BLOOMFIELDS CLOCK THE SAME

speedbinning suggests that the higher rated chips should overclock further.

all bloomfields are the same.
it seems that there is at least five threads a day on the same subject.

I don't think he was debating speedbinning, simply pointing out that you'd somewhat contradicted yourself. But yeah I can agree the number of topics on this is pretty stupid, I think the issue comes from when quite a few people originally had motherboards that had issues with 200 BLK, hence it seemed that the 960/975 chips clocked higher?
 
He could perhaps be a little clearer - they're not all absolutely identical - each individial processor is different. Some will clock amazingly but some will be dogs. Almost all will do 4GHz though. What he means is there's no speed-binning, so there's no statistical difference between the different 'models' 920, 930, etc etc. Buy a 975 (because you're insane) and you're just as likely to get a good or bad chip than if you buy a 920.

As I've said in another couple of threads - SAY NO TO INTEL GHz TAX!!! i7 975? £500+ quid over 930 price for Intel to give you a few hundred MHz that you can easily just set yourself in bios on a 920? Er, no thanks!
 
He could perhaps be a little clearer - they're not all absolutely identical - each individial processor is different. Some will clock amazingly but some will be dogs. Almost all will do 4GHz though. What he means is there's no speed-binning, so there's no statistical difference between the different 'models' 920, 930, etc etc. Buy a 975 (because you're insane) and you're just as likely to get a good or bad chip than if you buy a 920.

Yeah, so there's no statistical difference between all the different models. It's a game of probability ie: a 920 will likely reach the same speed as a 950, because there's not binning, and any speed difference found in the two is only exclusive to that particular chip and not the whole model range.
 
On a similar note, how come no-one's found any way of unlocking the multiplier on non-extreme i7 CPUs? How is that actually done, in BIOS code or something on the chip itself?
 
I dont get it?

He's saying that it depends on the individual chip rather than the product range as a whole.

ie: he has 3 920 chips which outclock a 950 chip. BUT he probably also has a 930 which outclocks the 920 chips. He might also have a 960 which outclocks them all, and a 975 which can't get close to the speeds of the other ones. What he's saying is that based on probability, a 975 will achieve the same clocks as a 920.
 
i consider outclocking to mean the top end achievable overclock or 'what the cpu is capable of hitting with lots of volts'

sure you get the odd chip that will do 4.6ghz+ but these chips arent going to be capable of sustaining that under conventional cooling.

24/7, all bloomfields overclock the same, within about 5% of eachother. 4.2ghz +/- a few hundred MHz
that figure will be very similar across every cpu in the range.
 
Thanks for this rjkoneill.

What your saying is that its not worth buying a 950 over a 920 as they will both clock the same and require the same volts to reach target overclock.
 
Thanks for this rjkoneill.

What your saying is that its not worth buying a 950 over a 920 as they will both clock the same and require the same volts to reach target overclock.

Essentially I think that's what he is getting at. 950 is about £25 more expensive than a 930 but both should overclock the same as they are based on the same chip. The only difference is the stock speed
 
If rjk is refereing to clocking with highend air, then the view that "they all clock the same" is perfectly true.
However, under water of phase, that view is not always the case.
I for instance use phase 24/7 and the 5ghz clock in my sig is what i use. As you can see, that's with a 975. The same SS will not take ANY 920 that i have ever used anywhere near to 5ghz, let alone on a 24/7 stable basis.
 
For the average user, who will either use a good air cooler or one of those H50 or H70 coolers (watercooling for the inexperience aka me), all I would want out of my i7 would be 4ghz... so in that regard a 930 wil clock the same as a 950?
 
If rjk is refereing to clocking with highend air, then the view that "they all clock the same" is perfectly true.
However, under water of phase, that view is not always the case.
I for instance use phase 24/7 and the 5ghz clock in my sig is what i use. As you can see, that's with a 975. The same SS will not take ANY 920 that i have ever used anywhere near to 5ghz, let alone on a 24/7 stable basis.

yes kfit
exactly that, obviously i am referring to conventional cooling.
there used to be a time where watercooling made a massive difference.
in my opinioin, custom water offers nothing over high end air cooling apart from silence and aesthetics.
obviously your SS unit will churn out more MHz from any cpu. but under a run of the mill heatsink, all bloomfields are gonna be rather similar

Essentially I think that's what he is getting at. 950 is about £25 more expensive than a 930 but both should overclock the same as they are based on the same chip. The only difference is the stock speed

hit the nail on the head there :)
 
without wishing to sound rude or condecending.

if you have a bloomfield

and you have a motherboard that does 200 BCLK

then you can get 4.2ghz with ease



200BCLK
21 multi

memory set to rated speed+timings

llc enabled
cpu vcore to 1.35v - 1.38v
dram volts to 1.65


done - no issues. these settings work on every single board

I tried the exact same settings on my EX58A-UD3R and it blue screened several seconds into prime, so I'm a bit confused :p

4.0ghz seems a bit more steady, but still crashes now and then - still tweaking it. 4.2ghz just dies instantly though, even with 1.38v.

I'm presuming you need to tweak the other voltages a bit more, as they didn't seem happy with auto :cool:
 
Well voltages change a lot depending on the luck of the chip. Mine can do 3.8 GHz at 1.225 V (maybe lower) but requires about 1.27 V for 4 GHz (if I remember correctly).
 
Last edited:
QPI is a large factor in obtaining a stable overclock. QPI is dependent on memory speed/size/frequency so there is more to it than what Rjk has stated. That said if you leave QPI to auto, it normally over compensates and the settings Rjk had suggest should work in the majority of cases. I would however advise caution on leaving voltages to AUTO as most manufacturers will set this level way too high e.g Gigabyte seems to think my rig needs need 1.58v when in fact 1.45 is enough for 24/7 stability.
 
My i7 920 4.2 when using a H50 > MX2 > Viper/Apache > Vcore 1.3125 > Vdrop >1.296, quite happy at 4.2, have no interrest in seeing what she will do...
 
Back
Top Bottom