Nvidia Fermi versus radeon 5800 benchmarks out!

[MrK6]

Is it getting hot in here, or is it just me?

I know I'm not alone in those who've been waiting for a +1GB offering. 5870 CF was blazing fast and allowed me to max all details in every game tested. However, I felt I was hitting the ceiling with 4x AA or higher. The GPU's could render anything happening in the scene, but pan and move, bam, accessing and churning of the buffer. A heavily debated topic in itself, but damn it I've wanted a larger frame buffer just to see.

So yes, there's still a lot left to for ATI to offer. It was rumored a 2GB card would launch shortly after 5870's debut. It didn't happen. And now, it's delayed until Fermi's launch window. All eye candy, +4 AA, still a lot to offer..

There's plenty more to offer in the gaming market, I'm not arguing against that. Hell, I'd love to run all my games with ludicrous amounts of AA at 60FPS+, that would be awesome. What I'm saying is that Fermi doesn't bring much to the table over what's available. Recent estimates point to the GTX480 as being ~15% faster than the 5870. Even if it overclocks the same (which I'm going to say it probably doesn't, although time will tell), if you're hitting 60FPS with a GTX480, you would have been hitting over 52FPS with a 5870. The extra frame buffer will be nice in some games (STALKER CoP, Crysis, off the top of my head anyway), but then there's going to be the price/performance issue, which really can't be argued until we know prices.

 

[MrK6]

Please I've been building my own gaming PC's since the mid 90's before 3dfx was on the scene. But you did a great job of avoiding my factual answers to your insinuations.

That's pretty sad then. I'll break it down then: the market doesn't give a shit about Fermi. We do, that's it. As far as we know, we have two extremely high end, flagship parts that will be released and will be competing in the $400+ range. That is such a tiny sector of the market it's not even on the radar of most folks. These people looking at a 5770, or even a 5850, aren't even considering a GTX470 (actually, most aren't considering whatever is not on the shelf in front of them or wasn't recommended by their "tech friend," but I digress). To quote Steam survey figures in this argument is ridiculous at best. Actually, the fact that AMD shipped something over 2 million DX11 GPUs before they even launched their low-end parts is pretty incredible considering the tiny size of the market, and also points to a large percentage of folks upgrading. All that said, to the small group of us that Fermi is actually relevant to, well, so is AMD's new architecture, which is reported to be coming in the next six months (who knows though). The point is, Fermi doesn't seem to be adding much to what's available on the market to the few of us at that level market level. Time will tell.

Terrible logic.......

Not everyone is reading BSN every day and keeping up with launches.

When you are in the market for a card, you do research and pick the card that you think suits you best. Who cares when it launched compared to a competitor?

That's some pretty damn simple logic, amazing that it escaped you. Stemming from my explanation above, we care. If you're in the market for a card, in that research you'll discover what's coming up (if you're worth a shit at research). How many threads are on this forum and others every day with titles along the lines of "Should I upgrade now?" No one wants to buy a 2010 model when the 2011's are right around the corner.

 

[PCboy]

http://www.youtube.com/watch?v=vpzxhr1ibj8

10/10

also, charlie has a mole so no one can't argue.

 

[Hauk]

Wish list:
Batman AA benches with and without PhysX
AVP bences with and without tesselation
lower res benchmarks
real benchmarks,
temps, power consumption, price,

and much more. It's been a long road huh?

 

[happy medium]

Wish list:
Batman AA benches with and without PhysX
AVP bences with and without tesselation
lower res benchmarks
real benchmarks,
temps, power consumption, price,

and much more. It's been a long road huh?

+Metro 2033 benchies

 

[Nemesis 1]

Is that really what you're going to do Nemesis? Is it really that "fun" and "sporting" for you to constantly disrupt the already volitile forums?

Hay keys you best go back look at the ATI threads . Talk about disrupting the forums . I at least am honest and straight forward about my actions . Plus the fact I have always liked ATI. Yes it is fun and sporting. Running down the decievers is fun . I linked to that june 09 topic NV to be first with DX11 parts. Yes it is good sport go back read that thread . Hell I am having a blast and its sporting . That dam r600 was no fun for my side of the fence but you had a good time . Now its ATI turn . Besides Keys I been wrong a few times but I am been right way more than wrong . Who was it that said the 4000 series was going to surprize all . Who was it said PH2 was going to be just fine and AMD would do just fine . When most others alreeady buried them . Na keys I do have fun and it is sport in a sense. But I also research . The best tho is still to come as imagination Tech becomes more and more familiar to this forum . The Force is strong in this one.

Keys I am going to show you guys soon enough my gamer. Than and only than will you truely understand me.

 

[akugami]

For six months now Nvidia guys were saying that DirectX11 was not important. Why the sudden infatuation with DX11/Tessellation? Oh wait, Nvidia is about to release cards that support DX11, awesome.

To a degree, one has to agree that DX11 is more important now than before. As more time passes, DX11 becomes more and more important. I'd have to say that it is a valid argument about DX11 being "irrelevant" to a degree.

