Forward: official GIGA 965P-S3/DS3/DS4/DQ6 beta BIOS

[Marvin_The_Martian]

Well, my X-Fi never disappears with F8a, but my overclock that was previously stable with F4 is now unstable with F8a. I tried to lower the FSB a bit and stability actually got worse. This continues the trend I've seen since F5. I don't know what they're tweaking in these newer BIOSs, but it's killing my stability.

I think I'm going back to F4 until I get confirmation of better overclocks and better D9 compatibility. At least I have confidence we won't see the X-Fi issue again in the future.

Indeed F8a does not improve but limits oc, I found that to

[markopoleo]

e.


and if i had ADD which just so you fully understand means Attention Deficite Syndrome. would mean i cant stick around to let the AS5 setup... and i am constantly removing the heatsink and reapplying the AS5 in hopes it will get better.

before you think about insulting someone you might take a second and stfu and realise to whom your speaking.


now back on topic..

I was unaware ADD stands for something you describe as ADS... The medical community, and Artic Silver can't figure out anything, but then you come along to correct the mistakes they make.

[ScythedBlade]

Here's a POSSIBLE quick fix for D9 Memory: Using the F8a BIOS for the DS3, set the Memory Performance Enhance option to TURBO.

[Marvin_The_Martian]

Here's a POSSIBLE quick fix for D9 Memory: Using the F8a BIOS for the DS3, set the Memory Performance Enhance option to TURBO.

[Lestat]

yeah every so often i play around with turbo.. havent been able to see any difference, infact just the opposite.


but whats been the best you have achieved ?
wht bothers me the most is the fact that, you take the same D9 ram on lets say a P5B and you can hit 1000+ 4-4-4 but on this one you cant even do it with 5-5-5.
last time i came across this was ,, i dont recall what board i think it was when i switched from a NEO2 Plat to the Ultra-D. the Ultra-D had a better memory control and voltage system so it allowed it to take the ram tighter and higher.

i have assumed this was the case all the while with the DS3.
just a poor memory voltage and subsystem.

i also have had but can not even come close to confirming that boards such as the DS3 and Biostar 965PT are simply using failed 965 chipset. chipsets that would normally go into the Deluxe version or in this case the 965 DQ6 but have failed testing high extreme clocks so they simply put them in the low end budget boards.
this is a very possible situation in which case there wont ever be a fix.

but who knows.

[Marvin_The_Martian]

yeah every so often i play around with turbo.. havent been able to see any difference, infact just the opposite.


but whats been the best you have achieved ?
wht bothers me the most is the fact that, you take the same D9 ram on lets say a P5B and you can hit 1000+ 4-4-4 but on this one you cant even do it with 5-5-5.
last time i came across this was ,, i dont recall what board i think it was when i switched from a NEO2 Plat to the Ultra-D. the Ultra-D had a better memory control and voltage system so it allowed it to take the ram tighter and higher.

i have assumed this was the case all the while with the DS3.
just a poor memory voltage and subsystem.

i also have had but can not even come close to confirming that boards such as the DS3 and Biostar 965PT are simply using failed 965 chipset. chipsets that would normally go into the Deluxe version or in this case the 965 DQ6 but have failed testing high extreme clocks so they simply put them in the low end budget boards.
this is a very possible situation in which case there wont ever be a fix.

but who knows.

Don't think they speedbin chipsets... do they

Best I got the ocz at was 499 4-4-4 2.5v ( warranted ) and it wasn't stable. Compare that to those cheap Geil's I got in my sig!

Afaik memory performance option changes subtimings, open up memset make a screenshot at normal reboot with turbo and compare ( to lazy atm.. was an all nighter yesterday )

Edit: just realized, they were stable, finished spi32m atleast. Should run memtest I know

[Richard Dower]

Lestat (and others)...need your opinion on this. In regards temps and performance, which settings are better?

425x8=3400
378x9=3400

Consider the RAM is 756 vs. 850...pure performance which would be better.
Now also factor in temps and how hard the mobo/CPU has to work, is it true the NB runs hotter due to the 8x multi, or is it the higher FSB?

I did my third application of AS5, the second application was actually pretty good on examination, and my IHS and Ninja appear to be flat, using the razor blade test.

Temps are roughly the same, Orthos testing varies from 55-62C depending on the test pattern being completed.

Naturally if i got uber high speed fans things would change, but i really like these 1200 RPM Scythe fans...very very quiet. My Vcore is set in bios to 1.4000V, around 1.3600V actual. MCH and FSB set to normal, DIMM set to +0.375V.

[Marvin_The_Martian]

Lestat (and others)...need your opinion on this. In regards temps and performance, which settings are better?

A. High fsb helps performance
B. Idk nb clocks change with lower multi but it sounds plausible
C. Test it and tell us!

[Lestat]

i tested temps with my XP-120 yesterday on the biostar board and wow hot running chips even at 1.4v still way too hot for me. so went back to water.

rich.. same cpu clocks but higher ram is best.
martian, i am not aware that cpu multi's change NB clocks,, but it is a fact that higher FSB will change the NB latency. thats why on the 975x boards it will auto switch to the 1333 strap. higher ram = looser NB timings.
thats why when using the 1333 strap.. they clock higher but performance suffers.

lower cpu multi doesnt change it. although it is highly possible. yup very possible but i havent seen any of the more knowledgeable big boys say that.

richard is that the limit of the board at 9x ? or you just dont want to increase the vcore ?

the difference in those two clocks is only going to be about 1 second or less in 1mPi so the choice is yours.
as long as everything can handle it fine then no need to really worry just run the 8x multi.

also on the thought of nb latencies and the strap its using.
it is possible altough i dont know how to check this, that the DS3 may be stuck on the 1066 strap so it cant go any higher on the ram as the latencies on the chipset are too tight and since it can not switch to the 1333,, then the ram cant go higher.
heck it might be stuck on the 1333 strap i dont know. i havent heard amny people talk about what strap the 965 chipset uses. only the 975 chipset.

total guess work on my part though.

