X48 Rampage Formula Preview.

[Zucker2k]

Thank you very much for the info. I will be pushing the FSB up to 467 and this will really help. BTW, I just ordered the 2x2Gb G.Skill 1100 kit to run 5:6 at 1120. Running 1:1 is killing my PL. I did some initial testing at 900 and I couldnt even run PL8, but once I set the ram to 1080 with 5:6, I was able to run PL7.

You should be abe to run PL 8 with cas 5 eg. 5-4-4- or even 5-5-5. Set DRAM static to enabled, and clock twister to stronger. Set DRAM skews to advanced 300 on A/B (very important).

[mikeyakame]

Solor,

Higher tREF only gives better results because increasing the turnaround between Self Refresh cycles frees up memory cycles which in turn reduces latency. This is a cheat way to decrease latency, and a good way to corrupt your memory for stable system usage. Self Refresh is a procedure to ensure that data stored in memory is available is kept valid and not lost. Otherwise you just consume more bandwidth on the bus having to request the data that wasn't paged where it was to supposed to be for the requesting device. On top of that it's also to make sure that all memory is refreshed successfully during the cycle. If its too short there is a chance that it wont be completed, if its too long there is a good chance that you memory ic's will lose charge before the next refresh and result in page faults or increase access time to fetch data, the time to fetch data across the bus requires far more cycles than it does to fetch from DRAM. It may seem irrelevant but if you are doing that often it will add up to a lot of wasted system cycles and bus time. Worst case is you some how end up with memory pages holding incomplete data from parts being lost due to expired memory or being rewritten with something completely different.

My point is it's not good to use it full time, its a way to drop latency in benchmarks and thats all its safe to use it for. By setting it so high you cripple your systems memory performance for tasks which run for extended periods of time, and get lots of page faults

I've already asked Felix if he can add to Memset, auto tREF value setting according to DRAM frequency.

Also put in request to Asus bios engineers to add self-tracking tREF in bios or manually set tREF whichever they rather do.

[mikeyakame]

You should be abe to run PL 8 with cas 5 eg. 5-4-4- or even 5-5-5. Set DRAM static to enabled, and clock twister to stronger. Set DRAM skews to advanced 300 on A/B (very important).

Do you see much bandwidth and latency difference between Moderate with 50-100ps skew, and Stronger with 300ps skew? I've still yet to figure out what else clock twister changes besides DRAM clock skew, data strobe skew possibly and address strobe skew.

[Zucker2k]

Do you see much bandwidth and latency difference between Moderate with 50-100ps skew, and Stronger with 300ps skew? I've still yet to figure out what else clock twister changes besides DRAM clock skew, data strobe skew possibly and address strobe skew.

From strong and below, regardless of skew, read speeds drops about 200 points in everest, and copy speed gets a bump of about the same (200). Write speed is unaffected.

A heads up to those running 1:1 with high fsb (450 and above) your best running on the 400 divider.

[spoof]

So, my 460FSB stability was total fake ...

After i didn't manage to get i stable @455 i decided to push a little bit and try there, with one step higher combinationa of voltages, that i thought were helping. And here is what i did with 460FSB - 600% full memtest, 5 Hours Prime95, 10/10 IBT - no lockups, no errors, nothing. After this was my evening Crysis Wars time and after 2 hours it just froze - damn lockup. Restart, without changing anything i ran Prime95 and guess what - lockup after 10 minutes.
Yesterday i was fitting with 460 again, but no succsess. This evening, after bios set-lockup-bios set-lockup i push it even more to 465, set voltages and guess - i was able to play Crysis 3 hours and than 2 hours browsing - no crashesh, no lockups. Happy i restart to save my OC profil, enter Windows, let Prime95 Benchmark to run for a while and after 5 minutes - my lovely lockups.

So, now i'm even more confused ):


Morning news: with the settings(same settings) that crysis was playable and Prime95 Benchmark causes lockups before i go to bed i ran Prime95 Small-FFT - after 8 hours still runing, no lockups and errors. Prime95Bench bad way of testing? Well, i'm no actually testing for stability, it's just my initial test after overclock, because the bench starts every time i boot.

