P5B Deluxe 0609 BIOS FIX Memory 4:5 Problem

[AndreYang]

1. Support Conroe XE CPU(ES version) can adjust ratio to 13 (original is 11)

2. Patch OCZ2G8002GK-6400 can't boot issue

3. Fix memory 4:5

4. Fix FSB 395~399 stable


http://www.iamxtreme.net/andre/P5B-ASUS-Deluxe-0609.rar

[hatakiller]

has any1 tested this yet

[bachus_anonym]

Thanks Andre

I flashed it 10 mins ago. Running SPi32M @ 400x9 4:5 as we speak. Next I will do 450x8 4:5 and post results afterwards.

[Supertim0r]

thanks a lot Andre

[aggybong]

Is this official or still a beta?

[AndreYang]

Is this official or still a beta?

It's beta.

[thelostrican]

how about temps is it better now than before?

[pcdoc1]

thanks alot Andre

[metro.cl]

tomorrow i'll have this mobo shipped

hope it ocs well

[shine121]

thanks, will try later

[Supertim0r]

also priming @ 400x9 DDR2 @ 1000
was impossible before

[ScottFern]

Is 4:5 a better (faster) memory divider than 1:1 ???

[Kashath]

Yeah

[Raz2133]

how about temps is it better now than before?

I noticed that since BIOS 0607 (maybe 0605?), the temps reported by the Asus Utility are a lot closer to those reported by Coretemp and the Intel Thermal Analysis Tool.

[bachus_anonym]

OK, so check this out...

I'm on stock heatsink at this moment and my CPU maxes out at 3.60GHz, 32M stable. So, I first run SPi32M @ 400x9=3600MHz, DDR2-1000 (667MHz ratio, 4:5) and got 13:42.797.
Above result is about right, since at 400x9=3600MHz, DDR2-1200 (800Mhz ratio, 2:3) and same conditions I get 13:15.437.

Then I dropped multi to x8 and it appeared to boot at max 440x8. As it was not very stable, I lowered FSB to 432x8=3456MHz, DDR2-1080 (667Mhz ratio, 4:5) and got 14:40.641. Whoa, that's quite drasatic 58s loss! That's at lower CPU clocks, OK, but higher memory frequency (at least, according to CPU-Z). There would be drop, but this is a big one!

Now, this was for a comparison. I upped mulitplier back to x9 and decided to do one more 32M run at the same CPU clocks as above, therefore lower memory clocks. So, I'm at 384x9=3456MHz, DDR2-960 (667MHz ratio, 4:5) and got 14:11.250, wth

So, now the only thing that came to mind was to run 432x8=3456MHz, DDR2-864 (533MHz ratio, 1:1), which got me 15:11.266.

I repeated 432x8 run again and got the same result as the first time... Some more testing must be done, but it appears to me that perhaps past 400MHz and ratio 667Mhz selected from BIOS, performance really goes down the drain Would that be possible that internal chipset latency gets altered at this point, causing such a drop in performance?

[astaris]

OK, so check this out...

I'm on stock heatsink at this moment and my CPU maxes out at 3.60GHz, 32M stable. So, I first run SPi32M @ 400x9=3600MHz, DDR2-1000 (667MHz ratio, 4:5) and got 13:42.797

So, I'm at 384x9=3456MHz, DDR-960 (667MHz ratio, 4:5) and got 14:11.250

Then I dropped multi to x8 and it appeared to boot at max 440x8. As it was not very stable, I lowered FSB to 432x8=3456MHz, DDR-1080 (667Mhz ratio, 4:5) and got 14:40.641

Some more testing must be done, but it appears to me that perhaps past 400MHz and ratio 667Mhz selected from BIOS, performance really goes down the drain Would that be possible that internal chipset latency gets altered at this point, causing such a drop in performance?

Maybe past 400 Mhz the bios select the 1333 chipset strap and gives you an effective ratio of 5/6.

[dinos22]

Bachus check memtest bandwidth between the two

if you see a big drop you'll beed to investigate timings (i've noticed Gigabyte DS3 do that)

[Raz2133]

Unfortunately, this BIOS makes my previous dual Super-PI stable 3.4Ghz overclock (E6400 @ 8x425) totally unstable, it errors out within a minute or so.