[Marvin_The_Martian]

afaik 965 usses only 1066.

I seen all this explained by Tony, I think it was on be I have to look it up to be sure.

[Lestat]

hey did anyone else notice that they changed the defualt vcore back to 1.23v ?

i just noticed that.


hey Rich,, talking about good AS5 application..
im sitting at 23c right now.. thats a good application lol

[Lestat]

this is the same results as my Biostar board
this is ONLY possible when using the 5:4 memory divider.
this is also 2.0v so not too shaby but its as loose as a hooker saturday morning.

but thats it,, any attempt to go higher it fails. voltage increase doesnt help either.

[ScythedBlade]

Okay .. new problem ... how the hell do I fix my PCI-Express running at x1 problem?

[Bromptoners]

I'm not sure if this helps with the new F8a bios for the D9 issues??

1) Load Performance Defaults instead of Load Fail Defaults
2) Press "Ctrl+F1" to enable the hidden BIOS features
3) Disable all unnecessary features in the Advanced BIOS Settings -- eg: Virtualization Tech, and other Intel so called smart technologies..
4) Turn off Smart Fan control in PC Health Status and let the fan spins at full speed, though it might be alot noisier
5) Set "System Voltage Control" to "Auto" ~~ Don't mess with the voltages first to verify my assumption..

As for me.. I'm able to get FSB 480 with RAM runs in 960MHz at 1:1 ~~ however, when I'm trying to check what chip was used in my Corsair CM2X1024-8500C5 1GB module, I accidentally ripped 1 memory chip off the stick when I'm removing the heat-spreader.. (Crying ~~~) The Chip used in my RAM is D9GMH ...

So.. I'm not sure if it's the chip difference or it's the method I used.. Cos I'm only able to run at single channel DDR now... Sigh.. I'm going to get broke again inorder to get a new memory module

[Lestat]

omg!!! you ripped off one of the chips ? !!!

wow thats some very expensive ram and now its ruined... wow

yeah the D9GMH and D9GKX are the two that have issues.

950-960 is usually around the area where it stops working. @ 4-4-4 5-5-5 sometimes get a little more,, and as i just showed using Auto SPD lets you get higher. higher but VERY loose. loose enough thats its not worth it.
but i have seen system voltage = auto have slightly better results.
so the "Auto" setting is doing something else with the mobo too.

what did you have your SPD timings set to ? 4-4-4 ? 5-5-5 ?

[ea6gka]

GMH is my ballistix and it does with 2.15V 1000 @ 4-4-4-8. Funny is if i set CL=5 it doesnt boot.

[Lestat]

ea6 is yours the DQ6 DS4 or DS3 ?
the DQ and the DS4 run the D9 ram fine from everything that i have seen.

[masse]

Hi. I'm using DS4 and F7b bios. When i set (G)MCH over +0.25 my PCI-Express run as X1. Can i fix this? How?
Thanks very much.

[Richard Dower]

Masse...email GB tech support about the issue.
In regards my own setup, when i left my PC this morning i was 18 hours Orthos stable at 3.4Ghz.

I am going to leave it run a minimum of 24 hours and likely 30 hours, thus i can assume it's 100% stable then?

[HamidFULL]

ea6gka
GMH is my ballistix and it does with 2.15V 1000 @ 4-4-4-8. Funny is if i set CL=5 it doesnt boot

whats your board? Gigabyte?

masse
Hi. I'm using DS4 and F7b bios. When i set (G)MCH over +0.25 my PCI-Express run as X1. Can i fix this? How?
Thanks very much.

only report it to Gigabyte TS for future BIOS

[Richard Dower]

I've ordered a second Scythe 120mm fan, i'll run two in push/pull at full speed...the fans are pretty quiet at 1200 RPM.

Thus i may get another 1-2C drop abd when the AS5 settles in it may drop another 2C, so i'll be within my 60C range.

Running Intel TAT and creating a lof file the temps never went above 64C and the majority of Orthos test patterns it was between 55-60C, some tests went to 62-64C.

It's looking good all round, hopefully a new bios update will allow me to drop the voltage further.

[Lestat]

your temps from what i am testing on my water(i have a really crappy pump so water flow is very low, temps would be better if i would get off my arse and install my good pump and 1/2" ID tubing) are pretty good actually not bad at all for air.
here is a shot of what i do.

notice the voltages (as i have a sh|t eating grin on my face) and yes the DS3 boots at 422x9 thatss why when people say buy the abit or the bad axe or some other baord thats 100-150 more,, i say why,, those baords are not any better at 9x.

[Richard Dower]

28 more posts and we hit 2,000!

:-)

[Lestat]

WOOT!

sad thing is this has become less and less about bios's and more about issues.
but its nice to see it all in one thread.


now if we can only get people to use the "search thread" function.

[Richard Dower]

I got another tech support email from GB about the SYS_FAN issue, they say they are going to test again, with my suggestions, and get back to me with the results.

My guess is they'll spout some crap that SYS_FAN being controlled by bios was removed as a feature and will no longer be available...wait and see!