375MHz FSB Brick wall on Gig 965p DS3 rev3.3?

[jsh712p]

Hi guys,

After some help from you guys a while back my system has been running at about 2.8Ghz for a while.

After fitting HS05 and A C Freezer 7 I though id try and reduce the multiplier, and increase FSB.

Ive got it all nice and stable running 375*8 - with everything stock except MCH +0.1v

However, if i try to get 380 I cant POST.

I have tried many combinations, including increasing vcore to 1.4, and increasing dimm voltage, fsb voltage and increasing mch to 0.2 (at the same time and separately).

I have also tried 380 with a 7 multiplier to eliminate the CPU from the problem - but nothing seems to let me get there.

So my question is this just the limit for my board?

Im happy with 3.0GHz from a 1.8 chip - but just wondered if anyone else has encountered this?

Thanks again.

Martin

 

[bfdd]

Probably the CPU, I have the same board I had a 3.73ghz oc on my previou e6600 which messed up when my PSU went poof, this can only get to 3.4ghz(378mhz fsb) not an inch further.

 

[jsh712p]

Ok,

So the processor can have a max fsb - not just overall frequency? Because the non post happens even with the setup as 380 with a 7 multiplier (2660MHz). I thought I would be ok procssor wise as long as the multiplier is kept low?

TIA

 

[loafbred]

This threadshould give you an idea of the effects of the strap changing at certain fsb ranges. This example is with an Asus board, but I would think it's typical for most 965P boards.

 

[Ravynmagi]

I have the same problem with my E4300 and the DS3 rev 3.3. I'm not sure if I tried going as low as 375 (think I'll give that a try). After I confirmed 3GHz was very smooth at 9x335, I tried jumping to 8x380 or 7x430. I tried everything. But could never get stable at 380 or various FSBs above. I finally gave up. Not sure if it's the CPU or the board though.

 

[jsh712p]

Thanks for that thread link. Thats really interesting. I wonder how many people know about the Northbridge clock calculations like that.

Ravynmagi - you will want to check that out as well - read down to see the explanations etc.

Very interesting - ill let you know how i get on

Thanks again,

 

[Ravynmagi]

Hmm... so the high multiplier of the 4300 is working against us. Hmph. I bought it thinking a high multiplier was a good thing. The article says just lowering the multiplier is raising my NBCC.

Does the 680i chipset have this issue as well? I wonder if the P35 behaves this way as well?

Oh well, I did find 8x378 is working real well for me. (9/8)*378 gives me a 425MHz NBCC. The boot strap is probably holding me back more than the NBCC. Unfortunately I can't do 8x401. 3.2GHz seems to be more than this CPU will do (can't do 9x356 3.2GHz either) and 7x401, aka (9/7)*401 is 515MHz NBCC, which seems to be more than my board can do. So I'm caught between the boot strap and the NBCC cap.

I guess if I've understood this, then it now makes since why I'm having the problems I am.

 

[MrK6]

The DS3 boards should be good for up to 500MHz FSB. I'm running my E4300 at 334x9 (3.0GHz) at stock volts, didn't want to push it further as this is just a hold-over CPU until quad core. I'm curious though, have you guys tried keeping the 9x multiplier and just cranked the FSB while leaving everything else low (especially RAM multiplier)? I've been up to about 3.25GHz in Orthos at only 1.4V, where she got really unstable, but I don't want to test further in the heat.

 

[Ravynmagi]

Bleh. I was wrong. That link didn't seem to help or explain my problems.

I wanted to test my NBCC limits, so I lowered the multiplier to 6. (9/6)*350= 525 NBCC. That ran for half an hour fine. I then rebooted and went to (9/7)*401 = 515 NBCC and can't even boot with it. 2.8GHz well below what my CPU can do, 515 NBCC below what I was able to do, and I'm above the boot strap change. Hmmm...

Yeah, I've left my multiplier at 9 and raised the FSB up to 356 to get me 3.2GHz. It boots, but isn't stable with Orthos, seems 3.15 seems to be the most I can do with conservative voltage. But in my case, I'm happy with my 3.0GHz, my goal was to raise the FSB so my memory runs faster and stays in sync with the FSB at 1:1. My memory can do 550, seems a shame I have to run it at 335.

Well, it does seem I'm hitting some sort of limitation either CPU or motherboard. Guess I'll just need to figure out if I should stay in sync with the 9x335FSB or run the memory faster, but async.

 

[jsh712p]

well thats thrown a spanner in the works hasnt it -

Ive just had a BSOD while running 8*375. this gives me a NBCC of 421 i think - which i though would have been ok - but the crash could be low voltage to cpu - as i lowered it to 1.3 on the dot. I might retry this with higher CPU voltage.

At the moment im running 335 * 9 - with the ram at 2.5 x 335 = 837. i have the timimg up a bit to compensate for the slight overclock.

This seems stable so far.

It is annoying that it does that calculation for northbridge clock - i have verifed this in sandra.

I also couldnt do 8*401 - even ramping cpu vocre to 1.4 (as high as i want to go), and putting all others to +0.2v - it still refuses to start. And there is no point trying 7*401 - as that pushes the NBCC too high.

I agree that now the lower default multiplier cpus are looking more attractive.

but like you I am happy with 3.0Gz from a 1.8!!!!!!

Cheers again for the info and assistance.

 

[loafbred]

I've done very little testing with my E6600 and Asus Commando, as I'm using the retail Intel cooler until my Freezer 7 arrives today. I ran it for a day at 9*300 w/1.25v vcore, and am now running at 6*450, which runs the RAM at DDR-2 866 (4-5-5-15, 2.1v vdimm). This gives me a Sandra memory bandwidth score which is comparable to a 975 board at DDR2-800, if I remember correctly. Those are the only two configurations I've tried.

I'm anxious to see what happens with a higher multiplier in the mid-to-high 400 range. I'm taking the advice given in that article, and staying away from the high-300 to low-400 range. Once I settle on something, without jacking the vcore up, I'll post some screenshots of the numbers read by Sandra and CPU-z.

 

[loafbred]

I got the new cooler installed and tried two different 3.0 GHz settings; 9*333 and 7*433. Both were stable for about fifteen minutes in Orthos, at which time I stopped it, with the vcore still at 1.25v (it's set to 1.28, but shows 1.25 actual). Sandra and CPU-z both give false readings, as you can see in this screenshot.

Another screenshot showing memory bandwidth and Orthos.

Both screenshots were made at 7*433, RAM at DDR-2 866, 4-5-5-15.