[mikeyakame]

Heres a lesson.

Whatever you do don't damage the contacts on the cpu socket, or you'll enjoy hours of trying to get your board to post correctly!

Worst part is don't even know how it happened, i've fixed it 70% at best but I'll be rma'ing it when I get X58 board. ;\

[SoLoR]

Also put in request to Asus bios engineers to add self-tracking tREF in bios or manually set tREF whichever they rather do.

i did, they will get back to me after they contact bios team. Altho they probably say same thing to everyone...

[mikeyakame]

Oh lol I meant we (myself and Alien Grey) already put a request in already, and bios engineers said that it will most likely be added and they had no problems adding it, but since they need to do testing and what not may not be in next beta bios which should include NB & CPU Clock Skew adjustment.

By the way my Rampage is running barely! I've been forced to run most options on auto since at least 2 of the cpu socket contacts are damaged pretty badly. If I even try and tighten latency it results in MCE BSOD's 0x3B which is apparently General Exception occuring during transition between Priviledge 3 and Priviledge 0 most likely from a SYSCALL where the EIP address is corrupted or similar. Sucks. Got it running at 3ghz (9*333) at least and 1111mhz memory heh. Sucks going from 52-53ns latency all the way down to 75-80ns! It'll only have to hold out for a week or two until I can get hold of all the parts I need for X58 build, then time to RMA it and hope I get a board with at least as good of a chipset as my current, if not better.

[Alien Grey]

Oh lol I meant we (myself and Alien Grey) already put a request in already, and bios engineers said that it will most likely be added and they had no problems adding it, but since they need to do testing and what not may not be in next beta bios which should include NB & CPU Clock Skew adjustment.

By the way my Rampage is running barely! I've been forced to run most options on auto since at least 2 of the cpu socket contacts are damaged pretty badly. If I even try and tighten latency it results in MCE BSOD's 0x3B which is apparently General Exception occuring during transition between Priviledge 3 and Priviledge 0 most likely from a SYSCALL where the EIP address is corrupted or similar. Sucks. Got it running at 3ghz (9*333) at least and 1111mhz memory heh. Sucks going from 52-53ns latency all the way down to 75-80ns! It'll only have to hold out for a week or two until I can get hold of all the parts I need for X58 build, then time to RMA it and hope I get a board with at least as good of a chipset as my current, if not better.

What did you do to srcew up your socket pins so badly?

[Brama]

A heads up to those running 1:1 with high fsb (450 and above) your best running on the 400 divider.

Do you mean the 400 MHz strap?

[EOD]

Update: I backed off the FSB OC since I needed too much NB voltage IMO. So I have it running 4.14GHz with 1.288v vcore (actual) and a 458 FSB. Thanks everyone for your help in dialing in my new CPU!

Prime Stable:




Temps and voltages:

[spoof]

Zucker2k, did you found solution for this problem?

I'm getting hard time cut the lockups at 465, but with pluging of the cable i can push it to 475 and maybe more. So, is there solution or i have to unplug the cable, push the start button to "unload" the motherboard and than power it on?

[Zucker2k]

Zucker2k, did you found solution for this problem?

I'm getting hard time cut the lockups at 465, but with pluging of the cable i can push it to 475 and maybe more. So, is there solution or i have to unplug the cable, push the start button to "unload" the motherboard and than power it on?

That's interesting yours starts so early on at 475, mine starts beyond 495 with multi at 8.5. No, I've found no solution. On the other hand, I got a Q9650 here that does not exhibit the same issue with the same board so

[spoof]

So far, what can i say is, that maybe now on 471, chipset is making troubles. In idle i have no lockups, which i think was the CPU. Small FFT for 6 hours, no problem, but Blend, IBT - lockup. I'm at 1.63 vNB, no matter what GTL (I can't seta 49x, 57x, 60x, it just refuses too boot). PL is 8, which is pretty normal for that FSB - i'm just out of any ideas.

