D975XBX Badaxe 1333fsb and 4:5

[eva2000]

I see some folks say 4:5 on 1333fsb isn't working and think it's 1:1 but.........

I thought reason why Badaxe 4:5 can clock so much higher is that it is due to the less agressivenss of the MCH settings at 1333fsb to allow fsb to clock higher at least on badaxe... ?

Why else would at 1333fsb strap 1:1 is slower than 1333fsb strap 4:5 for me at same fsb and mem timings on cpu and ram ?

I am on 1333fsb strap and 2x512MB OCZ PC5400 Plat XTC max at around 423-426mhz at 2.2v on badaxe, when i tried 4:5 with 346fsb for 433mhz ram it wouldn't boot so i think 4:5 is indeed working.

i.e.

2x 512MB OCZ PC5400 Plat XTC maxes at 2.2v @423-426mhz 4-4-4-8/15

2x 512MB Corsair PC5400ULv1.5 Fatbody D9 Micron so far at 2.2v @506mhz 4-4-4-15 and still haven't tried higher yet.

Click image for full timings and mem screenshot


At 3649Mhz 9x405fsb and 4:5 506mhz 4-4-4-15 only 8 seconds slower than @3809Mhz (9x423fsb) and ram @423mhz 4-4-4-8 of 15min 05s!

5400ULv1.5 now at 3816Mhz 9x424fsb 4:5 with ram @530mhz 4-4-4-15 at 2.2v



Crazy 530mhz 4-4-4-15 at 2.2v which is 33 seconds faster than @3809Mhz (9x423fsb) and ram @423mhz 4-4-4-8 of 15min 05s

I think reason Badaxe clocks higher is MCH voltage can be raised to 1.725v whereas even stock Asus P5W DH only allows max MCH volts of 1.6v. To me badaxe 1333fsb 4:5 is NOT running 1:1 but indeed 4:5


Update: 1067fsb definitely faster

I wonder if how high fsb you can clock on 1067fsb is memory dependent ? i.e.

on 1067fsb with 2.2v vdimm 1304 bios

2x256MB Centon PC4300 Advance Micron D9 fatties = max 382fsb

2x512MB OCZ PC5400 Platinum XTC = max 378fsb

2x512MB Corsair PC5400ULv1.5 Micron D9 fatties = max 421fsb!

I've hit the limit of 1067fsb strap for 2x512MB PC5400ULv1.5 at 4-4-3-5 at 2.2v. The max is 421fsb. I guess you can subtract 20+ seconds if i had 2x1GB modules hehe.

[Onepagebook]

good article, so it's not a whacked!!
Michael, you buy that yet?

[eva2000]

Kev you try as well do same as well 1333fsb strap, same fsb but one test with 1:1 and 4:5 at same timings... bet the 4:5 is faster on badaxe

now 38 seconds faster heh

533mhz 4-4-4-15 at 2.2v

[Onepagebook]

Kev you try as well do same as well 1333fsb strap, same fsb but one test with 1:1 and 4:5 at same timings... bet the 4:5 is faster on badaxe

Yes that's what I talking about, but I am just afraid that michael doesn't think so it definitely faster no doubt, unless he get a intel platform and run it . Thanks George!!!!

[eva2000]

No probs Kev

[alpha0ne]

Now the proof is in can anyone disprove your findings George

[bachus_anonym]

How about same benchmarks and 1066fsb strap? I don't buy it at all as it's common sense. Look at last ofcoolaler's threads with his PC2-8000 HZ... On Giigabyte mobo he manages DDR2-1000 5-4-4-3.0, which is fine and doable. He switches to BadAxe and bam - DDR2-1000 5-3-3-3.0 and probably even not 2.5v (he hasn't mentioned).

Do you guys have any idea how much voltage one would need for DDR2-1000 5-3-3-3.0? I'm not mentioning those "WR" of DDR2-1100 8-2-2-3.0(!!!) or DDR2-1200 8-3-3-3.0. Switch to P2WD2-E Premium or P5W DH or AM2 boards and you will see how far in reality RAM clocks. I simply don't trust BadAxe overclocks, period! And don't buy a theory of loosening MCH timmings, either.

