HD 3870 Benchmarks here

[aiya24]

Originally posted by: munky
What a card is supposed to cost is of no use to me when that card can hardly be found, and even then is jacked up to $300. I'm still going to wait for benches to see how these cards compare at 1920x1200 with AA, but I am not paying $60 for 10% more performance, and if the faster card can't even be found then fahgetabaudit! Also, the overclocking headroom may determine what the real performance is of these cards in an enthusiast system.

QFT

i'm also waiting for benches @ 1920x1200 w/AA. i would certainly purchase a card that is widely available and take a performance hit and pay less.

some guys over on XS with HD3870s are overclocking to 850/1300 stock cooling/volts and look at the cooler, no heatpipes and its doing 850mhz on the core with the fan on auto. i'm pretty sure these will hit 1ghz if not close to it with aftermarket/water cooling.

Edit: forgot to add that if your looking at the big picture and waiting for G92 GTS/GTX cards, your in for a long wait since the 8800GT is OOS everywhere and i'm guessing due to low yields so that means the other cards will be pushed back too.

 

[taltamir]

I would agree if that was all there was to it....
but merchants are already selling the ATI card for 260$... so my point is that right now the best thing to do is wait till the feeding frenzy is over, the dust settles, and stable prices are reached with both cards. And then make a decision.

 

[Hauk]

Originally posted by: bryanW1995
I wonder if the ip35 pro is set up like the asus or the msi mobo in that comparison, thilan29? the asus appears to do much better in xfire.

Good question, it's set up like the MSI board:

Expansion Slots:
2 - PCIe x16 (1 - x16, 1 - x4 electrical)
1 - PCIe x1
3 - PCI Slot 2.3

I was curious why Abit didn't advertise crossfire support. Likely same reason MSI didn't:

"MSI P35 Diamond (P35) board has an unbalanced x16 and x4 combination, which not only means there's an obvious bottleneck in the x4 bandwidth, but the data has to go into the southbridge and share the northbridge-southbridge DMI before meeting with the other card's data. Its latency is far greater making it the most inefficient solution, and because of this, even MSI doesn't advertise CrossFire as being a feature on its P35 Diamond. We're using it here simply because it can do it and it's also DDR3 capable so we can keep the systems as even as possible."

"We would recommend avoiding virtually every P35 board that sports CrossFire as a feature, but that's not to say you should avoid P35 boards as a whole though - it's just CrossFire on in this situation is financially stupid. Why pay for a cheaper P35 board and then fork out for two expensive graphics cards when you can't get the most out of them? "

 

[trajan2050]

Far from 8800GT scores


Radeon 3870, or RV670XT, and Radeon 3850 aka RV670PRO, will be launched tomorrow. We can tell you that Revival and Gladiator cards will be out and that you will see the cards with two DVI's, DVI and native HDMI.

The Radeon HD3850 512MB card will score some 20 ? 30 percent lower than a Geforce 8800GT 512MB and as we said before it will end up slower.

New Nvidia driver fixes Crysis complaint.

ForceWare 169.09 Release Information
ForceWare 169.09 was released today on NVIDIA?s website.

This driver:


Adds SLI support for Crysis

Fixes a water reflection bug in Crysis

Adds Control Panel anisotropic filtering support for Crysis
With this driver, SLI will deliver 40-50% speedup over a single GPU for Crysis. In addition, Crytek will release a patch in the next week or two that will improve SLI performance further. Here are some reference performance numbers obtained using the built-in benchmark:
GeForce 8800 GT 512MB:
1600x1200 High Quality - Single GPU: 24.05 SLI: 34.76 Increase: 45%

 

[Rusin]

http://www.iax-tech.com/video/3870/wic.jpg

They also updated their Crysis test and fixed that water bug. Max framerate dropped from 30 -> 20, but average framerate did rise. Also min fps did rise.
http://www.iax-tech.com/video/3870/38705.htm

There's also this synthetic benchmark:
http://www.iax-tech.com/video/3870/lightmark.jpg well synthetic results are useless if you are looking for best performer in games.

 

[JPB]

I for one plan on picking up 2-4 of these 3870's and a new X38 board when released. Wonder how well 4 of these would run Crysis ?

 

[aldamon]

Originally posted by: JPB
I for one plan on picking up 2-4 of these 3870's and a new X38 board when released. Wonder how well 4 of these would run Crysis ?

Since it will cost you over $1000, I hope very well for your sake.

 

[Rusin]

Does that 3-4 card systems work with other chipsets than AMD 790?

 

[bryanW1995]

Originally posted by: thilan29

Originally posted by: bryanW1995
I wonder if the ip35 pro is set up like the asus or the msi mobo in that comparison, thilan29? the asus appears to do much better in xfire.

What do you mean much better? vs. the other 2 boards? (it actually does worse except for one instance)

Here:
http://anandtech.com/mb/showdoc.aspx?i=3142&p=10

Performance seems to be very similar amongst the P35 boards so you'll probably get similar performance to the Asus if you went Crossfire.

If the Abit board was out when I bought my DS3 (Core 2 Duo launch...July 06?), I would have definitely bought that. Of course P35 wasn't out at the time

I meant from the bit-tech article http://www.bit-tech.net/hardwa...intel_x38_versus_p35/1 it showed the asus blitz extreme p35 mobo as being 8x8x instead of 16x4x. at the end of the review the blitz extreme was very close to the p5e3, and both were significantly better in xfire than the msi mobo.

 

[thilanliyan]

Originally posted by: bryanW1995
I meant from the bit-tech article http://www.bit-tech.net/hardwa...intel_x38_versus_p35/1 it showed the asus blitz extreme p35 mobo as being 8x8x instead of 16x4x. at the end of the review the blitz extreme was very close to the p5e3, and both were significantly better in xfire than the msi mobo.

Yeah I think Asus reconfigured the PCI-e lanes to make it 8x8. I'd say you can chalk up the MSI results to using 16x4. Although for ultimate performance you'd be probably be better off with Asus.

 

[taltamir]

so.... first release of GT. No crisys sli... 169.04 beta... 25% boost in sli. 169.09 beta 50% boost in sli and fixed water bug that should also improve single card performance... with a promise to make fixes to the game engine to make sli even MORE effective...

Gee I guess waiting until I game is actually finished might pay off in terms of it's performance this time... who knew! ;p

 

[SniperDaws]

Originally posted by: taltamir
so.... first release of GT. No crisys sli... 169.04 beta... 25% boost in sli. 169.09 beta 50% boost in sli and fixed water bug that should also improve single card performance... with a promise to make fixes to the game engine to make sli even MORE effective...

Gee I guess waiting until I game is actually finished might pay off in terms of it's performance this time... who knew! ;p

i always thought 50% was half of a full one, i mean if the SLI cards are performing at 25fps and a new driver comes out boasting of a 50% increase in performance then why isnt arent the sli cards performing at 50fps now, the only increase i saw with the review of Crysis and the new drivers was 2fps.

 

[taltamir]

50% boost = 1.5x of a single card.


So from 20fps it should go to 30... not 40. (your example was 25 go 50fps)

while with the 25% boost people reported 20fps on single GT and 25fps on two GT.

Your configuration might be such that you are cpu limited or some such resulting in such poor improvements...
Firstly, I would recommend you get more ram (have 4gb not 2).
Secondly I would wait until the game is released and has multi core support, that should drastically improve your performance.