A-Grey, i have huge request to you. Since i have your words with ASUS Tech Support, send them a noto to this problem, that Zucker2k discovared - http://www.xtremesystems.org/forums/...postcount=7594.

Tell the, that earlier version of s775 like P5K everytime on boot run for a sec, then stop and then run normaly. I'm 101% sure, that this will fix this annoying problem with lockups after certain FSB. Thanks in advance!

[Zucker2k]

Zucker2k, did you found solution for this problem?

I'm getting hard time cut the lockups at 465, but with pluging of the cable i can push it to 475 and maybe more. So, is there solution or i have to unplug the cable, push the start button to "unload" the motherboard and than power it on?

Actually, problem didn't go away. So any news on the new beta bios?

[Alien Grey]

A-Grey, i have huge request to you. Since i have your words with ASUS Tech Support, send them a noto to this problem, that Zucker2k discovared - http://www.xtremesystems.org/forums/...postcount=7594.

Tell the, that earlier version of s775 like P5K everytime on boot run for a sec, then stop and then run normaly. I'm 101% sure, that this will fix this annoying problem with lockups after certain FSB. Thanks in advance!

Is this the famous dual post that people experience with ASUS boards?

If this is what really happens there isn't much they can do about it. I don't know if there is a fix for that. This can happen to when an overclock fails.

I never had any problems with that. The strange thing is that some people already experience this on stock settings.

[Alien Grey]

Actually, problem didn't go away. So any news on the new beta bios?

Sorry, no news on the new beta bios. I hope to receive some news soon.

[Brama]

A-Grey received the beta BIOS 802 and posted here

[mikeyakame]

Our cries have been heard Hurrah Hurrah!

Boys now it's time to see if we can break the infamous 500mhz FSB mark with our quad cores

[Alien Grey]

WTF, why is the thread removed about the BETA BIOS?

I started a new thread for this BIOS so that we don't have to see problems with the ASUS Rampage BIOS that are already fixed in this BETA BIOS.

Ahhh... I see it it's moved to the Intel Info And Support forum.

[mikeyakame]

Well I confirm the clock skews are working I can even post at 400mhz FSB now! LOL It doesn't fix a broken cpu socket though but it seems they definitely help

At 400mhz fsb I'm using:

CPU Clk Skew = -100PS
NB Clk Skew = Normal

How did I know that seems right? Well Normal/Normal gives a longer delay on LCD Poster between CHECK CPU and VGA BIOS, 100/Normal gives a near instant response between the two

Seems skewing isn't too bad at this point.

Zucker,

You are going to have to be the GOD who breaks the magical 500mhz FSB mark, I'm having enough trouble right now breaking 400mhz

Edit:
PS C:\Users\mikey\linpack> .\linpack_xeon64.exe .\lp64.in
User Specified Datafile

Current date/time: Thu Feb 05 09:39:46 2009

CPU frequency: 3.600 GHz
Number of CPUs: 4
Number of threads: 4

Parameters are set to:

Number of tests : 2
Number of equations to solve (problem size) : 2000 14000
Leading dimension of array : 2008 14008
Number of trials to run : 6 4
Data alignment value (in Kbytes) : 4 4

Maximum memory requested that can be used = 1569180256, at the size = 14000

============= Timing linear equation system solver =================

Size LDA Align. Time(s) GFlops Residual Residual(norm)
2000 2008 4 0.189 28.3305 4.657913e-012 4.051814e-002
2000 2008 4 0.168 31.7649 4.657913e-012 4.051814e-002
2000 2008 4 0.165 32.2756 4.657913e-012 4.051814e-002
2000 2008 4 0.168 31.7843 4.657913e-012 4.051814e-002
2000 2008 4 0.161 33.1090 4.657913e-012 4.051814e-002
2000 2008 4 0.170 31.3822 4.657913e-012 4.051814e-002
14000 14008 4 44.732 40.9045 1.832709e-010 3.309677e-002
14000 14008 4 45.606 40.1207 1.832709e-010 3.309677e-002
14000 14008 4 43.486 42.0760 1.832709e-010 3.309677e-002
14000 14008 4 43.799 41.7756 1.832709e-010 3.309677e-002

