and how much I game(honestly no one games enough to justify 900(even more because i watercool) dollars on graphic card every years(good games that have great graphics just don't appear often enough).

Fermi with say an extra 20% performance along with its current lead would mean a card that is 80 percent faster than previous gen. Something that might justify an upgrade, but at the present time, no, not even close, especially with the power consumption.

[Chumbucket843]

anyone else find this odd that R600 got beaten down by g80 ... yet a couple years down the line and its giving a fair fight to nvidia's next gen arch

arch v. arch? not a valid comparison. check out the gtx480 running tessellation 3x faster. ATi got beat at their own demo.http://www.hardware.fr/articles/787-...e-gtx-480.html

the loss of power efficiency is from process yields. they cant bin chips for the 480 at the clocks they would want and be able to launch at a reasonable date or price.

[***Deimos***]

Computer history is whimsical and key pivotal moments seem to happen in August.
2002: 9700pro came out. $399. Same 100fps in Quake3 as GF4. Great AA performance. Nothing special. DX9 wasn't released yet. XP was about to ship.
2 years later, the elegance of the 9700pro bore fruit in Far Cry. ArtX, creators of GameCube, new design evolved into X800, X1800s and XBOX.

But the most key pivotal time in computer history I can imagine in MY lifetime - 1995.
- Win95 was launched catapulting Microsoft from millions into billions.
- IE and Netscape Navigator.. the web had begun.
- Playstation! No more blocky scrollers. 3D! Gaming opened to those beyond 10yr olds.
- Voodoo and Quake were being polished off to launch in 1996.
- and Pentium Pro was launched!

Wait, "whats that?" you ask. Surely it can't be as big a deal as Win95!?
It was power hungry. Huge die made on 500/350nm with novel offdie cache. Same 200Mhz max clock as Pentium.
And Win95/apps sometimes ran slower than on Pentium.

Pentium Pro
First x86 server processor. First OOO architecture. RISC core running CISC, with pipelined FPU - starting idea for SSE. Intel's 4th 32bit processor, but first actually optimized for 32bit performance. And it was the seed that evolved into famous Intel cash-cows like Pentium2, Pentium3, PentiumM and 11 years later, Core2.

IMHO, Fermi is NOT the great leap G80 was. Unlike the 9700pro and Pentium pro, Fermi is already MUCH faster from day1 and will only get faster. And DX11 games show amazing performance. All the unified cache, C++, GPGU etc was a bold leap forward by nVidia. They could have done like 5870, and just doubled up the G200 and saved themselves all the headaches.

In the near future, DX11 games and GPGPU application will reveal and demonstrate more of Fermi's design advantages. And ofcourse it will be basis for future designs. I wouldn't be surprised if 5 years from now, Unreal5 or Doom6 use OpenCL/C++ and requires Fermi architecture, just like Vista/Win7 require DX9 (9700pro) for Aero.

Although, A++ for architecture design, I still wont buy the product (GTX480). I just dont have $$$ and haven't even finished playing through pile of DX10 games. But, by the end of summer, kinks and availability should be fixed, and who knows, maybe even "200W low power" (as ironic as that sounds) versions. Maybe I'll make the Win7 plunge and pick one up. 5870 better for now, but Fermi definetly more future proof.

[Mk]

this thread now is called "bash Nvidia thread" there is nothing useful here any more.... Fermi is not news any more..... and all the things about GTX 470 & GTX 480 should be discussed in Nvidia Forum.
I hope mods close this thread.

[ajaidev]

5870 better for now, but Fermi definetly more future proof.

Agreed the fermi arc is very good but Nvidia's released card GTX 480/470 not so much.

Nvidia needs a shrink like GTX 280 -> GTX 285 but also need full 512 SP's. A GTX 485 with 512 SP's will be much better than anything ATi has with a single GPU. But ATi is suppose to have a refresh soon and also will have a next gen part coming soon. Nvidia took 6-7 months for the GTX 280 to GTX 285 convert i hope it does not take that long.

Fermi as a arc is really strong but Nvidia did not execute it well. I hope Nvidia goes with GF

[***Deimos***]

I want to put an end to the "dead horse beating".

