-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD "Never Settle" 12.11 Driver - benchmarks
- Thread starter Final8y
- Start date
Soldato
- Joined
- 16 Jan 2004
- Posts
- 2,662
- Location
- Workgroup
Yeah most likely the case, but if the bug isnt fixed it leaves me without the possibility of upgrading drivers in the future!
@MeMeMeMe: Yeah the 6770 is the top/master card.
6770/5770
you can't crossfire a 5770 with a 6770
it as to be from the same series
Last edited:
6770/5770
you can't crossfire a 5770 with a 6770
Yes you can.
http://sites.amd.com/PublishingImag.../WebBannerJPEG/AMD_CrossfireX_Chart_1618W.jpg
Last edited:
That's not true but the slower card can sometimes hold back the faster under certain circumstances.
A 6770 is a re-named 5770, they are the same card and do work in crossfire.
ok I see it now
In practical terms there is load balancing. I believe that is correct, the more powerful card will be under utilised hence matching the speed of the slowest card.
I have ran my cards out of Sync, i have ran one as low as 500Mhz while the others were at 950Mhz.
Running them all at 500Mhz was slower.
http://forums.anandtech.com/showpost.php?p=33974086&postcount=20
http://www.amddevcentral.com/gpu_assets/Harnessing the Performance of CrossFireX.pdf
Last edited:
Soldato
- Joined
- 16 Jan 2004
- Posts
- 2,662
- Location
- Workgroup
I know when I got the additional 5770, I used Kombuster to make sure everything was stable, and with the 6770 overclocked from 850/1200 to 910/1300 and the 5770 running at stock clocks it basically ended up with the 6770 at roughly 90% load and the 5770 at full load.
I overclocked both to the same clock speed and it was 100% load on each. The 6770 isnt a great clocker
so I havent even tried to see how far the 5770 will go. I may try it now if it actually makes a little extra difference.
I overclocked both to the same clock speed and it was 100% load on each. The 6770 isnt a great clocker
so I havent even tried to see how far the 5770 will go. I may try it now if it actually makes a little extra difference.I know when I got the additional 5770, I used Kombuster to make sure everything was stable, and with the 6770 overclocked from 850/1200 to 910/1300 and the 5770 running at stock clocks it basically ended up with the 6770 at roughly 90% load and the 5770 at full load.
I overclocked both to the same clock speed and it was 100% load on each. The 6770 isnt a great clocker
I would look at the fps and not the load, i have used profiles on my Quadfire that have givem me 80 % load across all the GPUs and have given me 30fps i have then used a different profile that has only used 25% loading across all the GPUs and given me a solid 60fps in the same game.
This is game dependent on how its all utilized and being on quad GPU gives CF more leeway on unbalanced clocks.
There are some games that don't even trigger max clocks or much loading on the other GPUs until there is enough going on and mostly because i Vsync.
Last edited:
I am familiar with the clocks being different, it is the load balancing. I suspect if i paired an old 3870 with my 7850 i would cripple the performance and it would be lower overall. Thanks for sharing the info about the different clocks being better than equal clocks, that looks promising as it shows load balancing is not equally distributed. I still think it would be possible to reduce performance by matching two cards that are vastly different.
I would say its better to have equal clocks, performance can be affects by the slower card, its just not absolute, sometimes the OC on one is worth it and sometimes its not.
Running one of my GPUs at 500hmz on purpose did reduce performance, but running all of them at 500Mhz was even worse, so its not a case equal is better as in speed but a case of one being higher potentially will gave more performance than all being the lower.
I would say forced unequal clocks could add more micro stuttering.
Last edited:
Thought this would be the best place to ask this:
To anyone that's using a 7970 is AB, is the voltage bump you apply in AB reflected in your AB voltage monitoring?
When I bump the voltage on my 7970 to 1.25v it stays at 1.187 in AB voltage monitoring.
To anyone that's using a 7970 is AB, is the voltage bump you apply in AB reflected in your AB voltage monitoring?
When I bump the voltage on my 7970 to 1.25v it stays at 1.187 in AB voltage monitoring.
Voltage droop, i get the same thing, when loaded the voltage drops anyway and uping it only makes a small difference once loaded.
at stock (1.218v ends up 1.187v and 1.256v ends up as 1.225v) that's in Sapphire TRIXX, MSI AB is even worse, in that 1.256v ends up 1.222v when loaded.
Last edited:
I'm not sure that accounts for it. I just did a quick test.
One run of Heaven at 1.131v: I get 1.187v in AB and GPU-z stays around 1.131v.
One run of heaven at 1.3v: I get 1.187v in AB and GPU-z, again, stays around 1.131v.
No difference in temperatures either.