Hashwell-e reviews

you are right, this is not a giant step up, it is gargantuan.

not only we have an increase in core count, but also in IPC. I see hints of better parallelism and inter-core data speeds as well:

CineBench R15 (CPU Multithreaded Performance):

Intel Core i7-5960X @ 4.0 GHz: 1508
Intel Core i7-5960X @ 3.0 GHz: 1270
Intel Core i7-3930K @ 3.3 GHz: 1096
Intel Core i7-4770K @ 4.4 GHz: 822


Also gone are the fears of a lousy overclocker, since 6.2GHz on LN2 is not too shabby.
 
Actually appears to be exactly as expected...

It will outperform 2 4770k's at the same clockspeed.
 
I really like that OC at only 1.2V, gives a little bit of headroom to squeeze another 100 or 200 MHz out of it (it seems consensus is that 1.28-1.3V is the safe stopping point before extreme chip degradation and lifespan shortening will set in for Haswell) - all on a H110 too at 55C. I'm sure Prime or Aida tests will run hotter but still, not bad for having 8 cores under the hood. Gargantuan indeed when compared to what SB-E to IB-E did...which was damn near nothing.
 
Papaj said:
(it seems consensus is that 1.28-1.3V is the safe stopping point before extreme chip degradation and lifespan shortening will set in for Haswell)
Click to expand...

Consensus based on what?

"Extreme chip degradation"? Having to bump vcore by a bit (~0.05v) a few days after stressing your new OC is normal. Unless you are pumping dangerous levels of voltage (>1.52v), or you have insufficient cooling, the degradation rate isn't anything to worry about.

Max safe voltage hasn't really changed since SB. 1.52v if you can handle the heat. This is specified by intel in their spec sheets, not some tribal knowledge. I've done a fair bit of research on this mark - anyone who's damaged their processor from voltage alone was using at least 1.6v, Nehalem architecture to now. If you have proof showing damage occurs lower than 1.52v i'd love to see it. All I've seen over the years is conjecture.

What HAS changed in this generation (Haswell-E) is we are now back to solder. So the problems with IB and Haswell not being able to transfer their heat out of the die due to shitty thermal paste is no longer a problem. Most people are not able to raise their voltage higher than 1.35-1.4 under water on IB/haswell because of the thermal paste problem. This is where the chip deaths come from - people not respecting the thermal limitations of the crappy paste, or damage/poor contact from delidding.

Solder makes everything cooler (if you have a beefy cooling setup), which makes everything more stable. Higher voltages are safer with solder over paste because you wont have as extreme of spikes in temperatures.


A good haswell overclocking guide with tons of data:

http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics

And a fearmongering thread about haswell degradation.

http://www.overclock.net/t/1409797/the-haswell-death-degradation-thread

The people blowing up their chips in this thread is from their own stupidity from what i can see. Not enough cooling or they damaged it while delidding.


Now, all that being said I probably wouldn't want to get near 1.52v on haswell-E unless I'm using a custom water cooling loop or better. Air cooling or AIO water, probably 1.4-1.45v max.


Edit:

Another thing to keep in mind is overclocking is an equilibrium between voltage, frequency, and temperature. If you are using too much voltage for a given frequency (like many have that killed their haswells) it can cause damage if left that way for an extended period of time. Using too much voltage causes excess heat, which leads to requiring EVEN MORE voltage to stay stable at those higher temps.

Many in that "death" thread are using way too much voltage. 1.44v for only 4.4ghz for instance (4.4 should only take ~1.2-1.3v with sufficient cooling). Or trying to cool 1.52v/4.8ghz with only a H110 (4.8ghz should only take ~1.4v with sufficient cooling). I'm just talking theorhetical safe maximum if you 100% know what you're doing and have a beastly cooling setup. If that isn't you, keep it <=1.4v.

Edit 2:

The haswell degradation HARDOCP reports here seems a little knee jerky to me: http://www.hardocp.com/article/2013...ro_lga_1150_motherboard_review/7#.U_9K3vldUdk

4.5GHz @ 1.2v is an insanely low voltage for that clockspeed. Most people require ~1.3v for that speed. I don't think what Kyle noticed here is anything outside the realm of that initial post-break-in voltage bump requirement.
 
Last edited:
Perhaps it was a poor choice of words on my part throwing the 'extreme' in there, as 1.3V won't imemdiately or even within 2-3 years fry a chip. It might take the lifespan if ran 24/7 like that to 4-5 years down from 7 or 10 years, but that'd be the extent of it. But yes, I agree its a balancing act and that the people in the death thread were doing some really stupid shit voltage wise - I however plan to keep my 22nm stuff into the next decade so my personal preference is to keep it 1.28V or lower and I know a couple MoBo manufacturer's agree...otherwise why limit their auto-tune OCing software to that voltage.

I'm sure nobody here discards dielectric breakdown or the hysteresis effects as being false, but I'm the "seek the best long-term bang for the buck" type guy, so if keeping it under 1.3V is going to net me a couple years of useful life for a loss of only a couple hundred MHz,I'll seek that route instead of spending money on custom cooling loops and having the inconvenience of dealing with them every so often.
 
Fair enough. Not trying to tell you what to do with your money! :D Just throwing it all out there for others like me, who are sitting on way too much cooling power waiting for the 5960x to arrive... :) God bless solder.
 
Zomoa said:
God bless solder.
Click to expand...

Quoted for truth my friend, you don't need a several hundred dollar cooling setup to appreciate a good solder job. If that cooling power keeps that chip at 1.4V happy for the better part of a decade I'd be a fool to bash it. But for us mere semi-[H]ard folks, being Mr. Cheap & Easy with an H110 and keeping the volts down makes us more than happy :D One day I'll join the [H]ard crowd with my U-tube HX primary/secondary setup I'm mulling over the design for (think nuclear PWR) - but that's not in my X99 plans :(
 
Wow, an unlocked 8-core....:rolleyes:

Thanks for passing this bleeding edge technological achievement to us Intel. You could have done it 30 months ago by unlocking the E5-2687W (and even up to 34 months by offering an un-neutered i7-3960X), but it's only a mere three year wait...:rolleyes:

And even better, you locked up the E5-2699 V3 good and tight...wouldn't want to give us grubby overclockers any reason to buy a $4k processor or actually provide us with a truly extreme piece of hardware now would you??:mad:
 
You must log in or register to reply here.
Back
Top