So, Ivy Bridge sucks at OC'ing

[blackbelter]

I am in a dilemma now. Ivy Bridge runs faster and cool at stock voltages and frequencies, but it sucks once oced. Should I go with sandy bridge, and oc all the way? What kind of performance can I expect from stock Ivy Bridge? Is oc absolutely essential, to Ed sims at least?

 

Need your comments guys... Thanks.

[EtherealN]

How do you mean sucks once OCed? I'm seeing Ivy's doing 4,9 quite happily at 1.35 volt. Sure sandy can go slightly higher, but IMO 4,9 is already past the OC sweetspot anyhow.

 

Overclocking is not "absolutely essential". Most of the time I run my sandy without OC since it plain isn't needed. As for expected performance from a stock Ivy Bridge? The best performance of any stock-configured machine available on the consumer market. :P

[blackbelter]

How do you mean sucks once OCed? I'm seeing Ivy's doing 4,9 quite happily at 1.35 volt. Sure sandy can go slightly higher, but IMO 4,9 is already past the OC sweetspot anyhow.

 

Overclocking is not "absolutely essential". Most of the time I run my sandy without OC since it plain isn't needed. As for expected performance from a stock Ivy Bridge? The best performance of any stock-configured machine available on the consumer market. :P

 

4.9 quite happily at 1.35 volt? According to the review that I have read (http://www.anandtech.com/show/5763/undervolting-and-overclocking-on-ivy-bridge), I imagine that the temperature will be terribly high, even with liquid cooling as was the case in the review.

 

I am sticking with aircooling.

 

My actual question is, with and without OC, will there be a substantial difference in performance in ED sims? How much actual fps can you squeeze out by OCing? Mind you that I just got a copy of Gigabyte 680 Windforce 2GB to go with my new rig. Since I intend to play it 3D, I will benefit from any amount of fps increase.

 

Thanks.

[EtherealN]

http://techreport.com/articles.x/22833

 

As for performance with and without OC, there is no straight answer. Assuming there are no other bottlenecks (the 680 should be good for that), then sure. But whether you'll notice depends on missions and situations. The most rediculously over-filled and complex missions - yes. But I'd say 95% of the "pay" happens when OCing into the 4,5-4,6 area, after that you start paying more in wear-and-tear etcetera than you get in performance.

 

And, as I said, it's fairly fine on stock.

 

3D will be almost entirely GPU-bound.

Edited April 24, 2012 by EtherealN

[blackbelter]

Inconsistency in the reviews probably can be traced back to inconsistently manufactured copies... Got to wait a little more then.

[EtherealN]

Well, there always is inconsistency in manufacturing, that'll never go away. I remember seeing reviews of i7-950's that did 4GHz while undervolted, others needing overvolt to do anything at all. There's always some play and always a bit of a "lottery" going on simply due to how these things are manufactured.

 

Still, being careful on a first launch is generally a good idea, yeah.

[GregP]

The benefit of CPU o/c'ing in DCS is, as EtherealN mentioned, very much system-dependent. What I can say, however, is that based on my experience as detailed here, with a single screen at 1920x1080 in an example mission my FPS increased 14% by going from a stock 3.3GHz up to 4.8GHz, while using 3 screens at 4680x1080 the same overclock increased my FPS 8%.

 

In other words, nowhere near a linear improvement in performance with overclock, but essentially a 'free' increase with no downside other than increased power consumption and heat generation, which are both probably non-issues for most of us --especially given how just about everybody, regardless of system, seems to be wanting more FPS in DCS.

[Pilotasso]

Its a question of luck with the manufacture batch, which can happen with either series 2000 or 3000.

My I5-2500K is overclocked at 4.5Ghz with 1.24V better than any of those tested 3770K CPu's :D

 

If you ask me these reviews all show me that its better to save the money and get a unit of the former, saving 100 bucks. The margin of perfomance can be easely covered by the overclock you might (or not) reach on either.

 

Simply put: bet on the cooling and buy the cheapest unlocked CPu you can get.

 

On another note, the reviews scientific rigour is just barely short of fraudulent. They used unlocked CPU's on default speeds to obtain an average score using sinthetic benchies running 8 threads with games that only use 4 or less.

 

by that token LGA 2011 would be better for the overall consumer because the average score would be higher with 6 cores even if slower in games! :lol:

Edited April 24, 2012 by Pilotasso

[blackbelter]

Thanks guys for the information!

[Sov13t]

I believe this quote from Anandtech summarizes Ivy Bridge pretty well:

 

While it's not enough to tempt existing Sandy Bridge owners, if you missed the upgrade last year then Ivy Bridge is solid ground to walk on. It's still the best performing client x86 architecture on the planet and a little to a lot better than its predecessor depending on how much you use the on-die GPU.

 

I, for one, am actually due for an upgrade, so will be playing around with Ivy sometime next week.

 

Source: http://www.anandtech.com/show/5771/the-intel-ivy-bridge-core-i7-3770k-review/22

[Pilotasso]

I might Haswell wait for the next generation :D

[Sov13t]

I might Haswell wait for the next generation :D

 

Keep in mind though, that OCing in general is going to become less and less viable with these newer architectures - as we are seeing with Ivy Bridge. Their thermal density keeps increasing, as well as their overall efficiency... so the need to OC will die out.

 

Now, of course clock speed will still matter for poorly coded applications... but let's hope developers continue pursuing perfection in their art.

[EtherealN]

Overclocking isn't about "poorly coded" applications.

[Pilotasso]

Keep in mind though, that OCing in general is going to become less and less viable with these newer architectures - as we are seeing with Ivy Bridge. Their thermal density keeps increasing, as well as their overall efficiency... so the need to OC will die out.

