First retail 6870 photo; now includes full Chinese leaks and Guru3D review discussion

[maddie]

Are you getting 2x HD6870s for those 3 displays?!



Your 5% performance advantage is inaccurate. Even your own link to TechPowerup shows 180% for GTX480 vs. 112% for HD5870 at 2560x1600 (that's 60% performance difference). Please check DX11 games where HD5870 can't even beat a GTX470 despite an $80-100 higher price; so I have no idea about your "little to show for it" comment.

No, I am not getting wet about pointing fingers about Cayman being more power consuming, cuz I could care less about power consumption when it comes to $500 enthusiast videocards; and I am not a hypocrite about it either. I overclock my CPU and don't complain about the extra power for it. But I still giggle at biased arguments over power consumption by *certain people* who constantly brought up power consumption in every thread for the last 6 months. I bet when their "precious" brand will be in the lead, they'll somehow figure out a way to justify the additional power consumpion.

See I have no problems with recommending either brand - I am not biased. That's why I'll recommend the Cayman XT with no reservations even if it consumes 300W. Given the "strong stance" against power consumption by ATI faithful, I will fully expect none of the same people who complained about it for the last 6 months to purchase the Cayman XT because power consumption was soooooo important to them during the last generation despite a 20-30% performance disadvantage in DX11. Makes sense right?

I really think you should link to the tests when Fermi first appeared. That was when the first impressions were made.

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_480_Fermi/32.html

 

[happy medium]

You are looking at total score, GPU scores tell a totally different story.

Thank you thats the answer i was looking for. Seems it is a good indicator of performance.

 

[blastingcap]

Are you getting 2x HD6870s for those 3 displays?!

Haha no, I avoid multi-GPU whenever possible. In fact, I opted for 3x22" rather than 3x24" in part because I doubted that single-GPUs could drive 3x24" that well. (I had a 24" Ultrasharp and a 22" Acer last year, so I could have gone either way by adding 2 monitors of either size.)

Looks like 6870 for me unless it's overpriced, in which case I'll settle for a 6850 or even a 5-series, depending on pricing.. if pricing is really bad, I'll ditch Eyefinity, sell/give away a monitor, and get a GTX460. I can't hold out till "late November" to get Cayman. My IGP overclocked at 1GHz can run Tales of Monkey Island at 1024x768 at medium settings without too much lag, and even L4D is semi-playable at minimum settings at 1280x800, but I'm going to finish ToMI soon and already finished my backlog of older games like World of Goo.

 

[happy medium]

This image has been resized. Click this bar to view the full image. The original image is sized 513x239 and weights 4KB.

I would think the gtx465 would score better then that? Its not that much slower then a gtx460

Thanks for the link

 

[Lonyo]

Thank you thats the answer i was looking for. Seems it is a good indicator of performance.

Not at all, especially when it comes to ATI vs NV.
When within 'an architecture' it can be 25% out (as in the case of GTX460 vs 470), comparing it across architectures is even more meaningless.

It just happens to manage to indicate an overall performance difference between the GTX470 and HD5850 which happens to tally with the real world, but that's more down to chance than any actual usefulness from the benchmark.

Like you said already, Vantage is pretty useless for telling us much about actual games.

 

[Skurge]

Yes they are equal in Vantage but the gtx470 is closer to 5870 speeds in real life gaming. Vantage is not a very good indicator of real world performance.

I agree that vantage is not a good indicator (Better than the older version mind you). Those real life games are also mostly canned benches, so I could say the same.

 

[blastingcap]

Not at all, especially when it comes to ATI vs NV.
When within 'an architecture' it can be 25% out (as in the case of GTX460 vs 470), comparing it across architectures is even more meaningless.

It just happens to manage to indicate an overall performance difference between the GTX470 and HD5850 which happens to tally with the real world, but that's more down to chance than any actual usefulness from the benchmark.

Like you said already, Vantage is pretty useless for telling us much about actual games.

If we're talking about games, this is the only actual game that I've seen allegedly benched with a HD6xxx series (probably a Cayman variant):

http://we.pcinlife.com/thread-1500103-1-1.html

(direct picture URL here:

)

 

[notty22]

One more comparison, I put the gtx 460@775, it looks about a draw, to the one posted.
http://tweakers.net/ext/f/3m0pOqyxtPwZPLJY1ZoVACGF/full.jpg

 

[Skurge]

One more comparison, I put the gtx 460@775, it looks about a draw, to the one posted.
http://tweakers.net/ext/f/3m0pOqyxtPwZPLJY1ZoVACGF/full.jpg

how high can you get that 460? Thanks for the tests btw.

 

[notty22]

how high can you get that 460? Thanks for the tests btw.

