If one is looking for the biggest overclock, would they be better off with several i7 920s or an i7 975

[Smartazz]

I actually have no intention of doing this, but I'm curious to know what would theoretically be the better option if someone was looking for the most extreme overclock using some form of extreme cooling(liquid or phase change).

If someone went the 920 route and bought three of them, then tried to discover which chip overclocks the highest, they could get lucky with at least one of them. On the other hand, if someone bought the i7 975, it's a better binned chip.

What's everyone's thoughts on this?

 

[Ben90]

264 to my knowledge is the world record bclock on a classified. 264*21=5544 world record 920
975s can reach up to just barely over 6ghz

Both of those are under ln2. If ur going water, 3 920s gives u a higher chance of a better chip. I dont really know too much how phase ranks up with ln2, but i think that it would start having problems after 5ghz (not 100% sure)

920s and 975s are basically the same thing last i heard... although i do remember there being some controversy because people were whining how someone elses 920 could do 4.6 on water, while their 975 could only do like 4.2. I havent been paying too much attention, but it is possible that Intel does bin their 975 more aggressively not, but i find that highly unlikely still, because it just adds costs

 

[Idontcare]

Originally posted by: Ben90
264 to my knowledge is the world record bclock on a classified. 264*21=5544 world record 920
975s can reach up to just barely over 6ghz

Both of those are under ln2. If ur going water, 3 920s gives u a higher chance of a better chip. I dont really know too much how phase ranks up with ln2, but i think that it would start having problems after 5ghz (not 100% sure)

920s and 975s are basically the same thing last i heard... although i do remember there being some controversy because people were whining how someone elses 920 could do 4.6 on water, while their 975 could only do like 4.2. I havent been paying too much attention, but it is possible that Intel does bin their 975 more aggressively not, but i find that highly unlikely still, because it just adds costs

The implication there is that Intel bins the 975's not only for stable operation at stock speeds but that they also bin them for overclocking prowess as well.

While I can understand how folks could convince themselves that this might be the case, there is actually no such "overclocking headroom validation" binning going on with the 975's any more so than it going on with the 920's.

You are guaranteed stock clocks, that is all that Intel is going to invest money into validating at bin and test. These aren't cherry-binned TWKR chips.

The "rationalization" trap that people get themselves wound-up in is the shmoo plot, or rather the fundamental device physics that give rise to the shmoo plot are also what gives rise to the concept of entitlement that enthusiasts tend to want to believe they have coming to them when they get a low VID chip or buy a high stock-clock chip.

But each IC has its own unique set of clock-limiting defects, entropy and the second law of thermodynamics makes sure of this, so while Intel does their job in establishing/guaranteeing one datum point along the cpu's shmoo curve it is up to us to determine where the rest of the curve lies, including where it ends. For some 920's it may very well end at a clockspeed higher than that of some 975's.

 

[palladium]

^ So IDC, if I get you correctly, on average there should be no difference in OC headroom between 920s and 975s on the same MB, right?

 

[Ayah]

Originally posted by: palladium
^ So IDC, if I get you correctly, on average there should be no difference in OC headroom between 920s and 975s on the same MB, right?

A 975 has an unlocked upper multiplier. You could run 40X200 if your chip would let you run that fast.

The unlocked multiplier does however tend to lead to easier and higher overclocks, but at a dramatic multiplication in price.

 

[aigomorla]

Originally posted by: Smartazz

What's everyone's thoughts on this?

LOL i think your trying to justify getting a 975.

http://i125.photobucket.com/al...enchmarks/newstats.jpg

And she can go a lot higher. 1.38 vs 1.55 which is max intel spec...
On water you could probably push 1.45-1.5
On Phase... 1.55

I just made your thinking process a lot harder now huh?

There is no guarantee tho.
But you do have a higher chance at getting a better cpu.

Originally posted by: Ayah

The unlocked multiplier does however tend to lead to easier and higher overclocks, but at a dramatic multiplication in price.

yes it does.

She was cruising on 30x

 

[Rubycon]

Or just get a W3520 and unlock it yourself.

 

[aigomorla]

Originally posted by: Rubycon
Or just get a W3520 and unlock it yourself.

