6870 Benchmarks Leaked

kllrnohj said:
This goes back to ATI having problems with CFX, which you are using. Single card drivers have been fine :)
Click to expand...

There are still a few having issues with single cards... I am one of them.


This new series looks good, I will keep an eye out.
 
35-40% faster than the 5870 even though it is supposed to be almost the same size, and on the same process? obviously fake. physically impossible.
 
bAMtan2 said:
35-40% faster than the 5870 even though it is supposed to be almost the same size, and on the same process? obviously fake. physically impossible.
Click to expand...

Only impossible if they kept all the same linear line of development with no changes. Rumor has it they improved/new tesselation engine, changed the ratio of a few units, etc. Size does not equal performance. Look at what nVidia has done. Bigger chip ~ same performance.

Faster ram, optimizations can make a big difference. Not saying it will as I am really wary of leaks, but it is possible.
 
kllrnohj said:
This goes back to ATI having problems with CFX, which you are using. Single card drivers have been fine :)
Click to expand...

UUmm....BF:BC2 has been affected by both single and crossfire issues. Metro 2033 also pops into my head as having issues in both configurations...

Don't count your chicken before they hatch.
 
I have two Matrix 5870 2GBs running on the 10.8 drivers with the 10.8a crossfire profiles (newest of both) and they are running extremely smooth in CFX / Eyefinity.... I have definitely had a few driver hiccups in the past but the new set seems to be working well.
 
Kangg said:
I could fucking care less about benchmarks. I don't play benchmarks, I game. let's see some actual gameplay info.
Click to expand...
You should write a nice letter to AMD politely asking them to release gameplay benchmarks to you only. I can't believe no one has thought of doing this yet.
:rolleyes:
Hels said:
As far as the rumours goes. ATI has increased the die size a bit to get a bigger boost, since they are not going to 28nm the way they planned. This means that the amount of power/temp/noise is a bigger question mark.
Click to expand...
Word on the street is this is 3/4 of what they had planned for 28nm (480SP instead of the full 640SP), so I imagine they did cut it back to keep power/cooling sane. It also explains why the performance jump isn't as big as we're used to.
 
bAMtan2 said:
35-40% faster than the 5870 even though it is supposed to be almost the same size, and on the same process? obviously fake. physically impossible.
Click to expand...

Could be fake, but it could be possible if they figured out some new architecture that rocks. Just saying.
 
Megalomaniac said:
It’s not impossible. I used to bullseye womprats with my T16 back home. They’re not much bigger than two meters.:rolleyes:
Click to expand...

LOL. Classic retort!

There's no dobut in my mind that the 6000 series will surpass Fermi. And it should. 6000 is just the evolution of 5000 on a now mature process plus new architecture that is a planned part of the 7000 series that will also feature a smaller process. It's a hybrid but it's still going to kick ass just because the tech and engineering has matured that much further. I expect that Fermi Revision 2 will also likely kick 6000 series' tail. But that's just how it goes back and forth. This isn't exactly a new trend.
 
vjcsmoke said:
LOL. Classic retort!

There's no dobut in my mind that the 6000 series will surpass Fermi. And it should. 6000 is just the evolution of 5000 on a now mature process plus new architecture that is a planned part of the 7000 series that will also feature a smaller process. It's a hybrid but it's still going to kick ass just because the tech and engineering has matured that much further. I expect that Fermi Revision 2 will also likely kick 6000 series' tail. But that's just how it goes back and forth. This isn't exactly a new trend.
Click to expand...

I don't think you're wrong about "Fermi 2" beating 6xxx but I highly doubt we'll see it until a year from now at the earliest. And that's not a joke about NV's ability to deliver products on time - the problem is that GF100 is hot and huge at 40nm. They don't have any headroom to play with and they probably can't do much until 28nm. Even Jen-Hsun Huang said that there wouldn't be a new architecture until 28nm. Unless they find a way to pull off some miraculous efficiency (perf/mm2) gains, they'll probably be stuck with GF100 as their fastest Fermi. I would not expect anything high-end from NV for a year except for a dual GF104 part.

AMD on the other hand is not nearly as constrained on 40nm. I'm sure they'd love to continue on the "sweet-spot" strategy but with cypress at 334mm2 it's entirely feasible to increase that up 50% or so and still be cheaper to manufacture than GF100. Charlie guesses Caymen to be around 390mm2. That's a 15% increase in die size already, and if the 4D rumors are true they should be able to fit quite a few more shaders in the same amount of space (this would however increase TMU count a lot too). So bAMtam2 it's really not impossible.
 
30-40% performance over the previous generation is about standard isn't it? Why would they release anything less than that?

The jury is still out for me though. Catalyst 10.6, 10.7 and 10.8 have been major fuckups. Ridiculous performance with the 2D clock speeds for several months of drivers and I can't even make an Eyefinity display group with 10.8, it's so messed up. I went back to 10.5 today as they are the last ones that work as advertised.
 
The Cobra said:
UUmm....BF:BC2 has been affected by both single and crossfire issues. Metro 2033 also pops into my head as having issues in both configurations...

Don't count your chicken before they hatch.
Click to expand...

what issue does single card have?
 
shansoft said:
what issue does single card have?
Click to expand...
Same, I've played Metro 2033 a few times and BF:BC2 many times through every drive revision discussed here and I've had zero issues. But then again, I haven't had any issues with my 5850, ever. I must be lucky.
 
bAMtan2 said:
35-40% faster than the 5870 even though it is supposed to be almost the same size, and on the same process? obviously fake. physically impossible.
Click to expand...

The 5870 and the GTX 480 are on the same 40nm process yet GTX 480 has a die size of 529mm² and 3 billion transistors compared to the 5870's 334mm² die size and 2.154 billion transistors.

