The Fermi Thread - Part 3

[SKYMTL]

How can you resist? So the Fermi is just sitting in a box or on the bench? I would not be able to control myself. Nice to hear you're taking the time to bench with the newest drivers possible.

Literally haven't opened the box. For all I know NVIDIA sent me an Optimus Notebook.

Honestly though, I can't get excited simply because there is too much to do. It's always best to benchmark the new stuff last so you can put its performance into context right before you have to write the article itself.

I usually take the "slow and steady" way when it comes to benchmarking to make sure I can catch any issues. I did however have a frantic search last night for my misplaced HD 5970.

[eric66]

I run 2560x1600, so to me it is extremely important that a high-end card can perform at high-res. As far as the avg's being "low", the minimums are important to me, in fact what drove me to buy a 5870 over my then-current GTX 280 even though the averages weren't as spectacularly high like the minimum gains were. As SKYMTL said of course, can't really judge a ton off of the FarCry 2 bench there without context, but if minimum gains like that turn out to be the norm in real benches across review sites, then it would be a smashing success of a card: I don't care if it can spike to 150fps while maintaining a 60 average... I care if it can stay above 45-50 while maintaining a 65-70 average, even if that means the max is only around 100 then . Max framerates and by extension averages can be easily skewed with that.

what i am saying is that they tried to reach limits of 5xxx cards if they benched at 2560res

[ToxAvenger]

Literally haven't opened the box. For all I know NVIDIA sent me an Optimus Notebook.

HAHA. But you must have nerves of steel to resist.

Honestly though, I can't get excited simply because there is too much to do. It's always best to benchmark the new stuff last so you can put its performance into context right before you have to write the article itself.

Makes sense actually.

I usually take the "slow and steady" way when it comes to benchmarking to make sure I can catch any issues. I did however have a frantic search last night for my misplaced HD 5970.

Is this a hint about the 480's performance?

[azza21]

Literally haven't opened the box. For all I know NVIDIA sent me an Optimus Notebook.

Honestly though, I can't get excited simply because there is too much to do. It's always best to benchmark the new stuff last so you can put its performance into context right before you have to write the article itself.

I usually take the "slow and steady" way when it comes to benchmarking to make sure I can catch any issues. I did however have a frantic search last night for my misplaced HD 5970.

So you post on the fermi thread nearly everyday, clearly taking a keen interest on all things Fermi and whatnot yet you MAY have a product that could be a 480 but you haven't checked it out as off yet as you like to go at a slow and steady pace with the benchmarking and you got too much to do...

Nice one

[SKYMTL]

So you post on the fermi thread nearly everyday, clearly taking a keen interest on all things Fermi and whatnot yet you MAY have a product that could be a 480 but you haven't checked it out as off yet as you like to go at a slow and steady pace with the benchmarking and you got too much to do...

Nice one

I usually post here when there is a benchmark running in the background or driver installing.

[highoctane]

what i am saying is that they tried to reach limits of 5xxx cards if they benched at 2560res

So all previous benchmarking ever done at 2560 was simply done to reach the limits of the 5xxx cards all along, even before 5xxx cards where out, whatever...

2560 has been being benchmarked for a long while now just zeroing in on that like its something "new" fits your obstinate anti nvidia agenda I guess.

[Mungri]

Has this been posted yet? Appologies if it has:

http://www.tomshardware.com/news/gef...-470,9926.html

GeForce GTX 480 : 480 SP, 700/1401/1848MHz core/shader/mem, 384-bit, 1536MB, 295W TDP, US$499

GeForce GTX 470 : 448 SP, 607/1215/1674MHz core/shader/mem, 320-bit, 1280MB, 225W TDP, US$349
Internal benchmarks reveal that GeForce GTX 470 is some 5-10% faster than Radeon HD 5850 and similiar for GeForce GTX 480 over the Radeon HD 5870. Interestingly, the TDP of GeForce GTX 480 is almost similar to Radeon HD 5970 which is a dual GPU card. Interestingly, our sources revealed that there are indeed plans for dual Fermi cards and the TDP of the card is probably gonna be mind blowing.

http://translate.google.gr/translate...241285024.html

Communication GTX470 card official recommended price of £ 299

I think it is looking like it will be £299 for the GTX 470, and £449 for the GTX 480, but that is just speculation.

