Asus P5B DX 56K WARNING

Printable View

Show 100 post(s) from this thread on one page

12-27-2006, 06:17 AM
Blauhung

Quote:

Originally Posted by heikis

i was using the 1333 strap all along (333MHz referene frequency). but it wont even POST at 266MHz reference frequency. i tried playing with the voltages and so on. anybody have this problem with the lates bios (2395)?

what FSB were you trying to use on the 1066 strap?

Quote:

Originally Posted by raj_kemp

Hi,

Recieving BX2 tomorrow and have downloaded 2395 BIOS from Intel site. Reading posts about not able to hit 400 FSB with this BIOS, can anyone post/link the previos BIOS 2333 or 2330 as I can't find them on Intel site.

Thanks

Edited: Found them on Intel site...

Still testing, but i think it's more an issue with people using the default reference frequency. I'm only drawing inferences from what others have said, but Intel probably fixed a few memory timing issues from 2333 --> 2395 and this boosted the possible performance on the 1066 strap but lowered its workable FSB range. Keep in mind this board is geared more towards high multiplyer CPUs and most likely the bulk of the BIOS departments efforts are going to be going towards improving timings and performance at sub400MHz FSB speeds where there is more then enough data bandwidth to feed the CPU and still hold tight latency. If you want to move 400+ on the FSB you will be using the 1333 strap, and I don't know how that got changed in newer BIOS revisions.
12-27-2006, 07:01 AM
heikis

Quote:

Originally Posted by Blauhung

what FSB were you trying to use on the 1066 strap?

400fsb...
12-27-2006, 07:54 AM
Tony

Guys

The tweak I quoted will limit you to 335 ish with a true conroe and 340ish with an allendale, so you REALLY do need a 6700 minimum as you need the high multi to get the CPU horsepower you need....this is why I want to test with a 6800.

Intel are looking to add more tweaks into bios, I asked for latency tweaks away from strap so you may see a bios that allows tighter latency with the 1333 ram table,this would make the board killer IMO.

I see a 3 horse race now with 975 boards, Badaxe2 Vs Infinity 975 Vs Aw9D max, out the 3 boards here the Infinity clocks the best in 1067, the badaxe2 has fully working 1333 and is able to force the 800 strap as does the AW9D....All 3 are awesome IMO with just price splitting them apart ;)
12-27-2006, 10:17 AM
xgman

Quote:

Originally Posted by Tony

If you have a 6800 you may want to try this.

Set 200 reference, set 533 for the ram (3:4) and 333fsb for the CPU. If you have good memory you should be able to set 4-4-3-8, TRAS dead low will actually hurt performance.

Now with the 6800 you will set a 12 multi for 4GHZ or 11 for 3.6+.

Access latency will be somewhere around 6ns faster in this config than running the ram 1:1 even if you are running well over 900 1:1 in 1067ref mode.

Im trying to source a 6800 to test here, im stuck with a 6700 so 10 multi is all i have but I have some data i can share.

Goldmem pro used for bandwidth, everest for latency.

Code:

memspeed FSB REF Strap ratio bandwidth latency 833 333 267 4:5 4736 61.4ns 833 311 200 3:4 4639 55ns 888 333 200 3:4 4965 51.5ns 888 355 267 4:5 5041 57.6ns

CPU speed was manipulated with the strap to force given ram speeds, even though 4:5 267 REF at 888 had the highest bandwidth its access latency is no where near as good as 888 200REF, which will make the 200REF setting faster.Now I know the CPU would be faster at 355fsb but we are looking at the memory here not the final CPU speed.

Remember C2D does not need tons of bandwidth, it just needs to use the bandwidth more efficiently. Lower access latency = more performance.

Tony

Tony, I can try this with my 6800 later, but since I am on phase and my cpu will run at 4500, how should I adjust your suggeted settings for that?

Also I am using 2gig of ocz flex 9200 with 2 more sticks on the way. How woud you suggest I adjust for running 4g sticks of the flex in the above scenario? Thanks.
12-27-2006, 10:25 AM
Tony

just set a 13 multi and off you go.