930/1050

 

[Tempered81]

If we're talking about games, this is the only actual game that I've seen allegedly benched with a HD6xxx series (probably a Cayman variant):

http://we.pcinlife.com/thread-1500103-1-1.html

(direct picture URL here:

)

Yeah a GTX480 with a 4.2g i7 does this at about 34-35 fps

 

[tviceman]

Just wont accept AMD possibly bringing out a better card :thumbsdown: So basically the 460 has to OC 20% to keep up according to this. Lets also not forget they still carry the same max OC and cost $20 more, so its pretty moot.

These scores don't mean anything in real world performance. A gtx480 scores about 2.5% higher than an hd5870 in 3Dmark, 50% higher in Unigine, and both cases are the very extreme compared to benchmarks that matter: GAMES. If the 6870 isn't faster than a gtx460 1 gig in DX11, then it'll be a laughing failure. Almost without a doubt though, it WILL be faster.

The more important battle vs. the gtx460 will be the hd6850. If it isn't faster more often than not and it's only $10 cheaper, then it won't be that great of a card if the only incentive to get it over a gtx460 is power draw (which the gtx460 is manageable in).

 

[SlowSpyder]

I'm sorry if I missed it somewhere earlier in this thread, but do we know when NDA lifts? I think next week sometime, but do we know the day? I assume the NDA is just for the 68xx parts, the 69xx parts will be later?

 

[Ares1214]

What about 720? If the 6870 stocks at 850 MHz (?) and this card is 915 MHz, and equal oc would be 720 MHz on the 460.

 

[MrK6]

Your 5% performance advantage is inaccurate. Even your own link to TechPowerup shows 180% for GTX480 vs. 112% for HD5870 at 2560x1600 (that's 60% performance difference). Please check DX11 games where HD5870 can't even beat a GTX470 despite an $80-100 higher price; so I have no idea about your "little to show for it" comment.

Stop right there and verify that you really think the GTX 480 is really 61% faster than the HD5870 as a true average, that way I can write off pretty much any other point you make as mindless drivel. You either A) don't understand how averages work or B) are purposefully misrepresenting data to prove your point, and in either case, well, that proves the point I said earlier about fanboys.

No, I am not getting wet about pointing fingers about Cayman being more power consuming, cuz I could care less about power consumption when it comes to $500 enthusiast videocards; and I am not a hypocrite about it either. I overclock my CPU and don't complain about the extra power for it. But I still giggle at biased arguments over power consumption by *certain people* who constantly brought up power consumption in every thread for the last 6 months. I bet when their "precious" brand will be in the lead, they'll somehow figure out a way to justify the additional power consumpion.

Yes, power consumption is justified by performance, something the GTX 480 sorely lacks in comparison.

See I have no problems with recommending either brand - I am not biased. That's why I'll recommend the Cayman XT with no reservations even if it consumes 300W. Given the "strong stance" against power consumption by ATI faithful, I will fully expect none of the same people who complained about it for the last 6 months to purchase the Cayman XT because power consumption was soooooo important to them during the last generation despite a 20-30% performance disadvantage in DX11. Makes sense right?

Again, you're generalizing to justify your purchase of a GTX470, as I said earlier. I already posted that it's not the power consumption that's the problem, it's the power consumption and the lack of the performance. However, you're conveniently ignoring this and continually chanting on about some inane argument I've already dismissed awhile ago. So, what's your deal?

Your welcome, I found your observations and conclusions rather funny.

Of course you did, you don't know a thing about hardware.


The flamebait comments need to stop.

Re: "you don't know a thing about hardware"

Moderator Idontcare

 

[Dark Shroud]

I'm sorry if I missed it somewhere earlier in this thread, but do we know when NDA lifts? I think next week sometime, but do we know the day? I assume the NDA is just for the 68xx parts, the 69xx parts will be later?

It's been posted as Oct 22, I'm not sure if this is confirmed or rumor.

 

[MrK6]

*pics*

Great finds tempered81, thanks for posting!

So it seems the 6870 @ 915MHz is ~GTX 460 overclocked to 775MHz in Vantage (thanks notty22). I'd be curious to see the extreme preset scores when available.

 

[SolMiester]

snip.

looks to me like its techpowerup showing the 60% avg performance advantage!?

 

[cusideabelincoln]

Your 5% performance advantage is inaccurate. Even your own link to TechPowerup shows 180% for GTX480 vs. 112% for HD5870 at 2560x1600 (that's 60% performance difference). Please check DX11 games where HD5870 can't even beat a GTX470 despite an $80-100 higher price; so I have no idea about your "little to show for it" comment.

It's quite obvious the 2560x1600 numbers are off. He must have mislabeled or typed in the wrong number somewhere. Without even crunching numbers the GTX 480 is not faster than the HD 5970, and if you look through all of the tests the 5970 wins by the vast majority, and sometimes by a huge margin. The only test where the 480 beats every card by a good margin is the Metro one, and obviously all of the cards run into VRAM limitations.

But I did take the liberty of crunching the numbers. Here are the results for all the games, even including the very skewed Metro tests. Excluded are the 3DMark and Unigine tests:

GTX 470: 100% (base)
HD 5870: 108%
GTX 480: 125%
HD 5970: 159%

looks to me like its techpowerup showing the 60% avg performance advantage!?