I'm starting to feel dejavu with many of these arguments about Fermi's tessellation/potential future performance vs. 5870's current performance. It reminds me so much of when the Geforce256 was first released and 3dfx was screaming that hardware T&L wasn't relevant yet.

How is 3dfx today, anyways?

EDIT: With a little bit of 5800fx in there too.

Except for the fact that the 3dfx situation is nothing like the current one we have now. ATI came out with a tessellator first and were championing one for a while now. We could make the argument that nVidia is only now catching up feature wise in this department, even if nVidia's tessellator seems superior at this point in time judging by leaked benchmarks. nVidia is late with a tessellator because they felt it wasn't relevant yet. ATI's Radeon 4800 series has a hardware tessellator and while I don't know how far back ATI has had a tessellator, ATI has definitely been championing one for much longer than nVidia. Does that mean that nVidia will go the way of 3dfx?

@McCartney....Just cause nVidia's GPU's are better for your company's needs doesn't mean jack to the rest of us. We are entitled to our opinions. Your post reeks of an elitist attitude.

 

[taltamir]

To a degree, one has to agree that DX11 is more important now than before. As more time passes, DX11 becomes more and more important. I'd have to say that it is a valid argument about DX11 being "irrelevant" to a degree.

its also a matter of computational power. It isn't relevant for older, weaker cards... you need a certain minimum amount of power for DX11 to be relevant... kinda like nobody would play DX10 games on something as pathetically weak as an IGP (or even things like the 8600GTS)

 

[zerocool84]

To a degree, one has to agree that DX11 is more important now than before. As more time passes, DX11 becomes more and more important. I'd have to say that it is a valid argument about DX11 being "irrelevant" to a degree.



Except for the fact that the 3dfx situation is nothing like the current one we have now. ATI came out with a tessellator first and were championing one for a while now. We could make the argument that nVidia is only now catching up feature wise in this department, even if nVidia's tessellator seems superior at this point in time judging by leaked benchmarks. nVidia is late with a tessellator because they felt it wasn't relevant yet. ATI's Radeon 4800 series has a hardware tessellator and while I don't know how far back ATI has had a tessellator, ATI has definitely been championing one for much longer than nVidia. Does that mean that nVidia will go the way of 3dfx?

@McCartney....Just cause nVidia's GPU's are better for your company's needs doesn't mean jack to the rest of us. We are entitled to our opinions. Your post reeks of an elitist attitude.

Well the reason why tessellation is barely being relevent is mostly Nvidia's fault. They wanted it out of DX10. They carry more power to get things put in or taken out than ATI does.

 

[akugami]

Well the reason why tessellation is barely being relevent is mostly Nvidia's fault. They wanted it out of DX10. They carry more power to get things put in or taken out than ATI does.

Well, my point was that tviceman was making a poor attempt at trashing ATI by making nVidia out to be a forward thinking company pushing new tech while ATI is saying that the new tech is irrelevant. Much like how 3dfx was saying hardware T&L was irrelevant and much like how 3dfx went the way of the Dodo bird.

Now, nVidia is a forward thinking company and it does attempt to move tech forward as does ATI for those who are unbiased. Both companies have different ideas on where we should be moving with technology but that's the great thing about competition. It promotes innovation.

In this case with tessellation however, ATI was there first, even if from leaked benchmarks it seems like nVidia is doing it better. Just pointing out to tviceman that for his comparison to be true, nVidia would have to flounder and go out of business and then be bought up by a rival.

 

[Silverforce11]

A few points others may or may not have mentioned that are important:

1. Fermi's polymorph engine handles the tessellation very well, only if it isn't taxed in other heavy shader tasks. This is evident in the Unigine bench that NV themselves release in comparison to a 5870, they ran it at 1920 res, with NO AA and low AF to really put the least strain on the fermi gpu as they can, so its tessellation can be impressive. This is further evident in the 4xAA and 8xAA results, where the 5870 actually equal and beat Fermi, respectively.

2. NV's original clock goals are in the 750mhz range. Now they are settling for 625mhz. The thermals are also very high, with many sources suggesting 280-300W TDP for the 480 (even FUD agrees with this), and 225W for 470. The 470 comes with 2x6pins, thus maxing out at ~225W dead on. The 480 has 1x6 and 1x8 pin, giving it ~300W. There is very little room for OC based on reference boards. Any OC will have to wait for custom boards.

3. VRAM limited benchmarks, at max res and mostly 8xAA, the 5870 1GB start to chug along, as you can see in some min fps. its single digit low, indicating main memory usage. Tone it down to 4xAA, situation will be different (ie. benches available everywhere for 5870 at 4xAA, numbers are good). Ofcourse, the opposite is also true, Fermi handles 8xAA max res better (but not in tessellation scenarios), but realistically, those games aren't at all playable at the low fps regardless. You need a 5970 or CF if you want max eye candy.

4. A 5850 can be had for ~$270. It will nearly always OC to 5870 speeds, in most cases easily reaching 900+ mhz (Mine does 950mhz and stays quiet). Nothing comes close to bang for buck, and if you want more performance, CF 5850 and OC both.

 

[Kuzi]

The force is strong with this one.