4 gigs may need slacker timings, thats all. 4-4-4 maybe
12-27-2006, 10:29 AM
xgman

thanks.
12-27-2006, 10:33 AM
xgman

Tony, you suggested using Everest for latency check. Are you not getting the same Everest vcore re-setting in bios issue that everyone else seems to be?
12-27-2006, 10:45 AM
Tony

im using an old build, one that does not crash my system.

you could also use sandra, but you will have to test to get your own base line figures.

T
12-27-2006, 11:02 AM
eva2000

why not use mbench for latency/bandwidth http://www.x86-secret.com/Download/mbench.zip ?
12-27-2006, 11:33 AM
mnemonik23

Quote:

Originally Posted by Tony

Guys

The tweak I quoted will limit you to 335 ish with a true conroe and 340ish with an allendale, so you REALLY do need a 6700 minimum as you need the high multi to get the CPU horsepower you need....this is why I want to test with a 6800.

Intel are looking to add more tweaks into bios, I asked for latency tweaks away from strap so you may see a bios that allows tighter latency with the 1333 ram table,this would make the board killer IMO.

I see a 3 horse race now with 975 boards, Badaxe2 Vs Infinity 975 Vs Aw9D max, out the 3 boards here the Infinity clocks the best in 1067, the badaxe2 has fully working 1333 and is able to force the 800 strap as does the AW9D....All 3 are awesome IMO with just price splitting them apart ;)

I wish I knew this before :( Bunch of posts were suggesting 975 for 6600 and up.

Tony, what would be the optimal setings with 6600?
12-27-2006, 12:16 PM
screwtech02

Quote:

Originally Posted by mnemonik23

I wish I knew this before :( Bunch of posts were suggesting 975 for 6600 and up.

Tony, what would be the optimal setings with 6600?

I aggree, seems that most if not ALL the reviews of boards now days use only the biggest/fastest/$$$$$ proc out there.... Kinda hard on the guys that want to use midline procs and overclock, there seems to be no info on these grounds...

Also, seems to be very few using 4x1024, and overclocking, or people are not coming forward with the results/settings ect.....

Maybe its just me......:rolleyes:
12-27-2006, 12:53 PM
Blauhung

Quote:

Originally Posted by mnemonik23

I wish I knew this before :( Bunch of posts were suggesting 975 for 6600 and up.

Tony, what would be the optimal setings with 6600?

The best performance at a given clock speed will most likely come from the lowest stable strap setting. From the sounds of things FSB will be held to ~335 on the 200 strap, ~380 on the 266 strap (a little higher on BIOS-2333 but with funky RAM timing issues), and mid 400's on the 333 strap. Probably the best way to work our your settings is to start on the 333 strap and move the FSB up until you are either above 400 or you have hit the wall on what your CPU can do. If you are above 400, stick with the 333 strap, if you are below, then move down to the 266 strap and look for stability at lower FSB speeds as i'm guessing there's something like a 20MHz gap in performance from a strap change (such as 380FSB on the 266 strap could be better performing then 400 on the 333 strap).

There is no "best setting" it all depends on what your individual CPU can do with the cooling you specifically have setup, and then finding the lowest timings on the chipset and memory that you can support at the FSB it takes to get that CPU speed.
12-27-2006, 01:11 PM
gtj

Quote:

Originally Posted by xgman

Tony, you suggested using Everest for latency check. Are you not getting the same Everest vcore re-setting in bios issue that everyone else seems to be?

You can use both everest and sandra to run the benchmarks, just don't try to read any sensor info. They give VERY different latency readings though. The RALPH machine in my sig gives a 47.2ns latency in Everest but a 62ns in Sandra.

If you start everest with the following in the everest.ini file (just those 2 lines, nothing else), it'll start up without hanging.

Code:

[Generic] HWMonEnableSensorIcons=0
12-27-2006, 01:21 PM
xgman

Quote:

Originally Posted by gtj

You can use both everest and sandra to run the benchmarks, just don't try to read any sensor info. They give VERY different latency readings though. The RALPH machine in my sig gives a 47.2ns latency in Everest but a 62ns in Sandra.

If you start everest with the following in the everest.ini file (just those 2 lines, nothing else), it'll start up without hanging.

