Modded ASUS P5Q/Pro/Deluxe/Premium BIOS

[DMF]

P5Q Pro 901m

Link is dead.

[ieatrawbacos]

Firstly, I'd like to thank Ket for the work he's put into these BIOSes, it's appreciated.

However, I'm wondering if anyone has gotten the memory tRD pull-ins to do what they are supposed to.

I've tried every official, and everyone of the mBIOSes for my P5Q-D and the pull-ins have never worked.

I've heard people report them working, but their results contradict them. If all pull-ins are enabled memory bandwidth scores should increase by a few hundred points, and the latency should drop by a full ~2ns. Most people that report a change are only seeing changes that are within the margin of error of their tests. The display of the phases in the BIOS, and in programs like everest should also change, but they do not.

Has anyone gotten them to work, for real?

Hey Ket.
Thanks for the work on the Bios.

Question about the bios and a specific setting. I have read when you overclock that you should set "SOMETHING" in the bios to 33mhz so that the FSB doesn't creep up on other hardware, as you increase FSB on the CPU.

Here's a screenshot, it's called : PCI Clock Synchronization
Link http://www.ocforums.com/showthread.php?t=515316
I believe it is the second actual picture from within the bios. I can't find this setting for the life of me in your bios. I'm sure it's called something else now. If you could just point me in the right direction please.

Also, a question about flashing. I use the Asus Update Utility from within the windows environment. Two things. Do YOU recommend another/safer way?

Also, that utility comes with with something built in called My Bios which more or less lets you change/create your own Bios Flash Screen. Do you have any opinions on this? Thanks again Ket.

BTW....if there is some magic "guide" out there for the 1406 Bios that shows you step by step, item by item of what everything does, I sure wish someone would point me to it. Thanks.

PCI clocks have been locked at 33MHz, regardless of FSB speed, on all major chipsets/boards for at least the last three to four years.

There is no need for a BIOS option to adjust or enable it. It's a given.

[concretefire]

PCI clocks have been locked at 33MHz, regardless of FSB speed, on all major chipsets/boards for at least the last three to four years.

There is no need for a BIOS option to adjust or enable it. It's a given.

Yeah, I knew that. Just testing you guys out to keep you on your toes. 3 or 4 years huh? Geez time flys when you're divorced.

Speaking of Bios. Back to the bios splash screen. On my p5K-pro using ket's bios....(and the original bios too) that Asus Splash screen disappears really fast. I mean like within 1 second flat. Is there any way to adjust that splash screen to stay up just a few seconds longer? I'm weird I know. But Ket or someone must know.

[DMF]

I can't figure out what the 'FSB Strap to NB' option is supposed to do. When set non-Auto it would seem to force a certain divider where Auto would calculate one from FSB and DRAM clock settings. But if that's true then 'FSB Freq', 'Strap', and 'DRAM Freq' should be mutually exclusive. You could set two of the three, but not all three - at least not to mutually exclusive values.

What am I missing?

[concretefire]

I can't figure out what the 'FSB Strap to NB' option is supposed to do. When set non-Auto it would seem to force a certain divider where Auto would calculate one from FSB and DRAM clock settings. But if that's true then 'FSB Freq', 'Strap', and 'DRAM Freq' should be mutually exclusive. You could set two of the three, but not all three - at least not to mutually exclusive values.

What am I missing?

Yep , yep.....add that one among the many, many other features/seatings on that page. I wish there was a step by step guide. Or rather, just a line by line definition with a brief explanation and maybe one or two "for instance"
type of things.

I've been screwing around with this Mobo/CPU for 2 weeks now and just tonight got the fuzzies to up the FSB to 333 , setting ram vcore at 2.1 and left everything else on auto. Booted right up. Temps were a tad high for me so I checked vdim and it was at 1.42
So I rebooted and set it at 1.325. It was a shot in the dark number but knew it was close based on what I have read so far. Again, booted right up....now temps 10 C cooler. Until I find out more about each setting. I'm just gonna leave it here for now and play around another week I guess.