THATS how you beat my 975... (thats cheating.. )

i was wondering how the hell.... :

Meh... i'll spank your W3520 now tho.

 

[Rubycon]

Originally posted by: aigomorla

Originally posted by: Rubycon
Or just get a W3520 and unlock it yourself.

THATS how you beat my 975... (thats cheating.. )

i was wondering how the hell.... :

Meh... i'll spank your W3520 now tho.

I seriously doubt it - lots of CPU WAITING for those slow disks you have over there!

The key is the uncore - always has been. I'm sending one of my 3520's that has been doing 4.5GHz on air 24/7 stable for months to someone to perform cascade plate testing. The goal is -100C load temps. I'm asking for the compartment to be mic'd as well in case something pops! :laugh:

 

[aigomorla]

as long as there is no disks involved... i have more guys doing more work on my system right now.

 

[YBS1]

Originally posted by: Rubycon
Or just get a W3520 and unlock it yourself.

Wait...What?? Do tell, first I've heard of this. Especially since I think I'm going back to water in order to get these temps on mine inline.

 

[n7]

Originally posted by: Rubycon
Or just get a W3520 and unlock it yourself.

Wat?

Please do tell.

 

[Zap]

Originally posted by: Rubycon
Or just get a W3520 and unlock it yourself.

Explain yourself, plz. :beer:

Originally posted by: Ben90
264 to my knowledge is the world record bclock on a classified. 264*21=5544 world record 920
975s can reach up to just barely over 6ghz

Isn't it pretty tricky to get most X58 boards to work at over 220MHz bus, and some can't even make 200MHz?

 

[deputc26]

Originally posted by: Rubycon

Originally posted by: aigomorla

Originally posted by: Rubycon
Or just get a W3520 and unlock it yourself.

THATS how you beat my 975... (thats cheating.. )

i was wondering how the hell.... :

Meh... i'll spank your W3520 now tho.

I seriously doubt it - lots of CPU WAITING for those slow disks you have over there!

The key is the uncore - always has been. I'm sending one of my 3520's that has been doing 4.5GHz on air 24/7 stable for months to someone to perform cascade plate testing. The goal is -100C load temps. I'm asking for the compartment to be mic'd as well in case something pops! :laugh:

Haha! Aigo and Ruby, I'd love to see the numbers you guys are competing on and how this plays out

 

[aigomorla]

Originally posted by: n7

Originally posted by: Rubycon
Or just get a W3520 and unlock it yourself.

Wat?

Please do tell.

only possible on the R2E and P6T6 from what im hearing.

Basically its an unlock in bios, only works on xeons, not on 920's.

BWAHAHAHAHA ruby your mailbox is gonna get nuked now...

Originally posted by: deputc26
Haha! Aigo and Ruby, I'd love to see the numbers you guys are competing on and how this plays out

upwards in 4.6ghz+ area..

however i spanked her with +2C
(if u get my drift on the +2C)

 

[Idontcare]

Originally posted by: palladium
^ So IDC, if I get you correctly, on average there should be no difference in OC headroom between 920s and 975s on the same MB, right?

Not really because the 920's and 975's don't come from an infinite pool of samples when binned.

Binning 975's removes a certain percentage of good-clocking chips (one's that already could clock to 3.33GHz at stock VID) from the pool of CPU's that are destined to be validated for 2.66GHz operation.

Basically for every 975 they pull from the binning distribution leaves one less available cpu to form the far-right hand tail of the distribution of clockspeed potential for the 920 samples.

Your statement is correct in the limit that the total number of 975 SKU's goes to zero or the limit that the total number of 920 samples goes to infinity. But in the middle-ground of reality the truth is that the 920 distribution is somewhat "clipped" as Intel carves out the 975 SKU's from it (but not ALL of them, some get left over after 975 binning and they become the sought after 920 "golden samples").

So what we can say is that a 920 can overclock just as much as a 975, but if we tested 10,000 samples of each we'd find the distributions don't overlap exactly, the 975 distribution will likely have a tail to it on the right-hand side (higher clocks) that is much higher than that of the 920 at any given clockspeed (since those 975's were stripped out of the same total sample distribution at time of binning and test).