It literally NOT possible for FatBody to run DDR-1100 8-2-2-3.0 or new Micron at 1200 8-3-3-3.0 Also, it's totally off that suddenly Fatbody on BadAxe clocks 100Mhz higher than on other boards and new Micron D9XXX clocks on BadAxe waaay higher than on anything else. We're yet to see anyone hit DDR2-1000 8-3-3-3.0 on other Intel boards or AM2 but no, BadAxe has DDR-1200 already Someting is off with this board and 1333fsb starp...

But can you run at 1066fsb strap and 1333fsb strap George, please. Compare same FSB/CPU speed/ memory clocks results? The key is a strap, not neccesarly 4:5 ratio...

EDIT: Now, after I look at it one more time... Have you tried, tighter latenicies? 8-3-3-3.0 ? Maybe that's what important as well. Your DDR2-1080 8-4-4-4.0 is doable on other platforms as well so this one is not a problem. I have a problem with those huge, out of this world OCs on BadAxe at 8-2-2-3.0 or 8-3-3-3.0

[eva2000]

1067sb has a lower max fsb limit.. you won't be able to get fsb at 1067fsb anywhere near the the mem max....

1067fsb strap limit is around 378fsb on unmodded badaxe... 1333fsb strap has around 429-430fsb limit for me...

my badaxe has no mods and i haven't even shorted the oc debug pin so no vcore options available in bios heh

what's max MCH voltage on P5WD2-E ???

yes 1067fsb strap is more efficient clock for clock on badaxe.. look at my initial clocks with 1067fsb strap at http://www.xtremesystems.org/forums/...d.php?t=105003 - specific post http://www.xtremesystems.org/forums/...1&postcount=41

Tony/Bigtoe has already stated on intel platform 1067fsb strap is more efficient clock for clock than 1333fsb strap.

And you can't really compare mem performance on AM2 vs Intel, it's like saying because DFI NF4 can do TCCD >300mhz 2.5-4-3-7/2.5-3-3-7 then the TCCD can do that on say Gigabyte NF4 board hehe

[bachus_anonym]

1067sb has a lower max fsb limit.. you won't be able to get fsb at 1067fsb anywhere near the the mem max....

1067fsb strap limit is around 378fsb on unmodded badaxe... 1333fsb strap has around 429-430fsb limit for me...

my badaxe has no mods and i haven't even shorted the oc debug pin so no vcore options available in bios heh

what's max MCH voltage on P5WD2-E ???

That's OK. I just want to see how much voltage you'd need for e.g. 375FSB and 4:5 ratio = ~DDR2-940 8-3-3-3.0 or 8-2-2-3.0. Also, isn't there 2:3 ratio ?

EDIT: Exactly my point... 1333fsb is what seems to be triggering those huge low-latency OCs.
I suspect AM2 is not much different and can be compared. People hit very high DDR2-1200+ 8-4-4-4.0, just waaaaaay lower @ 8-3-3-3.0 compared to BadAxe on 1333fsb. Besides, you don't need to look at AM2, just look at other boards for Intel

[eva2000]

well i only have 2.2v vdimm and even for PC5400UL i need 2.2v for 333mhz 3-3-2-8 LOL...

yes there's 800 mem option, haven't tried it yet as trying to figure out my max fsb first



Intel D975XBX Badaxe FSB and Memory clock calculator



heh

[bachus_anonym]

If BadAxe is indeed such a monster for memory OCing then I don't want to see DDR2-1000+ 8-3-3-3.0 @ 1333fsb strap and 400+FSB @ 4:5 BUT DDR2-1000+ 8-3-3-3.0 @ 1066fsb strap and 333+FSB @ 2:3.

But it would be much better to see other board, that doesn't have 1333fsb in BIOS, to do it

[eva2000]

Heh i'll do some more investigating... just after i find the max for my conroe cpus hehe

[Onepagebook]

1067sb has a lower max fsb limit.. you won't be able to get fsb at 1067fsb anywhere near the the mem max....

