Intel's Core i7-4770K Overclocked to 5.0 GHz on Just 0.904 V

Status
Not open for further replies.

eklipz330

Distinguished
Jul 7, 2008
3,033
19
20,795
"It is very possible that once a load is applied, the CPU destabilizes and the system crashes."
the computer would have to go under load when loading up windows, even if it is for a short while... i wonder HOW stable it is... if this is true, and the person got it to 5.0ghz, an upgrade from ivy bridge may actually seem worth it!
 
That is impressive. So if .9v at 5ghz is fully stable(lets assume) how much higher can it go with more voltage is the question.

I really want to see the temperatures of these things.
 


They mentioned a limit of or around 8Ghz because of the multiplier limitations.
 


Interesting, well I just hope this is true, this chip can have a lot of overclocking potential.
 

hakesterman

Distinguished
Oct 6, 2008
563
0
18,980
They mentioned an limit of or around 8Ghz because of the multiplier limitations. - Try about 5.5 GHZ Limit. And that may not even be totaly stable in all applications.
 

khaledegy200

Honorable
May 8, 2013
239
0
10,690
just a stupid calculation:-
1.maximum recommended volt/this one 1.5/0.9 = 1.66666666
2.1.666666 * 5=8
it looks like 8Ghz is really the max clock for this epic CPU
 
G

Guest

Guest
Question is, is it worth an upgrade for us that are still on Sandy Bridge (2600k) from 2 years ago I want to see that comparison.
 

InvalidError

Titan
Moderator

I bet those overclocks are done under idle condition and since Haswell has 10X lower idle power than Ivy Bridge, this means the CPUs will be at close to ambient temperature.

I am expecting most of those overclocking results to get thrown out the window as soon as you enable all cores, HT (where applicable) and give the CPUs an actual workload.

 

InvalidError

Titan
Moderator

~7% gain from SB to IB, ~10% from IB to Haswell so apart from software optimized for new instructions, we are talking maybe 20% more stock performance.

If I were you, I would wait for Skylake in about two years but even then, we may still be talking about only a ~50% gain.
 


it would be roughly a 17.7% gain clock per clock, but if you really can push 6+GHz stable then you are talking about a huge raw performance boost. Lets say that you have a SB @ 4.5GHz which is a normal and respectable OC without going crazy. An OC to 6GHz would be a 35% boost alone, plus being faster clock per clock of an added 17% would mean over a 50% performance boost in only 2 generations... which is crazy and rather unexpected.

Personally I will be waiting for the Sky series. Mostly because I am not made out of money and cannot afford to upgrade every 2 years, but also because my i7 2600 is frankly overkill for everything that I do. I would love some bigger/faster SSDs, I really need a GPU with more onboard memory, and I could do with an upgrade from my DDR3 1333 to some 1866... but really, the bottlenecks of the system are in those support parts, not the CPU itself. There is no added CPU performance that would be noticeable to me, so I am simply going to wait until we start seeing SATA4, DDR4, and some other faster busses to become standard before updating, because I will notice those upgrades much more than the upgrade of the CPU itself.
 

timil

Distinguished
Jun 3, 2011
341
0
18,790
It's amazing to see these advances in architecture but I feel sad, it's like the days of enthusiast socket CPU overclocking is coming to an end.
 

groundrat

Distinguished
Dec 11, 2012
952
0
19,160
If stable, this bodes badly for the FX fan boys. Not only does Hazwell overclock, it does it on less than one volt! How much voltage does an FX chip eat when clocked up to 5 Ghz? I have seen an FX 8350 at 5.2ghz, but eating 1.524volts. Not a huge difference, but as intel gets cheaper and cheaper to run AMDs share of the market will continue to drop..
 

Soda-88

Distinguished
Jun 8, 2011
1,086
0
19,460


Last time I checked, voltage to clock ratio wasn't linear in the slightest.
 

ohhgourami

Distinguished
Sep 6, 2011
126
0
18,690
.904V is the voltage taken at the FIVR! That is not VCORE. Get things straight Tom's Hardware! Stop spreading misinformation and posting 3 days old "news"!
 


So Tom's is wrong because it says "core voltage" and then says ".904v" and they say it has a vcore of .904v because of that?
 

InvalidError

Titan
Moderator

If Haswell chips could be pushed that high that easily with all features enabled under burn-in conditions, I think Intel would have factored most of that headroom in stock clock rates in the first place and the stock lineup would be 3.5-5GHz instead of 2.9-3.9GHz. The whole point of the non-K lineup is to burn up most of the high-yield headroom with multiplier-locked chips and force overclockers to buy the highest-binned chips. It would make no sense to give overclockers over 1GHz of extra headroom if yields really were that good.

Haswell is built on the same process as IB, has a wider execution pipeline, a deeper register file and deeper OoOE buffers all of which make it a substantially more complex chip and means almost certainly more combinational logic and propagation delays between DFFs in those critical blocks. In principle, those should actually make Haswell a worse overclocker than SB/IB under real-world scenarios.

That's why I'm saying most of the ridiculous overclocks - assuming they are real - are likely nothing more than artifacts of Haswell's exceptional idle caracteristics.
 

twelch82

Distinguished
Dec 8, 2011
182
0
18,680
I wish I could get excited about some new hardware, but a 4770K @ 5GHz would be a sidegrade from my 2-year old 3930K @ 4GHz.
I wish Intel was more ambitious with their improvements. If they've hit a wall in clockspeed, how about at least finding ways to give us more cores? Give me a 16, 32 core chip, and I'll write some threaded software to run on it. Deal?
 

mortsmi7

Distinguished
Mar 10, 2010
354
2
18,790
Everyone keeps talking about the theoretical 8GHz limit imposed by the multiplier. Am I the only one that remembers reading an article about the Haswell base clock being adjustable without screwing up the other internal clock rates?
 


The 8GHz limit is a factor of that; if you increase your base clock, your maximum multiplier reduces so that you can never get more than 8GHz total... which means that having an unlocked base clock is pretty much pointless.
 
Status
Not open for further replies.