[axy1985]

Temps were a tad high for me so I checked vdim and it was at 1.42

Perhaps you meant vcore? vdimm is for RAM, and I highly doubt there's any DDR2 RAM kit that actually boots at 1.42V.
And yeah, 1.42V on a 45nm CPU is something to be scared off with normal air cooling.

[Vazovskiiii]

hallo again ket i'm beck i have tried 3 mobos to choose the best one lanparty dk p45, lanparty dk x38 and biostar tpower so i've decide to get p5q-e as i see no more new bioses after 1406 why? and do you still will do the urban bios or you do not think about thin anymore?
and how do you think i've got 4x1 gb trnscend axeram 1200+ micron d9gmh do will this mobo and this ram well?
thank's a lot

[ieatrawbacos]

I can't figure out what the 'FSB Strap to NB' option is supposed to do. When set non-Auto it would seem to force a certain divider where Auto would calculate one from FSB and DRAM clock settings. But if that's true then 'FSB Freq', 'Strap', and 'DRAM Freq' should be mutually exclusive. You could set two of the three, but not all three - at least not to mutually exclusive values.

What am I missing?

All it does is limit the memory ratios you can use, and possibly loosens some of the subtimings that are still on auto.

In the olden days of yester year (well like two years ago), the strap settings used to control tRD before anyone knew what tRD was. Now it doesn't do jack on mobos with manual tRD control.

[truehighroller]

Firstly, I'd like to thank Ket for the work he's put into these BIOSes, it's appreciated.

However, I'm wondering if anyone has gotten the memory tRD pull-ins to do what they are supposed to.

I've tried every official, and everyone of the mBIOSes for my P5Q-D and the pull-ins have never worked.

I've heard people report them working, but their results contradict them. If all pull-ins are enabled memory bandwidth scores should increase by a few hundred points, and the latency should drop by a full ~2ns. Most people that report a change are only seeing changes that are within the margin of error of their tests. The display of the phases in the BIOS, and in programs like everest should also change, but they do not.

Has anyone gotten them to work, for real?

I'm not going to comment on the PCI Frequency thing but, I will comment on the pul ins. I can get them to work on the 0803 "Modded" BIOS for me and I have the 9450 in my rig#1 of my sig. I haven't really payed attention to if it is working or not with the 1406. I will check this out tonight when I get time and report back to you. I'm thinking it should be working in 1406 but, I might be wrong. I know for a fact that it works in 0803 "Modded" for me though. I still haven't tried the 0703 or 6 what ever the last number is yet. I need to test that one out and see if I can get to 4GHZ with it.

[ieatrawbacos]

I'll give the 803M bios another shot and see what happens.

What memory ratio are you using? 5:6?

[ieatrawbacos]

Pull-ins don't work for me with the 803M BIOS.

No change in benchmarks, no change in stability, no change in the BIOS tRD phase display and a "neutral" read delay phase adjust in everest; just like the 704, 803, 1201, 1306, 1406, modded and unmodded BIOSes I've tried.

[Big Lar]

Running 1406 bios here and just adjusted all pull-ins to Enabled. No change in Everest benchmark, guess they don't work with this bios, they did however make a difference with 1 of the older bios versions, can't remember which tho.

Larry

[concretefire]

Perhaps you meant vcore? vdimm is for RAM, and I highly doubt there's any DDR2 RAM kit that actually boots at 1.42V.
And yeah, 1.42V on a 45nm CPU is something to be scared off with normal air cooling.

Yeah, sorry. I meant vcore was at 1.42. I adjust Vdimm for the Ram to 2.1 as per the specs. My bad

[samsonJS]

In the olden days of yester year (well like two years ago), the strap settings used to control tRD before anyone knew what tRD was. Now it doesn't do jack on mobos with manual tRD control.

I can achieve much higher fsb frequencies on the 333mhz strap than with the others (my q9450's default fsb is 1333, but I'm not sure if this is why). That makes me think that it must be adjusting something because there are differences between the 333mhz and 400mhz straps at the same clock. Maybe the subtimings you mentioned?

----

Pull-ins have no effect for me either (q9450, mushkin ascent pc2-8500, p5q-e).

