X48 Rampage Formula Preview.

[Alien Grey]

The fact that Gigabyte has 50ps skews means nothing. What has already been said many times getting high fsb on quad + 600mhz on ram + pl6 is no go.

Try to lower PL, increase vDDR, vNB. Don't tell people that 50ps skews are the key while they aren't - you just can't make thing works so fast with so tight timings and so low voltage.

They are the key for stability. It doesn't matter what voltage you use when you are on the wrong skew. You will move your problem like I do with the 8 x 465MHz and DDR 1239MHz and tRD 6.

I can give as much voltage as I want it will never be 100% stable. You will reach a point that if you raise the voltage a little bit more that you will make it more unstable.

You just don't understand the importance of clock skew because you never needed it.

You see the screen shot with the 25 passes and still you don't believe that it's possible. When it isn't possible you will never reach the 25 passes, believe me. You will fail immediately or after a few passes.

[kuebk]

25 passes is nothing, run 100 or 200.
Anyway I highly doubt that I will ever meet such a problem since I'm not even going to try to run PL=6 at >600mhz on ram.

IMO the problem is not with the board but with people trying to run settings above hardware physical capabilities

[neo_rtr]

25 passes is nothing, run 100 or 200.
Anyway I highly doubt that I will ever meet such a problem since I'm not even going to try to run PL=6 at >600mhz on ram.

IMO the problem is not with the board but with people trying to run settings above hardware physical capabilities

I just cant resist asking

Can you tell me one application that will use the system the same way the claimed 100-200 passes . Leal life application i mean.

[mikeyakame]

Ok did some further testing and stressing. Right now I'm at 1161Mhz via 4:5 divider. This gives me noticable performance boost during boot and normal use.
I tried 1236Mhz aswell with PL9 but it didnt even boot windows, and dont want to give it too much vdimm.




Any ideas how to improve?

Looking good. Pretty close to what my RF was doign at 1160mhz.

I haven't got a good quality everest pic, seems I have deleted it. But here is the values I had at 483Mhz FSB, PL7 Phase pullin A1, 1160mhz, 4x1gb 5-5-5-16, Vdimm 2.34V, Vnb 1.53v, Q6600 @ 3.38ghz. (7x), AI Clock Twister Stronger, Dram static Read Enabled. Divider would be 12:10.

Mem Read = 9732Mb/s
Mem Write = 8754Mb/s
Mem Copy = 9504Mb/s
Mem Latency = 52.1ns
L1 Read = 54072mb/s
L1 Write = 53890mb/s
L1 Copy = 107223mb/s
L1 Latency = 0.9ns
L2 Read = 24581mb/s
L2 Write = 19181mb/s
L2 Copy = 27861mb/s
L2 Latency = 3.2ns

[mikeyakame]

25 passes is nothing, run 100 or 200.
Anyway I highly doubt that I will ever meet such a problem since I'm not even going to try to run PL=6 at >600mhz on ram.

IMO the problem is not with the board but with people trying to run settings above hardware physical capabilities

Much truth here. Board can only do so much, the VRM is the weak link on the board, and with age it only gets worse. Unless you have a 1000W or so PSU to help stabilize the input voltage at high FSB, etc it'll bounce voltages under load.

If A-Grey is saying Gbyte have 50ps increments, could be clock generator dependent then. Asus and GB use different clock gens, Asus and DFI on the other hand use the same ones, which funny enough both also have 100ps increments for clock / nb skew.

[kuebk]

Anyone knows where are the reading points for vCPU?

[mikeyakame]

Anyone knows where are the reading points for vCPU?

If you can locate a trace from the VR to the cpu socket landing pins (perhaps one of the mosfet pins?), you can probably find a suitable one to measure. I don't have my board here anymore so afraid I can't help all that much. Just look for one that gives you the closest voltage to what you expect it to be

[kuebk]

If you can locate a trace from the VR to the cpu socket landing pins (perhaps one of the mosfet pins?), you can probably find a suitable one to measure. I don't have my board here anymore so afraid I can't help all that much. Just look for one that gives you the closest voltage to what you expect it to be

I have no electric experience so its highly possible for me to something up.

