So, do we know anything about RV670 yet?

Those overclock results from the HardwareCanucks review are phenomenal.

Are those typical results? Does the G92 overclock well?
860MHz core is about an 11% OC for the 3870, you can buy factory-overclocked 8800GT's with a 17% OC (700MHz). When you consider their relative cooling solutions, transistor counts and process technology the 8800GT looks like the more impressive overclocker so far. But it's early days for both GPUs so time will tell.
 
860MHz core is about an 11% OC for the 3870, you can buy factory-overclocked 8800GT's with a 17% OC (700MHz). When you consider their relative cooling solutions, transistor counts and process technology the 8800GT looks like the more impressive overclocker so far. But it's early days for both GPUs so time will tell.

You have to factor the actual performance gains when considering the raw clock increases. So far we have seen G92 is more bandwidth and shader-compute bound than pixel/texture fillrate bound, so increasing the core clock brings very little performance gain (if any). This is not true with RV670, since the SPs are in the same clock domain as the RBEs/TMUs.
 
860MHz core is about an 11% OC for the 3870, you can buy factory-overclocked 8800GT's with a 17% OC (700MHz). When you consider their relative cooling solutions, transistor counts and process technology the 8800GT looks like the more impressive overclocker so far. But it's early days for both GPUs so time will tell.

One thing to take into consideration is that Nvidia's AIB's bump the core voltage as well. Something you can't do in CCC. You will have much better luck OC'ing RV670 in tray tools or ATiTool.
 
I guess I was just so impressed with how much performance was gained once RV670 was overclocked in general. I mean, there were very significant gains to be had and I haven't seen any articles regarding G92's potential in the same area.

Just curious.
 
I guess I was just so impressed with how much performance was gained once RV670 was overclocked in general.
I think the huge OC on the video memory (+500MHz) had more to do with those big gains (which were tested with plenty of AA) than the small OC on the RV670 (the GPU) itself.

I was referring more to the relative overclocking headroom on the GPUs themselves than the card as a whole. But you're right about the overclocking potential of the HD3870 as a complete card, pretty impressive.
 
I think the huge OC on the video memory (+500MHz) had more to do with those big gains (which were tested with plenty of AA) than the small OC on the RV670 (the GPU) itself.

I was referring more to the relative overclocking headroom on the GPUs themselves than the card as a whole. But you're right about the overclocking potential of the HD3870 as a complete card, pretty impressive.

HD2900 has more membandwidth to play around with anyway, so it doesn't explain it, not all of it anyway
 
Just got some interesting info from The Stilt at MuroBBS :oops:
He's been testing a HD3870, and like pretty much all reviewers who have OCd the card, reached ~860MHz.
He said, that ATI has already released a new BIOS to fix an issue with the current/review cards - they have the PLL VIO divider set wrong, which prevents the cards from reaching over 862MHz on core. However, most of the retail cards should carry the new BIOS, luckily.

He promised to report with the new BIOS on OCs tomorrow
 
Just got some interesting info from The Stilt at MuroBBS :oops:
He's been testing a HD3870, and like pretty much all reviewers who have OCd the card, reached ~860MHz.
He said, that ATI has already released a new BIOS to fix an issue with the current/review cards - they have the PLL VIO divider set wrong, which prevents the cards from reaching over 862MHz on core. However, most of the retail cards should carry the new BIOS, luckily.

He promised to report with the new BIOS on OCs tomorrow

Interesting. Is the core speed likely to visibly improve performance as much as the memory has shown to be capable of? Or is it looking like memory speed will be the determining factor?

Either way, it'll still be interesting to see what the core ceiling is with the new BIOS...

~kanya
 
Ati seems to have most of the problems with 4xAA/16AF performance... at this settings Nvidia 8800GT has highest advantage against 3870...(19%)
At 8AAx/16AF Ati's new baby beat the green goblin's superstar (aka 8800GT) with 5%(!), as was revealed in this excellent review :

file:///D:/New%20Txt/translate.htm ... ( translated from computerbase.de)

I think this give us food for thinking...lot of...
hope 'the experts among us' will analyze the situation deeper...
 
Hello Revan,

can you kindly share your D: partition over teh internet, give us your ip, so we can read the translation? kkthx.

:p

Does the first poster allow to put the url link and edit his error? :smile:

I remembered this same thing happened to me too :oops:
 
Interesting and yet confusing article with their relative percentages. If you look at the raw numbers you'll see it's a mixed bag.
 
Just got some interesting info from The Stilt at MuroBBS :oops:
He's been testing a HD3870, and like pretty much all reviewers who have OCd the card, reached ~860MHz.
He said, that ATI has already released a new BIOS to fix an issue with the current/review cards - they have the PLL VIO divider set wrong, which prevents the cards from reaching over 862MHz on core. However, most of the retail cards should carry the new BIOS, luckily.

He promised to report with the new BIOS on OCs tomorrow

So have you heard any update on this yet?
 
<offtopic>
Minor request here to a moderator, but can we get this thread split out somewhat -- at least the recent and very interesting algorithm discussions? It seems like they might be getting lost in the sea of RV670 discussions.
</offtopic>
 
Back
Top