ATI has a lot of room to rearrange and add transistors if need be but how specifically is beyond what the average consumer would know.
 
Kangg said:
I could fucking care less about benchmarks. I don't play benchmarks, I game. let's see some actual gameplay info.
Click to expand...

The Crysis bench is in-game. Not exactly real world gameplay since there isn't much action going on, but it's close enough. Now the Crysis flyby is bullshit.

Finkus said:
30-40% performance over the previous generation is about standard isn't it? Why would they release anything less than that?
Click to expand...

Depends. The 5870 is almost a doubling of the 4870 tho.

30-40% increase is about right and still within expectations, since they are still manufactured on the same process (40nm). I know I am not disappointed if the leaked benchmarks turn out to be true. Significant increases are not really possible unless they're on a smaller process otherwise the chip will produce too much heat.
 
Last edited:
Xzyrus said:
The 5870 and the GTX 480 are on the same 40nm process yet GTX 480 has a die size of 529mm² and 3 billion transistors compared to the 5870's 334mm² die size and 2.154 billion transistors.

ATI has a lot of room to rearrange and add transistors if need be but how specifically is beyond what the average consumer would know.
Click to expand...

yes but if it runs as hot as a gtx480 i'll wait till next year and 28nm
 
rUmX said:
Depends. The 5870 is almost a doubling of the 4870 tho.
Click to expand...
And the HD4870 was a full blown 250% of the HD3870 :eek: Well, mostly because the HD3870 GDDR4 was a lither slower than the HD2900 XTX :p
bill_d said:
yes but if it runs as hot as a gtx480 i'll wait till next year and 28nm
Click to expand...

Well, given the fact that ATi has already much earlier experience with larger dies on 40nm (RV740 - HD4770 vs nVidia's later, initially OEM-only g210), it's likely they won't have serious issues getting decent yield rates (and by a distant relation, lower power required due to higher quality dies) out of a yet-larger 40nm chip.

Just speculation, though. :D
 
There's no dobut in my mind that the 6000 series will surpass Fermi. And it should. 6000 is just the evolution of 5000 on a now mature process plus new architecture that is a planned part of the 7000 series that will also feature a smaller process. It's a hybrid but it's still going to kick ass just because the tech and engineering has matured that much further. I expect that Fermi Revision 2 will also likely kick 6000 series' tail. But that's just how it goes back and forth. This isn't exactly a new trend.
Click to expand...

They did this same thing with the hd2900 series when they launched the 3000 series. the 2000 was the initial release and then they used what they learned from it and applied it to the 3000 series.
 
Looks really good if it's true. Considering drivers are still beta I'd think an additional 20% performance boost can be expected from better drivers, which brings this product about 55% faster than previous gen. Not a bad upgrade at all. Also considering that 3x 30inch eyefinity needed just that amount of boost to run smooth in the most demanding apps I'd say AMD has a real winner on the way, if this is true ;)
 
AMD will never admit the 5 series GSODs are a hardware issue for financial and PR reasons. Lets hope they've isolated the problem for the 6 series.
 
sadbuttrue said:
AMD will never admit the 5 series GSODs are a hardware issue for financial and PR reasons. Lets hope they've isolated the problem for the 6 series.
Click to expand...

Wow dude if the card is defective return it. I don't understand you people and your GSOD's, --hell, having just had another 8800GTS crap out on me after driving me mad while troubleshooting, be thankful its that easy to diagnose.

Anyways, them benchies look really nice, but really its all dependent on that MSRP isn't it. I'm going to take a stab at $549, $399 for the HD 6870 and HD 6850 respectively.

PS where are the *60's?
 
MrWizard6600 said:
Anyways, them benchies look really nice, but really its all dependent on that MSRP isn't it. I'm going to take a stab at $549, $399 for the HD 6870 and HD 6850 respectively.

PS where are the *60's?
Click to expand...

Hope it comes in closer to an MSRP of $399 for the 6870 to $299 for the 6850 or less. Those prices seem about as high this current market will tolerate imo.
 
Where does the 6770 fall into the equation? Are we talking 5850 performance?
 
sweet, should be able to pick up an extra 480 on the cheap in a couple months.
 
I love how people only share the 3dmark score. They dont show anything that proves anything. I could modify a screenshot to say what ever I want, until I see proof, I would take this with a grain of salt.

It could be 2 5870's showing that screenshot too.
 
Looks like Im retiring my 4890, can't wait for official words on this card.

Was extremely tempted to upgrade to the 5 series but decided to hold off since at first I couldn't get what I wanted then became complete contempt with what I had. By the time the 5870 came into stock they were talking about the 2gig versions ... so I waited for that, then couldn't get in on that deal so now I'll wait for this.
 
4870 finally out. 6870 with the first K model Sandy bridge early next year. Not that concerned with what it costs either.
 
I cant wait for this series t come out. I need to replace my X800XL Radeon!!!! :cool:
 
Finally, a card that will give me a reason to replace my GeForce 3 TI200.
 
If you are having GSODs still then send your card back, because there's a problem with it. I had GSODs with my 5870 and I was annoyed. I do not any longer. Clearly they were able to fix it in drivers for at least most cards.

If you are having those problems don't sit around and bitch, get support because you shouldn't be anymore.
 
GSOD was a software problem and not a hardware problem. It has been confirmed many times before.
 
I would like crossfire numbers...

Deciding if it would even be worth considering upgrading from 2 x 5850's. I do play on eyefinity res. (5292 x 1050). What do you guys think?
 
Finally, a card that will be a good enough upgrade to replace my geforce 2 mx
Posted via [H] Mobile Device
 
You must log in or register to reply here.
Back
Top