Looks to me people should double check their sources.

Not sure what is wrong with the 2560x1600 numbers, but the other resolution summaries look fine. I'm not going to check them though, but they make sense. The 2560 numbers, though, do not, and I find it hard to believe no one took a double take of them. Instead you post them as indisputable facts!

 

[bryanW1995]

It's all the rage now. Where you been?

that's ok, intel measures it in Gjz.

 

[bryanW1995]

Based on AMD's history with graphics, and based on a few articles anand put up detailing AMD's strategy, it's implausible that they'll turn Cayman into a large die/ high TDP monster.

BTW, that "Single slot" fansink card in ATENRA's post... looks tempting.

"large die/high TDP monster" is relative. I wouldn't call a 400-420mm^2 card that consume ~ 225-250 w at load either of those compared to a 576mm^2 card that is closer to 300w at load.

they went larger b/c they saw that nvidia was struggling to execute their strategy on 40nm and knew that they could basically guarantee continued market strength though q4 2011 by doing it.

 

[Grooveriding]

If we're talking about games, this is the only actual game that I've seen allegedly benched with a HD6xxx series (probably a Cayman variant):

http://we.pcinlife.com/thread-1500103-1-1.html

(direct picture URL here:

)

If that bench does turn out to be correct the 6970 will be about 30% faster than a GTX 480 in DX10. And note that image shows the Radeon as being a 6870, but that was at a time before the new naming scheme was known and that bench was claimed to be from a Cayman part (6970).

I guess vantage is something to banter around about as its all we have, but with the cards coming next week and real benches being available, why bother.

Crysis on enthusiast with 4xAA, Metro 2033, BFBC2, Stalker and the Heaven benchmark are all I care to see about these new AMD cards and how they perform.

I'm not in the market for a 6870, got my eye on a 6970. But if I was, I would only care to see a stock 460 against a stock 6870 and an overclocked 460 against an overclocked 6870 in the above mentioned benchmarks.

 

[SolMiester]

It's quite obvious the 2560x1600 numbers are off. He must have mislabeled or typed in the wrong number somewhere. Without even crunching numbers the GTX 480 is not faster than the HD 5970, and if you look through all of the tests the 5970 wins by the vast majority, and sometimes by a huge margin. The only test where the 480 beats every card by a good margin is the Metro one, and obviously all of the cards run into VRAM limitations.

But I did take the liberty of crunching the numbers. Here are the results for all the games, even including the very skewed Metro tests. Excluded are the 3DMark and Unigine tests:

GTX 470: 100% (base)
HD 5870: 108%
GTX 480: 125%
HD 5970: 159%



Looks to me people should double check their sources.

Not sure what is wrong with the 2560x1600 numbers, but the other resolution summaries look fine. I'm not going to check them though, but they make sense. The 2560 numbers, though, do not, and I find it hard to believe no one took a double take of them. Instead you post them as indisputable facts!

Who is posting them as undisputed facts?

 

[SolMiester]

I wonder what the check box for x64 does, ticked on the Radeon, not on NV.....is there a x64 version of this game?

 

[RussianSensation]

Just wont accept AMD possibly bringing out a better card :thumbsdown: So basically the 460 has to OC 20% to keep up according to this. Lets also not forget they still carry the same max OC and cost $20 more, so its pretty moot.

Barts XT looks like a more powerful videocard Consumers will likely have to choose between the slightly slower GTX460 card such as the 780mhz MSI TwinFrozr for only $210 and the faster Barts XT for a little bit more $. On the whole, I can foresee NV dropping prices another $10-20 if Barts XT launches at say $219-229. If Barts comes in at $249+ though, then we may not have a price war.

I really think you should link to the tests when Fermi first appeared. That was when the first impressions were made.

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_480_Fermi/32.html

Ok, but it's October 2010, no? Since the initial launch, new drivers have improved performance significantly, and more and more DX11 games have come out where HD5000 series is not quite as fast. So what's the point of looking at 8 months old benchmarks?

How about these? It's pretty obvious that HD5870 is the most overpriced card right now. That doesn't matter though since HD6000 is coming out and it will have the best price performance in each segment pretty soon.

Like you said already, Vantage is pretty useless for telling us much about actual games.

I think Vantage is probably the worst in this regard. Not only is it outdated since it doesn't support DX11, but relying on 1 benchmark to gauge overall average performance in games can be very dangerous:

I mean if we just used 1 benchmark, we could just as easily conclude that a GTX275 is faster than an HD5850 - which is a wrong conclusion when you measure average gaming performance.

Let's wait for real world results.

Also, keep in mind that HD6850/6870 are being tested with initial/early drivers. We have seen how Fermi has improved its performance in the last 6 months. Therefore, I can only see HD6870 extend its lead over the GTX460 after AMD's driver team works on more optimized drivers 2-3+ months after launch.