Haswell pentium/celeron can do ram overclocking?

[xCero]

Hi,

 

With proper motherboard like Asrock Z87 OCF or Asus Impact, these cpus can do overclock? the IMC could support 2600+ MHz? (sorry for my english)

 

Thanks for the info

[TaPaKaH]

The low end chips clock memory very well, 1300 is usually easy on a decent board.

But, you don't have straps on these chips, so with 2933 being the highest available multi, ~1550MHz suicide is max you can hope for.

[xCero]

Thanks for the answer Sam

[l0ud_sil3nc3]

IMHO this is a poor way to test/bin ram because you are running such a low cache speed that mhz/timing combinations that would normally fail on a 4770K will run easily on these Pentium/Celeron IMC's (i.e. 2700+ 8-12-8-28)

 

Steer clear of any ram binned this way or your going to have a bad time at high frequency on an i7 setup when pushing uncore/cache.

[websmile]

IMHO this is a poor way to test/bin ram because you are running such a low cache speed that mhz/timing combinations that would normally fail on a 4770K will run easily on these Pentium/Celeron IMC's (i.e. 2700+ 8-12-8-28)

 

Steer clear of any ram binned this way or your going to have a bad time at high frequency on an i7 setup when pushing uncore/cache.

 

I agree, above 2666 results will need much more voltage on a K-CPU system - I have such a system myself, results above 2600 only show you the potential, on all my mems I have to do retests on Z77 at least with memtest to ensure the mems can take bandwidth and to see voltage combinations. On the other hand, Haswell is by no means as reliable memory binning system like Ivy was... - so normally I would advise you stay clear of haswell anyway

Edited March 24, 2014 by websmile

[TaPaKaH]

The problem with Haswell K-chips (especially retails) is that IMC capable of running PSC at 1333+ 8-12-8-28 on air even at Pentium core/cache speeds is ridicilously difficult to find.

Another problem with result reliability are the frequency-dependant voltage holes with PSC at 1333 and onwards. For example, you can have a kit capable of doing certain settings at 1.78-1.79 all day long, 1.80-1.84 will fail first loop for no apparent reason, 1.85+ will work fine again. When you're on air binning mems, you're usually not in a hurry so you can figure such things out. But when you're pushing other components and just want stable RAM clocks, this might indeed be a big nuisance.

 

In general, I think that PSC (and may be also some BBSE) are the only ICs running which at high speeds (1333+) with high-clocked i7 is noticeably harder than with low-clocked Pentium. So yes, low-clocked results might be misleading in a way but if you take that into account, you can still make a decent judgement on kit quality rather than shelling out hundreds of dollars/euros on forums/ebay for kits without a single listed result.

What comes to the other ICs - from what I've experienced so far on different kinds of Samsung and Hynix, you get a bigger voltage difference going from board to board than from going from Pentium to a 5GHz i7.

[ObscureParadox]

On a side note from this, is it likely that the new Pentium K series will be the best DRAM overclockers when they come out with they have the extra straps available to them?

[l0ud_sil3nc3]

The problem with Haswell K-chips (especially retails) is that IMC capable of running PSC at 1333+ 8-12-8-28 on air even at Pentium core/cache speeds is ridicilously difficult to find.

Another problem with result reliability are the frequency-dependant voltage holes with PSC at 1333 and onwards. For example, you can have a kit capable of doing certain settings at 1.78-1.79 all day long, 1.80-1.84 will fail first loop for no apparent reason, 1.85+ will work fine again. When you're on air binning mems, you're usually not in a hurry so you can figure such things out. But when you're pushing other components and just want stable RAM clocks, this might indeed be a big nuisance.

 

In general, I think that PSC (and may be also some BBSE) are the only ICs running which at high speeds (1333+) with high-clocked i7 is noticeably harder than with low-clocked Pentium. So yes, low-clocked results might be misleading in a way but if you take that into account, you can still make a decent judgement on kit quality rather than shelling out hundreds of dollars/euros on forums/ebay for kits without a single listed result.

What comes to the other ICs - from what I've experienced so far on different kinds of Samsung and Hynix, you get a bigger voltage difference going from board to board than from going from Pentium to a 5GHz i7.

 

 

I have tested more Hasfails, than I care to post but, there has not been one retail or ES 4770K that I have had that can't run 2666 8-12-8-28 on air on the Asrock Z87 boards. Depending on IC quality, voltages usually run from 1.87v -1.94v on most of my kits but I do run much tighter than profile on secondaries and tertiaries so it's to be expected.

 

Timings are obviously a huge part along with IC quality and namely IOA and IOD and VCCSA. If the balance is not right between everything then nothing seems to work right.

 

I have never tested any pentium/celeron imc's just the i3 4340 I have and it doesn't really like much more than 2666 with PSC or BBSE (air or LN2), doesn't matter. I think a lot of this has to do with the fact that bclk must be pushed on non K cpu's like the 4340 and that hinders mem oc on the top end.

 

Regardless I am going to stick to my binning method which is simple, buy the highest binned ram you can afford and then precede to test under LN2