Higher multi of i7 950

[aamsel]

Will the higher multiplier of the i7 950 likely help
produce an overclock equal to a 920/930 with
lower voltage, since less BCLK will be needed?

 

[Ben90]

It entirely depends on the situation and what part is holding you back.

Most motherboards are fine until ~190 Bclock, from there you need more voltage but I believe 200 is in reach of every x58 ATX board today.

A 930 at 200 bclock and a 22x multiplier will yield you 4.4ghz. Not very many processors will allow you to do this with a reasonable voltage on air keeping settings like HT on.

At 200Bclock, the slowest memory divider will translate into DDR-1200. Aside from the initial batches of DDR3 that were waaaay out of JEDEC spec, any RAM you can find on Newegg will hit 1200 with ease. I bought some budget DDR3-1066@9s based on Micron D9s and with some love they can do around 1600.

The final variable of the equation is your uncore. Because of the 2x DDR-speed requirement of Bloomfield processors, the lowest Uncore you can use is 2400Mhz. This is a 272Mhz bump from stock, and have yet to see an Uncore that cannot handle this speed.

The conclusion from this is that if your mobo can handle a 200 Bclock, which it should, you will most likely be voltage bound by your cores before anything else. The extra multiplier from the 950 will not help in this situation.

 

[aamsel]

I have tried several 930's that appeared to get to 4GHz, but high temperature (with Megahalems) caused a LinX or OCCT failure of one or more cores with Hyperthreading on.

My hope was that having higher multi might have the 950 hitting a more reasonable temperature at the same overclock, guessing that BCLK has been the issue to this point. I realize this is hit or miss, and chip by chip, but that was my thought anyhow.

I would like to have a Westmere, but the E5620 is the only one priced non-exorbitantly.

Any advice welcome.

 

[Ben90]

Unfortunately, higher multipliers do not yield lower temperatures. A 4000Mhz Bclock using a 1x CPU multiplier would have the cores running at the same temperature as a 1Mhz Bclock and a 4000x multiplier. Whatever generates the Bclock would be boiling in the first example;yet, the cores would be the same temperature.

I find it odd that you are overheating with a Megahalems though. I personally only use the stock cooler so my help will be limited, but perhaps Aigo/Someone with more experience could help you diagnose the problem. If you decide to seek help, try to include as much information as possible because diagnosing cooling problems can sometimes be somewhat of an art.

 

[Rifter]

what are your temps with that Megahalems?

 

[aamsel]

They get to the high 80's in LinX and then a core fails.
I am using 2 good fans in push/pull.
I have tried several types of TIM.

 

[Rifter]

ouch thats hot, what are your ambient temps?

 

[aamsel]

80f

 

[Rifter]

that pretty hot, and your temps are not alot higher than mine with my ambients at the same i will hit high 70's. Ive been up to low to mid 80's when my ambients got up to 85-90f(27-32c).

Never had an error in prime or linx yet though even at 84c.

It really sounds like maybe you've just been getting bum CPU's they wont all play nice at 4Ghz+

 

[aamsel]

...It really sounds like maybe you've just been getting bum CPU's they wont all play nice at 4Ghz+

I think you are right.

 

[bryanW1995]

I think you are right.

my ambient is ~ 80f as well, and my max oc of 4.2 yields temps in the 80's (high of 87 running 300 min linx followed by 48hrs seti w/ optimized apps). technically it's 13c delta to tjunction since I have tjunction set at 100c. is your tjunction also set at 100c? even at that high of a temp I didn't error out, and I'm using a TRUE in push pull. my vcore at max is 1.36875 btw.

1. did you follow the oc guide here?
2. what is your vcore at max? what are your other voltages?

 

[Rifter]

my ambient is ~ 80f as well, and my max oc of 4.2 yields temps in the 80's (high of 87 running 300 min linx followed by 48hrs seti w/ optimized apps). technically it's 13c delta to tjunction since I have tjunction set at 100c. is your tjunction also set at 100c? even at that high of a temp I didn't error out, and I'm using a TRUE in push pull. my vcore at max is 1.36875 btw.

1. did you follow the oc guide here?
2. what is your vcore at max? what are your other voltages?

you might be right, a little more voltage might help stay stable. Those temps are borderline though.

 

[aamsel]

I did follow the guide went up to 1.35 vCore, then called it off.
The others were increased only slightly from stock.
I will keep better notes with my next attempt.

 

[Rifter]

how high did you raise the QPI/uncore voltage? i found i needed it right up there at 1.35V same as the Vcore to be stable past 4Ghz.

 

[aamsel]

how high did you raise the QPI/uncore voltage? i found i needed it right up there at 1.35V same as the Vcore to be stable past 4Ghz.

I did find some notes.

The QPI/uncore got to 1.35v, just as you suggested.

Anyhow, tomorrow I will have a new W3520 to check out.
I decided that I would have a better shot at getting a decent chip
with it instead of a 930/950.

We shall see.

 

[MarkAtHome]

i7 930 @ 21 x 200 = 4200Mhz 1.35v w/Noctua NH-D14 both lapped on a ASUS P6X58D-E

Hi Rifterut -- what does the above mean in regards to lapped on?

I have the same cooler on order for my i7 950, intended to replace my Coolermaster V8.

 

[ehume]

Lynnfield chips tend to run hot.

Higher multi's do help. At 22x182 my cpu needs a lower voltage and produces less heat at 4GHz than it does at 21x191.

Cooling power helps. I have a Megatron but I went with my D14 because I needed better cooling.

I also have two 860's. I'm using the one that requires lower Vcore at high clocks. The other is in my non-oc'd machine.

 

[Rifter]

Hi Rifterut -- what does the above mean in regards to lapped on?

I have the same cooler on order for my i7 950, intended to replace my Coolermaster V8.

lapped, as in sanded to a mirror polish, to get the best metal to metal thermal contact.

 

[aamsel]

Lynnfield chips tend to run hot...

I have read this in several places.
What is the actual reason that Lynnfield runs
hotter than Bloomfield as a rule?

Higher multi's do help...

I assume that getting a chip that has a lower VID to start with may be a factor also?

 

[Rifter]

I have read this in several places.
What is the actual reason that Lynnfield runs
hotter than Bloomfield as a rule?

I've been told its due to the PCIe controller being on die on lynnfield, and having the voltage to run that as well as all the data for that bus going through the CPU. So then it increased voltage requirements and heat output. I dont know for sure if that true but it does seem logical.

 

[aamsel]

I've been told its due to the PCIe controller being on die on lynnfield, and having the voltage to run that as well as all the data for that bus going through the CPU. So then it increased voltage requirements and heat output. I dont know for sure if that true but it does seem logical.

I think you are correct.

They also made a switch where there is no QPI as with Bloomfield.

 

[ehume]

I read somewhere that the chip that is no longer separate - was it the Northbridge? - was the hottest part of the mb, and often needed its own cooling. My PCH - the replacement for various parts of the chipset - never reaches 40c, no matter how hard I stress my system. Usually it's in the low- to mid-30's. So whatever was in that hot chipset went somewhere. I suspect it's on-die in the Lynnfield.