 

[SolMiester]

Please I've been building my own gaming PC's since the mid 90's before 3dfx was on the scene. But you did a great job of avoiding my factual answers to your insinuations.

Behind you mate, however 3dfx was the gaming scene!!, their was no ATi gaming cards and no nVidia...

 

[SolMiester]

deleted

 

[solofly]

http://www.semiaccurate.com/2010/03/12/semiaccurate-wrong-about-nvidia-480gtx-power-use/

 

[v8envy]

Behind you mate, however 3dfx was the gaming scene!!, their was no ATi gaming cards and no nVidia...

Computer gaming started well before 3d accelerators. There were some very fun and addictive games for the Apple ][, commodore 64 and so on. Doom was completely software rendered. So was the original Quake. To say computer gaming started once Quake got hardware 3d acceleration is incorrect.

 

[SlowSpyder]

http://www.semiaccurate.com/2010/03/12/semiaccurate-wrong-about-nvidia-480gtx-power-use/

I cannot wait for launch, Charlie is going to end up looking pretty damn smart or have a LOT of egg on his face. Fermi is looking more and more like a flop (assuming these rumors are mostly true).

 

[zerocool84]

I cannot wait for launch, Charlie is going to end up looking pretty damn smart or have a LOT of egg on his face. Fermi is looking more and more like a flop (assuming these rumors are mostly true).

Only 2 more weeks to see if everyone owes him a bj, or he takes it in the ass from all of us.

 

[jaggerwild]

All that and i did not see one link to a valid score so this whole thread is for not?

 

[OCGuy]

He is getting more and more mad because his "broke and cant be fixed" article is going to be funny on launch day when it sells out.

It smells of even more deperation than usual.

None of those power requirements surprise anyone here. People bought 280s without even flinching about power.

If you dont think you can talk about performance, talk about crap that most people dont even care about.

Coming up next: Die size! Oh no!


Show me the MSRP, show me the e-tailers' actual price, and show me the benchmarks.

 

[1h4x4s3x]

Only 2 more weeks to see if everyone owes him a bj, or he takes it in the ass from all of us.

I'd rather do neither.

 

[konakona]

A few points others may or may not have mentioned that are important:

1. Fermi's polymorph engine handles the tessellation very well, only if it isn't taxed in other heavy shader tasks. This is evident in the Unigine bench that NV themselves release in comparison to a 5870, they ran it at 1920 res, with NO AA and low AF to really put the least strain on the fermi gpu as they can, so its tessellation can be impressive. This is further evident in the 4xAA and 8xAA results, where the 5870 actually equal and beat Fermi, respectively.

2. NV's original clock goals are in the 750mhz range. Now they are settling for 625mhz. The thermals are also very high, with many sources suggesting 280-300W TDP for the 480 (even FUD agrees with this), and 225W for 470. The 470 comes with 2x6pins, thus maxing out at ~225W dead on. The 480 has 1x6 and 1x8 pin, giving it ~300W. There is very little room for OC based on reference boards. Any OC will have to wait for custom boards.

3. VRAM limited benchmarks, at max res and mostly 8xAA, the 5870 1GB start to chug along, as you can see in some min fps. its single digit low, indicating main memory usage. Tone it down to 4xAA, situation will be different (ie. benches available everywhere for 5870 at 4xAA, numbers are good). Ofcourse, the opposite is also true, Fermi handles 8xAA max res better (but not in tessellation scenarios), but realistically, those games aren't at all playable at the low fps regardless. You need a 5970 or CF if you want max eye candy.

4. A 5850 can be had for ~$270. It will nearly always OC to 5870 speeds, in most cases easily reaching 900+ mhz (Mine does 950mhz and stays quiet). Nothing comes close to bang for buck, and if you want more performance, CF 5850 and OC both.

some very good observations. I definitely concur with #4, I don't think fermi will be anything close to competitive price-wise, unless nvidia is ready to take some BIG TIME loss. If someone's actually interested in buying one of these, it would be like a BF rush only x10 worse lol

If fermi 2 does end up a formidable competitor, then we could sell our 5850/5870s to upgrade. doesn't look like the first fermi is very promising from what we have seen so far.

 

[v8envy]

Wow, Charlie is reaching. Little systems with GTX480s, really? Only the cream of the crop "gaming" OEM boxes would be contenders for such hardware, and those bad boys come with 700+ watt Delta or better PSUs and full size towers packed with fans just like enthusiast PCs, not fanless microatx cases with no corner left uncut.

Article is fail, and simply rehashes things already covered.

 

[ginfest]

http://www.semiaccurate.com/2010/03/12/semiaccurate-wrong-about-nvidia-480gtx-power-use/

Who says AMD/ATI don't have their own viral marketing team

 

[1h4x4s3x]

Wow, Charlie is reaching. Little systems with GTX480s, really? Only the cream of the crop "gaming" OEM boxes would be contenders for such hardware, and those bad boys come with 700+ watt Delta or better PSUs and full size towers packed with fans just like enthusiast PCs, not fanless microatx cases with no corner left uncut.

Article is fail, and simply rehashes things already covered.

http://www.semiaccurate.com/2010/03/10/mobile-gtx480-shown-cebit/