i3 540 - Arghhh!

[marvin-miller]

Hi Folks!

New user here and first time overclocker (well, almost, the last computer I overclocked was an Apple //GS ).

I've got an i3 540 in an Asus P7P55D-E LX with a Corsair HD50 cooler on it. I've also got two sticks of DDR-1333 memory (7-7-7-21)

Before I bought all this stuff I did some research and pretty much everyone had good things to say about it - all of it. In fact, it seemed to me that it would be a very good candidate for overclocking - which is something I wanted to do.

To that end I read through the '3 steps to overclocking your Clarkdale' guide and followed along.

The max BLCK I was able to get was 218 @ 1.25 volts (cpu volts). I then went on to the memory section of the article and that's where things started to fall apart. I don't really understand any of it and I'm not getting anywhere but it seems to me that it's the memory side of things that is holding me back.

The best I've been able to do is 4.214 Ghz and it's not stable enough.

21x multipler
200 BCLK
1.35 CPU Volts
1.80 DRAM Volts
1.36 IMC Volts
I believe this was with the memory at 800 Mhz or so (DDR3 - 1600? I don't really understand the memory side of things...)

It doesn't seem to matter if I turn the CPU or DRAM volts up with those settings - I can't get Intel Burn Test to do 20 passes on the max settings. It will usually do 5-8 before a BSOD.

Most people on the net seem to be getting 4.5 or at least 4.2 with no issues or special hardware. So, I'm kind of wondering why I can't get 4.214 to work especially given that my temps seem to be at least 10c less then the folks who are getting higher speeds.

Can anyone help me get to that magic recipe?

 

[MattNo5ss]

to OCForums!

You shouldn't try running your 1333MHz RAM at 1600MHz, your RAM probably isn't stable with a 266MHz OC. Have you tried changing the RAM multiplier to 6 for 1200MHz RAM? What's the rated voltage on that RAM? Typically, it's 1.65v.

 

[marvin-miller]

Thanks for the welcome and the reply

OK, so stop aiming for 1600 Mhz memory? That's what I've been doing so that's probably one area where I'm way off. I've been trying (continually) to reach as close to 1600 as possible thinking that was the 'proper' value

The rated voltage on the memory is 1.5 volts.

I don't think I have a RAM multiplier (hence a lot of my confusion) but I can usually choose between numbers like 1200,1600,1800,2000 etc.

So, should I be aiming for numbers like 1200 or 1140 or the like?

 

[MattNo5ss]

It sounds to me like you are trying to OC both your CPU and RAM together. So when a crash or instability arises, you don't know which is the cause, CPU or RAM.

Those 1200, 1600, 1800, 2000, etc. are your blck * RAM multiplier. So, if you divide those speeds by your bclk, you get your RAM multiplier.

1200/200 = 6
1600/200 = 8
1800/200 = 9
2000/200 = 10

I would suggest using that 6 multiplier and OC your CPU until you get it maxed or at whatever you are shooting for. Then, you can try increasing your RAM multiplier, but I doubt your RAM will be able to handle anything higher than the 6 multiplier since it's only rated for 1333MHz.

 

[marvin-miller]

Perfect!

Thanks for explaining that to me. No wonder I couldn't get any further and had to keep turning up the ram voltage - I was aiming for something that was unattainable

I'll do another session and follow your advice - this time trying to keep that RAM at or around 6x. Hopefully this time I will get further along

 

[Grief]

you could also try lowering bclk, and useing higher multi for the cpu guess your ram could do 1400 or 1500 ram speed

 

[fishys87]

to OCForums!

You shouldn't try running your 1333MHz RAM at 1600MHz, your RAM probably isn't stable with a 266MHz OC. Have you tried changing the RAM multiplier to 6 for 1200MHz RAM? What's the rated voltage on that RAM? Typically, it's 1.65v.

I'm at a 37% OC (or 489mhz/stick OC) on my 1333 ram - 100% stable (1.68v). I've even gotten it as high as 981mhz, 100% stable. Any higher and my IMC can't handle it.

It's defiantly possible.

And yes, my stock voltage (VID) is rather high.. 1.176 for 2.93ghz (stock). The vcore is under load, btw, not idle.