Code:

[Generic] HWMonEnableSensorIcons=0

Thanks for the tip.
12-27-2006, 01:30 PM
xgman

Quote:

Originally Posted by gtj

You can use both everest and sandra to run the benchmarks, just don't try to read any sensor info. They give VERY different latency readings though. The RALPH machine in my sig gives a 47.2ns latency in Everest but a 62ns in Sandra.

If you start everest with the following in the everest.ini file (just those 2 lines, nothing else), it'll start up without hanging.

Code:

[Generic] HWMonEnableSensorIcons=0

I didn't have an everest.ini file in my Everest ultimate install directory. I copied one there containing what you posted but it still loaded all the tray monitors. ??
12-27-2006, 01:55 PM
gtj

Quote:

Originally Posted by xgman

I didn't have an everest.ini file in my Everest ultimate install directory. I copied one there containing what you posted but it still loaded all the tray monitors. ??

Weird. I just checked again and that's exactly what I have. I got this from the everest bug forum.

"c:\Program Files\Lavalys\EVEREST Ultimate Edition\everest.ini" right?
12-27-2006, 01:58 PM
xgman

Quote:

Originally Posted by gtj

Weird. I just checked again and that's exactly what I have. I got this from the everest bug forum.

"c:\Program Files\Lavalys\EVEREST Ultimate Edition\everest.ini" right?

right. hmm, well I guess if you disable/clear everything in the preferences/OSD section it has the same effect?
12-27-2006, 02:13 PM
Blacklash

No issues with BIOS 2395 here and a 402 FSB. So far I like it.
http://img469.imageshack.us/img469/2...dec2zr5.th.jpg
12-27-2006, 02:47 PM
Ph03n1x

I recieved BX2 few days ago...

Here are some results with stock cooling on northbridge and processor.

http://img178.imageshack.us/img178/4...pi1mbq7.th.jpg

http://img178.imageshack.us/img178/2...tionqe7.th.jpg

i like this board very much...with hr05 and sonic tower installed i think i could reach FSB480 easily
12-27-2006, 02:54 PM
bobov

Quote:

Originally Posted by malficar

No issues with BIOS 2395 here and a 402 FSB. So far I like it.

266 strap or 333 strap? Thanks.
12-27-2006, 03:38 PM
heikis

Quote:

Originally Posted by Ph03n1x

I recieved BX2 few days ago...

Here are some results with stock cooling on northbridge and processor.

http://img178.imageshack.us/img178/4...pi1mbq7.th.jpg

http://img178.imageshack.us/img178/2...tionqe7.th.jpg

i like this board very much...with hr05 and sonic tower installed i think i could reach FSB480 easily

what is your reference frequency set in bios?
12-27-2006, 04:10 PM
Blacklash

Quote:

Originally Posted by bobov

266 strap or 333 strap? Thanks.

266. I've only moved to 333 when I was briefly benching 3906MHz.
12-27-2006, 04:12 PM
Blacklash

Quote:

Originally Posted by Ph03n1x

I recieved BX2 few days ago...

Here are some results with stock cooling on northbridge and processor.

i like this board very much...with hr05 and sonic tower installed i think i could reach FSB480 easily

Good show.
12-27-2006, 06:37 PM
Tony

Quote:

Originally Posted by malficar

266. I've only moved to 333 when I was briefly benching 3906MHz with a 434 FSB and my memory @ 4:5 DDR2 542 4-4-4-4.

There is no 4:5 in 333ref mode on badaxe2. 667 is 1:1 and 800 is 5:6 ;)
12-27-2006, 09:31 PM
Blacklash

Quote:

Originally Posted by Tony

There is no 4:5 in 333ref mode on badaxe2. 667 is 1:1 and 800 is 5:6 ;)

I was rattling off what I saw with CPU-Z without thinking, and I am still not use to all the strap business with Intel overclocking:

http://img486.imageshack.us/img486/7...p1msar9.th.jpg

Until Conroe 99% of my CPUs were AMD. If I would have stopped and thought about it I would have realized your point. :p:

Show 100 post(s) from this thread on one page

Testing / comparing : Intel D975XBX2 / Asus P5B DX ***56K WARNING***

Testing / comparing : Intel D975XBX2 / Asus P5B DX 56K WARNING