Looks like its already been posted whoops.

[eric66]

So all previous benchmarking ever done at 2560 was simply done to reach the limits of the 5xxx cards all along, even before 5xxx cards where out, whatever...

2560 has been being benchmarked for a long while now just zeroing in on that like its something "new" fits your obstinate anti nvidia agenda I guess.

you ll see what i am trying to say when proper benchs will roll out mark my words i say

[Solus Corvus]

Well, even on the the DX11 game GTX480 has a leading of close to 20%.

HOWEVER, for NVIDIA to be performing that well in DiRT 2 (a game developed with the close participation of AMD) is a huge step in the right direction.

Good showing in DiRT2, although not enough info on the settings, but 20% up on an ATI happy game is not bad for early drivers.

Dirt2 is also a DX11 title with tessellation - something we have been told repeatedly that Fermi is particularly good at.

It looks like my 30% faster then 5870 guess may have been about right. It's something easily deduced from specs we knew ~5 months ago. It looks like the people expecting it to beat a 5970 were wrong and the people hoping it wouldn't significantly faster then a 5870 were wrong too. But I guess it's hard for those with a deep emotional investment in one outcome or another to reasonably moderate their expectations.

[Vardant]

If a fully fledged Fermi card hit the target clocks it would probably match or even beat the HD 5970, so I don't think anyone was wrong.

Also, you don't really need to match the performance of a dual GPU card, because even slightly slower single GPU card is always better.

[eric66]

lol 5970 people are getting high here ok keep on dreaming i say

[ElSel10]

Honestly, it isn't even in my system yet. I am trying to re-benchmark as many cards as possible with the new 10.3a and 197-series drivers before popping the new stuff in.

Nice that is a good idea! 197.13 or 197.15?

470's TDP is 216W, the information in the NV official PDF.

Is that PDF available anywhere?

Before anyone points out no Nvidia did not have true SSAA and no nhancer's SSAA mode was not a proper one.

Really and why is this?

[ToxAvenger]

It looks like my 30% faster then 5870 guess may have been about right.

Which card are we talking about here though? The 480SP or 512SP version? From what I've read there seems to be a lot of confusion about the actual specs of the GTX 480. If it's the 480SP version that is running ~15% up on the 5870, I'd say Nvidia is in good shape depending on power draw.

[eric66]

what is power draw of 480 ?

[Solus Corvus]

If a fully fledged Fermi card hit the target clocks it would probably match or even beat the HD 5970, so I don't think anyone was wrong.

Also, you don't really need to match the performance of a dual GPU card, because even slightly slower single GPU card is always better.

I was basing my estimates on a full 512 shader gpu at rumored target clocks. If GTX480 is indeed cut down or with significantly lower clocks then I'd estimate even lower. As for beating a 5970, you can believe that if you like but I doubt it. Even Nvidia's estimates, with full shaders and target clocks, was 40% faster then 5870. That's not enough to match 5970.

About the whole dual GPU vs single GPU thing, that's your preference not mine. Don't project your preferences on the whole gpu buying community. Thanks.

Which card are we talking about here though? The 480SP or 512SP version? From what I've read there seems to be a lot of confusion about the actual specs of the GTX 480. If it's the 480SP version that is running ~15% up on the 5870, I'd say Nvidia is in good shape depending on power draw.

I was estimating based on a full shader card. Less shaders would equal lower estimates.

[Sn0wm@n]

LOLL




official ferbi pics!!!

[Olivon]

[Andrew LB]

I'm half Korean. I'm a minor going to a high school. I live in the United States. It's so funny how the United States is so segregated about asians. Blacks and whites get along fine. Thats one race now. Asians...or "chinese" kids always get the **** now. I'm quite tired of it and if anyone needs to be be racist on a forum like that then fine. Dos has his wish. Some people humor me.

