Core i7-4790K Overclocked to 7003.38 MHz on ASRock Z97 OC Formula

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

CaptainTom

Honorable
May 3, 2012
1,563
0
11,960


Man it's a long list: BF3 & 4 , BFH, Crysis 3, Metro: Last Light, Watchdogs, and literally every big game that comes out these days. Go read the review of the new unlocked pentium. It was overclocked to 4.6 GHz and still narrowly lost to an i3 clocked substantially lower. Hyper-Threading gives a 50% boost in performance when used. But yeah go ahead and use double the energy instead lol.
 

CaptainTom

Honorable
May 3, 2012
1,563
0
11,960


Or trying to double justify their purchase of an i5 instead of an i7. I mean it is $100 cheaper for a reason people so don't be surprised in 1 year when your 290X CF becomes massively bottlenecked...
 

mapesdhs

Distinguished


Oc'ing the i5 can make up for the disparity to some extent of course, but yes there comes a point where
the gain from HT would help, and then the i7 can be oc'd too. 5GHz guaranteed = ASUS M4E + 2700K
(I've built five of these combos so far).

Having more proper cores would be even better though. That's what I meant by X58 not historically
feeling like an extreme high-end chipset because of the availability of several 6-core options, and
since oc'ing was mainly by bclk, lesser cost chips such as the i7 970 meant access to 6-core
wasn't crazy expensive.

Given the CPU loading of modern games, Z97 deserves a 6-core option IMO. No reason why
Intel couldn't do that.

Ian.

 

CaptainTom

Honorable
May 3, 2012
1,563
0
11,960


I would agree with you on that for sure. All they would need to do is remove the uses integrated graphics and place to more cores in their place. The TDP would probably be ~125-140w but who cares?

No they just want you to buy their crazy old "Premium" set-ups. The fact that their Ultra-High End line is always a year behind Architecturally puts them in direct competition with their cheaper and usually equally powerful lines.
 

mapesdhs

Distinguished
Almost as if they create their own product crossover conflicts. Sure, a 6-core S1150 chip would
conflict with X79/X99 options, but that'd be entirely Intel's fault since they should have moved
X79 onto 8-core anyway (and let's face it, the 3930K was an 8-core, just with 2 cores disabled).
IMO HW-E should be 8 or 10 core, not 6 or 8. It's a whole step behind.

Even worse for those I've been talking to, I fear the max RAM will still only be 64GB, which isn't
enough these days for various pro tasks.

Ian.

 

InvalidError

Titan
Moderator

HT on the P4 worked fine if you had software that actually used more than one active thread or were multi-tasking - on my P4, I was getting 20-30% extra performance from HT.

Intel's Haswell CPUs have twice as many execution units as my Northwood had and in some benchmarks, the i7 gains 50-70% extra performance from HT, which quite impressive.
 

game junky

Distinguished
So, the difference between the 8 and 9 series chipsets is more significant than a "meh". Guessing this is due to the new TIM. Going to pump the breaks on a new build until we see the next proc with using the Z97 boards but this is impressive
 

mapesdhs

Distinguished


Unfortunately, in other cases HT did slow things down (I tested this extensively with my Dell 650, dual-P4/2.66),
but even worse was the fact that so many P4 systems had rather slow RAM (266), a single-CPU Athlon64 with
400 RAM could easily beat a dual-P4, eg. see (tables later down the page):

http://www.sgidepot.co.uk/misc/mysystemsummary.txt

For 3D tasks (tested with Oblivion), it wasn't too bad at medium detail, but at high detail performance would
take a nose dive on the P4 Dell, and check the general tests, the dual-P4 is far behind. The HT tests I did
are not mentioned on that page, but I do remember it only helping in a few cases, hurting in others, mostly
the latter. Mind you, as someone pointed out, this was with XP Pro so perhaps the OS didn't help matters.

Ian.

 

InvalidError

Titan
Moderator

Everyone knows the P4 was a failure (for the most part) so comparing it with something else to try making a point about how HT helps or hurts performance in any given benchmark is pointless. It can only be compared to itself with HT on and off to make that determination. By the time Northwood with HT rolled out, DDR-400 was practically standard so DDR-266 for people who remotely cared about performance does not really apply.

In any case, we are not talking about HT from 10 years ago; we are talking about how much HT improved since then and today, there is practically nothing to gain from disabling it but a lot more to lose than before.
 

mapesdhs

Distinguished
InvalidError writes:
> Everyone knows the P4 was a failure (for the most part) so comparing it with something else to try
> making a point about how HT helps or hurts performance in any given benchmark is pointless. ...

It's not pointless at all, partly because the comparison shows the reality of what was going on at
the time (HT a waste of time since rival options were quicker overall anyway) and partly because
as I say I did do lots of HT tests. Clearly you've chosen to believe I'm lying just because that
file doesn't have the data. Feel free to believe that if you wish, but such a notion isn't remotely my
MO as anyone who knows what I do can attest.


> ... DDR-400 was practically standard so DDR-266 for people who remotely cared about performance
> does not really apply.