 

Now, of course clock speed will still matter for poorly coded applications... but let's hope developers continue pursuing perfection in their art.

 

They are not going to make any less efficient CPU's by shrinking the die. Whats hapening with Ivy bridge may have to do with the silicon layer tecnhology. Haswell should be faster or it wont sell.

[Sov13t]

Overclocking isn't about "poorly coded" applications.

 

:) That is not what I said. I said 'clock speed' - no mention of overclocking in the same sentence or paragraph.

 

They are not going to make any less efficient CPU's by shrinking the die. Whats hapening with Ivy bridge may have to do with the silicon layer tecnhology. Haswell should be faster or it wont sell.[/Quote]

 

Correct, what I said was they are making and going to make more efficient CPUs, thus the need to OC will become more and more extinct. Of course it has become an addiction to many, me included... but in general chips out of the box will run at lowest possible stable voltages with highest possible speeds +/- margin of error for safety of course.

 

And what is happening with Ivy Bridge, by the way? It is an undeniable step forward. Just from a simple thermodynamics perspective, the fact that it runs hotter should not be a surprise to anyone.

 

What is perhaps not as desirable to us is Intel forcing a GPU into the chip, would be cool if there was a separate chip line-up without the on-chip GPU - just look at how much space that HD 4000 gpu takes up compared to the cores... but, don't think we will see that.

[EtherealN]

Okey, then take my exact same question and substitute clock speed. My statement still stands.

 

Clock speed is a technicality, and a necessary one. (Though there are architectures that run asynchronously, but there's good reason why we don't see many of them.) Your post seems to indicate that clock speed is in any way linked to whether a given application is well coded or not. This is incorrect.

 

EDIT

Example to make my point: Is an automobile poorly designed because you go faster when you push the throttle? Errr... No. ;)

Edited April 25, 2012 by EtherealN

[Sov13t]

Each architecture is more efficient than it's predecessor - and efficiency != clock speed.

 

A piece of software that does not take advantage of multiple cores and threads and dumps everything on one core is a lot more dependent on clock speed than one that is optimized to perform in a multi-core/multi-threaded system. Otherwise we all would still be running Cedar Hills at 8ghz and would be 1337 kids.

 

Answer to your question... the driver is poorly designed for not utilizing the cars full potential.

 

And a counter-question to yours... a Yugo and a Ferrari are both traveling at 160mph... which car is better? :smilewink:

[EtherealN]

But that's not what we are talking about. Irrespective of which architecture we use, clock speed will ALWAYS be relevant to performance. If it wasn't, select a program you feel is "well programmed" and underclock your PC - 10MHz should be fine? Does it still run well? Of course it doesn't.

 

The thing one should never do is use clock frequency as a comparison tool between architectures, but that's a different thing.

 

Similarly, different applications lend themselves differingly to multithreaded operation. A perfect example is graphics - there's a lot of things you can do on a per-pixel basis where you don't really need to know anything about the pixel next to it. Thus, you could theoretically split it all up such that you run one thread for each pixel and then deliver the results to whichever stage eyou want to happen next. Simplified, that's exactly what modern graphics cards do.

 

Your quote:

"Now, of course clock speed will still matter for poorly coded applications..."

 

Is just plain wrong. Or well, it DOES matter for poorly coded applications, but it matters for well coded applications too. There is no difference between a well coded application and a poorly coded application, both will perform better at a higher clock speed on a given architecture. (But when swapping architecture you may be doing worse - for example, running Program X on a Sandy at 3GHz will often grant better performance than running it on a 3,6GHz Bulldozer. Again though - this applies to ALL software.)

 

Regarding Yugos and Ferraris - I'd say the Ferrari. I wouldn't want to hit something in the Yugo, and the Ferrari has superior stopping distance and handling. And of course, using your logic; in your example the Ferrari-driver is poorly designed for not utilizing the car's full potential. (Not that that matters for the example, but thought I'd point out your lack of consistency there. ;) )

 

Again, explain to me how clock speed is ever irrelevant to a well programmed application. You can't.

Edited April 26, 2012 by EtherealN

[Pilotasso]

Looks like we got the answer why Ivy bridge doesnt overclock like everyone expected:

 

http://www.overclockers.com/ivy-bridge-temperatures

 

 

ouch! A bit of a letdown. Until a few days ago I was hoping to upgrade as I expected this chip to hit over 5GHz. :(

[EtherealN]

Ouch indeed. A bit disappointed at that, but meh... It's money saved for me, I now don't have to struggle with my financial counscience to keep me from purchasing a new system right now. :P

[Pilotasso]

You wouldnt need to. The CPU would accept your Mobo with just a BIOS update, if you ever were considering it. Not even worth that IMHO. Your fine, and for me its a relief that i dont have to change anything either :D

[EtherealN]

True, but I would be tempted to build a PCIe3-capable system around it. Not that I "need" that, but it's geek bling... :D

[cichlidfan]

but it's geek bling... :D

 

Two good things. I also can stop thinking about upgrading for a little while and now I know what to call all the crap in my home office. Thanks EtherealN.!:D:thumbup:

[HiJack]

Good find there Pilotasso! Hmmm, will this make the new AMD come in to the light?

[EtherealN]

AMD still need a miracle, to be honest. :P

 

Remember, we're comparing Ivy to another intel product that is a year and a half old, NOT AMD's more recent stuff... That tells you something. And their recent restructuring doesn't make it look better for them in the enthusiast market. (They still have good stuff for the server and enterprise markets though.)