 

[MattNo5ss]

I'm at a 37% OC (or 489mhz/stick OC) on my 1333 ram - 100% stable (1.68v). I've even gotten it as high as 981mhz, 100% stable. Any higher and my IMC can't handle it.

It's defiantly possible.

And yes, my stock voltage (VID) is rather high.. 1.176 for 2.93ghz (stock). The vcore is under load, btw, not idle.

I didn't say it wasn't possible The RAM's overclock ability depends on the memory chips used, and it looks like you have some good ones.

For the sake of finding your best OC, it's wise to do it one component at a time so you'll know what's causing the instability.

 

[fishys87]

I didn't say it wasn't possible The RAM's overclock ability depends on the memory chips used, and it looks like you have some good ones.

For the sake of finding your best OC, it's wise to do it one component at a time so you'll know what's causing the instability.

I agree, and concur.

I honestly don't have "great" OC ram.. it's just what Newegg gave me in the bundle with my case..

It's your run-of-the-mill $50 4GB Ripjaws - 1333 @ Cl9. I've been running @ this OC/ram-OC since about day 3 or 4 now.. has been nothing but fantastic.

 

[MattNo5ss]

I would say you have some really good 1333MHz RAM. You're running at ~1800MHz with lower than rated timings even. Your not at 2000MHz+ that some RAM can do, but you're working with RAM rated at 1333MHz.

 

[Bobnova]

That makes me wonder if those are first gen Hyper IC based sticks, that's the only 1333 rated stuff i know of that will run that high.
Fragile though, sadly.

 

[marvin-miller]

It sounds to me like you are trying to OC both your CPU and RAM together. So when a crash or instability arises, you don't know which is the cause, CPU or RAM.

Those 1200, 1600, 1800, 2000, etc. are your blck * RAM multiplier. So, if you divide those speeds by your bclk, you get your RAM multiplier.

1200/200 = 6
1600/200 = 8
1800/200 = 9
2000/200 = 10

I would suggest using that 6 multiplier and OC your CPU until you get it maxed or at whatever you are shooting for. Then, you can try increasing your RAM multiplier, but I doubt your RAM will be able to handle anything higher than the 6 multiplier since it's only rated for 1333MHz.

The problem seems to be (other then the one between the keyboard and the seat) that when I change the multiplier or other settings then the memory automatically changes. It seems that I can't force the memory to run at 1200 all the time and then change other things?

I seem to have made some progress though as this looks like it's stable....

 

[jaymz9350]

You are currently running your memory at 1200 (well 1204) by that screenshot (the value from CPU-Z needs to be doubled for the actual speed.

Another question, when you tried for 1600 on the RAM did you loosen the timings as well or leave them at 7-7-7-21?

 

[fishys87]

That makes me wonder if those are first gen Hyper IC based sticks, that's the only 1333 rated stuff i know of that will run that high.
Fragile though, sadly.

Not sure.. I don't know the full story of the first gen Hyper's, but I have read on some accounts, them even failing before 1.65v. I've volted 1.90 volts through them at various points, just itching to get that 2000mhz plataeu. Though I completely understand benchmarks isn't about just speed; it's a combination of speed and the right timings.. but nevertheless, 2000 is a nice achievement for 1333 ram.

I might be replacing them with the ST "current" gen Hyper's and a Xeon X3440 next Friday (the Xeon for sure). Though if even in maxxmem, if the differences are negligible, I may return the ST's and keep my little ripjaws. I've seen the outstanding stuff the 2000 ST's put out (ie Cl5 @ 1600 with some volts).


Think I could see even better results with the 2000 ST's?

 

[marvin-miller]

You are currently running your memory at 1200 (well 1204) by that screenshot (the value from CPU-Z needs to be doubled for the actual speed.

Another question, when you tried for 1600 on the RAM did you loosen the timings as well or leave them at 7-7-7-21?

No, I left them at 7-7-7-21.

The reason I put the memory back down to 1200 was because it was suggested to me earlier in the thread that my problems were probably from trying to keep the memory at 1600.
BTW, the memory is G.Skill RipJaws DDR3-1333 7/7/7/21 @ 1.5 (factory specs).

 

[fishys87]

No, I left them at 7-7-7-21.