[ieatrawbacos]

I can achieve much higher fsb frequencies on the 333mhz strap than with the others (my q9450's default fsb is 1333, but I'm not sure if this is why). That makes me think that it must be adjusting something because there are differences between the 333mhz and 400mhz straps at the same clock. Maybe the subtimings you mentioned?

I have all of my subtimings as well as my tRD set manually, and I see no difference in FSB OCing between the 400 and 333 straps.

It's probably some change in timings that ASUS has linked to the strap setting, as well as the different memory ratios available, that have something to do with the differences you see.

It's also possible that the strap setting controls something that I'm not aware of that has more of an effect on your system than mine. I'm using an X3350 which is essentially the same thing as a Q9450, so I don't think the CPU is responsible for such a difference.

[DMF]

Googled a thread right here on the forum. I've done a little editing to make it clearer:

Your north bridge has an internal clock speed and latencies just like your CPU and memory. The FSB of your north bridge can be found by dividing your original CPU multiplier by your set CPU multiplier and then multiplying by your set FSB.

So if you are running a E6600 (2.4 GHz or 266 * 9) at 3.2GHz (400 x 8) your NB FSB is:

(9 / 8) x 400 = 450Mhz FSB (1800Mhz Total)

Just like your memory may be able to run at 4-4-4-12 at 1000Mhz but needs to run at 5-5-5-15 at 1200Mhz, your north bridge has a series of latencies which it must adjust in order to maintain stability at its FSB. These latencies seem to play a far more significant role in system performance than memory latencies. - emphasis mine

Intel has predefined specific latencies at specific NB FSB speeds. They are referred to as straps. There is a strap for when the NB FSB is 1066Mhz and under, 1333Mhz FSB and under, 1600Mhz FSB and under, etc. When you go from the 1066Mhz FSB strap to the 1333Mhz FSB strap, the north bridge's internal latencies loosen to allow for greater stability.

ASUS has redefined the NB strap so that the 1333Mhz FSB strap does not come into effect until 401Mhz FSB (1604Mhz). Other perimeters of straps are someone unknown.

Ponder this table from farther down in the thread. Note that ref freq (reference frequency) is another name for 'FSB Strap to NB'.

I haven't quite figured out the relationship between ref freq, mem freq, and ram speed. mem freq is the value selected for the 'DRAM Frequency' setting. Note that in all cases the CPU frequency is constant.

[DMF]

Here's another, from here, which seems to be describing the [Auto] setting:

The FSB "strap" is very similar to RAM SPD timings. On Intel's new chipsets, the northbridge has its own clock speed and latency timings, just like RAM. Below 400mhz FSB, your ASUS northbridge is running the 1066mhz strap setting. Once you go up to 401 mhz, it switches to the 1333mhz strap setting. However, just like RAM latency timings, when the clock speed is lower, the standard timings are also lower. I'll make up an example.

Let's say your mobo is set to keep the 1066 strap from 266fsb to 400fsb. And, lets say that the northbridge latency defaults to "1" at the 1066 strap setting. Well, as you overclock from 266mhz FSB to 267mhz FSB, your northbridge speed also increases, let's say from 1066 to 1070 (quad pumped). Your nb won't have any problem with that, but as you get closer and closer to 400mhz FSB, your nb speed is going to be going up as well, possibly as high as 1602mhz (my numbers probably aren't right, but the general concept is the same). Problem is, your NB might not be able to run at 1602mhz with a latency of "1". Well the good news is, once the NB changes to the 1333mhz strap at 401mhz FSB, your NB latency is set to "2". Again, just like RAM, your NB can run at a higher speed with a higher latency.

So, your RAM really isn't affected by your strap, but your northbridge speed is.

[DMF]

Whoa!! Here is an excellent thread, with pitchers!!

[ieatrawbacos]

I haven't quite figured out the relationship between ref freq, mem freq, and ram speed. mem freq is the value selected for the 'DRAM Frequency' setting. Note that in all cases the CPU frequency is constant.

In that table "Ref freq" is the 'strap', "mem freq" is the memory ratio, and "ram speed" is the final memory clock; pretty straight forward.