In numerous cases systems back then did not support 400 speed, so the idea that 266 data is
not relevant is just silly. That was the reality for numerous machines at the time. It's important
because I found that memory speed made by far a greater difference to overall performance
than even the best possible usage case of the P4's HT. Back then, there was a lot of marketing
pushing the HT aspect of the P4, when in reality customers were much better off just getting
something with the fastest possible RAM, and by the time I was doing those tests, an AMD. The
Athlon64 system I built for my brother left my dual-P4 Dell in the dust.


> In any case, we are not talking about HT from 10 years ago; we are talking about how much HT
> improved since then and today, there is practically nothing to gain from disabling it but a lot more
> to lose than before.

Yes, I entirely agree; I think the point some were making earlier is precisely that many have an
attitude towards HT today that's based on how poor it could often be many years ago with the P4.
They don't realise the tech itself is a very different design now, but it helps to understand why some
assume it can't be good for modern gaming, and hence help inform their opinions, update their info,
rather than just yell at them. :}

Ian.



 

InvalidError

Titan
Moderator

I am disregarding what you said because benchmarking a P4-HT against an Athlon proves absolutely nothing about HT that cannot already be explained by Netburst already being a lackluster architecture regardless of HT. Also, I do have a computer with a 3GHz Northwood in my inventory and did my own HT on/off benchmarking back then. That's where I drew my 20-30% performance gain while multi-tasking from. That may not be a lot but it did make the system noticeably more responsive while heavily loaded.


Northwood launched with the i865 series chipsets along with FSB800 and DDR-400 support. Anyone who cared to get the most performance out of their Northwood would get the FSB800-based model with i865 chipset and DDR-400 memory just like anyone who cares to get the most performance out of their Haswell today (without overclocking) would get a 9x series chipset and 1600MT/s memory.

Netburst was bad enough even under optimal circumstances, no point in adding unrealistic artificial handicaps. Only people who did not give a damn about performance would have paired a Northwood (or Prescott) with FSB533 and DDR-266... that's a completely unnecessary ~30% performance handicap right there.
 

The_Trutherizer

Distinguished
Jul 21, 2008
509
0
18,980
I'd like to know myself. How many hardcore OC'ers can there be in the world. Personally I've stopped buying K processors. I've only really ever done auto tuning kind of overclocks. And they've never made a real difference in my experience. Faster and more memory, SSDs, things like that... That's where I have seen real perceptible changes to my computing experience.
 

InvalidError

Titan
Moderator

Overclocking had nice benefits in the P3 days where many CPUs could be overclocked by 30-100% but with modern locked chips where a large chunk of the overclock headroom is already skimmed off the top by forcing people to buy K-chips which already are at the top of available speed bins, the leftover headroom is only 10-20%.

By the time I start feeling like my computer is too slow, even an optimistic 20% overclock would be unlikely to change that in any meaningful way.
 

leeb2013

Honorable


I've not seen that since switching from a I5-3570k 4.4ghz to a xeon E3 1230 3.4ghz. Xeon has much greater performance in recent aaa games.

 

agentbb007

Distinguished
Jul 27, 2006
291
3
18,815
Wow impressive, yes this does influence my buying decision. It tells me the 4790k is a great CPU to overclock and tells me the ASRock mobo does a great job as well.
 

Duckhunt

Honorable
Sep 22, 2012
339
0
10,810
R.I.P. normal electricity bill :(

There is your problem. The electricity bill goes out the window and then no more happy campers. I just wish they would work fitting or seating a heat sink to pull out more heat efficiently and the only way i can think of that happening is to use the case.
Use the desktop and make it more efficient then the laptop because of the space you have.
 

klepp0906

Distinguished
Apr 29, 2013
150
0
18,710
Wow, their is SO much misinformation in this thread. Toms hardwares popularity is it's own worst enemy. Chock full of retards.

Hyperthreading sucks?!? Wut?

Is it stable I wanna see some benchmarks (regarding 7k ln2 oc) srsly?!

I'm afraid the haswell e dram limit is going to be 64gb which is not enough even now .... Deeeeerp /sigh

It was at this point I stopped reading. Statistically that's likely only half the gems i would be able to pick out. Ignorance is bliss. Not sure who said it, but the guy was a frigging genius!
 

InvalidError

Titan
Moderator

DDR3 has a limit of 8GB per DIMM and two DIMMs per channel while DDR4 has a limit of 16GB per DIMM and one DIMM per channel. I imagine LGA2011 would be a fair bit less popular if it required slower, more expensive, server-oriented buffered DIMMs.

LGA2011 CPUs are not meant to replace the Xeon E5/E7 for serious workstation/server-style workloads
 

akula2

Distinguished
Jan 2, 2009
408
0
18,790
'When you buy a new motherboard, does this kind of information sway you in favor of the motherboard itself, the vendor, or does it not affect your buying decision at all?'

No, my business requirements drive me in planning for new Hardware.

My latest high-end horses (9) based on Z97 WS boards churning loads of work under VMs; three 27" displays for each machine.

OC is NOT preferred at my work places (policy), hence 'K' bad guys sit at my home! Oh BTW, I'm planning for a dozen more 'Pentium-K based' office PCs but pondering on the Z97 or H97 boards. Most likely MSI Z97 PC Mate (bundle offer).
 
Status
Not open for further replies.