Performance Summary (GFlops)

Size LDA Align. Average Maximal
2000 2008 4 31.4411 33.1090
14000 14008 4 41.2192 42.0760

End of tests

Poifect Couldn't even get this broken thing stable with GTLs, now its not throwing Linpack errors anymore above 333mhz FSB! With GTLs at 0.67x CPU and 0.63x NB.
Perhaps we can live without individual landing pin GTL adjustment after all Clock skews make a hell of a lot more difference for the reason I explained in the GTL thread.

GTLs alone only dynamically shift the problems of clock jitter but don't fix it, Clock Skews completely solve it

[Shiranui Gen-An]

Yay, now I can boot with >525FSB using my E8400! CPU Clock skews FTW!!

[Zucker2k]

CPU Clk Skew = -100PS
NB Clk Skew = Normal

How did I know that seems right? Well Normal/Normal gives a longer delay on LCD Poster between CHECK CPU and VGA BIOS, 100/Normal gives a near instant response between the two

Seems skewing isn't too bad at this point.

Zucker,

You are going to have to be the GOD who breaks the magical 500mhz FSB mark, I'm having enough trouble right now breaking 400mhz

Mike, you're right:

CPU Clk Skew = -100PS
NB Clk Skew = Normal

That's the magic number right there. All my 500FSB+ boot problems vanished into thin air. Of course I'm going to shoot for more this was my first attempt.

[mikeyakame]

Oh man

Thats brilliant mate !

I'll be RMAing my board on Monday, should have it back from my friend at Asus either same day or day after then we'll see what numbers I can pull

I'm actually impressed with the design of the board, I've seen P45 P5Qs needing a heap of CPU and NB skew to get stable at high 400s with a quad. 1-200 NB, 4-500 CPU. If you can do with only 100ps extra delay on CPU clock and normal base skew on NB that's big kudos to Asus for designing the RF quite well.

Now that our board isn't crippled anymore at 500Mhz FSB from the lack of NB and CPU clock drive skew adjustment, I think Asus may have a DDR2 Quad Core FSB overclocking champion on their hands


ALSO WARNING FOR EVERYONE.

CPU Clock Skew if you set it too far you won't post and no harm done.

NB Clock Skew if you set it too far and it manages to post and you get into the OS:
a) you risk corrupting it very fast
b) you'll cross strobes and end up wiping out registry sub trees or similar

so BE CAREFUL and don't play with NB Clock Skew unless you need to. It's now the most dangerous setting in the bios WOrse than too much Phase Pull in on PL !

Follow these instructions:

Set both to NORMAL
First try Delay 100PS on CPU and test everything.
Set GTLs to either 0.67x CPU and 0.63x NB for 65nm, 0.65x CPU and 0.63x NB for 45nm. Set up Vtt properly, and find the point you can't get stability with those GTLs and a Vtt voltage that isn't either too low or too high.

If this doesn't help and all voltages/GTLs are correct, then delay another 100PS on CPU.
If you hit 200PS Delay on CPU, then try delay 100PS NB with 200PS CPU and so on.

Keep them close, and don't go too far cause it seems from Zucker and I that the clock jitter isn't too severe on our boards, and it only takes a little skewing to stabilize near and above 500mhz FSB

If you don't have GTLs, Vtt, Vnb, Vcc setup correctly then changing NB and CPU clock skews won't give you much response to work with to fine tune. They should be set at NORMAL/NORMAL and only changed when the rest of the settings are where they should be.

Have fun.

[mikeyakame]

Even with damaged cpu contacts, just that 100ps delay on CPU clock drive Ive run a near 8hr Prime95 blend pass at 400mhz FSB stable. Awesome. Now to try and see how far I can push this broken board!