Whoa!! Here is an excellent thread, with pitchers!!

What's going on in that thread you linked to is exactly what I was talking about here:

In the olden days of yester year (well like two years ago), the strap settings used to control tRD before anyone knew what tRD was. Now it doesn't do jack on mobos with manual tRD control.

The P965 board tested there did not have any way to adjust tRD setting except through strap changes. In addition, almost no one knew what tRD was (it's the delay the chipset needs to move data from the memory clock domain to the FSB clock domain) at that time (over two years ago).

The strap settings do not do anything but limit the allowable memory ratios and auto adjust some of the subtimings on the P5Qs.

If you manually set everything, there is no difference in performance between the strap settings, all other things being equal.

[Ket]

Pull-ins do work, however nothing is a given. Many things are hardware dependent and it is indeed possible to have any number of hardware configurations where enabling pull-ins will make no significant measurable difference. Also don't forget that the principal of pull-ins is almost purely theoretical, and even in theory pull-ins do not do much for improving latencies - negligable at best.

[ieatrawbacos]

I have to disagree on pretty much every point.

The clock crossing operation Intel's chipsets perform is quite well documented at this point. The number of phases is directly linked to the FSB/DRAM ratio, and the delay for each phase is the tRD. The Effect of pull-ins are also well documented. On other boards, enabling them reduces the tRD on a particular phase of the clock crossing procedure, and the performance impact from doing so is directly related to the number of phases changed. If you enable them all, performance is exactly the same as if you have dropped the base tRD a full point. If you enable half of the available pull-ins of each memory channel, the performance is right in the middle of the two tRD notches in question. I've tested this myself on the DFI Blood Iron, and the DFI X48-T2RSB Plus. At 400MHz FSB, each tRD phase is 2.5ns of memory latency. If I'm running tRD 7 on the two boards I mentioned above (at 400MHz FSB) and drop the tRD to 6, OR enable all pull-ins, memory latency drops a full 2-2.5ns, and bandwidth increases by a few hundred MB/s all round, as it should. Furthermore, enabling the pull-ins can also have a significant effect on stability, when they work. If I am already near the limit of stability, enabling any of the pull-ins might cause instability. Enabling them all has the same effect on stability as dropping the tRD a full point, which is not surprising since it has the same effect on performance. This is why, that when tweaking, one, two, or more phases are left with the pull-ins disabled. The entire purpose of allowing tRD phase pull-ins is to allow access to what is essentially a fractional tRD. If tRD 7 is totaly stable, but tRD 6 un-bootable, chances are that you have plenty of room to enable some pull-ins for a very measurable boost in performance, if the pull-ins work correctly on the board in question. If one tRD level is on the verge of stability, enabling all pull-ins except for one can give nearly the performance of the lower tRD level without compromising system stability.

In my, and many other's experiences with other boards, where the pull-ins do work, they work regardless of hardware configuration. I've tried them with more than a dozen different sets of memory, and a half dozen CPU (ranging from E2140s to X3350s), at a huge range of FSB and memory ratio settings. On every board where the tRDs ever behaved as predicted, they always behaved as predicted.

I can only conclude, based on a great deal of first hand testing, and even more second hand research/experimentation, that ASUS has not enabled the tRD phase pull-in setting that they show in their BIOSes. I've tried almost a dozen different BIOSes with this P5Q deluxe, at FSB ranging from 200 to 520, using every memory ratio available, and ASUS' pull-ins have never done anything noticeable, for either performance or stability. Everest, which has the capability of detecting pull-in settings (I've tried it) also shows no change on the P5Q (D). Indeed, the very tRD by phase display in the BIOS settings page itself shows no change.

They don't work, not on the P5Q-Deluxe anyway.

Some articles on Anandtech about tRD, the clock crossing procedure, and pull-ins:
http://www.anandtech.com/mb/showdoc.aspx?i=3129&p=9
http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=5

If you like, I can take some pictures of the phase display in the BIOS of one of my Blood Irons, and run benches with various pull-in setting, to show what working pull-in settings should do.

