http://www.youtube.com/watch?v=FUyCBbZb3Y0 :censored:Quote:
Originally Posted by G.Foyle
Printable View
http://www.youtube.com/watch?v=FUyCBbZb3Y0 :censored:Quote:
Originally Posted by G.Foyle
that yours Pt1t?
Doing some tests I'm finding lower dividers are harder to hit higher + offer more performance at same clocks. That said hitting higher ram clocks at the highest divider isn't that surprising.
Pt1t where did you get that bios for the biostar. Cause i haven't seen it show up their site yet. Thats the only thing that makes me hate life about that board :(
Check the TW site. BIOS is still beta, released yesterday.Quote:
Originally Posted by TsunamiJuan
Not having any luck with it either, doesn't like any of the kits of mem I got here.
Yeah i saw that it was on the TW site, shortly after I asked that biostar sent me a link to it in my email :D
What I have noticed is that in general your going to have to play with the memory timings. Its not limited to the biostar board though the intial release would not read most memory correctly.
I am using some Dominator cmgtx3's 2400 2gbx2's and some mushkin redline hypers. So far the hypers clock about 100mhz further on this platform than the dominators do.
Gonna load up this new bios in a bit and check it out.
Yeah, this board has been kicking me in the nuts, while the gigabyte is a treat to play with.
I don't leave anything on auto, spent this morning doing timings one by one trying to even get 1866 to boot. Either my CPU is crap, or the board hates my mem, because the memory is just fine on the gigabyte. If you find something, please let me know, as I'm working on a review, and CANNOT even get 1866 @ stock on this Biostar board. I can't post a review when I can't even get stock.:p:
Last IIRC the biostarr will only run 1 stick of 1866, or if you do try to run it, the sticks will have to be in sockets 1 and 3, and you will have to play around some with memory interleaving.
easiest way to get it to work is to clock it down to 1600 and then bush the bclk up from there. Atleast that is what i have had to do.
I am working on a review aswell. I spent all of lastweek fighting with the biostar. Before I ordered the Asus f1a75-pro and the gigabyte ud4. Like you said the gigabyte seems to have the advantage right now. Though it does not support Multipliers past the chips set standard multi. The Asus will provide more voltage, but its allways high by .01 to .05volts, despite setting down the vrm controls. Which is the worst thing for these chips. The higher you have to crank the voltage on any of the Fusion chips the heat will increase more dramaticaly than on say a phenomII. I am guessing this is due to the IGP. The Asus board also currently lacks good IGP control.
Thanks very much. Sounds like we're in the same boat then, and that's very reassuring. I'll have to leave it be and move on to the next one and let Biostar sort it out then, as I kinda figured as much, seeing how both the A+ and M+ give the same problems(not like they are any different, anyway). Hopefully we'll be able to do like P1t1...I damn near died when he posted that. :ROTF:Quote:
Originally Posted by TsunamiJuan
:shrug:
Yeah I am busy chaseing max clocks right now. I am gonna try out the K10stat stepping changes tonight. I took 2nd lastnight on the bot. But Decided to stop when I realized I was starting to run into condensation in the ram sockets, AGAIN :\
If we the stepping changes work this will become legit.
http://img16.imageshack.us/img16/224/6900mhz.jpg
I doubt it will get that far, even c3 stepping only gained us about 500-600 from c2 on deneb. We will be lucky to be breaking 6 gig.Quote:
Originally Posted by TsunamiJuan
You mean 7gig?
No I mean 6 gig.Quote:
Originally Posted by eXa
C2 retails hit 6.5 to 6.6, C3 hit 7.2-7.3
Currently Llano is hard pressed for 5.5 and thats not due to lacking unlocked multis.
How much do you really expect from a stepping refresh?
1.5 gig?
You can't be serious.
These chips are not like deneb, not by a long shot.......
Im not expecting anything, i have no experience with sub zero cooling. But whats up with that 6.9Ghz screenshot then?
Edit: Nvm, i got it now.