I can't even get it stable at my previous 24/7 speed of 3Ghz (8x375).

I have a feeling that the new timings in this BIOS are messing with my crappy Samsung RAM. Arh well, I guess I will fall back to a previous BIOS, at least until I have some better RAM to try.

[bachus_anonym]

Maybe past 400 Mhz the bios select the 1333 chipset strap and gives you an effective ratio of 5/6.

That's my suspicion as well

Bachus check memtest bandwidth between the two. if you see a big drop you'll beed to investigate timings (i've noticed Gigabyte DS3 do that)

I can't find my Memtest CD, so I decided to give Sandra UnBuffered Memory Bandwidth benchmark a try... I think it confirms loss of memory bandwidth at 667MHz ratio and past 400MHz

384x9=3456MHz, DDR2-960 (667MHz ratio, 4:5) --> 5537 / 5602 @ 45% / 46%

400x9=3600MHz, DDR2-1000 (667MHz ratio, 4:5) --> 5702 / 5785 @ 45% / 45%

432x8=3456MHz, DDR2-1080 (667Mhz ratio, 4:5) --> 4903 / 4965 @ 32% / 32%

400x8=3200MHz, DDR2-1000 (667Mhz ratio, 4:5) --> 5449 / 5488 @ 38% / 38%

It blows...

[Rattle]

this bios also alows me 400x8 with 4:5 priming 1 hour now will let it go over night.

ram auto'd at 5-4-4-12 2.25vdimm nice ddr2 1000

[GAR]

ran into some issues with this bios, flashing back to 0507, poor performance as well.

[dinos22]

That's my suspicion as well I can't find my Memtest CD, so I decided to give Sandra UnBuffered Memory Bandwidth benchmark a try... I think it confirms loss of memory bandwidth at 667MHz ratio and past 400MHz

384x9=3456MHz, DDR2-960 (667MHz ratio, 4:5) --> 5537 / 5602 @ 45% / 46%

400x9=3600MHz, DDR2-1000 (667MHz ratio, 4:5) --> 5702 / 5785 @ 45% / 45%

432x8=3456MHz, DDR2-1080 (667Mhz ratio, 4:5) --> 4903 / 4965 @ 32% / 32%

400x8=3200MHz, DDR2-1000 (667Mhz ratio, 4:5) --> 5449 / 5488 @ 38% / 38%

It blows...

yeap it has to be something to do with >400 FSB

just manually set ALL the RAM timings if you haven't already

maybe they are forcing 1333 strap somehow past XXXMHz fsb

you able to do some 6xmulti tests with different FSBs

[mackanz]

Then the difference should be very visible between 9X399 and 9X400. Do a comparison with SPD enabled to take everything out of the equation. Theoretically, the difference shouldn't hardly be anything in either superpi or latency tests, but sandra might show a huge difference (as usual )

[bachus_anonym]

yeap it has to be something to do with >400 FSB. just manually set ALL the RAM timings if you haven't already. maybe they are forcing 1333 strap somehow past XXXMHz fsb. you able to do some 6xmulti tests with different FSBs

Then the difference should be very visible between 9X399 and 9X400. Do a comparison with SPD enabled to take everything out of the equation. Theoretically, the difference shouldn't hardly be anything in either superpi or latency tests, but sandra might show a huge difference (as usual )

I have everything set in BIOS manually - all voltages, timmings, ratios, etc. Also, it's clearly happening past 400MHz. For now, I flashed back to 0507 becasue I have another OCing Report to finish up first. But I will come back to 0609 in a few days

I'm actually now trying to remember If I haven't by any chance seen something similar on 0507 (and I guess earlier BIOSes) when using DDR2-800 ratio. I remember I run 16M @ 400x9 @ 2:3. Then I booted to 404x9 @ 2:3 and time was very noticeably slower than at 400x9... I was very surprised. I'll try to replicate that and post later.

[ent]

So this performance drop happens only with 0609?