[neo_rtr]

http://www.xtremesystems.org/forums/...d.php?t=179580
there is a picture about vDroop Mod / vCore Mod / EPU Mod
Is this what you refer to?

[Chemjsal]

Here is an on-the-fly pic of her doin' work. Note that there is an error message complaining about her being low on memory - courtesy of Microsoft. The reason why the Lavalys benchmark gives such low numbers in the first row is because I had been running Memtest Pro with all four threads before. So, I've finally settled on 466x9 with the memory running at 559 MHz.




and another one after her going back to normal



hope it helps

____________________

Rampage Formula @0902
4x2GB G.Skill PC2-8800 (1120 MHz, 5-5-5-15, 2N) at 1.90V (AI Clock Twister: Stronger, NB Voltage: 1.49V, PL7)
Q9650 ES(E0) @4.19 GHz (466x9) at 1.4125V (1.38V real, 1.30V load) on water, FSB Strap: 333, LLC disabled!
Asus 4870 TOP 512MB @880/1180 MHz on water
Areca ARC-1210, ATI Theater 550 Pro, Dell 2405 FPW
PC Power & Cooling Silencer 750W red - positioned outside the case
Lian-Li PC-V2000 black - both side panels off

[neo_rtr]

With regards to previous post.
Some issues with Everest not updating correctly Volts and temperature information while Prime95 Small FFT was running.
After changing the CPU & NB GTL Ref from normal to Auto everest is updating better.
100 or 200 or some other variations didnt work out.

also NB Voltage change from 1.67 to 1.74 had no effect on it
CPU PLL change (extreme range 1.50 to 1.64) show some improvement while boot 1st screen in win7 where gadgets loaded up almost instantly with higher cpu pll

[greg.m]

After changing the CPU & NB GTL Ref from normal to Auto everest is updating better.
100 or 200 or some other variations didnt work out.

You mean CPU, NB SKEWS

[neo_rtr]

You mean CPU, NB SKEWS

ahhh.. correct..

[greg.m]

Question:

Whats the fsbvtt limit for q9650?
Is 1.50(real) ok?

[kuebk]

Question:

Whats the fsbvtt limit for q9650?
Is 1.50(real) ok?

AFAIR VTT on 45nm cpus shouldn't be higher than 1.35 for 24/6 use.

[greg.m]

A first good attempt @ 500x8.5




ill try lower some volts later

[kuebk]

Too high VTT is going to kill your CPU faster than in your worst dreams.

[neo_rtr]

I have a dream ............ to do the same. and if my cpu then i have the reason to buy a new one.. hehe...

[greg.m]

Here some links about q9650s and voltages:

http://www.xtremesystems.org/forums/...9&postcount=67
http://www.xtremesystems.org/forums/...=202491&page=2




more Rampage extremeresults coming...

[greg.m]

One thing i realised is that with this memory G.SKILL F2-9600CL5D-4GBPI i can achieve much higher fsb clocks. Before with the dominator pc8500 i couldnt pass 480fsb.

[greg.m]

Greg.m presents.......:



.... i see smoke ohh maybe its my q9650....

[Armored_Kech1]

Impressive! Just curious, what is your room temp?

[mikeyakame]

Greg.m presents.......:



.... i see smoke ohh maybe its my q9650....

Very nice mate

I'm guessing watercooling the NB has a huge amount to do with it, since it seems you need ridiculously high Vnb for stability at high fsb with 2gb sticks and > 1200mhz DDR.

[freeagent]

you guys are getting some nice latency there

i have never been below 50ns with an intel setup, not yet anways

[greg.m]

Impressive! Just curious, what is your room temp?

Unfortunatelly my room temp is arround 23-25c