That ss is whats currently considered to be a bug with the current llano 3850 and 3650 unlocked chips. They Will show that as the clock speed but there is little to no performance increase. As well as you can watch the cores gate themselves more rapidly to prevent over heating(atleast it appears thats what they might be doing).
Thats the sole reason why this is not something thats been posted to the bot. Till I can show an actual performance increase before putting it up there.
Though that being said the multipliers do seem to have a cutoff point for what the system will boot at. Previously the best i had achieved on a single stage was just bearly 6ghz (bugged obviously). I could watch the core heat under load create instability.
As far as wisdom on the platform I yeild to your experience Chew. You have far more experience than I do, and more internal knowledge.
Yes Chew, youtube video is my one.
TsunamiJuan :
On retail one, rising multiplier above 29X = no gain of performance.
For those who use the latest Biostar Bios , yes 1866 doesnt work correctly yet and some other issues will be OK in new bios i asked. It will be normally release today :D
Chew: It is a bit strange...It is at Llano coldbug or not? Usually is new process better for OC potential (air cooled), if architecture is simillary...And aircoolied seems at 3800 MHz as top stable. Where could be a problem?
You see, the core has changed a little bit to improve CPC / power consumption and there's also a GPU on the die. Also board power delivery is weaker.Quote:
Originally Posted by FlanK3r
So next 3870 with unlocked multiplier will be on a different stepping?Keeping the same process and oc potential will just be non-sense...
Focus for Llano was area density because of huge transistor budget they needed to fit all the blocks. Denser transistor layouts on same process usually means compromised clock scaling. It is not without a reason why we have CPU cores containing only 900mil trannies taking around 330mm2 on 45nm and GPU's with much denser logic circuits containing 2.1billion trannies around same size on 40nm.
Simply high clock requires completely different approach to designing your circuits and sacrificing area density for clock speed.
The real deal :cool:
http://img29.imageshack.us/img29/8792/snapshot0006g.png
No there is no coldbug that i have seen, voltage scaling is just about non existant though. Look at the facts. The cpu competes with the GPU for power and to add more fuel to that fire they took away pins, pins that deliver power in order to get the socket/cpu into a lower TDP range.Quote:
Originally Posted by FlanK3r
Nice run Alex, i've been wanting to test 01, looks like 50K will be easy.
We have to remember, that Llano wasn't designed for uber overclocking. yes I know we still like to play with it, but I suppose they tweaked the transistors for power use not performance or clockability. I think BD will show us what 32nm from AMD is all about.
Thnks,stupid board wont clock mem as my maximus,which runs 2300 6-10-6 no problem.And not an inch over 147 fsb when memory is high,i can run 150 with lower divider but performance is crap.Been waiting for a month for a new bios but asus bastards only focus on 1155 boards,screw them :)
The llano's seem to be very sensitive about memory timing. I've had to experiment and run a very different set of timings than I have been running on my 1155 hardware.
At the same time if you notice the IGP pixelating tearing and artifacting before you you even start to really push. Theres a good chance that the memory timings are to sloppy for its likings. I have noticed this across several motherboards. I am gonna see if it does the same later on my biostar TA75A+. So far both my ASUS F1A75-V pro, and Gigabyte GA-A75-UD4 have both shown this behavior.
On the mention of a cold bug. I've been running a full pot of ln2 without problems.
Round 3 or something like that? I lost count.
01 quick run for a baseline.
http://chew.ln2cooling.com/qdig-file...%203d%2001.jpg
AM3 quick run for a baseline.
http://chew.ln2cooling.com/qdig-file...178K%20AM3.jpg
Then I went back to DX 9 benches and pushed the scores further.
35K 03 done.
http://chew.ln2cooling.com/qdig-file...2003%2035K.jpg
05 not much just a tad.
http://chew.ln2cooling.com/Llano/Llano%2005%2021436.jpg
06 12K done.
http://chew.ln2cooling.com/qdig-file...2012K%2006.jpg
Enjoy.
One of these days I might be able to afford some ln2 and do some proper scores ;)