[truehighroller]

I have to disagree on pretty much every point.

The clock crossing operation Intel's chipsets perform is quite well documented at this point. The number of phases is directly linked to the FSB/DRAM ratio, and the delay for each phase is the tRD. The Effect of pull-ins are also well documented. On other boards, enabling them reduces the tRD on a particular phase of the clock crossing procedure, and the performance impact from doing so is directly related to the number of phases changed. If you enable them all, performance is exactly the same as if you have dropped the base tRD a full point. If you enable half of the available pull-ins of each memory channel, the performance is right in the middle of the two tRD notches in question. I've tested this myself on the DFI Blood Iron, and the DFI X48-T2RSB Plus. At 400MHz FSB, each tRD phase is 2.5ns of memory latency. If I'm running tRD 7 on the two boards I mentioned above (at 400MHz FSB) and drop the tRD to 6, OR enable all pull-ins, memory latency drops a full 2-2.5ns, and bandwidth increases by a few hundred MB/s all round, as it should. Furthermore, enabling the pull-ins can also have a significant effect on stability, when they work. If I am already near the limit of stability, enabling any of the pull-ins might cause instability. Enabling them all has the same effect on stability as dropping the tRD a full point, which is not surprising since it has the same effect on performance. This is why, that when tweaking, one, two, or more phases are left with the pull-ins disabled. The entire purpose of allowing tRD phase pull-ins is to allow access to what is essentially a fractional tRD. If tRD 7 is totaly stable, but tRD 6 un-bootable, chances are that you have plenty of room to enable some pull-ins for a very measurable boost in performance, if the pull-ins work correctly on the board in question. If one tRD level is on the verge of stability, enabling all pull-ins except for one can give nearly the performance of the lower tRD level without compromising system stability.

In my, and many other's experiences with other boards, where the pull-ins do work, they work regardless of hardware configuration. I've tried them with more than a dozen different sets of memory, and a half dozen CPU (ranging from E2140s to X3350s), at a huge range of FSB and memory ratio settings. On every board where the tRDs ever behaved as predicted, they always behaved as predicted.

I can only conclude, based on a great deal of first hand testing, and even more second hand research/experimentation, that ASUS has not enabled the tRD phase pull-in setting that they show in their BIOSes. I've tried almost a dozen different BIOSes with this P5Q deluxe, at FSB ranging from 200 to 520, using every memory ratio available, and ASUS' pull-ins have never done anything noticeable, for either performance or stability. Everest, which has the capability of detecting pull-in settings (I've tried it) also shows no change on the P5Q (D). Indeed, the very tRD by phase display in the BIOS settings page itself shows no change.

They don't work, not on the P5Q-Deluxe anyway.

Some articles on Anandtech about tRD, the clock crossing procedure, and pull-ins:
http://www.anandtech.com/mb/showdoc.aspx?i=3129&p=9
http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=5

If you like, I can take some pictures of the phase display in the BIOS of one of my Blood Irons, and run benches with various pull-in setting, to show what working pull-in settings should do.

They work for me with Kets modded 0803 so they do work just by accident I think when he modded it.

[-n7-]

Yeah pull-ins worked quite well on my previous DFI; there was a measureable difference in benching.

I have yet to see them work with the newer bioses for my P5Q-D, though i'm pretty sure when i tested they were working on at least one of the older bioses...just cannot remember which one.

No difference in Everest means they are not working, as there will absolutely be a measureable difference when they are.

[Ket]

lol. Somebody thinks 2.5ns makes a significant difference in the real world

True; I was never looking to fix the pull-ins as the difference is so insignificant, but the fact they did get fixed is a bonus I guess

[hema]

Hi every one
I try to install this beta version of windows 7 on my p5q with no hope
befor the last step of installation the blue screen appear and the installation stop
try many dvds with variable speed
try the same dvd with gigapyte EP35-DS3R and the installation done well
replace the ram with the ram in EP35-DS3R and the same error
try many version of bios and the same
disable the component in bios
The onboard sound
the onboard network
the onboard IDE
surely some other stuff..
with no help
try another vega also
so I think we need abios to solve this prob.