3770K Good/Bad/Average?

[bwrlane]

So I got myself a 3770K, upgrading a 2700K, and would like to see how my overclocking experience rates.

Yes, I know from a performance perspective, there was absolutely no point in upgrading a 2700K. I have advised others of such on this forum. I just get this itch every few months and want to play with something new. I guess it's what having a hobby is all about.

After a couple of hours experimenting, I've kinda settled on an all-4-core overclock of 4.74 GHz. A higher overclock either blue-screens when I fully load the CPU or, with a higher voltage it runs stable, but not at temps I'm comfortable with.

How does that rank - good/bad/average? From my basic research it seems about average - noting that all 4 cores are running at this speed.

This compares with 4.84 GHz (again, all 4 cores) that I settled on with my 2700K.

Running Cinebench and Core Temp simultaneously I get the following results:

- Ivy Bridge at 4.74 GHz: During the test - VCore 1.328, temps hover around in the low 70s, score: 9.47

- Ivy Bridge at 4.84 GHz: During the test - VCore 1.4 (or thereabouts, didn't take a note), temps mid 80s, score: 9.65. NB this was experimental and once only. I'm not really comfortable with VCore of above 1.4 for Ivy Bridge or temps in their 80s. Nevertheless, it is interesting to compare with Sandybridge at the same clockspeed.

- Sandy Bridge at 4.84 GHz: During the test - VCore 1.46 (or thereabouts), temps mid 70s, score: 9.37.

Interesting that my clock-for-clock comparison yields a 3% performance advantage for Ivy Bridge.

This validates what most people are saying about IB vs SB. No point upgrading from a performance perspective. Achievable overclocks are higher with SB than IB, wiping out most of the already very small performance advantage.

My cooler is a Noctua N-DH 14

 

[bwrlane]

Seems pretty good- My 3770k is at 4.6 with a utility in windows that I have a profile for 4.7. Seems like the cpu temps not that bad but the core temps seem pretty high. I will definitely be in the high 70's on a stress test at 4.7. I have a an all copper tru-best air cooler I've ever used-they don't make em any more

Thanks: what do you mean "Seems like the cpu temps not that bad but the core temps seem pretty high" - I didn't mention core temps as distinct from CPU temps - just read the temperatures from the Core Temp app.

 

[bigj1985]

Y I debated the dame thing as you my friend. I currently sit on a 2600k I sit confortably at 4.5ghz only because there is no point whatsoever to go higher. Heck to be honest the performance of this CPU is so good there is pretty much no reason for me to even bring this thing past 4ghz. I just can't help myself.

I debated the 3770k just to have a new toy to play with. But I really love my 2600k so I wouldn't want to sell it; and and $300+ just for a new overclocking toy is expensive. If the IPC was improved to the point where 200mhz didn't closer the gap on any performance gain I would have jumped. But as it stands I'll be waiting for HasWell

 

[bwrlane]

Y I debated the dame thing as you my friend. I currently sit on a 2600k I sit confortably at 4.5ghz only because there is no point whatsoever to go higher. Heck to be honest the performance of this CPU is so good there is pretty much no reason for me to even bring this thing past 4ghz. I just can't help myself.

I debated the 3770k just to have a new toy to play with. But I really love my 2600k so I wouldn't want to sell it; and and $300+ just for a new overclocking toy is expensive. If the IPC was improved to the point where 200mhz didn't closer the gap on any performance gain I would have jumped. But as it stands I'll be waiting for HasWell

I know what's going to happen. I'll be saying I'll skip Haswell, and advise other to do likewise. It's nowhere near the step change in performance that would really make a difference, and make an upgrade a rational or logical thing to do. But I'll probably get the itch in a few months and get it anyway.

A slightly higher overclock, a slightly better IPC... these seem to be all that we can hope for these days, which is a shame. Plus, as you say, what we already have is fast enough for just about anything. Where it would make a difference of course, video editing, you really need to go 6+ core for any gain to be noticeable/tangible. Which leads to my next gripe...

I really hate the way Intel is dealing with its e-line of CPUs. They're a generation behind and don't include innovative features like quicksync. I'd get Haswell-E in a heartbeat if a 6+ core variant were available from the start, and had a way of using quicksync.

 

[IntelEnthusiast]

Most people dont understand that the our socket 2011 are our server processors (they can be Intel® Xeon® or Intel Core™ i7's) and they will have a longer life cycle than are normal desktop products. The general rules that desktop have a 1 year cycle while servers have a 2 year cycle. We have been runing this way since the start of our tick tock model.

 

[bwrlane]

Most people dont understand that the our socket 2011 are our server processors (they can be Intel® Xeon® or Intel Core™ i7's) and they will have a longer life cycle than are normal desktop products. The general rules that desktop have a 1 year cycle while servers have a 2 year cycle. We have been runing this way since the start of our tick tock model.

Don't get me wrong Christian, nobody's saying that Intel don't have the right to release exactly what they want, when they want. Of course they do: that's the essence of the free market and a free society. It's just frustrating for enthusiasts who see that there's technology out there, which isn't being optimally used. A 6 or 8 core socket 1155 or 1150 CPU would go some way towards alleviating this concern. It's the kind of thing that makes ordinary buyers like me, wish there was a little more competition at this end of the market.

As for the 2 year life cycle, this is all very well, but I understand that the plan is to replace Sandybridge-E with Ivy Bridge-E rather than Haswell-E. Clearly such a scheme isn't sustainable as it would see server parts getting further and further behind desktop parts.

What's kind of interesting is that Tom's Haswell benchmarks suggest that for some multithreaded applications, Haswell is almost exactly half-way in between the performance of a 4-core Sandybridge and 6-core Sandybridge-E. This suggests that a 6-core Haswell would deliver a level of performance that, to say the very least, would be interesting.