Here in Southern California and many other cities I've been, the Asian population typically segregates itself, not the other way around. And while I understand that much of this is due to the language and cultural difficulties the older generations experienced... the first and second generation Asian-Americans still have this tendency to group together. I'd love to have more Asian neighbors in my community since they typically are very family oriented, well educated, and keep their homes very nice. I can't say as much for many of my white neighbors (and yes... I'm white).

It would have been nice to see some Fermi benches by now in something other than a label-less graph straight from NVIDIA on the Heaven Benchmark....in DX11.

Agreed. Another week or so and we will.

I'd like to have seen some real stuff. Maybe they won't show it because Fermi isn't even up to par with the 5800 series...they had the lead with G80 by far, but no..."forget R&D and lets just refresh our architecture...it'll last 5 years for sure"

nVidia never shows off their products until the very last moment, which makes perfectly good sense from a business perspective. Why would they want to undercut sales and prices of current generation cards by releasing comprehensive benchmarking results months ahead of release? While the performance crown is an objective, nVidia would be stupid to release early data that would result in their third party card manufacturers being stuck with tons of product that they paid a premium for yet cannot sell for a decent profit margin.

[ToxAvenger]

While the performance crown is an objective, nVidia would be stupid to release early data that would result in their third party card manufacturers being stuck with tons of product that they paid a premium for yet cannot sell for a decent profit margin.

Didn't Nvidia release some Heaven benchmark scores not too long ago?

[Solus Corvus]

Would it be a fair comment to say.
The reason behind nvidia having such a strong NDA and being so tight lipped with there information on the 470s/480s, has alot to do with the fact that ati where able to keep the development
of eyefinity a secret from nvidia right up to the card being released?

It just seems to me like nvidia are making the ponit that they can keep things on the down low as well if they choose to.

nVidia never shows off their products until the very last moment, which makes perfectly good sense from a business perspective.

I hope you guys are kidding. Nvidia has been anything but quiet about fermi with one sneak peek, intentional leak, demonstration, and architecture review after another. They were so eager to start showing off fermi that they couldn't wait for a real card for their CEO to hold up in front of everyone.

What ATI did with 4000, 5000, and (so far) 6000 series was keeping a tight lid on things. What Nvidia is doing with fermi is the diametrical opposite of that.

[ToxAvenger]

I hope you guys are kidding. Nvidia has been anything but quiet about fermi with one sneak peek, intentional leak, demonstration, and architecture review after another. They were so eager to start showing off fermi that they couldn't wait for a real card for their CEO to hold up in front of everyone.

Very true, the Fermi architecture is the worst kept secret of any GPU launch probably ever. The odd thing is, next to no benchmark leaks, you'd think by now there would have been quite a lot of performance leaks.

[eric66]

Very true, the Fermi architecture is the worst kept secret of any GPU launch probably ever. The odd thing is, next to no benchmark leaks, you'd think by now there would have been quite a lot of performance leaks.

now lets think why there aren't except 60 sec of unigine and nice looking far cry 2 bench video

[ToxAvenger]

now lets think why there aren't except 60 sec of unigine and nice looking far cry 2 bench video

You're obviously implying that Nvidia isn't sitting on a 5870 killer, that remains to be seen right? But that's my point, we are so close to the Fermi launch, where are the leaks? (not cherry picked stuff from Nvidia, of course they are going to do this, same as AMD/ATI do)

[informal]

You're obviously implying that Nvidia isn't sitting on a 5870 killer, that remains to be seen right? But that's my point, we are so close to the Fermi launch, where are the leaks? (not cherry picked stuff from Nvidia, of course they are going to do this, same as AMD/ATI do)

If by 5870 killer you mean a card on average 10-15% faster ,consuming more power and OCing worse while costing much more to make(huge die),then yes, it's a 5870 killer alright.

[-Sweeper_]

If by 5870 killer you mean a card on average 10-15% faster ,consuming more power and OCing worse while costing much more to make(huge die),then yes, it's a 5870 killer alright.

and you, of course, have the two cards side by side to compare...

tell us more about the GTX480 pls...