R600 vs Fermi
both: hot, new DX, late
diff: Fermi is faster, not broken, amazing DX11 features/performance for future games, good MSRP (markup another story).

fx vs Fermi
both: hot, new DX
diff: Fermi is faster, not broken, no questionable "optimization" (yet) and doesn't sound like turbine

GeForce vs Fermi
both: hot, new DX, foundation for new technology lineup.

Basically,
ALL new DX cards are always hotter. That's the trend. Get over it.
Even RV870 was gotter than RV770. Its not somehow a new issue only Fermi has.

Somehow, in the rosy glamourous faded memories, many people forget 9700pro was ridiculed for requiring extra power connector with ATI's classic Win2000 driver issues and at best performance on par with GF4. Ofcourse a year later, with issues fixed and DX9 launced, everybody was buying one.

Agreed the fermi arc is very good but Nvidia's released card GTX 480/470 not so much.

Nvidia needs a shrink like GTX 280 -> GTX 285 but also need full 512 SP's. A GTX 485 with 512 SP's will be much better than anything ATi has with a single GPU. But ATi is suppose to have a refresh soon and also will have a next gen part coming soon. Nvidia took 6-7 months for the GTX 280 to GTX 285 convert i hope it does not take that long.

Fermi as a arc is really strong but Nvidia did not execute it well. I hope Nvidia goes with GF

Die shrink will be too late. 512 SP not important. Its like 5% difference.

There is 1, and only 1, critical "to-do".
DEVELOPERS.

If Fermi can get +20% in BattleForge, an "AMD game", image the performance if it was TWIMTBP!
Developer support is more critical than ever. All the consoles are DX9. There's little motivation to do "hard work" to make DX11 games.
It doesnt matter if its 480 SP or 512 SP, if nVidia isn't there to baby-hold and show 1-2-3 how to use it. Because if it never gets used, there's no benefit of having it.

[ajaidev]

Die shrink will be too late. 512 SP not important. Its like 5% difference.

There is 1, and only 1, critical "to-do".
DEVELOPERS.

If Fermi can get +20% in BattleForge, an "AMD game", image the performance if it was TWIMTBP!
Developer support is more critical than ever. All the consoles are DX9. There's little motivation to do "hard work" to make DX11 games.
It doesnt matter if its 480 SP or 512 SP, if nVidia isn't there to baby-hold and show 1-2-3 how to use it. Because if it never gets used, there's no benefit of having it.

More like 6.25% and BattleForge was not bad even under GTX 285. I would imagine TWIMTBP games would be reallly goood "Tessellation with only extreme settings"

Developer support is needed yes but DX11 is said to be simpler to work with than DX9 and hopefully new consoles will come soon.

EDIT: The fact that Nvidia's best card has only 480 SP's enabled from 512 SP's its a disabled card aka broken card.

[LiquidReactor]

IMO Fermi's scalability potential is upwards, not downwards.

We'll see soon, but considering the current ROPs are pretty inefficient (as you can see here Fermi has less fillrate than GTX285), don't expect the GTS 450 to handle 2560, and even in 1920 you might see bad omens.

Are the Fermi numbers in those tests result of bad drivers or are they architecture related?

[pentium777]

Enough of all the childish picture posts. Fermi is still news and is still waiting for the hard launch April 12th so the thread is still useful.

Check out overclock results 1165 core clock

http://www.legitreviews.com/news/7697/

[Macadamia]

Are the Fermi numbers in those tests result of bad drivers or are they architecture related?

I don't think you'd screw up something that fundamental such as fillrate with release drivers.

What drivers do optimize are probably shader code and texture formats, perhaps a bit of how games render here and there. Feature tests however should be extremely close to practical values from day 1.

[Chumbucket843]

More like 6.25% and BattleForge was not bad even under GTX 285. I would imagine TWIMTBP games would be reallly goood "Tessellation with only extreme settings"

Developer support is needed yes but DX11 is said to be simpler to work with than DX9 and hopefully new consoles will come soon.

EDIT: The fact that Nvidia's best card has only 480 SP's enabled from 512 SP's its a disabled card aka broken card.

well if you consider the 480 broken then an unknown number of 5870's and 4870's are broken. nvidia uses course grained redundancy so they turn of some sp's if they dont work. ATi adds redundant circuits to sp's so they can work if the other circuits dont. each has there pros and cons.