The reason I put the memory back down to 1200 was because it was suggested to me earlier in the thread that my problems were probably from trying to keep the memory at 1600.
BTW, the memory is G.Skill RipJaws DDR3-1333 7/7/7/21 @ 1.5 (factory specs).

I am not as experienced as most (probably 99% here), but try loosing the timings, say Cl9 - and try 1600 @ 1.66. If not, try Cl9 @ 1600 @ 1.68v.

What kind of QPI settings do you have?

 

[jaymz9350]

No, I left them at 7-7-7-21.

The reason I put the memory back down to 1200 was because it was suggested to me earlier in the thread that my problems were probably from trying to keep the memory at 1600.
BTW, the memory is G.Skill RipJaws DDR3-1333 7/7/7/21 @ 1.5 (factory specs).

I guess I misunderstood your previous post about the 1200 on the mem. Yea you do want to see what the chip can do with your mem at or below it's rated specs. After you find the cpu speed you want then you can try upping the mem and as fishys87 stated you'll have to loosen the timings to something more like 9-9-9-24 for the higher mem speeds and maybe up the voltage some.

 

[fishys87]

It should run at Cl7 @ 1333, however I have read in some instances it doesn't. So maybe loosing the timings, and bumping the multi to get to 1600(ish), maybe you can boot into Windows. If you can, run some stability tests, or whatnot. If it's working, then see if you can boot into Cl8 @ 1600.

You may have to bump the voltage a tiny.. like I had to bump mine up to 1.68, regardless if I was in at 1600 or 1962.

 

[marvin-miller]

OK, so the gist of it is that if it doesn't seem to want to run at the higher speeds then loosening off the timings may help?

I don't recall what QPI settings I had. Actually, I don't think I ever changed that. The only settings I was using was multiplier, blck, processor voltage and memory voltage.

The BIOS for this board is not like all the gigabyte screen shots I've seen so prevalently on the net. It seems to adjust the memory speed automatically when I make changes to the blck and the multiplier. So it's not like I can choose between 1066/1200/1600/2000 - the BIOS changes those values when I change other settings. I don't seem to be able to turn it off

I was up 'til 4am this morning trying to make it go faster. Right now it's recording some shows and as soon as it's done I'll climb back into the BIOS and try loosening the timings. In the past I very nearly got this working (it seemed to be a 'sweet spot';

21x200 @ 1.35 (cpu) 1.8 (ram) 1.36 (imc volts). That was with the memory set at 1600 with 7/7/7/21.

The issue was that it would not complete the burn tests completely. It seemed to be close so maybe if I increase the memory timings it might work.

BTW, during testing of the above 'recipe' I actually put the memory volts to 1.90 and the cpu to 1.375 and it was still no go. In fact, it almost seemed that the more voltage I threw at it the worse it got :-(

 

[fishys87]

OK, so the gist of it is that if it doesn't seem to want to run at the higher speeds then loosening off the timings may help?

I don't recall what QPI settings I had. Actually, I don't think I ever changed that. The only settings I was using was multiplier, blck, processor voltage and memory voltage.

The BIOS for this board is not like all the gigabyte screen shots I've seen so prevalently on the net. It seems to adjust the memory speed automatically when I make changes to the blck and the multiplier. So it's not like I can choose between 1066/1200/1600/2000 - the BIOS changes those values when I change other settings. I don't seem to be able to turn it off

I was up 'til 4am this morning trying to make it go faster. Right now it's recording some shows and as soon as it's done I'll climb back into the BIOS and try loosening the timings. In the past I very nearly got this working (it seemed to be a 'sweet spot';

21x200 @ 1.35 (cpu) 1.8 (ram) 1.36 (imc volts). That was with the memory set at 1600 with 7/7/7/21.

The issue was that it would not complete the burn tests completely. It seemed to be close so maybe if I increase the memory timings it might work.

BTW, during testing of the above 'recipe' I actually put the memory volts to 1.90 and the cpu to 1.375 and it was still no go. In fact, it almost seemed that the more voltage I threw at it the worse it got :-(

There's got to be something "QPI" related on your board. I had mine set to auto for a while. While I knew what it was, my settings were in "GT/s", and I never knew what I was changing. A helpful soul got me through it, and I understand it now, but I have a EVGA board.

So you actually were/are running 1.8/1.9v through the mem, and your still not posting; having stability problems?