1067fsb strap limit is around 378fsb on unmodded badaxe... 1333fsb strap has around 429-430fsb limit for me...

my badaxe has no mods and i haven't even shorted the oc debug pin so no vcore options available in bios heh

what's max MCH voltage on P5WD2-E ???

yes 1067fsb strap is more efficient clock for clock on badaxe.. look at my initial clocks with 1067fsb strap at http://www.xtremesystems.org/forums/...d.php?t=105003 - specific post http://www.xtremesystems.org/forums/...1&postcount=41

Tony/Bigtoe has already stated on intel platform 1067fsb strap is more efficient clock for clock than 1333fsb strap.

And you can't really compare mem performance on AM2 vs Intel, it's like saying because DFI NF4 can do TCCD >300mhz 2.5-4-3-7/2.5-3-3-7 then the TCCD can do that on say Gigabyte NF4 board hehe

Bingo, George, you push the right button.
that just give a true lesson on it;
well, for me, I rather buy the performance from intel, cuz Asus ...I just don't buy that either, no need to say other boards

[Metroid]

Kev you try as well do same as well 1333fsb strap, same fsb but one test with 1:1 and 4:5 at same timings... bet the 4:5 is faster on badaxe

now 38 seconds faster heh

533mhz 4-4-4-15 at 2.2v

What is the maximum that can you get with 1/1?

[eva2000]

max for unmodded badaxe rev304 with no oc debug pin shorted is around 423-427fsb 1:1

[gocchin]

Interesting stuff eva,

As I posted in the Intel section. I compared the P5W DH and Bad Axe with
identical hardware except the mobos of course and the Asus is far faster.

I was running the bad axe with 1333fsb and 4:5 ram.

Both at 3.5ghz with 1.4v and exact same ram and ram timings at 2.2v.

Asus:


Bad Axe:

[Gautam]

That's even more dramatic than I expected. I took a look in Memset for a bit to try and tighten some secondary timings and noticed them both way tighter than I'm used to seeing. If you get time, you think you could post a shot of Memset on each of the boards?

[eva2000]

try badaxe at 1067fsb and 350fsb 4:5

this was done at 2.2v vdimm 4-4-3-15 with 2x256MB Centon PC4300 Advance LOL with 1067fsb strap.. not sure how much a penalty 2x256mb would be heh



for conroe what's the diff between 2x256mb vs 2x512mb vs 2x1GB for 32m ?

also timings for tRAS differ it seems for your runs ?

[njkid32]

I am going to do some more testing and see if what your saying is true. So far I have only used the 1333 then clockgen up from there. I noticed that with this mem I was getting instability at 4-4-4-8. So I started playing around with some timings and figured out that these (d9 fatbodies) like to run at 3-3-3-8... I have to find the ss but I ran a 32m at 440fsb 1:1 with those timings. Thanks for the info George I'll give this some testing later.

[gocchin]

The differences are a little harder to spot with SuperPi 1M. 32m the differences are dramatic!

[njkid32]

Oh I know I am looking for the ss now.. Its on one of these comps Looks like a few min difference there.

[bachus_anonym]

I hope now you guys are seeing that there is someting "disturbing" about memory performance and/or clocks on BadAxe @ 1333fsb strap, just as I suspect This is why I think, forcing 1333fsb strap on BadAxe disqualifies it from serious and credible memory testing

[guess2098]

well
just get a vmoded p5wd2-e vs vmoded D975xbx

i use 1066 vs 1333 on D975XBX there is no problem actually, 1066 is slower than 1333
but the point is we are still using D975XBX to do the test rams

if we want to find out there is problem with D975XBX, we need to use different mobo!

[gocchin]

Here we go with 1067 all 3 screens together for comparo at 3.5ghz:

Asus P5W DH:


Bad Axe with 1333 bootstrap


Bad Axe with 1067 bootstrap

[bachus_anonym]

Any more questions? BadAxe @ 1333fsb vs 1067fsb and 1m16s difference This is a DRAMATIC change for 32M!!!